Australia’s Two Great Barrier Reefs: What Can ~360 Million Years of Change Teach Us?

Abstract

Coral reefs are among the most important marine habitats but face significant threats from anthropogenic sources, including climate change. This paper reviews and compares the modern Great Barrier Reef Province and the 360-million-year-old Devonian Great Barrier Reef of western Australia. Despite occurring at times with different climates, biota (both marine and terrestrial), weathering processes and marine chemistry, similar reefs were constructed under certain circumstances. Major differences in global temperature, marine carbonate saturation, sea level behavior and reef community constituents were evaluated. The comparison highlights the integration of, and interdependencies within, reef communities and the need for both carbonate producers and significant binders, whether skeletal or microbial, to construct a reef in a high-energy setting. Devonian communities with abundant corals and skeletal sponges were incapable of making modern reef types without competent binders to unify framework into rigid substrate. The current strong focus on corals and bleaching in modern reef conservation may be obscuring the equally significant issue of ocean acidification, which impacts on equally crucial framework unification, i.e., hard binding by coralline algae and microbialites and early cementation. The comparison also supports the idea that ‘empty bucket’ carbonate platform morphologies require increased accommodation from high-amplitude icehouse sea level oscillations.

1. Introduction

Coral reefs, despite their small area of the globe, are among the most biodiverse environments on Earth, potentially harboring as many as 25–33% of all marine taxa [1]. At the same time, coral reefs provide many important ecological services (ESs), including fisheries/food security, building materials, recreational centers, opportunities for economic development, and direct coastal protection, for as many as 500 million people in tropical regions of the world [2,3,4,5]. Significant economic value accrues from these services (e.g., [6,7]). Reefs and their habitable islands also serve as, and support, important cultural heritage for many indigenous populations (e.g., [8,9,10,11,12]). The great value of, and now payment for, reef ESs have been the focus of growing attention owing to the many threats faced by coral reefs from predicted climate change [13,14]. Coral bleaching from marine heat waves [15], increased storminess [16,17] and ocean acidification [18,19] threaten reef growth along with the services they provide. Previous anthropogenic changes to reef environments ranging from indigenous fishing and gathering [20] to regional industrialization and declining water quality from proximal land use change already have impacted significantly on many of the world’s reefs [21,22,23].

A major contributor to coral reef ESs, especially where coastal protection is concerned, is the characteristic of reef communities (including ancient reef communities that lacked corals) to be perhaps the world’s best natural engineers. These communities interact directly with local geography to modify proximal environments and construct rigid structures at sea level that directly support and benefit their constituent communities. These rigid, relief-bearing structures, defined by reef geomorphology, protect adjoining lagoons and coasts and thus uniquely support a variety of other marine and coastal ecosystems. Reef growth dynamics have been of great interest to ecologists and geologists since Darwin [24] and have been studied in a comparative sense (modern reefs as analogs for ancient reefs, and ancient reefs as analogs for the longer-term drivers of, constraints on, and anthropogenic effects on modern reef formation) since that time [25,26,27,28,29,30]. As builders, reef communities were the first life on Earth to create elevated structures large enough to be visible from space. That said, no two reefs today are exactly alike owing to differences in latitude, local climate, bathymetry, tectonic setting, basement substrate, local oceanographic settings, etc., and, when geological time is added, even marine and proximal terrestrial biological constituents and marine chemistry have changed significantly. Regardless, reefs in the ancient past commonly bear striking similarities to modern reefs in terms of geomorphology and, where studied in detail, even their local community structures and the distributions of morphological types and functional roles of organisms across different reef environments share many similarities [31,32,33,34,35,36,37,38,39].

As reefs are very sensitive to both climate and sea level, understanding them is even more important now with those factors projected to change owing to anthropogenic forcing. Additionally, as short-term emitters and long-term carbon sinks in the modern shallow ocean, reefs themselves play a role in the carbon cycle and their carbon dynamics are important to understand in that regard alone [40,41,42,43]. Understanding the developmental history of Quaternary reefs is one of the best ways to predict potential future effects of environmental changes that are now happening or are projected to happen (e.g., [25,28,30,44,45,46]). However, the broader constraints on, and patterns in, reef dynamics are relatively poorly understood and the immediate precursors of modern reefs are generally difficult to study. In areas of subsidence like the Great Barrier Reef Province (GBRP) in northeastern Australia previous (penultimate) reefs are buried beneath living reefs making them inaccessible without scientific drilling [45,47,48,49,50,51]. Elsewhere in the world where precursor reefs are not overlain by modern reefs, they commonly are uplifted and exposed to meteoric (freshwater) systems that destroy or alter original reef materials making them difficult or impossible to date and disturbing their environmental geochemical archives [52,53,54,55,56].

As reefs of the more distant past appear to have experienced many of the same types of oceanographic and climatic issues that face modern reefs, understanding how such reefs handled previous changes provides important information for forward modeling potential future changes in reefs [57,58] and possibly for predicting outcomes of some types of environmental change [26,27,29,59,60,61,62]. Although the need to understand broader reef dynamics through time is recognized, functional approaches to reef ecology (e.g., reef guilds as functional roles [32,33,63]; functional groups of herbivorous fish [64]; functional ecology of coralline algal crusts [65,66,67], coral community dynamics [68,69,70]) have lagged behind many other types of reef ecology and paleoecology studies [71,72]. Thus, the broader principles that guide reef communities through times of change are not adequately understood despite increasing comparative data being available [37,73].

Even the GBRP of eastern Australia, the largest reef complex in the modern world (Figure 1a), faces serious challenges [74,75,76,77,78,79]. The >3000 reefs of the GBRP contain many different reef types in different settings (onshore to offshore, higher or lower tidal ranges (meso- to macrotidal), greater or lower current velocities, differing water quality, etc. [80,81]). The latitudinal range of the GBRP is so large that the reef encompasses different climatic zones and the history of the reef itself differs significantly from north to south owing to the northward movement of the Australian Plate through those different climatic zones [82,83] (Figure 2). Few opportunities exist to investigate fossil reefs over such large and complex geographic distributions, but extensive reef growth during the Devonian Period provides possibilities [84,85]. Major Devonian reef provinces extended for more than 1000 km in many regions, peaking in the Middle Devonian [85]. Late Devonian reefs were less abundant but extensive reefs continued in western Canada (Alberta), NW Europe, NE Laurussia and eastern slopes of the Urals, South China and western Australia [84,85,86] (Figure 1b). Perhaps the best exposed Devonian reef for detailed study is in the Canning Basin, western Australia where the ‘Devonian Great Barrier Reef’ (DGBR) currently crops out over a lateral distance of 350 km [87,88,89,90] (Figure 1b). This ~360-million-year-old reef occurred mostly after peak global Devonian reef development (Givetian) but survived for more than 20 million years (Figure 2), through major changes in sea level and even one of the large mass extinction intervals that affected reef builders in particular [84,91]. The purpose of this paper is to review and compare these two major Australian reef provinces (the GBRP and DGBR) with their differing reef builders and settings to allow inferences to be made about reef building in general.

2. Operational Definitions—Methods, Perspectives, Concepts and Semantics

The present review is based on the literature and field work carried out on the Leonard Shelf between 1988 and 2010 and in the GBR between 1984 and 2024. The goal is to make the discussion accessible both to modern reef workers and paleontologists–geologists working on ancient reefs, but reef concepts vary widely between modern ecologists and geologists and even the definition of what constitutes a ‘reef’ or ‘reef framework’ varies between different practitioners [25,31,98,101,102,103]. Different reef semantics make better sense to different workers based on their background, experience, aims and scale of study [72,104,105]. The important thing, though, is to be clear about the semantics being used in an analysis so that the concepts can be directly compared, even if the terminology differs. For this review, a general definition is used [72,106] where reefs are defined as autochthonous structures that (1) have relief above the surrounding sea floor, (2) have adequate synsedimentary rigidity to withstand local wave action or potential wave action, and (3) were constructed by a specialized biological community that differs from that on the surrounding level (non-reef) sea floor. Although reef ecosystems today are dominated by scleractinian corals and crustose coralline algae (CCA), and corals are commonly considered the dominant and/or most significant feature of a reef, almost to the exclusion of other organisms [107], the living biology is but a thin veneer situated upon the substantial structure of the reef edifice itself. It was these massive structures that loomed up from deeper waters to provide hazards to shipping just below the ocean’s surface that first brought reefs to the attention of Western mariners (and no doubt many earlier Indigenous mariners) and then to modern science. The term ‘reef’ was most likely developed from an old Norse or German word ‘rif’, meaning a ridge or ‘Er Rif’, an Arabic term for submerged hills [102]. Using the present definition, a reef is a physical, self-supported edifice that includes both the so-called ‘framework’ and the particulate sediment that is protected behind it from wave energy (i.e., the shallow back-reef) along with the attached sediment that is gravitationally deposited on the reef slopes that helps create and maintain the relief. This definition avoids the difficult semantic task of differentiating a ‘reef core’ from the rest of the edifice, but regardless, there must be a demarcation between what is and is not part of the reef. Reef sediments transported away from the reef by currents or gravity (e.g., turbidites) to end up in different depositional settings, such as the deep sea, would not be considered part of the reef. In the case of modern linear, ‘ribbon’ reefs, reef sediments slope off the back of the windward reef margin and eventually intergrade with normal level bottom, non-reef sediments on the shelf. The demarcation could be chosen based on relief or by sediment source depending on the semantics chosen. For this paper, elevation above the level bottom is the key factor, but members of the reef-building community may intergrade continuously with those of the non-reef, level bottom community. In shore-attached platforms and fringing reefs, reef-derived sediments may grade into continental sediments at the same elevation. In that case composition would provide a necessary demarcation, but the elevation of even the continental sediments is supported by the reef rim and thus could be considered part of the overall edifice. Platform reefs are perhaps easiest to conceive, as they are elevations that are isolated on all sides by deeper waters, but every reef and reef system is unique in some way.

The thin veneer of living reef organisms makes these truly huge and impressive structures (and navigational hazards) given time. However, the nature of that ‘time’ has led to different perspectives being held by different reef workers [104] (Figure 3). In general, biologists–ecologists focus mostly on modern observations and instrumental records, but increasingly, written historical records are used to understand ecological processes on the surface of the reef. Geologists, on the other hand, more commonly study reefs internally as larger structures with longer histories. However, as our understanding of reefs has increased, the interface between ‘biological’ and ‘geological’ processes and time scales has merged [108]. The dynamics of reef environments through ecological time (e.g., life of a coral as demonstrated by annual banding [109], to intermediate disturbance regimes—years, decades or centuries [110]) have been well studied by modern ecologists. With modern radiometric dating [111,112,113] and geochemical proxies [30,114,115,116,117,118], such studies are being pushed back to multi-centennial and millennial scales and beyond. While whole reefs record the positions of sea level through time, coral skeletons and reefal microbialites can record geochemical archives of their local environments (temperature, salinity, pH, water quality, nutrients, etc.) from daily increments [119], to seasonal, yearly, decadal and millennial scales (Figure 3). Coral-based geochemical archives provide data on Holocene climate dynamics, such as El Niño Southern Oscillation (ENSO), Pacific Decadal Oscillation (PDO), and Indian Ocean Dipole, that are required for better climate modeling [120,121,122]. At longer time scales (10s to 100s of thousands of years; e.g., Milankovitch orbital forcing frequencies and icehouse–greenhouse climate phases), reefs have been studied more by geologists [25,26,39,46,123]. That work and interest in ancient reefs as hydrocarbon reservoirs in the subsurface [124] have focused most geological attention on older reefs. As a result, important reef-building processes, such as ‘lithification’ (turning reef products into rocks), are commonly considered the realm only of geologists and ‘geological time.’ However, construction of a rigid structure must occur on ecological timescales to provide adequate substrate for coral recruitment after disturbance [125,126,127,128]. Increasingly, the processes that control early lithification in reefs have been demonstrated to be biologically controlled or mediated over these shorter ecological time scales [129,130,131,132,133].

Although modern reefs are dominated by corals, CCA are crucial [72,134] and many other organisms contribute quantitatively to carbonate production [135,136,137,138] (Figure 4). Scleractinian corals are ‘keystone’ taxa in modern reefs and the diversity and range of morphology of corals directly supports much of the non-coral biodiversity for which coral reefs are noted. However, modern reefs generally are not composed mostly of corals preserved in growth position. Most reef corals are ephemeral and destined for destruction to form sediment, which, with other reef organisms, is transported and deposited to make up the vast bulk of a given reef edifice [41,139,140,141,142,143]. Ancient corals played major roles in reef building in some time intervals, but major reef systems also occurred in the complete absence of corals or even skeletons of any kind when microbialites (rocks formed by microbial biofilms) dominated [98,144]. Combined, carbonate producers, both skeletal and microbial, in a given reef community can be thought of usefully as a carbonate factory [145]. Members of a given carbonate factory are constrained by a variety of physicochemical factors (e.g., light, temperature, energy levels, pH, alkalinity, oxygen, etc.) in the local environment as well as nutrients [146] and have changed through time as skeletal and non-skeletal organisms have evolved to keep pace with changing marine chemistry, climate and each other [31,35,37,98,99,147,148,149]. No reef can form without a carbonate factory to make an adequate volume of carbonate sediment. However, for the volume of sediment to make an elevated structure in high-energy settings at sea level, the sediment must be well stabilized, and this is traditionally considered a function of ‘reef framework.’ In modern coral reefs, in situ coral skeletons provide a degree of strength, but they are essentially fragile (Figure 4), especially in the face of bioerosion. Some modern coral reefs are notable mainly for their lack of coral skeletons in growth position [141,142,143,150,151]. These rubble-rich reefs have gained considerable recent attention as coral bleaching and large storms have increased rubble production on some modern reef flats. Understanding how reef rubble becomes stabilized has become a conservation imperative [152,153,154].

Although every reef is unique in some ways [31] key ongoing processes (Figure 5) that generate and constrain reefs typically include:

• Construction of calcium carbonate by biomineralization (e.g., coral skeletons, foraminifer tests) or by induction of carbonate precipitation by microbial biofilms (microbialites); in situ carbonate production is the major role of the carbonate factory, and importantly, includes production of both skeletons in growth position and particulate sediment;
• Destruction of in situ calcium carbonate (e.g., today, coral skeletons) both by physicochemical processes (waves, storms, chemical dissolution) and by bioerosion (destruction by other organisms) create mobile sediment from initially immobile substrates;
• Transportation of sediment by gravity, waves, currents and wind;
• Deposition of sediment by settling, baffling and/or trapping;
• Unification, where skeletons in growth position, sediment (including coarse rubble) and microbial deposits are bound into rigid, hard substrate by (1) direct encrustation by other skeletal organisms, generally termed binders (e.g., CCA), (2) microbial precipitates within larger and smaller cavities and pore spaces and in some cases (3) early marine cement that precipitates naturally from seawater.

Carbonate factories have three key ecological functions in producing a reef and these roles can be mapped to ecological or community structures using the functional ‘guilds’ of Fagerstrom [31,32,33] (Figure 4 and Figure 5). Notwithstanding that the term guild also has different meanings and contexts for different users, a functional approach is particularly useful for comparing reefs of different ages. However, it is important to distinguish the broader reef ‘construction process’ outlined above, from the functional ‘constructor guild’ [31] (Figure 4 and Figure 5). The constructor guild includes members of the carbonate factory that create calcium carbonate in growth position, thus initiating framework development where they remain preserved in growth position (e.g., modern corals on the margin: upper reef slope, crest or rim). The construction process, on the other hand, includes all quantitatively important members of the carbonate factory, whether they are preserved in growth position or not (e.g., large benthic foraminifers—LBFs). Corals dominate the constructor guild in modern reefs (i.e., they initiate most framework construction), but they also play a major role, along with other organisms, in the broader process of constructing the bulk calcium carbonate that makes up the reef edifice. Other organisms may contribute to the constructor guild, including CCA, which can be volumetrically important framework constructors in some high-energy settings [155,156], and even encrusting foraminifers, vermetid gastropods and bryozoans may play a role. Different organisms have dominated the constructor guild in previous time intervals and the nature of reef growth and substrate depend significantly on the types of organisms in that role. As members of the constructor guild, many scleractinian corals have significant size and relief above the seafloor and thus enhance vertical growth (which leads to reef accretion). Such morphology is termed suprastral [157] as the constructors rise ‘above’ the sea floor stratum (Figure 6). Other organisms, such as CCA, some corals and sponges in the past, typically grew as thinner crusts on and parallel to the substrate, thus being termed constratal—‘with’ the sea floor stratum [157]. The ability to build vertical relief above the sea floor is in part a function of these constructor morphologies along with their growth and unification rates relative to ambient energy levels.

The binding guild has a critical functional role in reef frameworks and today is dominated by CCA at the surface (Figure 4), but microbialite-producing biofilms occur in cryptic settings (Figure 4). A given organism can overlap in functional guild (CCA are dominate binders but also may be volumetrically significant constructors in growth position [156] and dwellers producing abundant mobile sediment [134]). Binders are generally constratal in morphology and encrusting corals play a minor role as modern binders. A variety of non-skeletal organisms, such as sponges and algae, stabilize mobile rubble prior to hard binding (e.g., by CCA) [152,153,154]. Hence, they are binders, and potentially important contributors to the broader unification process [150,161], but they are not part of the carbonate factory. They may leave little or no record in ancient reefs. Regardless, initial binding performed by CCA and encrusting corals is a major step in reef unification in most modern reefs. The formation of microbialites within cryptic settings or in exposed benthic positions also gives biofilms a critical binder role, making them major agents of unification (Figure 4). The baffling guild today includes branching corals and algae, such as Halimeda (calcareous alga) and Caulerpa (non-calcareous alga) (Figure 4), which reduce proximal current velocities to allow settling of finer sediment than otherwise would be deposited. These same organisms help stabilize mobile sediment once it has been deposited, thus supporting the generation of relief by the carbonate factory. Bafflers do not play a dominant role in reef formation today, but were important at other times in Earth history, e.g., the Pennsylvanian when fragile phylloid algae helped construct large buildups [162].

The destroyer guild includes borers (today, bivalves, sponges and worms) and raspers (today, fish, mollusks and echinoids) and corallivores (today, gastropods, fish and crown of thorns starfish) (Figure 4n). Borers may directly create sediment in excavating their borings (e.g., silt sized ‘sponge chips’ made by boring clionid sponges [163]), but more importantly they decrease the physical strength of in situ substrates, thus enhancing the destruction process caused by non-biological agents (e.g., waves, storms) (Figure 4g). Microborers, such as cyanobacteria and filamentous green algae, also weaken substrates, but can increase local alkalinity causing localized cementation in some environments [164]. Raspers can turn a significant volume of rigid substrate into mobile sand [165,166]. Importantly, raspers, such as parrot fish, are not part of the carbonate factory. Although their feeding behavior produces sand, the sand consists of carbonate made by a different organism (coral, CCA, etc.). They produce no carbonate themselves other than perhaps their otoliths. The final ecological role includes a great variety of community ‘dwellers’, which have many other types of ecological functions and roles in reef trophic structure and reef services. Some of these functions are critical to modern coral reef development, such as the grazers that keep levels of fleshy algae down so that the slower growing calcifiers typical of the carbonate factory can accumulate [167], but these organisms may never be recorded directly in a reef’s preserved structure (Figure 4 and Figure 5). Filamentous algae and cyanobacteria also contribute to trapping and stabilizing sediments in ‘turf algae’. Other dwellers are very important members of the carbonate factory as they produce significant volumes of sediment. The most important dweller members of the carbonate factory in modern coral reefs are the baffler Halimeda, whose modular calcified thallus becomes sediment upon death, LBFs and mollusks [135,138] (Figure 4).

Concepts and semantics of, and perspectives on, reef ‘framework’ may vary even more than for the term reef [104,168]. For many, perhaps most, authors reef ‘framework’ consists only of skeletons in growth position [31,169], although Fagerstrom [31] treated ancient calcimicrobes as calcareous alga skeletons, rather than microbialites. Today, such a definition applies only to scleractinian coral skeletons for some workers. However, to discuss and compare reefs of the ancient past, a skeletal definition of framework is inadequate because reefs occurred literally billions of years before skeletons did [98,144]. The approach used below is to define the framework as the reef material that is unified, both skeletons in growth position and bound rubble, to make rigid reef substrate on ecological time scales [98,158]. Modern reef framework includes not only the corals, but the binding CCA and microbialites, and deeper stromatolites, where they occur [170]. Some authors [104,171] used a hybrid definition with skeletons in growth position being ‘primary framework’ and unifying organisms (e.g., microbial biofilms and CCA) providing a ‘secondary framework.’ However, the distinction becomes unnecessary from a functional perspective, because there is little difference between the rigid carbonate dome made by a biofilm on a hard, calcium carbonate stromatolite, for example, and the rigid calcium carbonate dome made beneath a thin veneer of living coral tissue in a Porites skeleton [98]. Both are rigid, relief-bearing structures made by biological members of a carbonate factory. Using a skeleton-in-growth-position-only framework concept, even many modern reefs would lack framework entirely. Some reefs occupy positions where hydrodynamically deposited rubble controls the position of the carbonate edifice more than the biological community itself [104,140,142,150,151]. These structures could be considered ‘reefs’ that lack framework or hydrodynamic sediment piles and not reefs at all. However, where the rubble is unified by binders in ecological time [140], which also occurs in reefs that have skeletons in growth position, it can be considered reef framework. Additionally, reefs dominated by microbialites would be considered as lacking any framework using a skeletal definition, despite meeting all criteria to be reefs, and that includes parts of modern reefs, such as the deglacial reefs of Tahiti, where microbialites make up as much as 80% of ‘framework’ volume [172]. The very different nature of carbonate factories and marine alkalinity in the past adds considerable complexity with the formation of ‘hybrid carbonates’ [169]. Morphological and ecological differences between reef-building corals and less well-understood Paleozoic stromatoporoid sponges and microbes, for example, led to four major types of reef margin framework being recognized in Late Devonian reefs in Canada: metazoan–microbial; sediment-laden metazoan; metazoan-dominated, and metazoan–marine cement types [168]. Thus, not only corals and microbialites had to be accounted for, but also marine cements [173]. Broader framework concepts [158] are used below to facilitate comparison of modern and ancient reefs that have different carbonate factories (Figure 6). This classification system applies to rigid boundstones [159] whether framestones, bindstones or bafflestones, but there has been no attempt to classify different types of skeletal morphology within the framework types (e.g., framestone, pillar stone, etc. [157]). All boundstones (reef frameworks) discussed below were adequately rigid in ecological time to make a hard substrate.

Reef framework is an example of a carbonate ‘facies’, which is a term (both singular and plural) that is applied in different ways and at different spatial scales to groupings of sediment or limestone based on a specific set of attributes [174]. For this paper, facies refers to a specified collection of sediments, be they particulate or in situ skeletons, along with cements, that are being, or have been, deposited together under a set of conditions that reflect a specific environment or sub-environment, in this case, on or around a reef. Each ‘geological’ reef facies generally has an analogous environmental–ecological connotation typically with a geographic or geomorphological distribution and Devonian and modern reef facies have previously been found to be comparable [34]. Hence, reef framework facies is part of a broader reef margin facies, which is part of a broader reef facies that also contains lagoon, reef flat, and back-reef facies, etc. The facies connotations are the same for modern and fossil reefs and terminology for reef zonation used in this paper is a combination of that devised for the DGBR [88] and GBRP [175,176] (Figure 7).

Another key concept required for comparing reefs of different settings or ages relates to stratigraphy—the layers of reefs built up through time (Figure 8). ‘Accommodation’ is a term for the vertical space available for deposition of sediment. For reefs, accommodation is the vertical space between the seafloor and ~mean low sea level. Modern reef growth is confined to shallow bathymetry owing to the need for light of the dominant photoautotrophs that drive the carbonate factory, but reef organisms cannot grow significantly above mean sea level owing to desiccation. Changes in local relative sea level (RSL) affect the accommodation for a given reef and can result from either eustatic (global) sea level rise or fall or to local subsidence or uplift. Major increases in eustatic sea level, such as the last deglacial, provided increasing accommodation for existing coral reefs such that they had to accrete vertically (aggrade) so as to ‘keep up’ with the rising sea surface [178,179]. Where reefs could not keep up with increasing accommodation, their carbonate factory could shut down and they could eventually ‘give up’ and become drowned with new reefs forming up-slope in shallower settings (i.e., back-stepping) [46,180]. Where sea level rise slowed to a halt, some reefs that were not able to keep up were able to ‘catch up’ to the new sea level [178,179]. The key control on a given reef’s behavior through time then is the ratio between the rate of change in accommodation versus the rate of vertical reef growth produced by its carbonate factory. Once caught up to sea level, accommodation has been filled, so reefs cannot continue to aggrade. They then grow laterally, causing the margin to ‘prograde’ at sea level over older slope deposits causing the areal footprint of the reef to increase. Where the slope in front of the reef is very deep and steep, such as on oceanic atolls or on continental shelf edges, seaward progradation may be severely limited [181].

Over long time frames, major cyclic changes in accommodation are caused by Milankovitch forcing of continental ice volumes, and thus eustatic sea levels, and these changes dominate reef behavior over scales of 10s to 100s of thousands of years [184]. During icehouse climates, such as today, large changes in ice volume at the poles create high-amplitude (~100 m+) changes in sea level that have resulted in a ‘layer cake’ stacking of reefs in regions with adequate ongoing subsidence, like the GBRP (Figure 8). Where there is little or no subsidence, the reefs are ‘stacked’ laterally between each glacial cycle. In southern Florida, for example, little to no recent subsidence has led to Holocene reefs growing laterally, seaward of the Last Interglacial (LIG) maximum sea level Marine Isotope Stage (MIS) 5e reef on older reefs that tracked lowering sea levels into the glacial maximum (substages 5a–c) [185,186]. The MIS 5e LIG reef currently is exposed above modern sea level to form the Florida Keys because LIG sea level was higher than today [187,188]. However, in times of greenhouse climate, sea level fluctuation has much smaller amplitude (m to 10 s of m) as there is less continental ice storage [183]. These lower-amplitude changes in sea level create less accommodation, so thinner layers result. These externally forced, high-frequency (in geological time) cycles dominate the stratigraphy of most ancient carbonate platforms and are called parasequences and their amplitudes vary between icehouse and greenhouse climates [189]. As renewed reef growth seeks to fill newly created accommodation with each new sea level rise, each parasequence generally shows a shallowing-up trend as the reef community grows up to, or progrades over deeper deposits at, the newly established maximum sea level.

3. Results

3.1. Location, Age and Tectonic Settings

3.1.1. Great Barrier Reef Province

The GBRP in northeastern Australia stretches from Lady Elliot Reef in the south to the Gulf of Papua in the north, a distance of 2300 km and ~14°55′ of latitude (Figure 9). The province’s NNW to SSE orientation gives it a latitudinal gradient from the subtropics (south of the Tropic of Capricorn) to the tropics north of 10° S latitude. It is bordered by the Coral Sea to the east and the Australian continent to the west. Although inshore reefs occur, the vast majority of the >3000 individual reefs occur within a 25–65 km wide zone on the outer margin of the continental shelf [181]. The zone extends as much as 300 km offshore in the Swains Reef complex where it is separated from shore by the Capricorn Channel (Figure 9e), but it is closer to shore in the north [181]. In the central GBR, south of Cairns, the reef zone lies mostly east of a NNW–SSE oriented structural ‘hinge line’ where structural basement deepens abruptly to the east (Figure 9a) [190]. As a mixed carbonate–siliciclastic domain, inshore areas commonly host significant siliciclastic sediment delivered by local catchments, which also helps restrict the most abundant reef growth farther offshore in a pure carbonate zone [181,191]. However, inshore reef communities have adapted to more turbid conditions [192,193]. Circulation in the Coral Sea is dominated by the southern part of the west flowing South Equatorial Current, which splits against the northern part of the GBR, flowing largely south to form the East Australian Current, but also to the north (North Queensland Coastal Current) where it forms complex eddies in the Gulf of Papua (Figure 9) [81]. The currents are complicated seasonally, especially in the north where they are affected by the Indonesian Australian Summer Monsoon. Dominant winds are from the SE for most of the reef for most of the year with a change to more NNW winds during the summer monsoon [81]. Tides are mostly semi-diurnal to diurnal and mesotidal with offshore Spring tidal ranges mostly >2.5 m, but tidal ranges increase towards the coast and are macrotidal in some coastal areas, such as Broad Sound (to 8 m) [80,81]. Throughout the province marine circulation has been heavily modified, or ‘bioengineered’ [81], by reef growth.

Modern reefs in the GBR are Holocene in age (less than ~10 kyr old) but offshore reefs grow directly on top of older reefs in a ‘layer cake’ fashion where shallow reef facies formed during interglacial high sea levels, like today, are separated by unconformities that represent low sea levels at glacial maxima [47,190,196,197] (Figure 10). The existing Holocene reef represents a single parasequence. Older, subsurface reefs have been difficult to date owing to their burial beneath the modern reef (i.e., poor access) and because the rare available carbonate materials obtained from cores have generally been altered by meteoric water, thus disturbing the geochemistry required for radiometric dating. There are many more dated rocks from the moon than from the base of the GBR [51]. The first deep GBR reef cores were recovered in 1926 (Michaelmas Cay [198]) and 1937 (Heron Island [198]) but the first geochemical and geophysical dating was done on a core recovered from Ribbon Reef 5 (RR5) only in 1995 [49,182] (Figure 9a). Although the Michaelmas and Heron cores sampled Pleistocene rock from beneath the Holocene reef [198], the cores could not be dated. A petroleum exploration well drilled on Wreck Reef, southern GBR in 1959 also penetrated the entire reef sequence, but also provided no radiometric dates [182,199] (Figure 9a). An exploration well drilled at the northern margin of the GBRP off Anchor Cay (Gulf of Papua) did not penetrate adjacent Holocene or Pleistocene reef rock [80] (Figure 9a). Prior to the 1995 RR5 core, seismic data and shallow reef sediments shed into deeper settings on the slope near Cairns recovered by IODP Leg 133 drilling suggested an initiation age as late as ~500 kyr ago [200]. Direct dating of the initial reef phases of RR5 in the GBR showed that shelf carbonate deposition initiated since ~770 kyr ago as it all represents the Brunhes geomagnetic chron (a time of consistent global magnetic polarity) [182]. Reef facies appear to have initiated around 400–600 kyr BP years ago, during sea level highstands at Marine Isotope Stages (MIS) 11, 13 or 15 based on packages of reef facies separated by unconformities [197] (Figure 10b) and poorly constrained strontium isotope (⁸⁷Sr/⁸⁶Sr) ages and limited open-system U-Th ages [182]. Based on current evidence, the reefs at the base of the modern GBR ‘layer cake’ initiated around 600 kyr BP and will remain relatively poorly constrained until it is possible to directly date reefal carbonates from the earliest reef package.

While the age of the ‘layer cake’ underlying the current GBR has been constrained to the Late Pleistocene–Holocene, the great latitudinal extent of the GBRP provides an additional complication. The northernmost part of the GBRP reached the tropics, and thus warmer SSTs, >30 My earlier than the southern part because of the slow northern movement of the Australian Plate and the nearly 15° latitudinal range of the reef [82,83] (Figure 11). Thus, the interplay of tectonic movement and global climate has controlled the development of the GBRP and the two factors are inextricably linked. The southern-most zone of the GBR currently straddles the Tropic of Capricorn, so reached that latitude only recently, much later than in the north. Although the northern GBRP reached the tropics during the Oligocene, the very warm conditions in the Early Cenozoic ‘Greenhouse’ meant that SSTs were adequately warm at higher, subtropical latitudes for carbonate development in the Eocene [83,201] (Figure 2, Figure 11 and Figure 12). However, very cool global temperatures with the onset of Antarctic glaciation at the end of the Eocene suppressed carbonate deposition through most of the Oligocene. With Australia’s continued movement north, subtropical and finally tropical carbonates formed again in the Miocene. Increased SSTs in the SW Pacific allowed reef facies to thrive on the Queensland and Marion plateaus in the Coral Sea from the Early to Middle Miocene [202,203,204] consistent with warm seawater surface temperature (SST) proxies (e.g., δ¹⁸O from foraminifers [205], foraminifer Mg/Ca ratios and organic molecule evolution, such as TEX₈₆^H [206]) and Sr isotope-dated shallow tropical carbonate sediments [200] from IODP cores (Figure 12). Late Miocene changes in oceanic currents [203], sea level [204] and/or cooling [207] led to platform demise or shift to carbonate ramps [202,204]. That and the onset of northern hemisphere ice sheets in the Pliocene led to a ‘Pliocene Reef Gap’ and continued low carbonate production in the offshore Coral Sea [207]. The plateaus only began to support shallow carbonate production again in the Late Pleistocene as temperatures warmed [205,206], presumably at the same time as tropical–subtropical reefs began growing on the continental shelf in the central [182] and southern GBR [208] (Figure 2 and Figure 12).

In the northernmost parts of the GBRP, Miocene reef carbonates also flourished on the continental shelf based on seismic data [83,201]. At the northern end of the GBRP shelf edge reefs east of the Torres Strait overlie a 1.5 km thick sequence of stacked, but intermittent, carbonate reefs–platforms that date from the Miocene to Pleistocene [201] but these older platforms reach thicknesses of nearly 3 km father north in the Bay of Papua [211,212] (Figure 12). Those reefs have been penetrated by few exploration wells and remain poorly known. Farther north in the Gulf of Papua Early Miocene carbonate shelf deposition appears to have initiated above Eocene carbonate ramps but may not have been continuous to the Pleistocene owing to changes in climate and sea level [211,212,213,214]. Broad older carbonate platforms are tilted to the NE owing to foreland loading from the Papuan Fold and Thrust Belt but younger carbonates remain near horizontal. Extensive siliciclastic sediments were delivered from the New Guinea highlands from the Late Miocene and reefs and platforms at the northern end of the GBRP were progressively buried by continental sediments shed from the New Guinea highlands [201,211,212] (Figure 12a). Current seismic exploration and drilling in the Gulf of Papua will eventually shed much light on the older parts of the GBRP that are now buried beneath thick siliciclastic sediments.

A recent review of earlier tectonic development of the GBRP and offshore Queensland and Marion plateaus in the western Coral Sea [215] showed opening of the Coral Sea began in the Late Cretaceous and terminated in the Eocene. Intracratonic rifting on the continental margin produced the complex basement relief that controls the position of the plateaus, coastal basins and modern shelf edge. Aside from spanning climate zones, the northward movement of the Australian plate through the Cenozoic led to collision and active convergence on the northern margin of New Guinea [215] with major implications for carbonate environments in the associated foreland basin [213]. The positions of most current reefs and carbonate platforms are controlled by basement highs, such as the edges of fault blocks, that resulted from that tectonic history [190,201] (Figure 12).

Today, northeastern Queensland is generally considered tectonically quiescent and an example of a passive continental margin in an intraplate setting [201,216]. However, much of the GBRP has experienced significant subsidence during the time of reef growth, thus differentiating it from modern passive margin reefs, such as those in Florida. Offshore Holocene reefs in the GBR represent only the current parasequence at the top of the Late Pleistocene stack [47,51,190,196,197]. Where cores that penetrate beneath Holocene reefs on the Queensland continental shelf have been dated using U-Th techniques, the immediately underlying reef is LIG in age [182,217,218]. The depth to the Pleistocene–Holocene boundary varies significantly around the GBRP [80]. This partly reflects localized differences in elevation on the underlying LIG substrates themselves [196,218,219] but substrates also deepen seaward of a structural hinge line [190] and along latitude, where they mostly deepen from north to the south [80], unlike older reefs farther north near Papua New Guinea. Some abrupt differences in depth to the LIG reef may be associated with active faults [220] and ongoing increased subsidence in the central region of the GBR is partitioned, north and south, by major NNE–SSW trending faults that presumably represent reactivation of older transform faults that formed much earlier during the opening of the Coral Sea [194] (Figure 9a). These NNE–SSW trending structures are associated with modern seismicity [194,221] although northeastern Queensland is generally thought to have relatively low seismic risk in general. In the northern part of the GBR near the Torres Strait, individual Quaternary reefs have been offset laterally suggesting active right-lateral strike slip movement since initiation of the Pleistocene platforms. The current right-lateral strike slip movement presumably occurs along reactivated NNW–SSE trending normal faults, also probably associated with the opening of the Coral Sea [222]. Movement on these faults is consistent with the current stress field caused by active convergence on the northern margin of Papua New Guinea, the northern edge of the Australian Plate [222]. Hence, although a passive continental margin, a degree of active tectonism still affects the GBR, but it is based on reactivation of older structures formed during the rifting that gave rise to the Coral Sea.

Although a degree of neotectonic (i.e., recent to current) activity in the GBR has been demonstrated relatively recently, the cause(s) of the significant subsidence that allows the Pleistocene reefs to be stacked into a ‘layer cake’ (a thickness of 260 m [190]) remains enigmatic. Minor (a few meters) but systematic differences in the modern elevation of LIG shoreline indicators around Australia suggested continent-scale subsidence to the north [223], which has subsequently been acknowledged as a major ongoing tilting of the Australian continent of ~250 m up in the SW to ~250–300 m down in the N since the mid-Miocene [224]. The rate of tilt was calculated at ~15–20 m/myr and was attributed primarily to vertical movements related to underlying mantle flow (dynamic topography) associated mainly with subduction beneath Papua New Guinea to the north [224]. However, that subsidence rate cannot explain the depth to LIG reef substrates in the central GBR [80], which, all other things being equal, would require ~10× the rate. Even where the LIG substrate is shallower (e.g., 5 to 8 m depth [80,217,219]) it would require > 3× the continent-scale subsidence rate. Significant subsidence occurred in northern Australia (Gulf of Papua, Arafura Sea to the Indian Ocean) in the Oligocene and Miocene owing to thrust and sediment loading as part of the foreland basin south of the Papuan subduction zone [213]. There, it accumulated more than 2000 m of carbonate shelf sediments (Darai Limestone) during the Miocene [213]. However, that subsidence was too early, oriented in the wrong direction (i.e., more E–W) and was too distal to account for ongoing subsidence through most of the offshore GBRP. However, it has implications for the northernmost part of the GBRP, which are discussed below.

Although tectonic subsidence occurred (and is occurring) in northern Australia since at least the Miocene, it is important to remember that the 260 m reef thickness in the GBR [190] formed only within the last ~700,000 years. This is particularly enigmatic, because it reflects a major difference in subsidence between onshore and offshore positions across the reef. Onshore LIG sea level indicators are above present sea level owing to the high sea levels of the LIG [188]. However, offshore LIG reef top elevations beneath modern reefs are significantly lower [80,225,226]. Karst removal of LIG reef was suggested as a possible reason for the difference in elevation [196] but was rejected owing to inadequate rates of karst removal to account for many of the depths to LIG substrates [227]. Karst was rejected in the southern GBR where coring has allowed the best study, because: (1) shallow water facies occur at the top of the LIG reef succession; (2) calcrete–caliche at the unconformity suggests a more arid climate (i.e., less dissolution); and (3) the depth does not systematically increase northward as would be expected for reefs transitioning into the wetter tropics [218,219]. Karst removal could not account for 30 m-thick Holocene reef successions resting on LIG reefs in any case. Sediment compaction is unlikely to explain the difference either. Although coastal LIG shore indicators occur farther onto the ~stable basement rocks of the continent, and have less overlying loading than the reef substrates, reef limestones, at least at the margins, are not as compactable as fine siliciclastic sediments owing to their early rigidity and are unlikely to account for so much difference. Dynamic topography, glacial isostacy and sediment loading also have been rejected as the cause [225,226]. Hence, minor neotectonic reactivations of normal faults formed during the opening of the Coral Sea are possible and are consistent with deepening east of the NW–SE ‘hinge’ line [190] (Figure 9). Unfortunately, adequate modern seismic data to test such a hypothesis are not available as appropriate surveys have not been allowed since the GBR Marine Park was declared in 1975 and few older seismic lines exist of adequate quality.

3.1.2. Devonian Great Barrier Reef

The DGBR in western Australia crops out over a length of 350 km with a width of as much as 50 km on the Lennard Shelf along the southwestern margin of the Kimberly Block, an ancient shield area of Proterozoic metamorphic and igneous rocks that make up the North Australian Craton [88] (Figure 13). However, the reef extends into the subsurface to the northwest and originally may have bordered the Kimberly Block for 1000 km turning back to the north and east to the Bonaparte Basin [88,90,228,229]. The reef also may have extended to the east and south along the edges of the Fitzroy Trough and Gregory Sub-basin and then turned west again along the shallower Crossland and Broome platforms [90] where similar Late Devonian reef facies have been recognized in seismic data and intersected in boreholes [230,231,232,233]. Hence, it may stand as one of the larger Devonian reef complexes, not the smallest of twenty [85]. Playford and others [90] provided a detailed history of study of the DGBR. Although first studied geologically in 1883 [234], the limestones were not interpreted as fossil reefs until 1924 [235]. They were investigated for their resource potential from the 1930s to 1950s culminating in major regional and integrative studies by the Geological Survey of Western Australia since the 1960s [87,88,90,236]. They are associated with minor hydrocarbon development [233,237] but host significant base metal sulfide deposits [238] and many hundreds of papers have been published on various aspects of the reefs over the last century.

The Fitzroy Trough was an active, extensional intracratonic basin during the Devonian and now forms the northernmost region of the broader Canning Basin [239,240]. The northern part of the Late Devonian, western Australia occupied the southern tropics located in the northeastern part of Pangea, thus representing similar latitudes to the current GBRP [210] (Figure 1b). However, the orientation of the exposed portion of the reef complex on the Lennard Shelf was near parallel to latitude, so represented a tighter latitudinal range at a given time than does the GBRP. Although a variety of rifted tectonic terranes occurred to the north and west of Australia at the time, faunal evidence suggests that the Canning Basin was open to Tethyan Ocean circulation to the west [241], but the basin was partially restricted [233] (Figure 13). The eastern coast of northern Australia also occupied the tropics during the Devonian and hosted shallow carbonate environments, but they were mostly Early and early Middle Devonian in age [85,242,243]. Givetian (upper Middle Devonian) shelf limestones occur in Queensland in the Broken River [243] and small patch reefs with corals and stromatoporoids occur in the Fanning River [69,244,245] groups, but major reef complexes similar to those in the Caning Basin have not been recognized [242,246]. However, stromatoporoid–coral buildups to 30 m thick and 2 km long occur in the lower Devonian part of the Broken River Group [247]. Older Lower Devonian limestones near Chillagoe to the north also contain stromatoporoids and corals, but no major reef complexes have been described [248]. Limestones were particularly rare in Queensland during the Frasnian and Famennian, the main age of the DGBR, and reefs of that age are unknown there [84].

The DGBR reef facies range in age from late Middle Devonian (late Givetian) to Late Devonian (Frasnian to late Famennian) (~385–355 million years ago) [100] with a total thickness of more than 2.5 km [88]. The Late Devonian was an interval of major global change [249] wherein ‘greenhouse climate’ temperatures (Figure 2) decreased following an end-Givetian thermal peak only to rise to a late Frasnian peak before shifting to somewhat cooler conditions through the Famennian [96,250,251]; terrestrial floras diversified with the introduction of forests with deep rooting trees and deeper soil development with implications for terrestrial weathering and coastal sedimentary environments [252,253,254,255]; fish underwent major diversification [256,257]; the first tetrapods invaded the land [258]; and major extinction events occurred: Givetian–Frasnian, Frasnian–Famennian, and end-Famennian [259,260,261,262].

Representing more than 20 million years of active growth, the DGBR reef complex is divided into a stratigraphic ‘supersequence’ or cycle consisting of two major depositional phases: the Givetian to late Frasnian transgressive back-stepping Pillara phase and the late Frasnian to late Famennian regressive prograding Nullara phase [263,264,265,266] (Figure 14). The dominant transgressive regressive supersequence most likely represents a global sea level signal (i.e., eustasy) as it has been recognized in other regions (Alberta, Canada and South China) that have differing local tectonic settings [265,267,268]. Currently, two Givetian, seven Frasnian and more than three Famennian (the largest part being undifferentiated) 3rd order sequences have been recognized through the reef complex [269,270,271] with ages based on biostratigraphy (primarily conodonts and ammonoids [272,273] but also microvertebrates [274]) and chemostratigraphy [269]. Higher-frequency, meter-scale parasequences dominate stratigraphy in parts of platforms [275], primarily being attributed to Milankovitch orbital forcing [276,277,278] but differentiating cycles is complex [183] and many lower-frequency cycles are associated with tectonic forcing, especially apparent in coastal sections where siliciclastic and carbonate sediments interfinger [278,279,280]. Interfingering of siliciclastic and carbonate facies in proximal basin deposits also has been attributed to sea level cycles with carbonate sediments delivered past the shelf edge during highstands and siliciclastic sediments bypassing the reefal carbonates or overtopping them during lowstands (reciprocal sedimentation) [281].

Reef growth in the DGBR was syntectonic [87] (i.e., occurred at the same time as major basin extension) and tectonic movements controlled RSL, and thus accommodation, in parts of many depositional sequences [90,278,280,282,283,284,285]. Some shelf edge collapse events that led to the deposition of very large olistoliths composed of shallow reef margin facies into deeper basinal settings (e.g., McIntrye Knolls) may in part be the result of coeval seismic activity [286,287]. However, margin collapse in general was associated with oversteepening and changes in relative sea level, that also were controlled partly by active subsidence along faults [270]. Neptunian dykes (fractures that ~parallel platform margins) are common in the reef complex owing to early cementation/unification and high rates of progradation of margins over less well-indurated slopes [177,287,288,289]. Since deposition, the reefs have experienced relatively little subsequent structural deformation and no metamorphism and many whole platforms have consistent low structural dips [89]. Currently, parts of the reef complex have been exhumed with near original topographic expression so that local bathymetric relationships have been preserved [87,290]. This in part reflects the position of the Leonard Shelf on the edge of the North Australian Craton where it has not been subject to as much deep burial as the basin to the south, but also to the removal of some overlying sediments during the Late Paleozoic glaciation events when the reefs were in some cases exhumed by continental glaciation from the south [291]. Paleogeographic relationships are also enhanced by the great disparity in relative erodibility of both underlying and overlying rocks. The readily eroded shales and sandstones of the overlying Mississippian Fairfield Group have allowed reef limestones to stand in high relief relative to their contemporaneous basins allowing one to walk up the exhumed reef slopes in some areas and onto the elevated reef margins and into back-reef facies and lagoons. At the same time, the very durable quartz-rich metavolcanics and metasediments of the underlying Proterozoic rocks still retain much of their original Devonian relief so that the relationship of the reefs with contemporaneous continental paleoislands and rocky shores are well preserved [92,290,292,293].

3.2. Global Constraints—Climate and Water Chemistry

3.2.1. Great Barrier Reef Province

The closely linked climatic–tectonic interactions through most of the Cenozoic [83,201] (Figure 11 and Figure 12), including the Eocene ‘Greenhouse’ climate, are mentioned above with the tectonic setting (Section 3.1.1). Today’s GBRP experiences an ‘icehouse’ climate spanning nearly 15° of latitude from fully tropical conditions with a well-developed Indonesian–Australian Summer Monsoon (IASM) in the north and more arid subtropics at the southern end. Thus, it experiences both spatial and temporal (Milankovitch) changes in climate. Both terrestrial and marine temperatures generally decrease from north to south but are partly ameliorated by the south-flowing EAC, which carries warm tropical waters south along the Queensland coast (Figure 9). Rainfall and fluvial discharge are significantly higher and more consistent in the north. South of the Wet Tropics, and encompassing the two largest catchments in the GBR (Burdekin and Fitzroy river systems), discharge is lower, but much more flashy and flood-prone [294] with disproportionate impacts on water quality in the GBR lagoon [295]. Cyclones (hurricanes) have major impacts on the GBR in terms of significant physical damage to the reef, but also with increased terrestrial runoff and water quality declines inshore and increased sedimentation and re-suspension of sediments in the lagoon [75]. Cyclone frequency increases eightfold in a near linear latitudinal trend from around ten cyclones in the south to eighty in the north at Torres Strait based on data collected since 1906 [296].

Although the increase in marine heatwaves and attendant coral bleaching is of major concern for coral reefs today, the greatest climatic effect on the modern GBR is the larger scale temporal oscillation of sea level with each glacial cycle that has resulted in the layer cake stacking of highstand reefs. During each glacial interval, the reefs of the GBR migrated seaward where appropriate substrates were available (currently subsea terraces) and then grew until sea level again lowered, exposing them [46] (Figure 10). The reefs of the Last Glacial Maximum (LGM) grew farthest seaward and then aggraded so as to keep up or catch up with sea level during deglacial time [297]. Then as sea level continued to rise, reefs drowned, and communities back-stepped onto the previously abandoned MIS3 reef terraces where not hampered by siliciclastic sediment flux, until finally reestablishing themselves on the previous highstand platforms during the Early Holocene [46,297] (Figure 10). Overlain on these parasequence level changes are smaller changes in climate [30]. Unlike some areas of the world (e.g., the Caribbean) the Holocene local sea level curve (RSL) is complex for the GBR with sea level initially reaching more than a meter higher than present by 8–7 kyr BP and then falling to the present sea level through the Holocene [298,299]. However, the fall was not gradual, and small but synchronous sea level oscillations at 5.5, 4.6, 2.8 and 1.2 kyr BP regulated various reef turn-on and turn-off events in both inshore and offshore reefs [300,301,302,303,304,305]. Some of the hiatuses in reef growth lasted more than 1000 years and were coeval with reef hiatus events in reefs in other parts of the world [306], but their exact relationship to global climate is not known. Inshore reef behavior is complicated by sedimentation, water quality, and potentially by localized subsidence from compaction. The high tidal ranges for GBR reefs make measurements of previous sea levels more difficult, but even offshore reefs show well-dated hiatuses in growth. The broad shut down of reef flat growth between 3.6 and 1.6 kyr BP occurred in offshore GBR reefs throughout its latitudinal range with the only exception being the central region that was undergoing greater ongoing subsidence [194,305]. Hence, eustatic sea level may be implicated in that case, suggesting a more global driver.

Although latitudinal gradients in climate existed throughout a given parasequence, the temperature gradient for SSTs along the GBR was steeper during the LGM and deglacial than it is today [307]. The change was attributed to the expansion of cooler subtropical waters northward associated with changes in oceanic circulation, but core data are not available to test broader changes to biodiversity that may or may not have accompanied the changes. SSTs increased dramatically between the LGM and Holocene but there were also smaller temporal perturbations. Immediately before the Holocene the Younger Dryas cool event did not cause cooling where it has been tested in deglacial GBR cores (IODP Expedition 324) [308]. Modeling based on associated coral δ¹³C also showed that rapid reef accretion at that time had negligible effect on atmospheric CO₂ [42]. Deglacial reef accumulation was heavily dependent on local bathymetry slope angles and proximity to fresh terrestrial or re-suspended particulate sediments [288], thus complicating broad carbonate budget modeling.

Within the Holocene the GBR experienced a mid-Holocene thermal optimum as did the rest of the western Pacific [309,310,311,312]. Data from the southern GBR suggests SSTs were cooler than present during the Holocene both before and after the optimum when they were as much as 1–2° higher between ~6 and 7 kyr BP [309]. Snapshots of mid-Holocene climate (~4–6 kyr BP) have been documented from geochemical archives in corals showing differences in monsoon patterns, seasonality and terrestrial flooding based on coral luminescence bands and δ¹⁸O [313,314]. However, longer records of secular change in Holocene climate of the GBR are scarce [30]. Major changes in climate and Pacific Ocean circulation potentially associated with a shift in El Niño Southern Oscillation (ENSO) are recorded in changing ¹⁴C reservoir corrections [315,316]. Major frequent changes in the correction occurred prior to stabilization at ~5.5 kyr BP, when more modern ENSO behavior engaged, altering the degree of upwelling of deep older waters that reached the GBR. Halimeda bioherms in the GBR lagoon record changes in δ¹⁵N that reflect changes in terrestrial versus upwelled water sources that also correlate well with ENSO frequency and intensified El Niño activity after ~5 kyr BP [317]. The data also correlate with monsoon (IASM) reconstructions based on independent speleothem data [318]. Longer secular records of trace element geochemistry preserved in reefal microbialites record similar differences in Holocene terrestrial runoff in the GBR [319] with major controls in precipitation being the IASM and ENSO. Limited analyses of coral δ¹⁵N and Ba/Ca ratios also show much higher terrestrial influence on offshore GBR waters from ~8 to 6 kyr BP, with peak terrestrial nutrients associated with significantly lower coral extension rates and deeper, more turbidity-resilient taxa [320]. Cyclone abundance and intensity varied through the Holocene with the GBR experiencing very active times (>5 to ~3.5 kyr; ~2.5 to ~1.6 kyr; ~600 yr to today) interspersed with periods of lower activity [321]. It is not clear how the cyclone record relates to global or regional climate systems [321], but the intervals of activity and inactivity vary regionally. The record is not adequately long in the GBR to compare to other water quality data, but one reef hiatus in the central GBR was interpreted to reflect intense cyclone activity that did not allow reef accretion in some inshore areas for 5000 years between ~6.5 and 1.5 kyr BP [322].

The GBR today occurs during an interval of ‘aragonite seas’ with high Mg/Ca ratios, low pCO₂ and low temperatures [94,323,324] (Figure 2) thus favoring aragonite precipitation by organisms such as scleractinian corals and Halimeda [325]. Most of the GBR is a classic example of an oligotrophic system, favoring low-nutrient, open ocean seawater [146]. Modern tropical seawater pH and alkalinity promote an adequate carbonate saturation state for abundant carbonate precipitation. However, lower atmospheric CO₂ during the LGM and deglacial, relative to the current interglacial (i.e., the Holocene), appear to have increased pH and saturation, and thus, the precipitation of reefal microbialites [148], which as non-obligate calcifiers are greatly affected by ambient water chemistry [98,132,170].

3.2.2. Devonian Great Barrier Reef

The Middle to Late Devonian was a time of global greenhouse conditions [96] with average global temperatures mostly 5–10 °C higher than today [326], but these high global temperatures fluctuated through the Middle and Late Devonian. Cooling during the Famennian is associated with minor continental glaciation at very high paleolatitudes in South America [327,328], but there is little evidence for large-scale glaciation at that time [329], and it occurs against a backdrop of very high temperatures based on most lines of evidence [96,330,331]. Continued cooling into the Mississippian eventually led into the Late Paleozoic icehouse phase. Local paleoclimate in the DGBR was consistent with the tropical to subtropical latitude with arid conditions in the Early Devonian [332], and this continued into the late Givetian where the oldest part of the exposed reef complex, the Cadjebut Formation, contains abundant evaporites [90,333]. That is consistent with the shallow intracratonic setting within the southeast trade wind belt with evaporation greater than rainfall at that time. However, modeling suggests higher humidity in the Late Devonian Canning Basin [250] with tropical monsoons [326] and cyclones tracking from the northeast [334]. Storm deposits occur in inner and mid-ramp settings in the Hull Range region [285], but it is unclear to what extent event beds represent storms, seismic activity or stochastic platform collapse. However, Late Devonian paleoclimate data are complex over the entire Canning Basin. Although evaporites are not known from Frasnian or Famennian platforms on the Lennard Shelf [90], they did occur, at least in the Famennian, in the southeastern part of the basin on the Barbwire Terrace just across the Fitzroy Trough [230,232] (Figure 13).

The common, high-frequency parasequences, including alternating siliciclastic and carbonate deposition, throughout DGBR stratigraphy are generally attributed to Milankovitch orbital forcing of minor sea level fluctuations [277], although the driving forces for cyclicity are complex with other climatic and tectonic forcing potentially dominating some sequences [278]. Late Devonian southern hemisphere monsoons were modeled to peak during precession maxima, thus leading to intervals of higher terrestrial runoff at those times [335]. Weaker monsoons during precession minima would have led to lower continental runoff and thus, might have supported greater carbonate production at those times. The alternation of carbonate and siliciclastic sediments would be enhanced during more arid times in general [285], and thus climate may have modulated aspects of the carbonate–siliciclastic contribution to parasequences beyond regulating eustatic sea level.

3.3. Carbonate Factories

Carbonate factories are intimately related to local conditions, such as bathymetry, energy levels, oxygen and nutrients and typically define the spatial distribution and productivity of particular facies and their zonation across carbonate environments [174,336,337]. Maximum carbonate production typically occurs on reef margins facing the open ocean (Figure 15). That is also where reef framework constructors are concentrated. Where those constructors are suprastratal and adequately unified, they can support much greater aggradation than the factories in other environments leading to barrier reef morphologies and back-reef lagoons where adequate accommodation is available. The degree of aggradation also can be limited by the rate and amplitude of accommodation generation (Figure 15b). Inadequate accommodation can dampen aggradation, thus equalizing the potential across different factories. Although generalities exist in carbonate production potential between carbonate factories of different ages, their constituent members have changed significantly over time and in some cases, the aggradation potential is more even across an entire platform (Figure 15c).

3.3.1. Reef Builders of the Great Barrier Reef Province

The GBR is known for its vast biodiversity [338,339] but the dominant, and ‘namesake,’ reef builders of the GBR are scleractinian corals. As keystone taxa they have been the key focus of most recent conservation studies [15,22,72,75,79] and have been considered the main or only constituent of framework by many workers [31,340]. What we know about the broader GBRP carbonate factory is based almost entirely on surface sediments [135,175,341] and current living substrates [217,342,343,344,345], supplemented only in places by underlying Holocene reef cores [80,196,219,346]. Broader biodiversity has been sampled and modeled using a variety of techniques [338,347,348,349,350], but the large scale of the province has put a major focus on satellite proxy data [351,352,353,354,355]. Very few cores penetrate the immediately underlying Late Pleistocene sections [80,197,217,356]. Data are much more limited for older (late Eocene to early Pleistocene) carbonate platforms of the northern GBRP, which are known primarily from seismic stratigraphy [190,201,212]. Where recovered from exploration wells or dredging from exposures on the sea floor typical older back-reef platform facies have been identified with, for example, abundant LBFs (e.g., nummulitids) and CCA rhodoliths in Miocene strata [211]. Reef margin facies have never been adequately described and may not have been encountered in core. Extensive drowned Miocene carbonate platforms also occur offshore from the GBRP on the Marion and Queensland plateaus [202,203,204], but of necessity this review focuses on the Late Pleistocene sections, i.e., the modern GBR, where data are better constrained. Modern (i.e., Late Pleistocene) carbonate factories differ greatly from those of their poorly known Tertiary and early Quaternary progenitors based on scant existing evidence. However, based on that evidence, older carbonate factories of the GBRP are consistent with better known, similarly aged equivalents on the northwest shelf of Australia where there has been more subsurface exploration [357,358,359].

On the Queensland continental shelf proper, the deepest reef cores (Michaelmas—1926; Heron Reef—1937; Wreck Reef—1959; Ribbon Reef 5—1995) (Figure 9a) did not encounter reef facies older than the Late Pleistocene [182]. The succession at RR5 was divided into three major sections. The lowest section consists of mudstones, packstones and grainstones interpreted at the base as a deep slope or ramp environment [182]. Carbonate sand and gravel were transported downslope from shallower settings and include coral, CCA, LBFs, Halimeda and bryozoans, but the muds represent deeper facies and contain planktic foraminifers. Higher in the section an in situ carbonate factory of corals, CCA, bivalves, bryozoan and echinoderms was interpreted to represent a deeper and cooler, non-reefal carbonate setting [182,197,356]. The middle section is dominated by in situ deeper CCA rhodolith beds amid wackestone and packstone. It contains a few thin intervals of shallow reef facies but is mostly interpreted as a non-reefal, shallow shelf. Only the upper 96 m of core, the youngest section, consists of true late Pleistocene reef facies [182,197,356]. The core did not reach crystalline basement, but shows a general shallowing-up trend [182], with CCA assemblages showing a pronounced shift from deeper melobesioid assemblages at the base to shallower, cool-water lithophylloid assemblages at the top of the middle section to shallow tropical mastophoroid assemblages in the true reef section at the top [356]. Six cycles of reef development separated by exposure surfaces occur in the uppermost section, each typically having lower-energy, massive coral assemblages at the base grading into shallower, higher-energy assemblages at the tops [197] (Figure 10b). Each cycle is interpreted as a parasequence.

Deep cores farther south at Heron and Wreck reefs have not been so well studied. The Wreck Reef petroleum exploration well reached silicified conglomerate basement at 547 m deep, but was not fully cored and did not sample any Holocene reef facies [199]. The top 121 m originally interpreted as ‘Recent’ is likely Pleistocene/Holocene reef facies overlying Pleistocene quartz sandstone. The underlying sequence primarily represents deeper calcareous marine sediments thought to be Pliocene in age down to 288 m with additional marine carbonates and siliciclastics of Miocene age below that to 547 m depth based on the occurrence of LBFs including Cycloclypeus and Lepidocyclina [199]. No reef facies were encountered lower in the core and the upper section of reef facies was not described. The Heron bore consists of an upper 156 m of stacked reef facies, similar to those at RR5, underlain by calcareous quartz sandstone and foraminiferal facies [198]. It did not reach hard basement at 223 m depth where drilling stopped. The 154 m thick Late Pleistocene reef sequence at the top contains six cycles, as at RR5, similarly interpreted as parasequences associated with glacial—interglacial cycles [47,360]. Although it was reported previously that the lower parts of the carbonate succession were Pliocene in age, based on foraminifers [201,361], the entire reef sequence is now considered Late Pleistocene, but radiometric dates are lacking.

Quantification of the carbonate factory is very difficult for a variety of reasons. Although some authors have suggested that corals alone are responsible for as much as 95.5% of carbonate production on the GBR (see case study 2 in [107]), such analyses are based primarily on modeled coral growth rates and coral cover data from reef slopes. They do not take into account the broader carbonate factory, including cryptic communities, and or sediment producers from other parts of the reef, like reef flats. They also generally fail to measure the carbonates that make up the reef edifice as preserved in reef cores. The broader modern GBR carbonate factory is dominated by scleractinian corals, CCA, Halimeda, LBFs, mollusks [134,135,137,362,363,364] and microbial biofilms (that make microbialites) [129,132,158,170] (Figure 4;

Table 1. Main carbonate factories and their functional roles for the Great Barrier Reef Province (GBRP) and exposed Devonian Great Barrier Reef (DGBR).

	Modern GBRP	DGBR CF1–2	DGBR CF3	DGBR CF4–5
dominant carbonate factory	scleractinian coral, crustose coralline algae, Halimeda, large benthic foraminifer, mollusk, microbial biofilm	massive, branching stromatoporoid, rugose coral, tabulate coral, brachiopod, crinoid, microbial biofilm	massive, branching stromatoporoid, calcimicrobes, microbial biofilm, rugose coral, brachiopod, crinoid mollusk	calcimicrobes, microbial biofilm mollusk
functional constructors *	scleractinian coral, crustose coralline algae	massive, branching stromatoporoid, rugose coral, tabulate coral	massive, branching stromatoporoid, calcimicrobes	calcimicrobes, micritic microbialite, tabular stromatoporoid
binders *	crustose coralline algae, scleractinian coral, micritic microbialite	none	calcimicrobes, marine cement micritic microbialite	calcimicrobes, marine cement micritic microbialite
bafflers *	branching Acropora Halimeda	branching stromatoporoids and corals	branching stromatoporoids and corals	None (?)

* Dominant forms boldface.

). Coral diversity decreases from north to south along the GBRP [365,366] but has had some medium scale (multi-decadal) monitoring in places [344,355]. Assemblages vary from lower-energy, more turbid, nutrient-rich, inshore to more exposed, open marine, offshore settings [347,367] with specialized coral communities adapted to more turbid inshore environments [192,193]. Corals in mesophotic settings (30–150 m depth) are becoming better known as these environments are evaluated as potential refugia in the face of increased coral bleaching [368,369]. In more open, clean settings, bathymetry and energy levels are the key determinants of coral ecology [197]. Algal assemblages also vary inshore to offshore [370] and non-calcareous macroalgae assemblages show complex patterns of distribution based on a variety of substrate and local environmental conditions [371]. Distributions of CCA have been poorly studied across the GBR as a whole [370,372], but bathymetric trends are well established for shallow reef-building forms [217,356] and CCA death assemblages correlate well with those of corals [46,217,288]. For the southern GBR, windward and leeward coralgal assemblages were defined for 0–6 m, 6–20 m, and 20–30 m depth [217]. The shallowest habitats are dominated by robust branching and corymbose Acropora, Isopora and Pocillopora corals with Porolithon onkodes and a variety of other CCA giving way to platy acroporids and massive Porites in deeper settings with more variable CCA [217]. Importantly, CCA thickness decreases significantly below 6 m depth, as does the abundance of vermetid gastropods encrusted within the thalli [217]. In the back-reef lagoon, extensive Halimeda bioherms occur in water depths of 20 to 50 m and with an area of >6000 km², they occupy more than twice the area covered by coral reefs [317,373]. The bioherms reach as much as 20 m in height and initiated ~10,000 yr ago, when the shelf was first flooded during the deglacial [374]. Other non-coral members of the carbonate factory have been studied relatively little in terms of distribution, e.g., [107,342], production rates or volumes through time.

Interest in reef sand has increased owing to its significance in carbonate budgets [41,364,375,376] and as evidence of bioerosion rates [377,378], but its constituents have rarely been quantified directly [135], commonly being estimated using hydrochemical, census-based or bulk accumulation estimation techniques [378]. Abundance of particular carbonate producers in sand varies such that less mature reefs have greater proportions of coral and CCA, whereas development of broader reef flats leads to a greater proportion of the Halimeda, mollusks and LBFs that live in those shallow environments [175]. Some mature reef flat sands are dominated by LBFs [137,175,379]. Hence, composition data from reef sands may be biased to a particular stage of reef development or sub-environment (e.g., mature versus immature reef flat), in a similar way that coral cover analyses on reef slopes may bias carbonate production data towards coral. However, sands do provide a means of sampling a broad range of the carbonate factory owing to today’s high bioerosion rates, where many larger carbonate producers are reduced to sand, e.g., [377], and the sheer volume that sands make up in the reef edifice. Reefal microbialites are a commonly neglected aspect of carbonate production on modern coral reefs, primarily because they are mostly cryptic in occurrence and only became widely recognized in the mid-1990s [172,380]. However, where quantified in Holocene sections in the GBR they account for 7.2% of total framework volume (including cavities) and ~13.6% of preserved carbonate in shallow, leeward reef framework, compared to ~32.4% corals in growth position, ~30% CCA and ~21% rubble (mostly coral, CCA and mollusk) [176] (Figure 16). Although generally not quantified, even where recognized, microbialites have been recovered in many if not most GBR Holocene reef cores [319,346,381]. During the LGM and deglacial, GBR microbialites were much more abundant, including both typical shallow-water cryptic forms and deeper benthic forms that lacked coral substrates and were not restricted to crypts [132,170]. Where quantified as percentage area on cut core surfaces microbialites account for ~3.1% to 81.6% of LGM and deglacial reef frameworks, with microbialite dominating mainly deeper slope settings [170]. Although some microbialite formed at the tops of successions as individual reefs drowned, most of it was coeval with the coralgal framework facies with which it occurs [170].

Typical high-energy GBR reef frameworks prograding at sea level are classified as skeletal–microbial frameworks consisting of robust branching, corymbose and digitate corals encrusted by CCA at the exposed surface with growth cavities and hollows filled by coral rubble, itself encrusted by CCA, and in some cases sand before encrustation within larger cavities and penetration of finer pore spaces in skeletal remains by cryptic microbialites (Figure 16). However, the percentages of corals in growth position, coarse rubble and microbialites vary [176,197,217,346]. Frameworks aggrading from greater depths, or prograding at greater depths on slopes, contain more massive and platy corals with coarse rubble and less CCA encrustation [197,217,346]. Additionally, the abundance of microbialites in high-energy windward settings is less well constrained, and although they occur in younger framework sections (e.g., 1–2 kya BP), they have not been adequately quantified in the rare cores of those ages. During the LGM and deglacial, reef frameworks were mixed skeletal–microbial or microbial, containing abundant microbialites (to >50% volume) with corals in growth position and some deeper slope settings were dominated by massive or laminated microbialite frameworks [46,132,170,288].

Over the last six interglacials, coralgal assemblages in RR5 reassembled after each successive lowstand [197] (Figure 10), in a similar way to coeval reef communities in northern PNG [59]. Deeper or cooler assemblages at the base of each cycle were dominated by massive Porites, faviids and Montipora, which gave way to robust Acropora, Isopora, Stylophora, Pocillopora, Goniastea and Platygyra in shallower, higher-energy settings at the top of each cycle. During glacial intervals, reefs tracked the lower sea levels down to the glacial maxima and then tracked deglacial sea levels back up, at least for MIS5 to MIS1 [46] (Figure 10). Similar coral assemblages occurred in the glacial reefs [288], but cycles commonly went from deeper at the base to shallower and then deeper again before individual reefs drowned where aggradation did not keep up with deglacial sea level rise [46]. However, in areas with abundant terrigenous flux, the reef facies were less well developed and had more turbidity tolerant species [288]. Drowning of reefs was associated both with rapid sea level rise and intervals of higher turbidity and nutrient flux from adjacent continental runoff (Figure 10). Hence, the Pleistocene reef carbonate factory has been very resilient through the last glacial cycle. However, detailed comparison of Holocene and underlying warmer LGM reef facies shows some significant differences. Although the carbonate factories are basically the same, and the LGM reef reached sea level with a similar catch-up succession to that in the Holocene [197,218,346], initial phases of growth suggest that once the platform was colonized by corals a rapid deepening up sequence occurred resulting in dominance of Halimeda and deeper-turbid coralgal assemblages [217]. Timing of the event suggests that the rapid sea level rise was associated with meltwater pulse 2B (MIS 6) of the penultimate deglacial [382]. That, and/or increased nutrient supply from upwelling may have led to the near drowning of the reef before the carbonate factories recovered adequately to catch up with the stabilizing sea level [217]. Hence, each parasequence was not identical and the warmer, higher-sea-level LIG differed from the current Holocene based on limited existing cores.

Aside from the carbonate factory itself, many other dwellers including vertebrates (e.g., fish, cetaceans, turtles) and invertebrates (e.g., arthropods, annelids, shell-less mollusks) occur in GBRP waters. More than 16,500 species occur across the habitats, including 450 scleractinian corals, at least 6000 mollusks and 630 echinoderms [339], and that does not include calcareous or silicious tests of LBFs and plankton. Biodiversity for many groups (e.g., corals, some fish) increases from south to north with temperature [339]. The understood biodiversity, in particular for corals, is probably underestimated, as modern genetic studies identify cryptic species in what were once thought to be very well-known and well-studied taxa (e.g., Stylophora pistillata [383]). Regardless, the overall habitat is highly differentiated and some taxa prefer more inshore, turbid or open shelf, inter-reef settings [338,347] whereas others are restricted to shallow reefs or mesophotic depths depending on individual requirements and along known trends (e.g., nutrients [146]). The destroyer guild is very well developed with scarid fish being the dominant raspers [377] but polyplacophorans are also abundant. Rasping echinoids are not generally acknowledged as major destroyers in the GBR but boring bivalves and worms are abundant [384]. Excavating sponges are ubiquitous [385] as are endolithic cyanobacteria, green algae and fungi [384]. Corallivores, such as Acanthaster planci (Crown of Thorns Starfish), do not actually damage carbonate substrates, but kill significant amounts of live coral during outbreaks, leaving the exposed skeleton more susceptible to bioeroders.

3.3.2. Reef Builders of the Devonian Great Barrier Reef

Perhaps the most comprehensive general descriptions of carbonate petrology and reef facies from the DGBR remain those of Kerans [236] and Copp [386], despite many more recent works focused on more regionally or temporally constrained parts of the complex, e.g., [265,266,270,285,387,388] and the major review by the Geological Survey of Western Australia [90]. Key members of the DGBR carbonate factory were summarized by Playford [88] and include stromatoporoid sponges [389], corals [390,391] (Figure 17), calcimicrobes [236,392], microbial biofilms (makers of stromatolites, thrombolites, many oncoids and probably ooids) [87] (Figure 18), receptaculitids, calcareous algae [392], brachiopods [393,394], crinoids, gastropods and bivalves [99,236,395] (Figure 19). Stromatolites [282,396,397] and siliceous lithistid [398] and soft sponges [236] also were significant on reef slopes and in deeper settings (Figure 19h) and encrusted broad event surfaces in the middle to late Frasnian [283]. Soft and siliceous sponges, although not carbonate producers, enhanced carbonate production by promoting carbonate precipitation through microbial metabolic processes [399], thus localizing microbialite formation as in Mississippian and Jurassic reefs [98]. Carbonate factories changed significantly through the Late Devonian, with the most drastic changes associated with the Frasnian–Famennian extinction event [84], but also the earlier end-Givetian event [70], which coincided with the end of the peak of Devonian reef growth globally [85]. In general, three major sets of organisms dominated the carbonate factories through DGBR time: Givetian–early Frasnian stromatoporoids and corals; middle–late Frasnian microbes and stromatoporoids; and Famennian microbes [90] (Figure 17 and Figure 18). However, as with the GBRP, the reef complex was extensive with complex geography and significant and abrupt changes in accommodation due to active tectonism leading to significant temporal and/or regional environmental differences in factories [236,267,400]. These major differences allow more or less distinct carbonate factories to be recognized (Figure 14). The carbonate factories specified below represent a much larger spectrum and combine some of the previously defined lithofacies and lithotopes [236] but are intended to represent the generalized temporal succession. However, difficulties in correlating shallow facies in the DGBR render differentiation of exact temporal versus spatial–geographic relationships difficult [269,271,401]. The exact relation of margin framework facies, for example, to coeval back-reef facies is hampered by incomplete exposure and subsequent erosion. Some margin facies are known primarily from debris on associated slopes, where more complete exposures may occur, e.g., [265,266,270,402].

Initial carbonate deposits (Carbonate Factory 1: CF1) were confined mostly to low-relief platforms and banks that initially may have had ramps, rather than steeper shelf edges, on the seaward sides [280,403]. These mid-late Givetian to early Frasnian platforms typically display m-scale, mostly subaqueous parasequences in their interiors [90,275,276,277] (Figure 20 and Figure 21) and in some places, at their margins [267] (Figure 22). Higher in section, and/or possibly in more open settings, Early Frasnian CF2 margins of low- to steeper-relief platforms became more stromatoporoid-rich with more abrupt, but still low-relief laminar stromatoporoid buildups at their margins. Back-reef areas had similar variable carbonate producers to CF1. In the early middle to late Frasnian, CF3 margin facies changed considerably with a major increase in the calcimicrobe Renalcis (used in the broadest sense for similar renalcid-like microfossils that can be divided into separate taxa, e.g., Renalcis, Shuguria, Izhella, etc. [404,405,406]) (Figure 18a,b) and different stromatoporoids, including much more abundant branching Stachyodes on the reef flat and proximal fore-reef slope and laminar S. australe on the margin [88,236,407]. Early marine cement became much more abundant at this time and increased aggradation led to larger framework cavity systems and higher relief with steep, upright shelf edges [88,236,267]. These three factories all occurred within the transgressive Pillara phase of the supersequence (Figure 14). Towards the end of the Frasnian, the abundance of stromatoporoids decreased in the CF4 margin to be replaced increasingly by low-domal stromatolites [88,236,267,402]. This may represent the beginning of the regressive Nullara phase. Stromatoporoids and colonial corals were almost entirely removed at the Frasnian–Famennian boundary extinction event following the previous major collapse at the end of the Givetian [84]. Subsequent early Famennian CF5 margins, where studied, were constructed primarily by microbes [88,236,388,408]. Key members for the different carbonate factories are shown in Table 1 with dominant ecological roles of reef margin and reef flat organisms.

Carbonate Factory 1 is highly variable with significant ecological shifts, e.g., [70,407]. Its age is late Givetian probably to early middle Frasnian (conodont zone 5) [270], but it is very difficult to correlate exposures even within the Pillara Range [409]. Basal CF1 carbonates are Frasnian to the north in the Hull Range [270,280] reflecting transgression and coastal onlap. Although the CF1 carbonate factories were highly varied they are included together here as they built similar low-relief carbonate banks. The earliest parts, to the south in the Pillara Range, were dominated in the east by the delicate branching stromatoporoid Amphipora (Figure 17e), which occurred mostly in more protected platform interiors (back-reef settings) [403,409]. That transgressive unit occurred non-conformably over Precambrian basement relief with feldspathic sandstone flooring paleovalleys and abundant siltstone units in the lower parts. The unit is less dominated by Amphipora to the west [407] and is absent farther to the north and landward in the Hull Range [285]. Succeeding bank facies are cyclic with stromatoporoid (Actinostroma) and fenestral facies or just fenestral facies to the west that then become more coral-rich with abundant massive Argutastrea (Figure 17f,g) before passing into more stromatoporoid-dominated cycles up section [409]. In offshore platforms (e.g., Pillara Range), the cycles represent low-relief biostromes in the bank interior that typically alternate between shallower, more resistant stromatoporoid and intertidal facies and recessive, deeper coral facies [70,267,275,276] (Figure 20). A composite cycle at Menyous Gap (Figure 21a) includes eight typical facies from base to top: (1) rare recessive carbonate mudstone–wackestone; (2) recessive coral facies dominated by the corals Disphyllum, Argutastrea, Alveolites, and/or Thamnopora in growth position, but crushed, with minor thin stromatoporoids in a muddy matrix; (3) more resistant tabular stromatoporoid facies characterized by tabular to low-domal constratal stromatoporoids in growth position in muddy matrix (platestone—equivalent to floatstone of [236]); (4) much more resistant domal stromatoporoid facies with more abundant, large (60 cm) stromatoporoids in growth position in variable wackestone–packstone matrix; (5) resistant Stachyodes facies consisting of abundant Stachyodes (stromatoporoid) floatstone with peloidal packstone–grainstone matrix; (6) Amphipora (stromatoporoid) facies consisting of Amphipora floatstone in peloidal wackestone–packstone matrix; (7) resistant peloidal facies consisting of well-sorted peloidal packstone–grainstone; and (8) resistant fenestral limestone facies consisting of peloidal packstone–grainstone containing abundant fenestrae (see [276] for more detail). Elsewhere cycles proximal to the margin may be capped by a high-energy oncoid facies (Figure 21b, facies 9), whereas shore proximal cycles can contain coastal siliciclastic units (facies 11) at lowstands and even fan deltas (facies 10) (Figure 21c) [292]. Exposure surfaces are very rare at the tops of Pillara cycles [275] and many cycles are incomplete, never having had the shallowest facies deposited [267,276]. Cycles generally shallow to the north with those in the Guppy Hills and parts of the southern Hull Range [280] being deeper to shallow subtidal, similar to those in the Pillara Range [276], but with fewer deeper coral facies and many more intertidal facies at cycle tops in the northern Hull Range [280]. Bank margins range from carbonate ramps [403] to more abrupt roll-over margins with high-energy oncoid facies and rare lagoonal facies protected behind thicker margin buildups [267] (Figure 22). Cycles are interpreted as parasequences resulting from Milankovitch forced minor eustatic sea level oscillations [277,278]. Shore attached cycles with major siliciclastic content are more complicated to interpret and have additional tectonic and climatic controls superposed on eustacy [90,278,280].

In perhaps the most detailed biological study of the cyclic successions [70] biotic membership in the back-reef platform interior was quantified for as many as 70 successive parasequences at Menyous Gap, Pillara Range (Figure 20a,b). Those data show how variable the biostromes were, but cluster analysis supported the major differences between deeper and shallower communities (i.e., typical parasequences). Overall, they found that cycles tended to have relatively stable community memberships over short time scales, but that over longer time scales taxonomic shifts occurred. In particular, they noted three major temporal shifts in broader associations of shallow biostromes from base to top: (1) branching rugose coral (Disphyllum) massive stromatoporoid (domes to >1 m in diameter); (2) stromatoporoids largely lacking corals, but with greater diversity of stromatoporoid morphology (massive, irregular and tabular); and (3) branching tabulate (alveolitids) and massive rugose (Argutastrea to >1 m diameter) coral and massive stromatoporoid. They [70] provisionally attributed the shifts in coral communities to changing sea surface temperatures near the end of the Givetian [410] but noted that the succession is not adequately dated for firm conclusions. Owing to the focus on corals and stromatoporoids in growth position, it is difficult to quantify the contributions to the carbonate factory of more fragile branching taxa, such as Stachyodes, Amphipora and Disphyllum, but the data uniquely highlight the long-term biostrome community dynamics while demonstrating some of the spatial and temporal heterogeneity (i.e., complexity) within the setting.

A retrograding CF1 platform margin is exposed at the northern end of Menyous Gap for the youngest few cycles [267]. Cycles at this margin near Menyous Gap may contain some of the highest energy facies (e.g., the oncoid facies), although subaqueous throughout, and entire cycles roll over the margin intact (Figure 22a,f). Although the current topographic expression of the northern margin of the limestone is eroded into a relatively abrupt, more or less straight outcrop (Figure 20a), the platform margin was irregular, with seaward projecting extensions potentially like buttresses exposed far to the east at Teichert Hills (Figure 23). Those structures were interpreted as spur and groove morphology [411] and contain a constructional margin community dominated by sloping tabular tabulate corals with minor low-domal stromatoporoids (Figure 23). The proximal fore-bank slope tabular tabulate facies at Menyous Gap [267] is very similar (Figure 22e). Shallowest intertidal and oncoid facies were not developed at the exposure within Menyous Gap itself [267], but nearby the deeper recessive facies underwent erosion above wave base beneath overlying shallower margin deposits when sea level fell (Figure 22a,h). This erosional truncation of recessive beds at the margin provides evidence that the cycles were indeed driven by changes in sea level [267] and were not autocyclic (formed by progradation alone—the Ginsburg model [183]). The shallow margin thickened to varying degrees to have some relief relative to the back-reef facies and consisted of constratal tabular stromatoporoids floating in a matrix of peloidal grainstone forming a platestone facies. This facies is not to be confused with the recessive ‘tabular stromatoporid’ facies of other authors [70]. It is not clear to what degree it may have been ‘wave-resistant,’ as the observed examples were poorly winnowed and clearly subtidal. At the margin the facies passes laterally into massive proximal slope facies of carbonate mudstone (microbial?) to packstone containing isolated, sloping tabular tabulate corals (alveolitids) in growth position (Figure 22e). Cement-filled cavities occur beneath overhanging alveolitids on the slope and coarse debris from higher on the platform is abundant. Recessive (deeper) facies continue over the margin largely unchanged except where pinched from below by higher-relief shallow margin facies (Figure 22a) and in places eroded by the succeeding shallower environment (Figure 22g,h) [267]. Other members of the carbonate factory, such as crinoids, bryozoans, brachiopods and the microbes that made peloids (and potentially ooids [412,413,414]) were also active, but their contributions have not been quantified. Where mounded margins accreted higher than platform interiors, microbes may have been significant, as skeletal frameworks were not developed and the isolated, non-encrusting stromatoporoids were constratal. It is unclear if they developed any significant rigidity, but in some cases protected deeper muddy (lagoonal) facies containing a mollusk fauna behind them (Figure 22b,c). Deeper facies at the toes of the banks include microbial mounds containing debris from up slope in clotted mudstone with subaqueous fenestrae [285], and sponges and stromatolites were likely present.

Carbonate Factory 2 represents a shift to greater tabular stromatoporoid abundance near the margin. Within the Pillara Range it may represent part of the lower-stromatoporoid unit, may be equivalent to the lower-reef-cap member in the western Pillara Range or may extend into the upper-stromatoporoid members [407]. Difficulty in correlation renders its spatial versus temporal distribution unclear and it may be distinctive only at the margin. Regardless, it occurs early in the prominent early Frasnian back-stepping phase, but before the major shift to aggradation associated with the CF3 association (Frasnian platform phases 3 or 4; conodont zones 4–5) [270] (Figure 14). CF2 margins are dominated by stacked constratal tabular stromatoporoids (generally less than 10 cm thick), in some cases reaching total thicknesses >2 m (Figure 24). These low relief frameworks are similar to the laminar stromatoporoid framestone of low-relief reef margins in the middle Frasnian Alexandra Reef Complex #2 of the Northwest Territories, Canada [168] and are classified as skeletal frameworks. Reef flat facies are difficult to define owing in part to the lack of abundant Renalcis, which is an indicator of reef flat facies higher in section (CF3) [236], but proximal back-reef to reef flat facies include grainstones with scattered constratal tabular stromatoporoid platestones and abundant brachiopod beds, crinoids and peloids. At Glenister Knolls exposures of the margin are commonly missing owing to margin collapse. The arcuate nature of the present margin of the knolls in map view combined with abundant debris in associated slope deposits and clear failure surfaces that cut into back-reef facies prior to encrustation by Rothpletzella (Figure 18c) suggest that some margins were quite rigid and even back-reef facies were well indurated prior to collapse events. However, the stratigraphy is very complex in three dimensions with younger, deeper facies overlaying older, shallower facies [415]. Some apparent in situ margin facies represent steep collapse scarps and it can be difficult to discriminate original margin slopes from winnowed or collapsed slopes. They contain very little Renalcis where margins can be traced directly into coeval adjoining slope and where a laminar stromatoporoid skeletal framework facies does not occur at the margin, non-encrusting marginal tabular stromatoporoids rarely have original dips greater than 22° relative to associated geopetal horizons. In other places, the substrate was not well indurated at these margins and sediment eroded from beneath tabular stromatoporoids and tabulate corals causing them to collapse seaward, steepening them, and forming shelter cavities beneath (Figure 17c). The nature of the outcrop makes it unclear as to where in a parasequence the less indurated facies is in respect to the thick laminated stromatoporoid facies but the two margin types may represent the deeper and shallower parts of a parasequence, respectively. An additional complication comes from the back-stepping itself, as it superposed the deeper facies over older, relief forming shallow facies formed earlier during the transgression [415]. Regardless, CF2 is associated with major reef drowning and back-stepping events in the early Frasnian. The skeletal stromatoporoid margins were not able to aggrade and keep up or catch up to sea level. Although Rothpletzella occurs (especially on slopes) throughout the Late Devonian succession, drowned reef margins were commonly encrusted by laterally persistent columnar Rothpletzella crusts as much as 50 cm thick, consistent with its abundance in deeper slope settings [88,90]. Some layers may mark individual drowning events.

Carbonate Factory 3 is entirely Frasnian in age (conodont zones 4?–11 or 12 [269], Frasnian sequences 2 to 5 [265]) (Figure 14) and differs from previous factories most prominently at the margin where suprastral branching stromatoporoids were more abundant and commonly supported by the calcimicrobe Renalcis and abundant early cements. This allowed them to aggrade and keep up better with increasing RSLs leading to a shift from the earlier low-relief bank facies to higher-energy, upright reef margins [236,267]. Margin facies tend to be massive without obvious bedding but may be thin zones laterally of just a few to tens of meters width (Figure 25a). Generalities are difficult to make as margins frequently experienced collapse. Key constructors at the margin were low-domal to complex stromatoporoids or more laminar forms, such as Stachyodes australis (Figure 25c,d), supported and suspended above the sediment–water interface by Renalcis and early marine cements. In other areas thicker tabular stromatoporoids were well bound by Renalcis and cements commonly holding open very large framework cavities, to be filled subsequently by cycles of internal sediment and cement (Figure 25f). Spatial or temporal trends in dominant CF3 margin constructors are difficult to constrain, but they include the constructors of the following lithotopes [236]: coral–stromatoporoid framestone; laminar stromatoporoid framestone; and Renalcis–Spharocodium (i.e., Rothpletzella)–stromatoporoid bindstone. Cemented oolitic grainstone shoal margins also occur in places (e.g., at Nardji Cave [265]). Proximal reef flats could contain similar facies but with framework cavities partially filled by peloidal microbial fabrics and minor bioclasts (crinoid, brachiopod and mollusk debris). More commonly, branching Stachyodes encrusted by Renalcis dominated the proximal reef flat (Figure 25e). The strength of the framework allowed subsequent filling of framework cavities by a variety of platform sediments, including siliciclastic sands from the mainland [270], and where pieces of reef margin and reef flat facies collapsed into the adjacent basin (e.g., McIntyre Knolls) the final internal sediment fillings were commonly basinal muds. Such collapse was particularly frequent in the earliest phases of Frasnian sequence 5 as highstand deposits initially built over basin facies that could not support progradation of new margins [265]. Some high-energy windward settings also had large oncoids and round, potentially mobile bulbous stromatoporoids floating in mud-lump (micritized grain) grainstones just behind the reef crest [290,292]. In other areas, probably later in the Frasnian, large Renalcis-bound peloidal sediment masses occurred (Figure 25g), but it is unclear exactly where they fit within the stratigraphy, as they mostly represent debris collapsed from the margin. Similar facies become more abundant in CF4 and CF5.

Cycles are not easy to discern in massive margin facies, but back-reef areas alternate between shallower and deeper facies and with increasingly prominent bedding, the reef flat passes into back-reef facies with the loss of abundant Renalcis [236]. Reef flat facies may pass into Amphipora rudstones and in deeper parts of cycles the branching coral (Disphyllum) [292]. Typical high-energy back-reef cycles have scattered massive stromatoporoid biostromes at the base and shoal up through mud-lump and intraclastic grainstones containing oncoids with common gastropods to terminate in shoaling oolitic grainstones with fenestral fabrics at cycle tops [290]. Where ooids do not occur, fenestral algal laminites made of mud-lump grainstones with linked fenestrae and exposure surfaces with desiccation cracks occur [292]. More protected distal reef flat facies contain peloidal and mud-lump grainstones with domal stromatoporoids, branching Stachyodes and Amphipora and a diverse fauna of mollusks with abundant gastropods [293].

Proximal fore-reef slopes contain a variety of boundstones, platestones, rudstones, floatstones, grainstones and packstones containing predominantly branching stromatoporoid debris [236]. Stachyodes is very common in the fore-reef slope where it occurs encrusted by Renalcis in places or as open, well-winnowed and cemented rudstones providing significant rigidity to proximal fore-reef slope environments (Figure 25). Where the Stachyodes was cemented adequately, these facies may be considered biocementstones (Figure 25h). Although tabular–laminar stromatoporoid platestones occur in places within grainstone (Figure 19c), Renalcis and Rothpletzella were important on some slopes and receptaculitids and lithistid sponges may be important [236] (Figure 19g). Farther down slope stromatolites and soft sponges became more abundant as well (Figure 19h). Debris flows of shallower shelf facies are common in many horizons on the slope [88,265,270].

Carbonate Factory 4 is most abundant towards the end of the Frasnian (Frasnian sequences (?)5–top, 6 and 7 [265,269]) and has increased microbial input and decreased skeletal reef builders [416], particularly missing massive and tabular stromatoporoids [402], but also with fewer branching rugose and tabulate corals and branching sponges [236]. At steep margins (Figure 26), microbe-dominated frameworks include microlenticular–cyanobacterial and Renalcis–fenestral bindstones [236]. Early cements still provided significant rigidity for margin facies, but although margin collapse was not as common as earlier in the Frasnian, margins are commonly missing or difficult to access, so are poorly known [265]. Dominant margin frameworks are low-relief calcimicrobial and peloidal cementstones.

CF4 reef flat and back-reef facies are dominated by microbial deposits, but they are poorly known owing to erosion [402] and their exact relationship to margin facies is challenging. Reef flat and back-reef facies developed at the top of Windjana Gorge may represent the end of CF3 or CF4 (Figure 26a,b). The Actinostroma community [387] contains a variety of very large (to 5 m diameter), but isolated stromatoporoids (Figure 26c,d), commonly encrusted by Renalcis and other microbialites in largely constratal settings in bedded grainstones to packstones. They supported some cavities containing early marine cements, microbes and internal sediments. The previously described Stachyodes–Shuguria (i.e., Renalcis) community [387] most likely represents the younger CF3 factory, but the fenestral ‘microbialite-forming community’ and ‘Rothpletzella stromatoporoid community’ [387] are characteristic of CF4 reef flats. Cyclicity becomes more obvious in back-reef facies [89,90] with more isolated stromatoporoids in the deeper parts of cycles shoaling up to fenestral microbialites in the shallowest parts (Figure 20c). At Windjana Gorge, abundant siliciclastic sediments from the proximal coast to the north interfinger with the microbe-dominated facies. The proximal fore-reef slope community is similar to that of CF3 with some remaining laminar stromatoporoids and lithistid sponges and abundant calcimicrobes in stromatactis-rich microbial facies [236]. Small sponge–rugose coral mounds occur in the fore-reef slope in the Napier Range [282] and some of these small mounds continued into the Famennian [397,402].

Carbonate Factory 5 represents the prograding Famennian Nullara cycle [88] after the Frasnian–Famennian extinction ‘event’ where most, but not all, stromatoporoids were removed from the DGBR [88,389]. Stromatoporoid abundance was already decreasing in CF4 [402,416] and CF5 intergrades with CF4 to a large degree, rather than representing a sharp boundary. CF5 margins were still abrupt (Figure 26e) and dominated almost entirely by microbes with diverse calcimicrobes, but calcareous algae, encrusting foraminifers and problematica also occurred [236,388,392,408]. Microbialites include micritic (i.e., non-skeletal microbialites [158]) microdigitate thrombolites (clot-shaped) and stromatolites (laminated) forming bedded and mounded structures and calcimicrobes are dominated by Renalcis, Rothpletzella and Girvanella (Figure 26h). Calcareous algae include both solenoporoids and Parachaetetes. Minor laminar stromatoporoids still occurred in the margin, at least in the lower Famennian [236,408], but their role in the framework was incidental. Lithistid sponges were more abundant. Minor bryozoans, encrusting foraminifers and problematica contributed to very diverse local factories, and account for at least four major lithotopes, some of which were similar to those in CF4 and the most dominant of which is a Uralinella–Renalcis–Rothpleztella bindstone [236]. Most frameworks would be classed as micritic microbialite or mixed calcimicrobial–micritic microbial frameworks. In all cases frameworks were massive, very porous and early marine cements were important and helped maintain relief on the high-energy margins [236]. Individual surfaces were low-relief mounds and lenses (Figure 26f), in many cases interspersed with lenses of siliciclastic sand [236] and low-relief, elongated mounds–buttresses probably extended from the margin onto the proximal fore-reef slope [266,408].

Reef flat facies graded back from the margin with decreasing cemented framework, but a columnar–fenestral stromatolitic bindstone facies is common [236] (Figure 26g). These columnar fine-grained stromatolites contain abundant carbonate or siliciclastic sand, but only minor calcimicrobes. Similar, but larger mounds occurred in Frasnian back-reef settings where there was abundant siliciclastic influence [285]. The discrete stromatolites also graded in places into more irregular thrombolitic textures, but cavities were still abundant and could be partly filled by Renalcis [388]. Other major members of the carbonate factory include the microbes that formed peloids and micritized grains and possibly the abundant ooids. Large megalodontid (Figure 19d) and lithiotid bivalves and gastropods also occurred [236]. The coarse, well-winnowed grainstones were dominated by microbial peloids, ooids, grapestones (grains made of several attached ooids), mud lumps, rounded intraclasts and minor skeletal debris and occur in discrete erosive channels that traverse the margin facies [282] or fill topography between stromatolite heads [267,388]. Back-reef facies are dominated by oolitic and intraclastic grainstones with many intraclasts themselves being oolitic grainstone, and agglutinated stromatolites have been observed [417]. Landward, a variety of peloidal, pisolitic and microbial grainstones and packstones occur with minor skeletal contributions from bivalves and gastropods. They commonly contain fenestral fabrics and interfinger with terrigenous siliciclastic sediments at the coast [236]. High-frequency Famennian cycles are recognized in slope deposits [266] and back-reef (Figure 20c) [267] but remain poorly known. Near the coast the carbonates interfinger with siliciclastic sediments [236], and siliciclastic sediments commonly extend to the margin [388], but the degree to which a given instance represents climatic, tectonic or eustatic forcing is poorly constrained. Regardless, the development of shoaling grainstones and microbial fenestral fabrics along with desiccation cracks observed in back-reef facies [417] are consistent with Milankovitch band cycles as in the Frasnian.

Fore-reef slope facies are basically continuations of the margin facies [236,266]. They consist of mounded microbial boundstones [408] (Figure 26f) with grainstones derived from the platform top and coarse debris generated from occasional margin collapse and smaller scale brecciation on the slope itself [266]. Some of the grainstones are turbidites [282]. The dominant in situ carbonate factory consists of the marginal Uralinella–Renalcis–Rothpleztella bindstone lithotope [236]), but brachiopods, gastropods and crinoids occur as in situ dwellers [402]. Cycles of platform collapse occurred throughout the sequence [90]. Distal slope boundstones and grainstones interfinger increasingly with well-bedded silty peloidal wackestones to packstones [266].

The destroyer guild is poorly represented in the DGBR [31], but massive stromatoporoids contain scattered Trypanites and Gastrochaenolites-like borings. The abundant micritized mud-lump grains in back-reef settings also presumably represent endolithic cyanobacteria and green algae, but micrite rims are not prominent on coarser skeletal allochems. Apart from the carbonate factories, the DGBR had extensive biodiversity amongst its non-carbonate producing dweller guild. Although many of these types of organisms are rarely preserved in shallow reef facies, a significant Frasnian lagerstätte developed in the deeper inter-reef environment, especially well-exposed in the Gogo Formation in Bugle Gap [418,419]. This inter-reef basin deposit provides significant insights into the co-occurring DGBR dweller guilds. Early diagenetic carbonate concretions grew around organic remains just below the seafloor in deep shelf pelagic facies preserving everything from conodonts, radiolarians [420] and arthropods [421], including eurypterids [422], to fish [256]. Like the GBRP, the DGBR was host to a diverse range of fish with more than seventy species identified in the Gogo Formation [418,423]. Fish teeth and scales also occur in the shallow platform facies [274], but pale in comparison to the lagerstätte. More typical pelagic carbonate fossils (ammonoids, nautiloids) (Figure 19i) are also preserved, but the vertebrate preservation is especially impressive, including three-dimensional preservation of delicate skeletons and preservation of many different types of soft tissues [424,425]. The deposit has revolutionized our understanding of early fish biology and evolution [418]. It is probable that the deposits formed during transgression-related Early Frasnian basinal anoxic episodes [423] that also preserved the organic matter responsible for oil generation from the Gogo Formation [233]. Localized anoxia also explains the lack of benthic scavenger activity that allowed articulated skeletons and soft part preservation. Devonian source rocks in the Gogo and younger Fairfield formations formed owing to eutrophication during transgressions associated with very high primary productivity from planktic algae [233]. Although it is unclear whether the nutrients that drove the productivity were derived from terrestrial runoff or more from upwelling from the deepening Fitzroy Trough over the entire age range of the Gogo Formation, biomarker studies are consistent with terrestrial pulses leading to photic zone euxinia at times and lower-water-column anoxia [426]. Hence, few of the larger preserved organisms in the Gogo nodules were local benthic fauna as the bottom waters were anoxic [421]. Most of the preserved fish were pelagic taxa and some other fauna may have been washed into the setting from shallower benthic settings, such as the eurypterid [422]. There are also ten genera of lungfish [423], which were adapted to stirring up shallow muds in a well-oxygenated setting [419]. Thus, the deposit is in some ways similar to the Jurassic Solnhofen Lagerstätte in Germany where shallow and even terrestrial taxa (e.g., the vertebrate Archaeopteryx and plants) were washed into deeper anoxic lagoons protected by sponge–microbial and coral reefs [427,428]. This setting allowed sampling from a much broader geography than the depositional environment itself [429].

3.4. Comparing Reef Geomorphology and Palaeogeography

Both the GBRP and exposed portion of the DGBR developed on continental shelves proximal to terrestrial terrigenous sources but with some reefs isolated on offshore fault-controlled basement highs (either exposed or subsurface) that formed during rifting of continental crust. Hence, they both represent mixed carbonate–siliciclastic settings with shore attached and more isolated, cleaner carbonate environments. Both reef provinces are confined to continental crust and true atolls on oceanic crust are absent. A few modern reefs occur on oceanic seamounts in the Coral Sea just east of the GBRP as part of the Tasmantid Seamount chain (e.g., Cato, Wreck, Kenn reefs) [430], but they are not part of the GBRP itself. Fringing reefs occur in both settings attached to the mainland and around continental islands composed of crystalline basement; barrier reefs isolate lagoons behind them, with isolated platform reefs occurring amidst deeper shelf settings. Both complexes faced deeper basins seaward, the Coral Sea, a true ocean, for the GBRP and the Fitzroy Trough, an epicontinental sea, for the DGBR. Owing to the size and complexity of the two provinces comparisons must be made at different scales. At the broadest scale (i.e., carbonate platforms) the two provinces vary considerably. The carbonate banks of CF1 and possibly CF2 in the DGBR do not have close analogs in the modern GBRP but may be similar to some of the Miocene platforms developed on the Marion and Queensland Plateaus or attached to the continent farther north (Figure 12). Equally, the vast prograding CF5 Famennian platform of the DGBR has no close analog in the GBRP. Regardless of those difference, where platforms occur at the edge of major relief, such as the ribbon reefs of the GBR facing the Coral Sea or the Frasnian and Famennian platforms of the Napier Range facing the Fitzroy Trough, great linear reef belts were formed (Figure 27a). However, the best DGBR analogs for modern GBRP reef geomorphology are the Frasnian CF3 shelf edge reefs (Figure 27b) and isolated platform (i.e., ‘pinnacle’) reefs (Figure 27c) surrounded by deeper shelf depths that formed during back-stepping. These platform-scale similarities and differences relate to the abilities of the carbonate factories to fill accommodation through time and are discussed below.

At smaller spatial and temporal scales, intra-reef geomorphology is far more comparable as the very different carbonate factories reacted in similar ways to the same types of coastal processes. Sedimentologic evidence documents many of the same types of coastal processes in the DGBR that are well known in the GBRP. The DGBR includes evidence for coastal erosion, waves, storms, tides and currents. The Middle to Late Frasnian (CF3) Mowanbini Archipelago (Oscar Range) provides a good example of Devonian reef palaeogeography [92,290,292,293] (Figure 28). The current Oscar Range stands in relief on the modern landscape as it is composed of very resistant siliceous Proterozoic metasediments and metavolcanics. During the Late Devonian virtually identical relief (90 m) formed islands that were buried progressively through time by the Frasnian reef complex [89]. Today, the upper part of the reef complex (Famennian and latest Frasnian) and overlying strata have been removed by erosion in places to expose a nearly coeval Frasnian paleogeography from the basin to the south across the reef fringed continental island and carbonate shelf to the mainland continent to the north. Examples of fringing and barrier reefs, high-energy and more protected, lower-energy lagoons, and high- and low-energy rocky shores are all well exposed. Although the archipelago was larger than Lizard Island on the GBR (Figure 9d), that too represents a group of continental islands connected by fringing and barrier reefs to enclose a shallow lagoon [431] (Figure 29). Lizard Island is surrounded by shelf depths within the broader GBR lagoon. The Mowanbini Archipelago, o the other hand, occurred on the edge of, and facing, a deep basin, the Fitzroy Trough to the south (Figure 28a–c). The western part of the largest island is encased by a fringing reef where reef limestone occurs directly against the metamorphic basement (phyllite) that makes up the low-lying southern part of the island (Figure 28d). Just to the east the reef separated from the island to become a barrier reef and a shallow, high-energy lagoon opened up behind the margin. A drainage flowing off the elevated island reached the coast near this point and fluvial sediments were reworked into a small shoal-water delta [292] (R1 in Figure 28a,e). Modern sediments being delivered by a stream most likely occupying the same channel in the very hard basement is compositionally the same as the sediment in the ancient delta, but there the sediment is better sorted, more rounded and occurs with grain-oriented fabric sloping seaward consistent with deposition on a beach (Figure 28f). The reef margin itself was not very wide and was dominated by typical CF3 Renalcis and stromatoporoids. The reef flat to back-reef shows well-developed cycles wherein the delta and other coastal siliciclastic sediments interfinger with tidal flat facies near shore, but pass into high-energy cycles just behind the margin with domal stromatoporoids and Disphyllum biostromes in the deeper parts of cycles overlain by reworked stromatoporoids, oncoids and ooid shoals in the highest energy sections (Figure 17h and Figure 21b,c) [292].

To the north along the edge of the lagoon the island has higher elevation held up by quartzite and metasediments. The southernmost quartzite ridges formed a prominent rocky shore to the lagoon [290] (Figure 29a–d). The well-bedded, nearly flat-lying lagoon sediments lapped up around individual sea stacks and immediately proximal beach sediments consist of seaward sloping grain-oriented beach breccia made up mostly of proximal quartzite clasts (Figure 29c,d). Although wave energy was not adequate to round the very hard quartzite clasts, its distribution and the orientations of toppled elongate stromatoporoids offshore showed consistent longshore current to the east, consistent with waves breaking over the margin to the south and onshore directed winds expected from heating of the continental mass to the north [290]. Smaller islands within the lagoon generally provided smaller sediment input to the shore-face, but longshore current directions were consistently to the east on the south sides of the islands. The shallow lagoon supported domal stromatoporoids as well as Stachyodes and Amphipora, but the last two taxa occur only as transported debris. Oncoids also occur in some horizons as well as abundant gastropods, which themselves became the nuclei for oncoid growth (Figure 18h and Figure 30a). Around the eastern end of the large island (Figure 28a, R3), flat bedded cyclic lagoonal facies continue with some high-energy shoals (ooids and oncoids) suggesting high hydrodynamic energy was funneled between the island groups (Figure 30b).

Across the large island to the north a small island (Figure 28a R4) crops out near Christopher Bore [293]. The island is composed of Precambrian metasediments, including metaconglomerate and phyllite, similar to the larger island to the south (Figure 30c,d). The shallow protected lagoon had a diverse fauna of Amphipora, Stachyodes, gastropods and bivalves and even scaphopods (Figure 30e–j). Although wave-worked and distributed, many of the Amphipora are well preserved. Farther offshore domal stromatoporoids occur in growth position within bioclastic grainstones, floatstones and fenestral algal laminites (Figure 30k). The sediments are commonly partially dolomitized. Although waves from the southeast dominated the protected environment, storms and larger waves occurred sporadically as indicated by toppled stromatoporoids and intraclasts in the sediment. Siliciclastic sediment also eroded from the island to be intermixed with the shore-face carbonates. Although detailed biofacies analysis has been done [293], the nature of parasequences cannot be explored owing to the very low dip of the strata and flat outcrop.

4. Discussion

Comparisons between modern and ancient reef systems suffer from biases inherent in studying the two different types of deposits (i.e., living communities with poorly accessible stratigraphy versus scattered sections of stratigraphy and paleogeography with varying degrees of preservation bias). Modern reefs are much better known in terms of everyday processes, both biological and geological, from seasonal to longer-term, stochastic event scales that drive intermediate disturbance [110]. The knowledge represents a very limited time span in the history of a carbonate system (i.e., stratigraphy). Almost all we know about modern coral reefs represents only the surface of a single ~10 kyr interglacial episode within the broader 100 kyr+ cyclicity that has driven modern reef growth over the last 500+ kyr. That knowledge has been supplemented by coring, and seismic surveys, but we had little understanding about what happened to the GBR in the majority of the time between interglacials until IODP Expedition 325 drilling in 2010 [46,288,432] (Figure 10). That said, the extent of the modern GBRP across climate zones causes the biodiversity and reef constraining factors (temperature, cyclone frequency, terrestrial runoff, etc.) to vary considerably along its length. Generalizing about the modern GBR as a single entity carries significant risk [30]. Although we now understand that the modern GBR formed over the last several Late Pleistocene glacial cycles, we know surprisingly little about the older Tertiary and Quaternary reef systems in the subsurface farther north. However, isolated modern reefs are growing on carbonate platforms that initiated in the Miocene if not earlier in the northern part of the province [201,211] (Figure 11). There may have been breaks, but the >20 million year history of the GBRP is of the same scale as the >20 million year history of the DGBR. We just have very little knowledge about most of it.

In studying fossil reefs, even the incredibly well-exposed DGBR, we have the inverse problem. We have detailed knowledge about aspects of its variation through stratigraphy (e.g., for fore-reef slopes [265,266]) with only scattered snapshots of a very complicated paleogeography for some intervals and places as allowed by existing exposure and exploration drilling (e.g., the Mowanbini Archipelago [292]). More detailed paleogeographic and paleobiological investigations are hampered by the vagaries of exposure and erosion, difficulty in correlating between different exposures and the sheer amount and age range of existing outcrop. Most biostratigraphic correlation for the DGBR is based on conodonts and ammonoids [272,273], which are restricted mainly to deeper slope and basin environments. Correlation of individual sections of shallow platform facies is still difficult, although microvertebrates [274] and chemostratigraphy [269] will help in time. Even assuming adequate correlation, the range of geographic settings provides similar problems as in the GBRP and makes it unwise to overly ‘lump’ the DGBR into a single ‘reef’ [283]. Sequence stratigraphy is very useful at different scales. The supercycle concept [88] and attempts to break the succession into shorter temporal units with different carbonate facies or factories [236,267,283] as attempted above, still leaves uncertainties as to what represents a purely temporal or spatial change in some cases (e.g., CF1 and CF2 margins). Matching to higher-frequency cycles—parasequences, e.g., [275,277] could provide very high temporal resolution but such cycles are notoriously difficult to correlate between platforms. Additionally, the exposed DGBR represents but a small part of a much larger reef system, of which we know very little as most of it is in the subsurface (Figure 12). Current exploration on the Barbwire Terrace may elucidate the reefs on the other side of the Fitzroy Trough, but existing cores appear to intersect less informative back-reef environments [233]. Despite those issues, the reef provinces described above can be compared meaningfully using the concepts of key processes, functional guilds and unified jargon (see introduction). Devonian reefs have things to tell us about modern reefs and vice versa.

4.1. Comparing GBRP and DGBR Reef Builders

Little if any overlap exists in the organisms that built these two reef systems (Table 1). Corals occurred in both provinces, but the scleractinian corals of the GBRP are most likely no more closely related to the rugose and tabulate corals of the DGBR than to modern non-skeletal anemones, although discussion continues, e.g., [433]. However, it is a testament to evolutionary convergence how similar many of the morphologies are between the different orders of corals as the skeletal needs of relatively simple cnidarian polyps are highly constrained [434]. The different coral orders also played similar functional roles in many cases in their respective carbonate environments [31]. Some of their morphologies, especially for the tabulates, include features suggestive of photosymbiosis, as with modern scleractinian corals [72,410,435]. However, it is likely that the trophic structure was quite different between the two reef provinces (more on that below). Hence, corals filled constructor roles in both provinces with massive forms, such as modern faviids and ancient Argutastrea (Figure 31a,b), and branching forms, such as living Acropora and the extinct tabulate Thamnopora (Figure 31d,e) occupying similar roles in some cases. Corals were important constructors in older Middle Devonian reefs although, there too, they were generally, but not invariably [436], subordinate to stromatoporoids [84]. However, corals in the Late Devonian DGBR were not major reef builders and appear never to have dominated shallow margins as do modern scleractinian corals. Rather, they were best developed on platforms in the deeper phases of subtidal parasequences (CF1–3), at platform margins in some places (CF1), or less abundantly in more protected back-reef settings and lagoons where margins had aggraded to sea level (CF3). They became rare or absent in CF4 and CF5. Platy stromatoporoids and alveolitid tabulate corals formed overhangs on fore-reef slopes in CF1 to CF3, so provide morphological analogs for some of the deeper platy corals in the GBRP fore-reef (Figure 31h,i), but they did not reach the level of benthic cover that modern corals do and generally were not self-encrusting although limited encrusting stromatoporoids and tabulate corals occurred.

The best functional analogs for scleractinian corals in the DGBR are the stromatoporoids, with encrusting, laminar, tabular, massive and branching morphologies (Figure 17a–e) that align closely with those of modern scleractinians and to a lesser extent with their ecological roles. However, even there, there are key differences. The vast majority of massive stromatoporoids in the DGBR were constratal tabular to low-domal forms and many if not most lived free on the sediment substrate without encrusting it or each other. They do not seem to have required a hard substrate to encrust for long-term recruitment as do modern reef-building scleractinian corals [437]. Platestone facies with loose stromatoporoids in growing position are abundant on some margins, reef flats and slopes that were not well cemented (CF1,2) and occurred on grainstone-dominated slopes of CF3 in places as well. Although some stromatoporoids reached very large dimensions [388], they were still basically constratal and no analogs for the very large multi-meter, free-standing Porites colonies of modern reefs have been noted. Branching Stachyodes was suprastratal and very similar in morphology to some of the less robust branching Acropora in the GBR; they generally did not reach the same size as many Acropora colonies but occurred in very similar fore-reef and back-reef settings and were responsible for similar beds of debris in those settings (Figure 32e,h). Hence, apart from CF4 and CF5 in the DGBR, the skeletal constructor guild was well developed in both provinces. The micritic microbial constructor guild was well developed in the DGBR, with micritic stromatolites abundant on deeper slopes throughout the Late Devonian but also building frameworks in shallow margin settings and reef flats in CF4 and CF5 (Table 1). Although less well-known, the benthic micritic microbialites (some stromatolitic) that formed in deeper reef slope environments of the LGM and deglacial reefs of the GBR [170] are analogous but appear not to have made benthic framework in shallow settings. Massive benthic structures built by calcimicrobes (Renalcis) in a constructor functional role were also abundant in the DGBR through CH3 and CH4 but were lower in relief in CF5 (Figure 26h). In most cases, they occurred with significant amounts of early marine cement. Such calcimicrobes have no real analog as constructors in modern reefs. Modern CCA (the dominant functional binder today) cross over as constructors on some high-energy oceanic reef margins [156]. Those constratal CCA crusts that reach great thicknesses in those settings (e.g., >5 m in Martinique [438]) may be analogous to laterally extensive calcimicrobial crusts like Rothpletzella (Figure 18c) or constratal stromatoporoids on margins in CF2 (Figure 24), but such thick CCA crusts are more common in open ocean reefs than in the GBR.

Regardless, CCA are the dominant skeletal binders in Holocene reefs, but have no skeletal analog in the DGBR, where encrusting types of tabulate corals, stromatoporoids and calcareous algae were uncommon. DGBR calcareous algae [392] generally did not form lateral crusts, occurring rather as small, commonly isolated, thalli (e.g., various solenoporoids) in more complex microbial frameworks [236,408]. They are best known from CF4 and CF5, but that could partly reflect their small size or scarcity in earlier factories, or the difficulty in recognizing them in frameworks containing other abundant skeletons and cement. They stand out better in more micritic frameworks. The dominant binders in the DGBR were the calcimicrobes, especially Renalcis (CF3–5) and micritic microbialites, both stromatolitic and thrombolitic (CF4,5). Rothpletzella was also common in shallow settings in CF5. In both cases, early marine cements also were very abundant [88,236], enhancing the binding role and providing significant strength to the framework. Whereas modern highstand reef frameworks in the GBRP are basically skeletal with a minor, but important component of cryptic microbialite [176], the lowstand and deglacial frameworks have much greater volumes of microbialite [132,170]. Although still largely cryptic in occurrence, in those cases, it overlapped with the constructor guild. Frameworks were variable in the DGBR [236], but were basically lacking in CF1, which primarily built low-relief banks and ramp margins. Thickened margins occur (e.g., Menyous Gap), but were formed from microbial sediments with abundant, but non-encrusting stromatoporoids in platestones (stromatoporoid bank margin facies) (Figure 22). CF2 constructed fully skeletal, but constratal, frameworks consisting almost entirely of tabular stromatoporoids (Figure 24). CF3 has the greatest diversity of framework types ranging from skeletal frameworks composed of tabular stromatoporoids with minor Renalcis and cement to a variety of mixed skeletal–calcimicrobial frameworks with stromatoporoids, Renalcis and abundant cement. These frameworks vary between laminar and tabular stromatoporoids with Renalcis and cement on margins to encrusted branching Stachyodes, more common on the slope and reef flat (Figure 25e and Figure 31g) (and see [236]). Fully microbial frameworks with varying percentages of micritic and calcimicrobial components and minor but locally important skeletal builders like laminar stromatoporoids or calcareous algae round out the framework types in CF4 and CF5.

The baffling guild includes branching taxa, such as Acropora and Stachyodes, which presumably baffled sediment in their respective provinces, but bafflers similar to modern Halimeda are difficult to identify in the DGBR. A possible candidate is the delicate branching stromatoporoid Amphipora, which is typically preserved as fragments a few mm in diameter and a few cm long. Although not known to be modular, like Halimeda, the skeleton was fragile and when erect, would have baffled currents in protected back-reef areas but became a major sediment source when broken. It is assumed that fleshy brown and red algae, such as are common in the GBRP, would have been present in back-reef settings of the DGBR, but there is little evidence to confirm it as such taxa do not fossilize well. From a theoretical basis [146], settings with high terrestrial sediment and nutrient flux should have favored such algae, and some lagoons had abundant grazing gastropod faunas and oncoids (Figure 30). Fleshy algae and other soft algal turfs may have occurred in inshore settings throughout the Late Devonian but may have been more important in the terrigenous-rich, regressive CF4 and CF5 facies, such as in the Napier Range. It is also probable that high nutrient loads would have favored micritic microbialite deposition at those places and times [439] and calcimicrobes also may have required elevated nutrient flux [440] (see below).

Easier to constrain are the bioeroders. Microborers (e.g., cyanobacteria, fungi) are known since the Archean and larger metazoan borers in carbonate substrates already existed in the Paleozoic, but increased dramatically in the Mesozoic [35,97,441,442]. Lithophagid clams appeared in the Eocene, but large Gastrochaenolites (bivalve borings) were already abundant in rocky shore settings by the earliest Pennsylvanian [443,444] and similar, but smaller, examples have been noted in the DGBR. Acrothoracican barnacle borings already occurred in the Devonian as well as Trypanites [444] and as many as eleven different ichnogenera were documented in Upper Devonian reefs [445]. Of those, Trypanites is the most abundant form in the DGBR [387] but is not responsible for significant carbonate removal where it occurs. Although boring sponges existed in the Devonian [446], they were not pervasive and the earliest definitive clionid sponge style borings (Entobia), which are pervasive in modern reefs, did not occur until the Jurassic [444]. Raspers are more difficult to evaluate. The abundant gastropods in back-reef settings in the DGBR presumably grazed on algal turfs, but their ability to rasp and excavate hard substrate has not been demonstrated. Polyplacophorans (chitons) were present in the Cambrian, but their rasping abilities through the Paleozoic also remain unknown and they have not been documented in the DGBR. The abundant modern rasping groups that reduce hard substrate to sand (scarid fishes and echinoids) expanded or appeared only in the Mesozoic [35]. The increased grazing pressure that they then applied may have disfavored the less well-calcified solenoporoid algae, such as those in the DGBR, leading to the rise of the better calcified CCA that dominate binding in reefs today [447].

The dweller guild was well developed in the carbonate factories of both provinces, but with major differences. Dwellers of the modern carbonate factory are dominated by mollusks, Halimeda and LBFs. Among mollusks, gastropods and bivalves were well represented in the DGBR back-reef, but benthic mollusks were subordinate to brachiopods in most settings. Nektic nautiloids and goniatites were abundant, commonly preserved in deeper inter-reef settings throughout the DGBR factories (Figure 19i). Of those, only nautiloids (the chambered Nautilus) remain today in the GBR where they occupy deeper off-reef settings. Bryopsidales like Halimeda were not abundant in the DGBR, but small solenoporoid algae were abundant in CF4 and CF5. Significant LBFs had not yet evolved and only small encrusting foraminifers occur in DGBR frameworks [236], where they were far less abundant than are the common cryptic encrusting foraminifers in the GBR today (e.g., Homotrema). Among echinoderms, crinoids were more abundant and major sediment producers in the DGBR, whereas they are very minor sediment producers in the GBR today, overtaken among the echinoderms by asteroids and echinoids. Non-carbonate producing dwellers are relatively poorly known in the DGBR, except for the great diversity of fish and arthropods preserved in the Gogo Formation Lagerstätte [419]. Although many preserved taxa were pelagic, some clearly came from well-oxygenated benthic settings on the reef margins and platforms [419]. The microvertebrate record also suggests diverse fish, including chondrichthyans, in shallow reef settings [274]. The in situ bottom fauna of the Gogo Formation was restricted, owing to low oxygen, with most preserved taxa introduced from shallower reef settings [421]. Regardless of the differences, a swim along any of the Frasnian reefs of the DGBR would probably have provided a recognizably ‘reefy’ experience for a modern reef goer. It is less clear how similar the experience would have been swimming along a Famennian microbe-dominated reef margin.

Modern tropical reefs are dominated by ‘photozoan’ assemblages where photosymbionts enhance biomineralization in typically oligotrophic settings [146,448]. As both reef provinces described above were constructed by tropical factories [145] it is easy to assume that they are both driven by photoautotrophs restricted to relatively shallow settings. However, the trophic structure of the Late Devonian factory is distinct from the modern one. For example, tabulate and rugose corals may, or may not, have had photosymbionts based on photoresponsive morphology, phenotypic plasticity and ecological distributions [410,449,450,451,452] but there is no clear consensus, and they were not major frame-builders in the DGBR in any case. Stromatoporoids, on the other hand, were important in DGBR frameworks but were filter feeders with no clear evidence of photosymbionts [453,454]. Reef-building calcimicrobes may have been mixotrophs rather than obligate photoautotrophs [455] and many occur in both exposed and clearly cryptic settings. Hence, the effective depth range of binding was greater in the Late Devonian owing to the greater bathymetric ranges of calcimicrobes and micritic microbialites. Although modern red algae can live deeper than 200 m [456], CCA are not abundant in GBR reef frameworks below ~6 m [217]). Devonian calcimicrobes, especially Rothpletzella, extended to depths where they occurred with toe-of-slope ‘mudmound’ sponge–microbialite factories that are less dependent on light [149], but they did display phototropic behavior, commonly growing vertically upward even from steep slopes (Figure 18c). Hence, although the DGBR was a tropical reef system, we cannot make standard assumptions based on modern reefs as the trophic structure was certainly different, even if not well understood. Some Cenozoic tropical carbonates also had different trophic structures than classical photozoan communities, such as more nutrient-rich SE Asian reefs that have more heterozoan content [457]. Those types of communities may have been more typical of the Miocene carbonates of the GBRP.

4.2. Tectonic Constraints on Reefs

Although both reef provinces formed in intraplate settings with geologic structures controlled the distribution of reef growth, there are major differences in tectonic forcing. Northward movement of the Australian Plate restricted the oldest growth in the GBRP to the north, where it currently rests in the subsurface (Figure 11 and Figure 12). The latitude–parallel orientation of the currently exposed portion of the DGBR changed little due to plate movement through the Late Devonian [210]. That latitudinal stability through time provided it with a more uniform stratigraphy along its exposure, but where it extends into the subsurface far to the north and south it would have encountered different climate zones. An example is the more arid and restricted setting to the south across the Fitzroy Trough on the Barbwire Terrace (Figure 13), which may have been affected more by the subtropical high-pressure zone. At shorter temporal scales, the two provinces differ greatly in the degree of synsedimentary tectonism. Although the GBRP experiences minor reactivation of old structures with little seismic activity the Fitzroy trough was actively rifting during deposition of the DGBR and this caused both subsidence, tilting and uplift of active carbonate platforms with major effects on local stratigraphy [280,283,284,285]. Earthquakes associated with movement on major faults also likely initiated at least some of the significant margin collapse episodes made so evident by debris in the slopes [265,402].

4.3. Climatic Constraints on Reefs

Global climate (i.e., greenhouse versus icehouse) constrained both reef provinces but also provides an overarching factor to differentiate them. The earliest carbonate deposits in the GBRP, high-latitude Eocene carbonates (Figure 11 and Figure 12), were able to form only owing to the very warm ‘Paleocene–Eocene Hothouse’ climate [96] near the end of the last global greenhouse climate phase (Figure 2). The shift to icehouse climate in the Oligocene shut down most carbonate production [83] before Miocene warming helped re-establish it at more tropical latitudes. Major late Miocene to Pliocene cooling then greatly restricted Pliocene carbonate development and the current GBR is fully the product of icehouse climate at mostly tropical latitudes. The DGBR on the other hand formed entirely during a very warm ‘super greenhouse’ [85] ‘Late Devonian Hothouse’ [96] world (Figure 2). Although cooling occurred with minor polar glaciation in the Famennian [85], it was not equivalent to Ordovician or Late Paleozoic glacial phases and temperatures were very warm by ‘icehouse’ (e.g., modern) standards [96].

Apart from the effects of global temperature on the carbonate factories themselves (e.g., potential breakdown of photosymbiont relationships as in modern coral bleaching [72,410,458]), which remain speculative for DGBR reef builders for reasons discussed above, the amplitude of high-frequency (parasequence) sea level change was much lower in the Devonian owing to the absence of major polar ice caps at that time (Figure 8d). Whereas modern parasequences in the GBRP reflect >120 m of sea level change for each glacial–interglacial cycle [459], those in the DGBR may have been only meters to tens of meters in amplitude [276]. Difference in sea level cycle amplitude between icehouse and greenhouse climates place fundamental constraints on reefal carbonate factories, sedimentary dynamics and resulting preserved limestones and parasequences [336,460]. Much greater sea level falls in the Late Pleistocene caused each parasequence beneath existing reefs to be bounded by an unconformity with some degree of exposure-related karstification–calcretization of the underlying reef. Associated offshore lowstand reef cycles are also stacked, with deglacial (transgressive) reefs having formed and drowned on top of the reefs that formed during the falling sea level (regressive) stage [46,288] (Figure 14). Many Devonian parasequences were entirely subtidal, meaning that sea level never fell below the edge of the platform [267] (e.g., Figure 22). Those cycles represent a relatively small difference in bathymetry, although some settings were deeper overall and some shallower (Figure 21). In any case, deeper and shallower deposits are stacked together one above the other. Although intertidal, and even supratidal facies are not uncommon at cycle tops in some intervals, especially near paleoshore, karst surfaces are rare at parasequence tops in the DGBR [275], and in some cases are restricted to areas with synsedimentary uplift [284,461]. Thus, much deeper karstification occurred in the GBRP, as the entire existing reef edifice on the continental shelf was exposed to meteoric water and diagenesis during each lowstand. The result is significant carbonate alteration and the formation of caves in parts of the GBR, where collapse has led to sinkhole formation with the surface expression being, for example, the ‘blue holes’ of the Pompey Reefs [462]. Although the southern part of the reef was in a more arid climate setting during glacial times, which limited karstification [217], the aragonite and high-Mg calcite (HMC) skeletons of the carbonate factory are particularly subject to dissolution and recrystallization. Hence, preservation of the Pleistocene GBRP is relatively poor, and one of the reasons that early radiometric dates from cores remain so elusive [51]. Only some LBFs (e.g., Amphistegina) and mollusks (e.g., oysters) in the GBR carbonate factory construct more stable low-Mg calcite (LMC) skeletons.

Parasequence-related karstification in the DGBR, where exposure even occurred, would have been much shallower, affecting fewer underlying parasequences, and important parts of the carbonate factory initially made skeletons of the more stable LMC (e.g., rugose and tabulate corals, brachiopods). Calcimicrobes, other microbialites and early marine cements also would have consisted of LMC. The original mineralogy of Paleozoic stromatoporoids remains enigmatic. They appear not to have been aragonite or LMC, but may have been HMC wherein they converted readily to LMC during early diagenesis [463]. Deep rooting trees and other soil-developing terrestrial flora, although increasing in the Late Devonian [253], were far less advanced than in the Pleistocene, meaning that karst-enhancing organic soil acids also were less abundant during DGBR lowstands. Thus, DGBR frameworks were less susceptible to dissolution in general and DGBR limestones are better preserved in many cases than those of the pre-Holocene GBR despite the latter experiencing significantly less time and burial. There are exceptions. Some DGBR limestone was subsequently or synsedimentarily dolomitized, and some of that was later de-dolomitized [232,236,464]. Dolomitization is not abundant on the Lennard Shelf but is associated with the economic base metal sulfide mineralization [236,464]. It is also common at non-conformities [290] and some reef margins [267]. However, Famennian dolomitization was more prevalent to the south on the Barbwire Terrace where it partly represents synsedimentary formation within an arid sabkha environment [232]. Deeper burial also caused preservation issues for some DGBR limestones. Well-cemented margins commonly did not undergo intense compaction, but initially less well-cemented back-reef facies underwent significant pressure solution (chemical compaction) during burial, losing as much as 25% of their initial thickness [88]. This resulted in relatively poor preservation. Regardless, only limited karst horizons occur in DGBR stratigraphy although significant unconformities have been recognized [90,283,284,461] and are not to be confused with significant post-depositional sub-ice dissolution that occurred following exhumation during late Paleozoic glaciation [291]. Certain DGBR reef flat features called ‘sand tubes’ appear very similar to subsoil karst solution pipes, but are interpreted as growth features made by microbialites and are considered pseudokarst [236,291]. As a complicating factor, microbialites could easily form within true solution pipes once re-inundated. Calcimicrobes like Renalcis were common in crypts and stromatolites even grew in intergranular spaces within siliciclastic conglomerates [90] (Figure 32b). Microbial deposits also grew in Neptunian dykes in the DGBR [289] as in Mississippian reefs in eastern Australia [465]. Regardless, karstification is a much less significant feature of DGBR stratigraphy than in the modern GBRP.

Higher parasequence amplitude combined with adequate subsidence (i.e., increasing RSL) also led to smaller, more restricted, particularly barrier, carbonate platforms in the GBRP. There, carbonate factories managed to fill the entire accommodation only at the margins or on pre-existing highs formed by previous reefs. Back-reef carbonate factories, even where isolated by distance and longshore currents from terrigenous siliciclastic flux, could not fill all accommodation provided by each cycle (Figure 15a). This resulted in a deeper back-reef lagoon on the shelf, i.e., an ‘empty bucket’ morphology [466,467]. As an example, the DGBR CF3 back-reef lagoon at Mowanbini [290] (described above) contains cycles with shallow subtidal stromatoporoids at the deepest and ooid shoals or intertidal fenestral algal flats at the top. Most available accommodation was filled to very shallow shoaling or intertidal depths for each cycle (Figure 15b). Deeper back-reef settings in the GBR contain discrete carbonate factories, like the Halimeda mounds that can reach 20 m in height [468], but none were able to aggrade to sea level along with their associated reef margins. The comparison supports the contention [460] that ‘empty bucket’ reef morphologies may better represent icehouse than greenhouse climates.

The best partial geomorphological analogs in the DGBR for many modern GBR reef types are the back-stepping and upright margin platforms made by CF2 and CF3 factories that struggled to keep up with increasing accommodation during the early part of Pillara supercycle transgression. That eustatic rise provided adequate accommodation to isolate individual reefs on the deeper drowned shelf (e.g., Lloyd Hill in Bugle Gap) (Figure 27c). However, even there, back-reef settings appear to have kept up better with the margins so that deeper back-reef lagoons (‘empty buckets’) are not the norm. Kerans [236] used the abundance of Renalcis behind the reef margin as an arbitrary, but definable, means of differentiating reef flat from back-reef facies owing to the gradual nature of the facies changes without major bathymetric change.

Over longer time scales parasequence amplitude also may have affected the evolution of the carbonate factories. The great accommodation imposed on reef growth (keep-up or catch-up) by Late Pleistocene parasequences put increased pressure on newly established reef communities to aggrade, or they might give up and drown, as may nearly have occurred in the southern GBR during the recent deglacial [46] and the previous LIG, e.g., [217]. This ‘pressure’ to aggrade may have favored the very fast growing and disturbance resistant branching acroporids that dominate most Indo-Pacific reefs today [27]. Suprastratal branching acroporids were present in reefs throughout the Cenozoic but took on their present dominance only in the Late Pleistocene when the amplitude of sea level change for each parasequence reached its very high current levels [27]. Morphological analogs in the DGBR, branching tabulate corals and Stachyodes, were not associated with major aggradation until the introduction of abundant Renalcis and early cement in CF3, which provided branching (and laminar) Stachyodes adequate strength to make rigid framework on high-energy platform margins, slopes and back-reefs. Branching tabulate corals, although prolific at CF1–CF2 highstands, made only thin biostromes, but may have lacked the high extension rates of modern branching Acropora. Prior to the provision of adequate binding and cement to cause framework unification, the late Givetian–early Frasnian (CF1–CF2) margin communities were incapable of aggrading rapidly enough to keep up with sea level rise during the Pillara transgression and drowned and/or retreated and back-stepped.

The potential enhancement of branching Acropora in Late Pleistocene parasequences of modern reefs occurred against a backdrop of general biological stability between cycles. As shown in other regions [59,68], highstand reef communities in the GBR reassembled in a very similar way after each glacial exposure interval, despite every new reef having to initiate independently on the previously exposed, dead substrate [197,217,219,356]. The same basic carbonate factory even re-established in the reefs through the glacial times as the reef tracked sea level down and back up [288] (Figure 14). Similar ecological stability occurred in CF1–CF2 communities in the DGBR [70]. Over short time scales—a few consecutive parasequences, communities remained stable, but larger changes occurred over longer time spans, presumably in reaction to broader environmental trends [70]. Although such quantification has not been accomplished on younger cycles in the DGBR, the recognition of cycles there in part reflects the repeatable ecological membership in similar geological circumstances through a longer succession. Hence, it is likely that such ecological stability was the norm over similar limited time spans.

This highlights the comparability issues between modern and fossil reefs imposed by differing temporal and spatial scales of study (Figure 3). At small spatial and temporal scales modern and DGBR reef communities reacted in similar ways producing similar facies zonation in similar settings with comparable geographical expressions resulting from external forcing, such as waves, tides and currents (Figure 32). However, the modern GBRP that we know well represents only a single parasequence highstand. Much larger changes may occur in reefs and reef communities at the longer temporal scales of supersequences, and these affect larger scale reef geomorphology (e.g., the difference between CF1–CF2 banks, CF3 upright margins and CF4–CF5 prograding platforms). Much of this depends on the differing rates of accommodation creation relative to the aggradation potential of the carbonate factories. The major driver of long-term DGBR stratigraphy was eustatic sea level change [266] with similar supercycles in Upper Devonian reef systems in Alberta [469] and South China [267,268] although each region has local tectonic differences [262,400,402]. Parasequences too would have been global if driven by Milankovitch dynamics. Aggradation of CF3 communities in the DGBR during a time of rising sea level generated geomorphological and facies similarities to the GBRP, because significant aggradation occurred with each parasequence. The main difference was that the GBRP experiences relatively slow subsidence (slow RSL rise) overlain by high-frequency, high-amplitude sea level fluctuations, whereas the DGBR experienced a low-frequency, high-amplitude sea level change (supercycle) overlain by high-frequency, low-amplitude parasequences (Figure 8d). Although the framework types differed significantly in the two provinces (specifically, Late Devonian mixed calcimicrobial–biocementstone–skeletal frameworks compared to primarily skeletal frameworks today), the key factor is the ability to aggrade adequately to keep up with accommodation at the margin. However, this same factor led to a major difference between the two reef systems, in that deeper back-reef lagoons of the GBRP hosted carbonate factories that could not fill required accommodation for each parasequence (Figure 15a). The following question then arises: Did increased accommodation of the Pillara transgression in the DGBR promote calcimicrobes like Renalcis, which then allowed CF3 reef margins to aggrade and keep up with accommodation in an analogous way to high-amplitude Late Pleistocene parasequences favoring fast-growing, branching Acropora? The answer to that question requires investigation of other types of external forcing.

4.4. Marine Chemistry Constraints on Reefs

Most high-energy DGBR reef margin facies (at least CF3–CF5) are dominated by calcimicrobes and early marine cement [88,236]. The abundance of microbial carbonates and marine cement has varied through the Phanerozoic, particularly in reefs [98,99,158,470,471]. Calcimicrobes are non-obligate calcifiers, meaning that they do not intentionally precipitate CaCO₃ skeletons. Rather, they may promote, localize and/or ease nucleation of carbonate minerals and/or locally increase alkalinity as a byproduct of their metabolism, thus incidentally increasing precipitation [98,472]. Hence, their presence/absence partly represents taphonomic processes (i.e., preservation potential) [404,473,474,475], and these depend on ambient water chemistry (alkalinity, cation availability, pH and temperature). Although concerns about ocean acidification (OA) causing reduced biomineralization in modern carbonate factories are valid [18,19], many reef-building organisms can still calcify in lower-pH waters at an additional metabolic cost [476,477]; there is a limit [478]. Both micritic microbialites and calcimicrobes, on the other hand, are completely dependent on advantageous water chemistry to form and cannot form widely, if at all, in waters with adverse chemistry (i.e., low alkalinity, low pH, etc.) unless partially isolated in a microenvironment [130,479]. Simple filamentous calcimicrobes, such as Girvanella, occur through most of the Phanerozoic, but their distribution partly reflects adequate preservation [475]. Where rapidly calcified in waters with abundant dissolved carbonate, identifiable calcimicrobes were formed, but where calcification was slower or less complete, micritic microbialites or mud were formed. Hence, microbialite occurrence and type should provide an independent proxy for marine calcification potential [98]. However, critical interrelationships between the metabolic inputs, influences and constraints of skeletal and microbial communities on each other during reef building are poorly understood [480].

Changing microbialite distribution through the Phanerozoic was attributed to the combination of fluctuating marine carbonate saturation and competition with skeletal metazoans as reflected by biodiversity curves [99,481], the latter partly regulated by atmospheric oxygen levels [482]. Marine alkalinity and carbonate saturation should be primary controls on calcimicrobe abundance, but major anomalies exist in their correlations [482] (Figure 2). Marine alkalinity and buffering capacity are highly complex, even in the modern ocean [478,483], but depend on basic chemical equilibria that should allow modeling through time [484]. This modeling is increasingly complicated with less secure proxies the farther back it is attempted [485]. Although modeling Phanerozoic saturation states is beyond the scope of this paper, several observations can be made relative to the comparison of Late Devonian and modern reefs. The general decreasing trend in Phanerozoic microbialites [471] is undoubtedly correct, but the initial data analyzed include earlier studies that commonly overlooked microbial carbonates in reef frameworks owing to a greater focus on, and easy recognition of, associated skeletal organisms (see [158,465,486,487]). Microbialites are still not considered part of the reef framework by many workers despite the large role and volumes of micritic microbialites in deglacial and Holocene reefs [132,439]. Hence, existing microbialite abundance curves [37,471] may not take micritic or cryptic reefal microbialites adequately into account. However, that cannot account for the major differences between modeled saturation states and abundance of reefal microbialite and especially reef cement [97].

Through time, modeled equilibrium saturation states may not adequately consider carbonate removal rates from changes in biomineralizers [482]. Differing biomineralization processes affect how organisms respond to changes in pH and carbonate removal [478]. A temporal mismatch between the rate of anthropogenic CO_2atm increase and natural removal by chemical weathering leads to the current OA problem [488], thus representing a temporal disequilibrium. Major anomalies between modeled carbonate saturation state and empirical proxies for ease of marine carbonate precipitation (e.g., marine cement) could in part represent temporal disequilibria in the broader carbon cycle or fundamental changes to mass balance. The largest anomaly in modeled saturation state versus microbialite abundance occurs from the mid-Mesozoic, when very high calculated saturation states were not associated with abundant microbialites or cement (Figure 2). This major difference was previously attributed to the evolution of calcareous plankton [98], which were able to expand carbonate precipitation into open oceanic settings, where nutrients allowed, and remove and sequester alkalinity to deep ocean waters and deep-sea sediment. Prior to calcifying plankton, most marine carbonate biomineralization occurred only on continental shelves. Such major changes in the marine carbonate system could have fundamentally, and unidirectionally, altered the dynamics of the marine carbonate system. Hence, the diversity of organisms through the Phanerozoic may be less critical than the source, volume and rate of carbonate removal they accomplished. Increased efficiency of calcification amongst some skeletal organisms (e.g., photosymbiosis in scleractinian corals, LBFs and mollusks) and expansion of biomineralization into new open ocean habitats by plankton could have caused major shifts in shallow marine carbonate saturation dynamics. Something must account for the massive decrease in biocementstones in reef frameworks since the Early Mesozoic [97,98] (Figure 2). Regardless, that decrease represents a major empirical difference in calcification potential between the DGBR and GBRP. Thus, the lack of evidence that OA was a problem in Late Devonian reefs despite the very high pCO_2atm [489] provides no security for modern reefs.

Regardless of broad Phanerozoic calcification patterns, the rise of abundant calcimicrobes and cements to dominate CF3 reef frameworks in the DGBR requires explanation, as modeled carbonate saturation states were high throughout the Devonian [482]. So why did Renalcis and associated calcimicrobes bloom when they did in the Late Devonian reefs of the Canning Basin? It is unlikely to represent the occurrence of new microbes. Calcimicrobes like Renalcis, as with Girvanella, occurred both before and after the Late Devonian (Cambrian [490]; Ordovician [491]; Mississippian [492]) but were commonly missing in intervening stratigraphic intervals where adequate substrates were available and other members of reef carbonate factories flourished (e.g., corals and stromatoporoids in CF1 in the DGBR). Calcimicrobes became abundant in reefs around the world during the Frasnian transgression, but were not as abundant everywhere as in the DGBR. They occurred in transgressive Frasnian reefs in Alberta, and the Northern Territories, Canada [168,493,494], in some cases as cryptic binders with abundant cement in aggrading and prograding margins as in the DGBR [469]. Frasnian reefs in South China had abundant calcimicrobes, but they were less pervasive, although more diverse, than in the DGBR [267,268]. The abundance of calcified algae, cyanobacteria and microproblematica increased by 34% in South China from the Givetian to the Frasnian, but then declined by 63% in the Famennian, as micritic microbialites dominated [406]. Local paleogeography controlled carbonate factory distribution in Late Devonian South China with microbial reefs primarily developing on isolated platforms, not attached to the continental mass [268]. The steep aggrading stromatoporoid–Renalcis frameworks of the DGBR are not common in South China and stromatoporoid–coral communities back-stepped during Frasnian transgression [268]. Calcimicrobes, including Renalcis also occurred in Frasnian reefs in Belgium [495] and in the Urals from Timan-Pechora [496,497,498] to the pre-Caspian of Kazakhstan [84,86,499]. Some of the larger (300–700 m thick) Frasnian buildups in Timan have stromatoporoid–microbial frameworks with Renalcis and abundant cements [496]. There, a general trend of decreasing skeletal frame-builders with increasing microbialite occurred through the Frasnian, but biocementstones were even more abundant during the Early Devonian, being somewhat reduced through the Frasnian before dropping off in the Late Famennian [498]. In all regions, margins were dominated by micritic microbialites (stromatolites and thrombolites) at or prior to the end of the Frasnian with major carbonate platforms continuing through the Famennian.

The Late Devonian increase in micritic microbialites in Alberta was attributed to increased nutrients associated with lowstands, but the aggrading calcimicrobial–skeletal frameworks of highstand margins were considered to represent oligotrophic conditions [469]. Other authors considered only coral–stromatoporoid frameworks to represent oligotrophic settings, with calcimicrobe-rich facies and micritic microbialites forming in mesotrophic settings and megalodontid bivalve assemblages representing eutrophy [440]. It is difficult to apply modern nutrient gradients to Devonian reef builders [500]. There are no modern analogs for calcimicrobes like Renalcis and the dominant skeletal constructers, stromatoporoid sponges, were most likely filter feeders [454] that may have required more nutrients than modern ‘photozoan’ reef communities. Regardless, it is critical to separate calcimicrobes and micritic microbialites in terms of their environmental implications and proxy value [98] and to discriminate between benthic and cryptic microbialite settings [469,479]. Modern micritic microbialites may represent higher nutrients [439,501], but cryptic varieties, which are common in modern and Devonian reefs, may still occur in oligotrophic settings because they form in microenvironments where nutrients and alkalinity are governed by co-occurring organisms, rather than ambient seawater [130,479]. Additionally, all microbialites are not the products of cyanobacteria. Calcification in many microbialites is driven mainly by heterotrophy, and in particular sulfate-reducing bacteria [472,502], which also have been implicated in modern cryptic reefal microbialites [503]. Whereas benthic microbialites may have cyanobacteria and algae as their primary producers with heterotrophs driving calcification beneath them, cryptic microbialite heterotrophs probably consume eukaryote organic matter produced by other cryptic organisms like sponges [399] or ascidians [132]. Calcimicrobes are less well understood, but Epiphyton (which has ecological similarities to Renalcis) is thought to be a cyanobacterium, and was recently shown to have carried out oxygenic photosynthesis [455]. However, its common ability to thrive in fully cryptic habitats suggests that it was a mixotroph [455]. Hence, calcimicrobe communities may have adapted to a wide range of nutrient settings and bathymetry. In the DGBR, Renalcis (cryptic and benthic) and cryptic thrombolite shared trace element geochemistry (rare earth elements, REE) with associated early marine cements, which are consistent with well-oxygenated, shallow seawater [504]. Early marine cements are the best proxies for ambient seawater trace element distributions as they incorporate most of their signature from dissolved ions, rather than from trapped and bound detritus [505]. Comparison to modern open ocean seawater suggests that Leonard Shelf seawater had greater terrigenous input [504]. Average CF3 stromatoporoid–Renalcis frameworks formed in waters equivalent to normal salinity, inshore waters in the GBR lagoon (based on light REE depletion levels—a seawater indicator: mean (Nd/Yb)_n = ~0.43 for DGBR early cement compared to mean (Nd/Yb)_n = ~0.40–0.50 for well-mixed estuarine seawater in the modern GBR lagoon [506]). For comparison, Holocene cryptic microbialites from Heron Reef, southern GBR, which occurs >70 km offshore near the continental margin, have much greater light REE depletion typical of more open marine (oligotrophic) seawater (mean (Nd/Yb)_n = 0.236) [507]. Inshore DGBR microbialites that occur with siliciclastic sediments, on the other hand, have trace element signatures dominated by terrigenous inputs as expected and show no light REE depletion [504]. Thus, Renalcis-containing Devonian reef margins may have required more nutrients than modern reefs, thus being less susceptible both to nutrification and drowning [440].

Even Devonian stromatoporoid–coral reefs from platform edges may not represent the oligotrophic settings that modern reefs prefer, as that interpretation is based largely on the tabulate corals having established photosymbionts, e.g., [72], a hypothesis that remains controversial. However, quantitative relationships between nutrient flux and terrestrial input may have differed in the Late Devonian, as terrestrial floras and weathering rates and processes were different than today [253,254]. Late Devonian continental runoff, although containing greater nutrients than in the Early and Middle Devonian [253] may have contained fewer nutrients relative to modern runoff entering the GBR lagoon. Limited palynological data suggest that the coastal lowlands bordering the DGBR were occupied by a well-developed terrestrial vegetation [90] and adequate nutrients supported a diverse plankton including abundant radiolarians [508], acritarchs, microalgae and cyanobacteria [233]. Givetian to Frasnian oils generated from Lennard Shelf source rocks (Gogo and Fairfield formations) were sourced from plankton and suggest episodes of water stratification with anoxic deposition due to eutrophication during transgressions with high primary productivity from eukaryotes (e.g., algae) [426]. Stratification is indicated by biomarkers for bacterivorous ciliates that fed at the oxic–anoxic interface, and that interpretation is supported by the normal marine fauna preserved in the anoxic to suboxic deposits that hosted the Gogo Formation Lagerstätte [419]. Across the Fitzroy Trough to the south, oils obtained from the Barbwire Terrace are dominated by bacterial sources in a more continuous suboxic environment that may reflect restriction from more open waters of the Fitzroy Trough behind the reef complexes developed along its edge [426] (Figure 12).

Somewhat paradoxically, although the prominent association between CF4 and CF5 micritic microbialites and terrigenous sediment is well documented and provides strong support for increased nutrients through the regressive (Nullara) part of the supercycle, the DGBR lacks evidence for the abundant shallow anoxic facies [509] that characterize the interval and Late Devonian extinction events elsewhere in the world. One of the major models for Late Devonian extinction events involves eutrophication and pervasive shallow anoxia within the photic zone resulting from the very high influx of terrestrial nutrients associated with the newly deeper rooting trees and greater soil thicknesses and weathering profiles through the Devonian [261,510]. The lack of shallow anoxia in the DGBR seems paradoxical being set within a relatively enclosed epicontinental embayment and argues against extreme eutrophication from the terrestrial nutrient flux through the Famennian. The global dominance of microbial communities in reefs after the Frasnian–Famennian extinction of most skeletal reef builders is well documented [511], but there was also a potentially global decrease in the amount of reefal calcimicrobes during the Famennian [236,406]. More importantly, early marine cement volumes decreased, and thus so did biocementstones [498].

Early marine cement declined in the DGBR from CF3 to CF4 and CF5, when terrestrial flux was empirically greatest. Alkalinity in the form of HCO₃⁻, like terrigenous nutrients, results from continental weathering and is delivered in runoff, so they might be expected to correlate. They do not correlate well in the DGBR. It is even possible that increased CF3 marine cement resulted from lower nutrients, wherein there was a lower volume of active microbial organic matter to promote nucleation, leaving abundant alkalinity available for thicker cement. Could the much more frequent nucleation of micritic carbonate in CF4 and CF5 stromatolites and thrombolites effectively remove adequate alkalinity to suppress other cements? Alternatively, the reduced CF5 cements may reflect the decrease in larger framework cavities to host large cement volumes. The lack of abundant early marine cements in CF1 facies does not support the first hypothesis, but comparisons of carbonate volumes through time are needed to test the concept. Regardless, other factors may have been significant. Salinity is a major control on shallow marine alkalinity [512,513] and would be subject to temporal changes in climate (both precipitation/evaporation balance and circulation). The arid hypersaline deposits across the Fitzroy Trough extended into the Famennian, so temporal changes to broader basin circulation could have altered salinity gradients, even without changes to local precipitation on the adjoining continent. Unfortunately, conodont-based δ¹⁸O isotope data are absent for the DGBR [514], but could help indicate relative salinity, and thus inform intervals of aridity versus terrestrial runoff [251]. Regardless of whether the abundant reefal microbialites and biocementstones of the DGBR were constrained more by nutrients or local alkalinity, the current OA trend poses a serious risk to the role of cryptic microbialites in unification of modern reefs. At the same time, modern reefs are probably far more susceptible to nutrient-invoked drowning (sensu [440]) than were Late Devonian reefs.

4.5. Evaluating the Analogs

The concept of uniformitarianism [515] is well known in Earth history. The present is the key to the past, and increasingly, the past is the key to the future. This is especially the case for climate studies, which depend more and more on numerical modeling to make climate change predictions. The conservation of coral reefs is intimately tied to understanding their past and present as their futures are similarly tied to changing climate. One of the unique contributions of geology to such studies is the introduction of real Earth system time scales where cyclic phenomena, such as Milankovitch orbital forcing and broader tectonic cycles can be recognized and differentiated from more stochastic events, such as bolide impacts, and unidirectional changes, such as increasing continent size [516], increasing solar luminosity [331,517] and changes associated with the evolution and diversification of life. The reef systems compared above show clear evidence of Milankovitch-scale sea level cycles but are overlain on different phases of the broader icehouse—greenhouse climate cycle and the calcite versus aragonite seas cycle (Figure 2) (

Table 2. Comparison of environmental forcing attributes of the Great Barrier Reef Province (GBRP) and the exposed Devonian Great Barrier Reef (DGBR).

	GBRP	DGBR
plate tectonic setting	intracratonic rift	intracratonic rift
	movement through latitude	latitudinal stability
	post rifting deposition	syn-rift deposition
latitude	mostly tropical	tropical
climate	mostly icehouse	greenhouse
	mainly low	very high
parasequence amplitude	>100 m	m—10s m
preferred mineralogy	aragonite	calcite
modeled calcite saturation	low	high
nutrients	oligotrophic	mesotrophic

). At the same time, the biology of the reef builders has changed unidirectionally with extinctions and radiations imposing very different carbonate factories on the two reef provinces (Table 1). As the evolution of oxygenic photosynthetic metabolisms in cyanobacteria unidirectionally changed the redox states of shallow Earth systems, the evolution and diversification of different biomineralizers may have altered the marine carbonate system such that even marine cement and calcimicrobe abundance have been significantly reduced since the Devonian.

Through Earth history, successive reef communities developed integrated systems, adapting to a steadily changing Earth, that is, when given time. It is best not to think of them piecemeal. Modern coral reefs are more than just healthy coral communities, more than just corals and CCA. They are the end products of a suite of closely integrated processes: construction, destruction, sediment transport, sediment deposition and unification. Reef communities dominate some processes and impact on each of the others, but every process has constraints imposed by external forcing that potentially affect sustainable reef growth. The ability of modern and ancient reef communities to consistently reform after major sea level changes with each parasequence is a testament to their resilience [70,197], but also to their integration and interdependencies [62,518]. Regardless, the communities must respond to environmental forcing [70]. Although reef communities engineer and construct the environment where they live, that environment must be maintained constantly through time or the entire system may break down, i.e., undergo a phase shift to a very different ecosystem [108,167,519]. To be successful, reef communities have required huge adaptability through time. A possible key to success for major intervals of reefs in the Phanerozoic is the combination of skeletal constructors, typically, fast-growing hypercalcifiers, with effective binders, such as CCA and microbialites, the key agents of unification [62]. This combination provides both capacity for aggradation and for rapid acquisition of rigidity. CF3 margin frameworks show that abundant marine cement can be added to the encrusters and microbialites. However, reef systems can collapse and be terminated or changed significantly when external forcing pushes constructors or key unifiers beyond their tolerances. The DGBR and a few other major Late Devonian reef systems (Figure 1b) were the last holdouts from a much ‘reefier’ Middle Devonian world, where many equally large and larger reef systems were distributed widely across the globe [85]. Even within the more depleted surviving DGBR the major changes in reef communities (CF1 to CF5) and resulting reef environments and geomorphology that occurred through time would have appeared ‘catastrophic’ if viewed from any one endmember to the next. However, even major changes, such as the replacement of most skeletal organisms by microbes that characterizes the Frasnian–Famennian extinction event, may have been more gradual than ‘catastrophic’ when viewed in higher resolution in the DGBR [402,408,416]. Each parasequence and lower-frequency sea level fluctuation causes at least a local ‘catastrophic’ change, making the recognition of gradual change more difficult (e.g., a temporal scale issue [70]). Regardless, Frasnian–Famennian changes followed other major changes in Devonian reefs, particularly affecting corals, near the end of the Givetian [85]. Long-lived, low-relief CF1–CF2 carbonate banks with rich, widespread coral and stromatoporoid biostromes shifted to back-stepping, and then increasingly higher-relief CF3 margins dominated by stromatoporoids, calcimicrobes and early marine cement. CF1 contained abundant hypercalcifying constructors, but lacked effective binders and abundant cement, thus failing to unify. The margins of most cycles never aggraded to sea level. Only abundant marine cement and rapid calcimicrobe calcification provided the unification that allowed CF3 margins to aggrade through the increasing accommodation of the Pillara transgression. The shift to prograding, low-relief CF4 and CF5 margins dominated by micritic microbialites and calcimicrobes had few hypercalcifiers, unless microbialites take that roll, but they prograded more than three km basinward through the lower accommodation Nullara regression. Each carbonate factory differed in important ways, but in each case, they engineered, constructed and maintained an environment that provided habitat for countless organisms, both on the reefs and in associated coastal areas behind them. Reef communities have changed even more through the Phanerozoic, but reef forming processes have built similar reefs when and where possible [35]. The way the carbonate factories of the GBRP and DGBR ‘engineered’ and constructed their reefs is similar, producing similar geomorphology at times (Figure 27 and Figure 28), and with ample evidence of comparable storms and waves producing both reef flat and fore-reef rubble (Figure 32a–f), relatively high tidal ranges (Figure 32g,h), and typical coastal currents (Figure 29).

Even with such basic similarities, to study reefs as ‘timeless’ entities it is important to focus on the functions and processes that govern their construction (e.g., Figure 5). However, a common tendency to evaluate ancient reefs against the ‘gold standard’ of modern reefs can lead to an (over)emphasis on skeletal contributors. The ‘pull of the recent’ is very strong, despite the great complexity and diversity of modern reef systems and the very different biology and external forcing through time (climate, marine chemistry, etc.) (Table 2). Bias against microbialite reef framework is discussed above, but many ancient reefs in high-energy settings simply do not contain adequate skeletal organisms to make a skeletal framework, e.g., [465], yet they were in every other way analogous to modern reefs. This perceived dominance or primacy of skeletal frameworks also has led to a preoccupation with corals, almost to the exclusion of the critical binders, CCA [61], never mind cryptic microbialites. Unification must happen for reefs to form hard substrates at sea level; carbonate production by constructors, even in situ skeletons, currently is not enough (e.g., Figure 32a,c,e). However, many studies on modern reefs focus primarily on skeletal organisms, commonly just corals, to predict if current reefs are healthy and likely to flourish. At the same time, some of the best recent ecological studies aimed at reef evolution in the Devonian, e.g., [69,70,72,410] also have focused primarily on corals. In the case of the DGBR, late Givetian to early Frasnian CF1–CF2 corals were abundant and thrived through each parasequence highstand, yet they did not make a reef that could aggrade adequately to keep up with sea level (Figure 21). Rather, they were completely subordinate to stromatoporoids, which occupied the shallower, higher energy facies and were capable of aggrading at the margin adequately to protect a lagoon facies near Menyous Gap (Figure 22c). Corals were not major reef builders at the time (in the DGBR). Constratal stromatoporoids made significant skeletal frameworks (CF2 margins) but still could not keep up with the Pillara transgression. Aggrading reefs geomorphologically similar to modern reefs occurred only when CF3 communities consisting mainly of constratal massive tabular and delicate laminar and branching stromatoporoids were joined by potent binders (calcimicrobes) and abundant marine cement. The reefs were made, not by hypercalcifiers, but by a consortium of, in some cases delicate constructors, but with adequate binders and effective early cementation to ensure unification in ecological time. Thus, rigid substrates (frameworks) were constructed upon which additional reef growth could occur in high-energy settings. Corals had little if anything to do with it, whether or not they had associated photosymbionts. The more likely controls on reef evolution in the DGBR were marine chemistry to support, or not, early and voluminous cementation, and thus allow for the formation of calcimicrobe binders [98,475]. This degree of unification supported the steep upright CF3 and CF4 margins that grew at sea level as evidenced by the extreme winnowing of well-cemented Stachyodes fore-reef slope rudstone–biocementstone (Figure 25h). Despite the morphological and taphonomic similarities between branching Acropora and Stachyodes, there are no modern equivalents to such Late Devonian rudstones–cementstones. Despite the Givetian and Frasnian corals preferring deeper settings, they did not thrive during the Pillara transgression when cycles were getting progressively deeper. Rather, the new CF3 stromatoporoid–calcimicrobe carbonate factory dominated the margins and relegated the more limited corals to back-reef environments, such as at Mowanbini, where Disphyllum continued in a limited role to make biostromes behind the margin.

Another major difference is nutrient flux [440,469], as the Leonard shelf waters were likely to be more nutrient-rich than are offshore GBR reefs today. This may have better supported the filter-feeding stromatoporoids and microbial communities, but its effect on corals remains speculative. Regardless, the restriction of corals to shore attached lagoons does not suggest that nutrients were the immediate cause of their decline. Increasing nutrients through the regressive CF4 and CF5 times may have reduced both corals and stromatoporoids leaving calcimicrobes and other microbialites to dominate the reefs. Increasing nutrients and local alkalinity associated with lowering sea levels during Famennian cooling are more likely causes for the decline of stromatoporoids and corals than the changing temperatures themselves, which despite falling in the late Famennian, were still very high by today’s or any icehouse climate standards. Additionally, local alkalinity may have increased on the Leonard Shelf owing to higher salinity in the Famennian. Unfortunately, the isolation of Leonard Shelf reefs within the intracratonic Canning Basin renders correlations to global processes more difficult. The degree of circulation with the Paleotethys Ocean is poorly constrained [233] and even the global shallow anoxia events of the Kellwaser and Hangenberg events are not apparent in the DGBR [509]. Thus, minor climate changes that drove precipitation patterns and upwelling in the basin could have altered nutrient flux, delivery of HCO₃⁻ from terrestrial weathering and salinity, all of which would have affected the carbonate factories. The uncertainties are exacerbated by the very thick stratigraphic succession and difficulties in correlation across the paleogeography. Additional detailed studies are required, especially to put both stable isotope and trace element geochemistry into the context of different carbonate factories and global trends and events, and this is true both for the DGBR and the GBRP (particularly the older buried parts to the north). There is much more to learn by comparing reefs such as the GBRP and DGBR.

5. Conclusions

Coral reefs are among the most significant modern habitats in terms of biodiversity and the ecological, economic and cultural services they provide. The reefs of the modern Great Barrier Reef Province (GBRP) are built by corals and crustose coralline algae in oligotrophic settings within a regime of high bioerosion, high-amplitude sea level change, relatively low temperatures and low marine carbonate saturation. Reefs and banks of the ancient Devonian Great Barrier Reef (DGBR) of western Australia were built by stromatoporoid sponges, microbialites, corals and marine cements within a regime of higher nutrients, low bioerosion, low-amplitude sea level change, higher temperatures and high marine carbonate saturation. Modern and fossil reefs share significant commonalities but also differences that can be related to specific external constraints. The study of one can usefully inform the other. Comparison of these two reef provinces allows the following conclusions to be drawn.

5.1. Reefs in General

• Biological reefs (today, coral reefs) are not just biological communities and ecosystems, they are the relief-bearing structures that are bio-engineered and constructed by those communities. Those structures provide many ecological services that extend far beyond the reefs themselves.
• Repeated reassembly of biological communities after successive disruptions, (e.g., between parasequences, even tracking sea level down and back up through glacial–interglacial cycles) in both reef provinces suggests that community integration and interdependency is a general characteristic of reef-building communities and provides further evidence for their specialized nature. Reef communities are more than random associations of organisms that can coexist in the same shallow carbonate environments.
• Despite bio-engineering their own geomorphological habitats, reef communities are subject to numerous external constraints (climate, sea level, wave energy, nutrients, marine alkalinity, to name a few). These constraints occur as both temporal and spatial unidirectional trends (e.g., latitudinal gradients, increasing continentality through time) and heterogeneous and stochastic distributions (e.g., intermediate disturbance to glacial cycles). In the face of these constraints, reef building depends on communities managing or balancing five key processes (Figure 5): (1) construction of calcium carbonate by a carbonate factory; (2) destruction of existing structures by physical or biological erosion; (3) transport of particulate sediment; (4) deposition of particulate sediment; and finally, (5) unification, where both in situ and transported carbonate skeletal material is bound and cemented into a hard reef substrate. It is that hard substrate that allows continued reef growth and underpins reef geomorphology and thus, survival of the broader reef system. Changes to the external constraints on, or biological roles in, any of the five key processes may cause major changes within the other processes and, thus, to the entire reef system.
• Very different reef communities can make very similar reef geomorphologies. However, they also can create very different structures and platforms where communities and/or external constraints differ. Understanding the differences can provide information on fundamental aspects of reef communities that might be applicable to reefs across time and space.

5.2. Modern Reef Workers and Conservationists

• The study of ancient reefs provides significant insights into how modern reef communities have come to be, not just at the assembly level, but the evolutionary level. For example, the dominance of many (but not all) Indo-Pacific reefs by fast growing acroporids has been attributed to the high amplitude of modern icehouse sea level oscillations [27]. Over longer time scales the rise of rasping bioeroders, like scarid fishes, may have favored the evolution of modern CCA at the expense of the earlier, less well-calcified solenoporoid algae [447], which occurred in the DGBR.
• Abundant carbonate production in an appropriate shallow setting is inadequate to build a reef. The abundant growth of CF1 corals and stromatoporoids constructed only low-relief carbonate banks. What was lacking was adequate unification to allow the banks to aggrade to sea level at their margins. Subsequent CF3 communities on the other hand, had far fewer and less massive skeletal constructors, but had effective binders in calcimicrobes and early marine cement to support unification. They constructed steep, high-energy reef margins similar to those of modern reefs from less volumetrically abundant, fragile stromatoporoids with virtually no corals at all. The example highlights that successful reef growth requires both construction, although not necessarily by so-called ‘hypercalcifiers,’ and unification from binders [62]. Importantly, these two processes may be constrained independently from each other by external forcing.
• While coral bleaching is a major threat to important members of the carbonate factory today, it is far from the only threat. Many skeletal organisms can manage calcification in lower pH waters, but the larval and juvenile forms of many taxa in reef carbonate factories are at risk, although under-studied [520]. These include CCA, which is more sensitive to carbonate saturation state than are corals [521,522]. The effects of OA on binders, both CCA and microbialites, could severely restrict modern reef unification, and thus development of the hard substrates required for recruitment and renewed coral growth after even small disturbance events. The scale and frequency of disturbance events are projected to increase, with increased wave energy and cyclone activity and higher sea levels allowing more wave energy across reef platforms. Critically, modern reefal microbialites that increase the strength of porous reef carbonates by filling pore space at many scales are non-obligate calcifiers; their occurrence and distribution is almost completely constrained by ambient water chemistry. Reduced marine pH associated with anthropogenic OA could inhibit their precipitation, thus removing a major component of modern reef unification.
• From a conservation perspective, it is critical to remember that it takes more than a coral community to raise a reef. Corals are without doubt the most important skeletal framework constructors on modern reefs, but all five of the key reef-building processes must balance for reefs to be maintained. The role of bioerosion and source and fate of reef sediment is increasingly acknowledged and studied (from conservation and carbonate budget–carbon cycle perspectives), but much of the reef conservation and monitoring effort is directed only to corals and commonly only coral cover on the slope. This partly reflects the relative ease of measuring coral cover remotely, but it is not the only measure of reef health and studies of other critical reef growth processes have lagged behind. Through modern reef coring we are only beginning to grasp the importance of cryptic microbial communities in framework unification and know very little about their metabolic processes, trophic structures or even rates of growth. Fortunately, reef cores through the last glacial cycle have provided a natural laboratory to understand the effects of changing pCO₂ on modern reefal microbialite growth [148], but the results are alarming in the face of OA.

5.3. Geologists and Paleobiologists

• Ancient reefs provide a means to explore temporal differences in marine chemistry, such as carbonate saturation state, as it provides major external constraints on carbonate factories, reef unification and carbonate platform development. Marine alkalinity is also intimately related to Earth’s climate through time. While direct comparison of modern and Late Devonian unification processes and modeled marine alkalinity appear to be consistent (explaining the lack of abundant calcimicrobes and marine cement today), major inconsistencies in expectations are evident in other time intervals since the Mesozoic suggesting major gaps in our understanding of marine alkalinity through time.
• Comparison of the GBRP and DGBR provides strong support for the hypothesis that ‘empty bucket’ carbonate platform morphologies are favored by, if not limited to, icehouse conditions [460]. This association likely results from the interplay between the amplitude of accommodation creation versus the relative aggradation potential of the carbonate factory across the geographical profile (Figure 15). Where the amplitude is high, geographically variable carbonate aggradation rates can lead to deep lagoons and steeper barriers and margins, as in the GBRP. However, low parasequence amplitude provides a vertical ‘equalizer’ that does not allow the greater aggradation potential at the margin to reach fruition. Higher-aggrading communities are limited by sea level, spilling sediment laterally to help fill accommodation with lower-aggrading communities in the back-reef. This may have limited development of deeper lagoons behind aggrading CF3 margins in the DGBR. Where aggradation potential is equivalent across the geographic profile, the role of accommodation is decreased, but still, no deep lagoon should form. The CF1 banks of the DGBR had similar aggradation potential across the bank mostly even to the margin and thus, did not create deep lagoons even as accommodation increased through the early Frasnian.
• The DGBR is not a single reef geographically or through time any more than is the modern GBR. To understand it, it must be separated into different temporal units that reflect the changing carbonate factories and external constraints.
• Despite the strong ‘pull of the recent’, corals are not the answer to understanding the DGBR or other Late Devonian reefs. Stromatoporoids were the best functional analogs for modern reef corals, but their trophic structures and general biology are less well understood. Importantly, the key binders in the DGBR, calcimicrobes, are even less well understood. Recent work suggests that some were mixotrophs capable of photosynthesis where exposed, but other types of metabolism in cryptic settings [455]. We know little about them but cannot afford to treat all reefal microbialites as the products of cyanobacteria.
• Trace element geochemistry suggests that the DGBR occurred in a higher nutrient setting than the outer GBR today (i.e., not oligotrophic). However, the source of the high alkalinity that drove cementation and calcimicrobe growth in CF3–5, but not in CF1, is less easy to identify.
• Isolation of the exposed DGBR in an intracontinental setting and absence of abundant shallow anoxia [509] suggests that local marine chemistry could have differed in many important ways from that of the ‘global ocean’ where coeval reefs around the world occurred. However, broad stratigraphy shows that most or all Late Devonian reefs were affected by a major eustatic sea level cycle through the Frasnian and Famennian. Hence, the DGBR may be typical of reefs of its time in some ways, but not in others.

5.4. Recommendations

• Conservation of the GBR and other modern reefs requires transdisciplinary study that addresses threats that affect changes to any and all of the five key reef processes (Figure 5).
• Additional deep reef coring is needed in the northern GBR and in the offshore plateaus to better understand how the older reef history (Miocene to Pleistocene) was constrained, as that history led to the reefs we know today.
• Additional Holocene and Pleistocene reef coring in the GBR is needed to provide direct analogues for modern reefs that had changing local conditions, and to recover both paleoclimate data to inform modern climate models and projections and data of the specific responses of individual organisms, communities and whole reefs to past environmental–climatic changes. All will inform conservation strategies for modern reefs.
• Current understanding of the DGBR is limited by difficulties in correlating the shallow successions. Hence, increased shallow biostratigraphy (e.g., microvertebrates) and chemostratigraphy (especially stable isotopes) are needed for addition to the ever-improving sequence stratigraphy.
• Additional geochemical investigations are also needed in the Canning Basin to better understand the patterns and causes of changes to marine alkalinity and nutrients that affected carbonate factories, hydrocarbon systems and base metal emplacement. Trace element geochemistry including REEs, redox sensitive metals and halogens along with stable C and N isotopes derived from carbonates, shales and organic matter would be useful.
• Additional study of subsurface examples of the DGBR (offshore western Australia and Barbwire Terrace) and the exposures in the Bonaparte Gulf would also shed more light on this large reef province.