A Decade of Optical Remote Sensing Applications in Marine Biodiversity and Benthic Habitat Monitoring: A Systematic Review

Highlights

What are the main findings?

• Optical remote sensing (ORS) applications for benthic biodiversity are primarily based on multispectral satellite sensors, with an increasing use of unmanned aerial vehicles (UAVs) and hyperspectral approaches. Pre-processing and validation methods remain highly heterogeneous.
• Studies mainly focus on broad benthic habitat classifications and dominant coastal ecosystems such as coral reefs and seagrasses, in tropical and subtropical regions.

What are the implications of the main findings?

• ORS is a key tool for multi-scale, repeatable monitoring supporting marine conservation and policy frameworks.
• Improving methodological standardisation, taxonomic coverage, and global research equity is essential for enhancing comparability and operational use.

Abstract

Monitoring biodiversity in coastal and marine ecosystems is essential for supporting conservation strategies, sustaining ecosystem services, and meeting policy commitments at multiple scales, including the European Union’s Habitats Directive, Sustainable Development Goal 14 (SDG 14, Life Below Water), and the Kunming–Montreal Global Biodiversity Framework (GBF). However, many benthic habitats remain insufficiently mapped or monitored due to the spatial, temporal, and logistical limitations of traditional field-based approaches. Optical Remote Sensing (ORS), based on the use of optical sensors to retrieve spectral information from shallow-water environments, has emerged as a powerful tool for mapping and monitoring these ecosystems. This study presents a systematic review aimed at providing a comprehensive synthesis of above-water ORS applications for benthic biodiversity and habitat monitoring over the period 2014–2023. A total of 179 peer-reviewed studies were analyzed to identify temporal trends, geographic patterns, target ecosystems, and methodological workflows. The review considered observation platforms including satellite, airborne, unmanned aerial vehicles (UAVs), and field spectrometry systems, together with key preprocessing procedures required for reliable benthic detection, such as atmospheric correction, water column correction, and sunglint removal, alongside validation using independent measurements. The analysis reveals a rapid expansion of ORS applications, with a strong geographic concentration in tropical and subtropical regions. Studies focusing on specific benthic groups predominantly target coral reefs and seagrass ecosystems, although many adopt integrative benthic habitat classifications that incorporate multiple benthic components at the habitat level. However, significant limitations persist, including inconsistent preprocessing workflows, limited reporting transparency, and the underrepresentation of several ecologically important taxa (e.g., annelids, mollusks, echinoderms). Despite these challenges, ORS has become a cornerstone of large-scale and repeatable coastal monitoring. By analyzing methodological practices, ecological targets, and geographic biases, this review provides a critical foundation for improving the robustness, scalability, and global applicability of ORS in benthic habitat mapping, biodiversity monitoring, and ecosystem-based management.

1. Introduction

Biodiversity monitoring in coastal and marine ecosystems is essential for guiding conservation efforts, ensuring the sustainability of ecosystem services, and fulfilling global policy commitments such as the European Union’s Habitats Directive, the United Nations Sustainable Development Goals (particularly SDG 14, Life Below Water, focussing on the conservation and sustainable use of marine resources), and the Kunming–Montreal Global Biodiversity Framework (GBF), which establishes global biodiversity targets for conservation and ecosystem restoration [1,2]. Yet, many benthic habitats, particularly complex rocky and sedimentary systems, remain poorly mapped or monitored due to logistical, technical, or financial constraints. Traditional field-based approaches, such as diving surveys and in situ sampling, while ecologically detailed, often lack the spatial and temporal coverage needed to detect large-scale ecological changes and long-term ecosystem dynamics [3]. In recent decades, remote sensing (RS) technologies have emerged as powerful tools to overcome these limitations. Among them, Optical Remote Sensing (ORS) has shown particular promise in benthic ecology through the detection and characterization of shallow-water habitats based on spectral information. Advances in sensor technology, spanning satellite, airborne, UAV-based, and field spectroscopic systems, have improved the capacity to assess habitat extent, ecological condition, and spatio-temporal change across coastal and shallow marine environments. As a result, ORS is becoming an essential component of ecosystem-based management strategies, biodiversity monitoring programs, and habitat restoration initiatives in coastal and marine environments [4].

Building on the rapid expansion of remote sensing technologies, a substantial body of review literature has emerged examining their application to marine habitat mapping and ecological assessment [3,4,5,6,7,8,9]. Broad methodological syntheses have examined the evolution of shallow-water remote sensing and satellite-derived bathymetry, with an emphasis on optical physics and radiative processes [4], as well as satellite-derived bathymetry approaches focused on depth-retrieval methodologies and their limitations [9]. Other contributions have focused on habitat structural complexity derived from remote sensing data [6] or on long-term trends in benthic habitat mapping methodologies [3]. In parallel, several reviews adopt taxonomic, habitat-based, or regional perspectives. Coral reef ecosystems have received particular attention, including analyses of machine learning approaches for coral habitat classification [10], comparisons between traditional in situ ecological surveys and remote sensing methods [11], and foundational reviews showing how optical remote sensing has evolved from localized reef mapping toward operational frameworks for coral reef monitoring and climate-impact assessment [5]. Tropical seagrass systems have likewise been reviewed with emphasis on suitable remote sensing technologies [7], and regional syntheses have applied systematic review frameworks to assess habitat mapping in specific biogeographic contexts, such as Macaronesia (North Atlantic region including the volcanic archipelagos of Azores, Madeira, Canary Islands, and Cape Verde) [8].

Although these works provide valuable technical, taxon-specific, and regional insights, they do not offer a comprehensive cross-taxa and cross-platform evaluation of ORS applications for benthic biodiversity. Consequently, the literature still lacks a global, structured synthesis dedicated specifically to ORS as a methodological framework, and remains fragmented across platforms, ecological targets, and geographic settings, limiting cross-comparability and a clear understanding of methodological strengths and gaps.

This systematic review aims to fill that gap by analyzing peer-reviewed studies using ORS to map or monitor benthic biodiversity in shallow aquatic environments, including coastal and intertidal systems such as mangroves. Specifically, this study seeks to: (1) identify temporal, geographic, and thematic patterns in the application of ORS to benthic biodiversity research; (2) evaluate the platforms, sensors and preprocessing workflows employed in habitat mapping and biodiversity assessment; (3) characterize the ecosystems and biological groups most frequently targeted; and (4) synthesize methodological challenges, research gaps, and emerging directions discussed in the literature to guide more robust, reproducible and inclusive applications of ORS in marine biodiversity science. While classification algorithms, including machine learning, deep learning, and object-based image analysis, play an important role in benthic habitat mapping, they were not systematically evaluated in this review. Likewise, studies primarily focused on photogrammetric reconstruction or Structure-from-Motion workflows were considered outside the scope of the analysis. Instead, this review emphasizes observation platforms, sensor characteristics, preprocessing procedures, validation practices, and ecological uses of ORS data, including habitat mapping, biodiversity monitoring, and restoration assessment.

2. Materials and Methods

2.1. Literature Search Strategy

To systematically review the use of ORS techniques in benthic biodiversity research, a comprehensive literature search was conducted in accordance with PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines [12]. The search was performed across two major academic databases, Scopus [13] and Web of Science (WoS) [14], covering the period from 2014 to 2023. Only English-language peer-reviewed journal articles were considered. The search query was designed to capture a broad range of ORS applications in shallow water and benthic environments, focusing on biodiversity assessment and monitoring. A structured Boolean query was developed to retrieve relevant literature from the SCOPUS and WOS databases. The search string was organized into five conceptual groups: (i) remote sensing technologies, (ii) shallow-water marine environments, (iii) ecological or mapping applications, (iv) biological or ecological targets, and (v) exclusion terms. Within each group, terms were combined using OR, while groups were combined using AND. For example, combinations such as (“remote sens*” OR satellite OR UAV) AND (“shallow water*” OR benth* OR seafloor). Exclusion terms were applied using NOT (

Table 1. Boolean search terms used in SCOPUS and WOS.

Group	Search Terms
Remote sensing technologies	remote sens*, drone, UAV, UAS, satellite, multispectral, hyperspectral, radiometer, spaceborne, airborne
Shallow-water marine environment	shallow water*, benth*, seafloor
Ecological and mapping applications	mapping, monitoring, identification, cartography, distribution, observation, localization, characterization, restoration, conservation
Biological or ecological targets	habitat, ecosystem, communit*, population, biodiversity, forest, reef, species, alga*, seaweed*, coral, phanerogam, seagrass, cnidaria, polycha*, mussel*
Exclusion terms	lake*, ice, river, fresh water, freshwater, oil

The asterisk (*) represents a wildcard character used to retrieve word variants sharing the same root (e.g., ‘benth*’ includes benthic, benthos, and benthonic).

).

2.2. Screening and Filtering Process

After retrieving the search results, all records (673 in Scopus and 572 in WOS) were exported and compiled into a single database. Duplicate records were identified and removed manually. The remaining articles underwent a two-stage screening process: (1) title and abstract screening to exclude clearly irrelevant studies and (2) full-text review, consisting of an independent full reading by at least two authors to assess final eligibility. The articles were included if they applied ORS techniques (e.g., multispectral, hyperspectral, drone- or satellite-based imagery) and were focused on benthic or shallow marine environments and relevant to biodiversity or ecosystem-level assessments, including mapping, monitoring, species or habitat characterization, or conservation. Studies were excluded if they primarily used non-optical methods, such as acoustic sensors or numerical modeling without remote sensing input, or focused on non-marine or freshwater systems (e.g., rivers, lakes, ice-covered areas, or oil-related studies) or pelagic species (turtles, planktonic or fishes). Studies centered on underwater photogrammetry, Structure-from-Motion workflows, or imaging systems operating exclusively below the water surface were also excluded from the review scope. The selection process is illustrated in a PRISMA flow diagram (Figure 1), which details the number of records identified, screened, excluded, and retained for final analysis. A total of 179 articles met the established inclusion criteria, forming the basis for the database from which variables were extracted and analyzed in this study.

2.3. Data Extraction and Categorization

In addition to the journal statistics obtained from academic databases, detailed information was extracted from each eligible study to support quantitative and qualitative synthesis. The following variables were recorded: maximum depth of the study area, bathymetric range, type of remote sensing platform (e.g., satellite, UAV, aircraft or others), the sensor used, spectral information (Multispectral, RGB, Hyperspectral or Panchromatic), geographical location (country/region), and location type (continental, island, atoll or coastal lagoons). Technical details, such as spatial resolution and year(s) of data acquisition, were also extracted (

Table 2. Description of the variables extracted from the literature review included in the PRISMA analysis, indicating their definitions and the categories or ranges applied in this study.

Variable	Description	Categories/Range
Publication year	Year in which the study was published.	2014–2023
Maximum Depth	Maximum water depth (m) analyzed from RS imagery as reported in the study; categorized as “non-defined” when not specified.	0–25 m or non-defined
Bathymetric range	Depth interval covered (in meters) in the study area, categorized into three classes according to the maximum depth reached.	0–10 m, 0–20 m, and 0–30 m
Platform	Type of platform used to acquire ORS data.	Aircraft, Satellite, UAV, Others
Sensor	Optical sensor or instrument employed for data acquisition.	e.g., Sentinel-2-MSI, Landsat-OLI, WorldView, Aerial cameras
Spectral information	Spectral bands or wavelength ranges used in the study.	Hyperspectral, Multispectral, True color RGB, Panchromatic
Country	Country where the study area is located.	Reported per study
Location	Specific geographic area or site of the study.	Region, archipelago, ocean or sea
Location type	General classification of the study area.	Atolls, Continent, Island, Coastal lagoon
Benthic category	Type of benthic habitat, taxonomic group or substrate investigated. The term “benthic habitat” is used here as an integrative category for studies mapping multiple benthic classes within a single classification framework, including biological groups and/or abiotic substrate types.	e.g., Coral reef, Seagrass, Algae, Sand, Rock, Mixed, Benthic habitat
Taxa/group	Main taxonomic groups or biological assemblages analyzed.	e.g., Ulva spp., Posidonia oceanica, Acropora spp.
Objective	Primary aim of the study based on its stated purpose.	Conservation, Ecological knowledge, Management, Mapping, Methodological improvement, Monitoring, Restoration

).

In parallel, information on methodological procedures was recorded, including the application of atmospheric, sunglint (surface sun-reflection effects), and water column corrections, as well as the use of validation with independent in situ data. These corrections address sensor- and environment-induced distortions common in aquatic environments, while validation provides an empirical basis for assessing the reliability of the remote sensing outputs. Recording these elements allowed for the evaluation of common practices and methodological rigor across studies.

Ecological variables recorded included the main benthic habitat categories, such as algae, seagrass or coral reef, along with taxonomic groups, highlighting the target species whenever mentioned. Studies that involved the mapping of multiple biological groups or that combined biological and substrate or geomorphological features (e.g., sand, rock, reef, rubble) were collectively categorized under the label `benthic habitat’.

As an additional step in classifying the literature, the studies included in the PRISMA analysis were categorized according to their main objectives, which were grouped into seven thematic categories: conservation, ecological knowledge, management, mapping, methodological improvement, monitoring, and restoration. These categories were defined a priori based on the aims explicitly stated in the articles, encompassing studies focused on protecting biodiversity, improving ecosystem understanding, supporting management and spatial planning, generating habitat maps, enhancing methodologies, monitoring temporal changes, and guiding restoration initiatives. Since many studies addressed more than one purpose, two columns were created in the database: one for the primary objective, corresponding to the most directly stated or central aim, and another for a secondary objective when applicable. This procedure allowed the multidimensional nature of ORS applications in marine biodiversity to be captured, while ensuring a structured and comparable categorization across studies.

Data extraction was conducted manually using excel and Rstudio to ensure consistent recording and classification of variables across all reviewed studies. Categories were standardized when necessary to harmonize terminology across studies (e.g., aligning various habitat terms under broader benthic categories). These variables formed the basis for further analysis of trends, methodological patterns, and research gaps in the application of ORS for benthic biodiversity.

3. Results

3.1. Overview of the Studies

3.1.1. Journal Statistics and Long-Term Citation Performance

The 179 reviewed articles were distributed across 73 journals, reflecting the interdisciplinary nature of ORS applications in benthic biodiversity research. The ten most productive journals accounted for 95 publications (approximately 53% of the total;

Table 3. Top 10 JCR-indexed journals by number of publications and cites.

Journal	Articles	Cites	Average Cites/Article
Remote Sens.	33	844	25.6
Remote Sens. Environ.	13	932	71.7
Int. J. Remote Sens.	10	149	14.9
Front. Mar. Sci.	9	142	15.8
Geocarto Int.	6	40	6.7
Coral Reefs	5	380	76.0
J. Coast. Res.*	5	59	11.8
Estuar. Coast. Shelf Sci.	5	147	29.4
IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens.	4	50	12.5
Remote Sens. Ecol. Conserv.	3	134	44.7

* It was included in Journal Citation Reports until 2021.

). Remote Sensing (MDPI) was the leading journal with 33 articles (18%), followed by Remote Sensing of Environment (13 articles) and International Journal of Remote Sensing (10 articles). Most studies (91%) were published in JCR-indexed journals, with nearly 90% published in first-quartile (Q1) or second-quartile (Q2) journals according to Journal Citation Reports (JCR) [15]. In addition to specialized remote sensing journals, contributions also appeared in marine ecology journals such as Coral Reefs [16,17,18,19,20] and Frontiers in Marine Science [21,22,23,24,25,26,27,28,29]. Overall, the reviewed articles accumulated 4721 citations during the study period. Citation performance varied substantially among journals (Table 3), with Remote Sensing of Environment and Coral Reefs showing the highest average citation rates per article. These patterns suggest that ORS applications in marine biodiversity monitoring are increasingly recognized across both remote sensing and marine science communities.

3.1.2. Authorship and International Collaboration

Authorship in the reviewed literature spanned 40 countries, though research activity was highly concentrated in a few. The United States led with 44 publications, followed by Australia (26), Indonesia (22), France and China (13 each), and Italy (11). Collectively, the 14 most represented countries accounted for approximately 78% of total authorship.

International collaboration was present in 39 studies (21.6%). The United States was the most frequent international partner (17 collaborative papers), followed by Australia (11), the United Kingdom (8), Japan (6), and France and Indonesia (5 each). The international co-authorship network (Figure 2) reveals strong collaborative links among major research hubs, particularly between Australia, the United States, and the United Kingdom.

3.1.3. Temporal Trends

As illustrated in Figure 3, publication output increased markedly over the study period, rising from 9 and 6 studies in 2014 and 2015, respectively, to a peak of 32 studies in 2020 (18% of the total). This trend corresponds to an average annual growth rate of approximately 23%, estimated using the Compound Annual Growth Rate (CAGR), which describes the mean annual rate of increase over a defined period [30]. Although publication numbers declined slightly after 2020, research activity remained consistently high during 2021–2023, indicating sustained scientific interest in ORS applications for benthic biodiversity monitoring.

3.1.4. Geographical Trends

The geographic distribution of studies reveals marked disparities in research activity across countries and regions (Figure 4). The United States accounted for 17% of all studies (30 out of 179), followed by Indonesia (12.3%) and Australia (11.7%). France, China, Italy, and Spain contributed smaller but consistent proportions of the reviewed literature, reflecting the concentration of ORS applications in countries with extensive coastal ecosystems and established marine research infrastructure.

Figure 4 also illustrates the geographic setting of study sites within each country. Continental environments dominated in countries such as the United States [31,32,33,34,35] and Australia [19,36,37,38,39], whereas island settings prevailed in several Southeast Asian nations, including Indonesia [40,41,42,43], China [44,45,46,47], and Malaysia [48,49,50,51]. Atoll environments were comparatively less represented and mainly associated with coral reef systems in the Pacific and Indian Oceans [17,52,53,54]. Coastal lagoon environments appeared only in a few studies [55,56], notably in Italy and Spain.

At the regional scale (Figure 5), Southeast Asia represented the largest concentration of study sites (28%), followed by Oceania (23%), the Caribbean (12%), and the Mediterranean (9%). In contrast, Central America, Western Europe, and East Africa collectively accounted for less than 9% of study locations.

3.2. Optical Sensors and Platforms: Distribution and Characteristics

3.2.1. Sensors and Platforms

The reviewed literature included a wide range of ORS sensors deployed across multiple observation platforms, including satellites (201 records), crewed aircraft (33), uncrewed aerial vehicles (UAVs) (26), and other systems such as handheld spectrometers or parasailing-mounted cameras (17). Although some studies combined multiple sensors or platforms, single-sensor approaches predominated, representing 125 of the reviewed articles. Figure 6 summarizes platform usage across two-year intervals from 2014 to 2023. Satellite platforms remained dominant throughout the study period and reached their highest use during 2020–2021. Aircraft-based applications showed relatively stable but limited representation, whereas UAV-based studies increased steadily after 2016–2017, reflecting their growing adoption for high-resolution shallow-water mapping. In contrast, the use of other non-conventional platforms remained comparatively limited and sporadic.

3.2.2. Spectral and Spatial Resolutions

To facilitate analysis, the optical sensors identified in the reviewed studies were grouped into three categories according to the spectral information recorded: multispectral, hyperspectral, and RGB/panchromatic (PAN) systems.

Table 4. Multispectral remote sensors: spatial resolution and spectral characteristics.

Satellite/Sensor	Spatial Resolution (m)		Bands	Spectral Range (nm)
	MS	PAN
ALOS AVNIR–2 [48,70,71]	10	–	4	420–890
Dove (PlanetScope)	3–4.1	–	4–8	430–890
Dubaisat–2 [72]	4.0	1.0	4	450–890
FORMOSAT–2	8.0	2.0	5	450–900
GeoEye–1	1.65	0.41	5	450–920
IKONOS	3.2	0.82	5	445–900
Landsat–1 (MSS)	79	–	4	500–1100
Landsat–2 (MSS)	79	–	4	500–1100
Landsat–4 (TM)	30 (120)	–	7	450–2350 (+TIR)
Landsat–5 (TM)	30 (120)	–	7	450–2350 (+TIR)
Landsat–7 (ETM+)	30 (60)	15	8	450–2350 (+TIR)
Landsat–8 (OLI)	30	15	11	450–2350
Landsat–9 (OLI–2)	30	15	11	450–2350
MERIS–ENVISAT [73,74]	300–1200	–	15	390–1040
Micasense RedEdge–M on UAV at 120 m high	0.08	–	5	465–860
Pleiades–HR	2.8	0.7	5	450–915
QuickBird	2.4	0.61	5	450–900
RapidEye	5.0	–	5	440–850
Sentinel–2 (MSI)	10, 20, 60	–	13	443–2190
Sentinel–3 (OLCI)	300	–	21	400–1020
SkySat - C Gen.	1.0	0.72	5	450–900
SPOT–1 (HRV)	20	10	4	500–890
SPOT–2 (HRV)	20	10	4	500–890
SPOT–3 (HRV)	20	10	4	500–890
SPOT–4 (HRVIR)	20	10	5	500–1750
SPOT–5 (HRG)	10, 20	5	5	500–1750
SPOT–6	6.0	1.5	5	450–890
SPOT–7	6.0	1.5	5	450–890
WorldView–2	1.84	0.46	9	400–1040
WorldView–3	1.24 (3.7)	0.31	17	400–1040 (+SWIR)
Zi Yuan–3A	5.8	2.1	5	450–890

summarizes the multispectral sensors identified in the literature, including 30 satellite systems and one UAV-mounted camera. Frequently used platforms included Landsat, Sentinel-2, SPOT, WorldView-2/3, Dove [57,58,59,60], RapidEye [61,62], QuickBird [53,63,64], and GeoEye-1 [65,66,67]. Most commercial satellites provided spatial resolutions below 5 m, supporting fine-scale coastal and benthic mapping, whereas UAV-mounted systems such as the Micasense RedEdge–M achieved centimetric resolutions suitable for ultra-high-resolution applications in shallow environments [50,55,68,69]. Despite differences in spectral coverage, most studies focused on visible wavelengths (approximately 400–700 nm), which are most suitable for underwater observations due to their greater penetration through the water column.

Hyperspectral systems (

Table 5. Overview of hyperspectral imaging systems: spatial and spectral specifications across platforms.

Sensor	Platform	Spatial Res. (m)	Spectral Range (nm)
AHS	Aircraft	2.0–6.0	430–12,800
AisaEAGLE–1K	Aircraft	0.3–4.0	400–970
ASD FieldSpec	Other	N/A (point spectrometer)	350–2500
AVIRIS	Aircraft	4.0–20.0	380–2500
CAO–2	Aircraft	1.0–2.5	380–2510
CASI	Aircraft	0.5–2.0	380–1050
HyMap	Aircraft	3.0–10.0	450–2500
Hyspex VNIR–1600	Aircraft	0.5–2.0	400–1000
MIVIS	Aircraft	2.0–20.0	430–12,700
NASA/JPL PRISM	Aircraft	0.3–1.0	350–1050
Pika XC2	Other (Translation system)	Setup-dependent	400–1000
Pika–L	UAV	Setup-dependent	400–1000
PRISMA	Satellite	30	400–2500
SAMSOM	Aircraft	1.0–5.0	380–970
SOC710 [86]	Other (Tripode)	Setup-dependent	380–1040

AHS (Airborne Hyperspectral Scanner); AVIRIS (Airborne Visible InfraRed Imaging Spectrometer); CAO-2 (Carnegie Airborne Observatory–2); CASI (Compact Airborne Spectrographic Imager); HyMap (Hyperspectral Mapper); MIVIS (Multispectral Infrared Visible Imaging Spectrometer); NASA/JPL PRISM (Portable Remote Imaging SpectroMeter); PRISMA (Hyperspectral Precursor of the Application Mission); SAMSOM (Spectroscopic Aerial Mapping System with On-board Navigation hyperspectral imager).

) were primarily airborne, with the HySpex VNIR–1600 sensor being the most frequently used [24,75,76,77,78,79,80]. Other systems, including AVIRIS, CASI, AisaEAGLE–1K, and CAO–2, appeared less frequently, while PRISMA represented the only spaceborne hyperspectral mission identified in the review [81]. These systems capture hundreds of narrow contiguous bands across the visible and shortwave infrared ranges, enabling detailed spectral characterization of benthic habitats. A distinct case was the ASD FieldSpec spectrometer, reported in 11 studies [24,34,47,68,75,78,80,82,83,84,85]. Rather than producing imagery, this instrument was mainly used to acquire reference spectral signatures for benthic substrates and biological components, supporting spectral library construction, sensor simulation, and image interpretation.

RGB and PAN systems (

Table 6. Summary of RGB and/or panchromatic (PAN) imaging systems by platform.

Sensor/Product	Platform	Type	Spectral Range (nm)	Spatial Resolution (cm)	Reference
Aerial orthophotography	Aircraft	RGB	NA	10–100	[78,93,98]
Integrated camera on DJI UAVs	UAV	RGB	∼400–700	<10	[16,29,38,43,47,50,69,88,89,90,91,92,93,94,95,96,97]
FluidCam NASA	Aircraft	RGB	380–720	Setup-dependent	[26]
Google Earth Imagery	Satellite	RGB	NA	NA	[34,101]
GoPro	UAV	RGB	NA	<1	[35]
Grasshopper GRAS-14S5M	Other	PAN	∼400–1000	0.5	[100]
Grasshopper GRAS-14S5C	Other	RGB	∼400–700	0.5	[100]
Digital photography	Other	RGB	NA	<1	[102]
Sony NEX-7 HD	Other (Parasailing)	RGB	370–730	NA	[87]
Sony RX0 1.0	UAV (Fixed-wing)	RGB	∼400–700	<4.62	[103]
WorldView-1	Satellite	PAN	450–900	50	[99]

) included satellites, aerial orthophotography, UAV-mounted cameras, and other non-conventional imaging systems such as parasailing with a Sony NEX-7 HD camera [87]. UAV-based RGB cameras, particularly DJI integrated systems, were among the most frequently employed sensors [16,29,38,43,47,50,69,88,89,90,91,92,93,94,95,96,97], reflecting the growing accessibility of centimetric-resolution imagery for site-scale coastal monitoring. Historical aerial orthophotography was also used in several studies [78,93,98], demonstrating the value of archival datasets for long-term environmental assessment. Satellite-derived RGB imagery, including Google Earth products, was occasionally employed, although incomplete metadata often limited reproducibility. PAN sensors were less common, with examples including the high-resolution WorldView-1 satellite [99] and the Grasshopper GRAS-14S5M camera [100]. Emerging approaches based on airborne fluid lensing, designed to reduce wave-induced distortions in shallow-water observations, were reported only once in the reviewed literature [26].

3.2.3. Ranking of the Most Frequently Used Sensors

Figure 7 ranks the 14 most commonly employed sensors across the reviewed 179 studies. Satellite platforms clearly dominated, particularly WorldView-2 (38 studies), Landsat-8 (35), and Sentinel-2 (29). The Dove microsatellites by Planet Labs (11) further highlight the growing role of small satellite platforms. When grouping by satellite families, Landsat missions accounted for 61 studies (Landsat-1 to -9), while Maxar Technologies’ suite, now operated by Vantor, (WorldView-2/3, GeoEye-1, QuickBird) appeared in 62 studies, underscoring their prominence. The Pleiades-HR [78,104,105,106,107] and SPOT-7 [82,108,109,110,111] sensors by Airbus Defence and Space (5 each) represent additional high-resolution alternatives. Among UAVs, DJI’s integrated cameras (17 studies) were the most frequently used, reflecting the increasing adoption of low-altitude, centimeter-scale mapping approaches. Airborne deployments favored the HySpex VNIR–1600 (7 studies), whereas the ASD FieldSpec (11 studies) was the principal sensor within the “Other” platform group.

3.3. Pre-Processing, Correction Methods and Validation Procedures

3.3.1. Processing Levels and Input Products

Image processing levels describe the degree of radiometric and geometric correction already applied to remote sensing products before analysis. In general, Level 1 products correspond to geometrically corrected imagery with limited radiometric processing, whereas Level 2 products typically include surface reflectance retrievals after atmospheric correction. Higher levels may incorporate additional processing steps depending on the sensor provider [112,113].

A total of 234 sensor instances from satellite and airborne platforms were identified for this analysis, including 201 satellite sensors and 33 airborne systems. UAV-based and other non-conventional platforms were excluded because the processing level of the imagery was rarely reported. Among airborne sensors, only two studies specified the use of Level 1B products, corresponding to CASI [114] and HySpex VNIR-1600 [75]. Most reported processing levels referred to satellite imagery (51 instances). Level 1 products were the most frequently used (33 cases), followed by Level 2 products (15 cases). Sentinel-2 Level 1C imagery was particularly common, being explicitly reported in 15 studies [42,45,56,71,115,116,117,118,119,120,121,122,123,124,125]. However, the processing level was not specified for approximately 75% of all satellite sensors (150 out of 201), revealing limited methodological transparency across the reviewed literature. A distinct case involved PlanetScope Dove imagery, for which three studies [47,126,127] reported the use of Level 3B products. According to PlanetScope specifications, these products correspond to orthorectified top-of-atmosphere radiance imagery broadly comparable to Level 1C products from other satellite missions. One study [47], however, explicitly referred to Level 3B analytic surface reflectance products, indicating that both orthorectification and radiometric correction had already been applied.

3.3.2. Atmospheric Correction

A total of 277 sensor–method combinations, primarily from satellite and airborne observations, were analyzed to evaluate the application of atmospheric correction procedures. In total 64 cases explicitly stated that no atmospheric correction was applied, while 49 provided no information on this aspect. A total of 30 instances reported the use of atmospheric correction without specifying the method.

Empirical or simplified approaches were used, with the dark object subtraction (DOS) method (34) as the most common, followed by the empirical line method (4) and COST (Cosine of the Solar Zenith angle) (2). Model-based radiative transfer approaches were also widely employed, including FLAASH (Fast Linearly-adjusted Adjusted Radiative Transfer Code) (36), ATCOR (Atmospheric and Topographic Correction) (16), 6S (Second Simulation of the Satellite Signal in the Solar Spectrum) (14), ATREM (ATmospheric REMoval) (7), Sen2Cor (Sentinel 2 Correction) (7), ACOLITE (Atmospheric Correction for Operational Land Imager ‘lite’) (5) and ICOR (Image Correction for Atmospheric Effects) (2).

Processing chains specifically designed for aquatic or coastal environments were less common but considerable, comprising C2RCC (Case 2 Regional Coast Colour) (3), SeaDAS (SeaWiFS Data Analysis System) (3), and the POLYMER polynomial-based algorithm (1). Less frequently mentioned tools included MIP (Modular Inversion and Processing System), libRadtran (library for radiative transfer), TAFKAA (The Algorithm Formerly Known As ATREM), and a case-specific implementation following [128].

3.3.3. Water Column Correction

Water column correction aims to reduce depth-related attenuation effects in optically shallow waters to improve the retrieval of bottom reflectance and benthic features. It can be applied to optical imagery acquired from satellite, airborne, UAV-based, and other above-water observation systems.

The review of water column correction methods indicates a clear predominance of empirical approaches over physically based, model-driven techniques. In a substantial proportion of cases, no water column correction was applied (42%), while in 16% of instances the methodology was not reported.

Among studies implementing a correction, the empirical approach proposed by Lyzenga was by far the most widely used, accounting for 51 applications (18%). This method, originally introduced by [129,130], combines spectral bands into depth-invariant indices to reduce the influence of water-column attenuation on bottom reflectance under the assumption of spatially homogeneous water optical properties, thereby improving the discrimination of benthic features. A further 17% of cases reported the application of a water column correction without specifying the underlying method. Physically based approaches were less frequent, with methods from [131,132,133] collectively accounting for eight applications. Other correction techniques, including those proposed by [134,135], as well as several singular or study-specific implementations, together represented a minor fraction (4%) of the total reported applications.

3.3.4. Sunglint Correction

The analysis of sunglint treatment across the selected studies reveals that explicit correction or mitigation strategies were applied relatively infrequently. In most cases, no sunglint correction was performed (138 occurrences), while a substantial number of studies did not report whether any procedure was applied (52).

When sunglint handling was indicated, the specific method was often not described (38 times). Among studies that did report an explicit approach, the empirical method proposed by [136] was by far the most widely used (29 occurrences), making it the dominant technique for sunglint correction in shallow-water remote sensing. This method exploits the assumption of negligible water-leaving radiance in the near-infrared over deep waters and relies on the statistical relationship between visible and NIR bands to estimate and remove the glint component.

Less formal strategies were also reported, including the visual inspection of imagery to exclude glint-contaminated scenes (11 occurrences). Other methods appeared only sporadically, such as a modified Google CloudScore algorithm [137], MIP [138,139], the approach of Lee et al. (2001) [140], POLYMER [45], or the use of a circular polarizing filter during image acquisition [29].

3.3.5. Validation Procedures

The review reveals substantial heterogeneity in how accuracy assessment is conducted and reported across studies, which prevented a consistent categorization of validation approaches. Nevertheless, some form of validation was reported in more than 90% of the reviewed studies, indicating that accuracy assessment is a common component of ORS applications, despite the lack of methodological consistency. While many studies reported validation using independent datasets derived from in situ surveys, underwater imagery, or UAV-based observations, others relied on internal validation strategies such as cross-validation or random partitioning of training data. In many cases, the distinction between independent and non-independent validation datasets is not clearly defined, further limiting comparability and reproducibility.

3.4. Benthic Habitats, Biodiversity Patterns, and Research Applications

3.4.1. Benthic Categories

Among the 179 reviewed studies, 97 articles (54%) were grouped under the broad category “benthic habitat” because they addressed multiple benthic components, habitat types, or integrated classifications rather than focusing on a single ecological target (Figure 8). This category often encompassed diverse environments and biological elements, sometimes combining components such as microphytobenthos, mangroves, oyster beds (e.g., Crassostrea spp.), and substrate features within a common classification framework [74,75,82]. Of these 97 studies, 55 (30.7% of the total) combined biological and geomorphological elements, including sand, reef, rock, or rubble, whereas the remaining 42 focused exclusively on biological components. These two subdivisions correspond to the outer ring of Figure 8. The remaining 82 studies (46%) focused on specific benthic targets or ecosystems. Seagrass ecosystems represented the largest category (20%), followed by coral reef studies (19%). Coral studies were assigned to this category only when coral reefs represented the explicit primary target of the study, whereas studies integrating corals within broader habitat classifications, mixed benthic assemblages, or combined biological–geomorphological frameworks were retained within the “benthic habitat” category. Algae represented 4% of the studies, followed by less represented targets including annelids [24], mollusks [141], foraminifera [17], echinoderms [38], and mangroves [142].

These patterns reflect the predominant use of broad habitat-based classifications in ORS studies. The following section further explores the biological diversity represented within these studies through the taxonomic composition of the reported organisms.

3.4.2. Taxa List

Building on this classification framework, the reviewed studies targeted a broad range of benthic organisms belonging to 13 major taxonomic groups (phyla or divisions; Figure 9), which are further subdivided into 26 orders and 48 families. Despite this apparent taxonomic breadth, there is a marked concentration of studies on Tracheophyta (vascular plants with 100 articles), primarily driven by seagrass-focused applications, although a smaller number of studies also addressed mangrove ecosystems, and on Cnidaria, largely associated with coral reef mapping and monitoring. These groups are frequently treated as broad functional categories, particularly “seagrass” and “coral”. In terms of taxonomic resolution, 87 studies (48.6%) did not report any genus or species level identification. Among the remaining 92 studies, 112 distinct species belonging to 69 genera were identified.

Within Tracheophyta, most research focused on seagrasses, reported in 46 studies (25.4%), while mangroves appeared in seven (3.8%). Commonly cited species included Posidonia oceanica (Mediterranean seagrass; 11 studies), Thalassia testudinum (turtle grass; 10 studies), Halophila ovalis (paddle grass; 10 studies), and Thalassia hemprichii (9 studies). Other Cymodoceaceae species, such as Cymodocea nodosa, Halodule uninervis, Halodule wrightii, Syringodium filiforme, and Syringodium isoetifolium, were each reported in more than six studies.

Cnidarians include reef-building corals, sea anemones, and related anthozoans. In 70 of these (39.2% of the total dataset), the term “coral” was used generically, referring to hard, soft, live, or dead reef-forming corals without further specification. When identified at genus level, the most frequently reported taxa were Porites (massive reef-building corals; 14 studies) and Acropora (branching or table corals; 11 studies), both dominant reef-building genera in tropical systems. This pattern reflects the strong emphasis placed on coral reef habitat mapping, although species-level resolution often remains limited.

Macroalgae were recorded in 62 studies (34.3%), although in 26 of these the group was referred to simply as “algae” without taxonomic detail. Brown algae were most frequently identified, particularly Sargassum species (rockweeds; seven species across six studies; [47,83,143,144,145,146]). Green algae were mainly represented by Caulerpa (three species in five studies) and Ulva (sea lettuce; two species in four studies). Red algae appeared in 13 studies; in five cases they were described as calcareous red algae, rhodolith-forming communities or maërl [36,54,147,148,149].

3.4.3. Biodiversity and Depth

Regarding the environmental context of these studies, depth distribution shows a clear predominance of shallow-water applications. In 79 articles (43.9%), the study depth did not exceed 10 m, while in 49 articles (27.2%) it remained within 20 m. In total, 28 studies (15.6%) extended up to 30 m, and only 3 studies were conducted exclusively within the first 5 m. Six articles focused on intertidal environments, and in 15 cases (8.3%) no bathymetric information was provided.

When examined by benthic category, clear depth-related differences emerged. Benthic habitat spans all predefined depth intervals, reflecting its use across a wide range of shallow and moderately deep environments. In contrast, categories such as Algae and Seagrass are predominantly associated with intertidal and shallow-water zones, reflecting both their ecological distribution and the depth limitations inherent to optical remote sensing methods.

Analysis of maximum depth shows variation among benthic categories. Studies classified as Coral reached the highest average maximum depths (22–26.7 m). In contrast, categories involving Seagrass or Algae were associated with lower mean maximum depths, generally below 16 m. Benthic habitat exhibited average maximum depths comparable to those reported for Coral.

3.4.4. Research Objectives in Benthic Biodiversity Studies Using ORS

Among the 179 scientific articles selected through PRISMA criteria, the most frequently cited research objectives related to the application of ORS to marine benthic biodiversity were Mapping (75%) and Monitoring (31.7%) (Figure 10). These were followed by Methodological improvement (30%), Ecological knowledge (10.5%), and Conservation (9.4%). Other cited objectives included Management (6.1%), Identification (5.5%), and Restoration (3.3%).

Most of the articles (72%) address more than one research objective, revealing through the chord diagram a dense and interwoven network of co-occurrences across the reviewed studies (Figure 10). Mapping stands out as the central objective, co-occurring with Methodological improvement (36 studies), Monitoring (34) and Conservation (14). Monitoring also appears with nearly all other objectives, but it is mainly the secondary objective behind mapping. Less commonly cited objectives, such as Management, Identification, and Restoration, tend to appear in conjunction with broader goals.

3.4.5. Cross-Analysis of Sensors by Benthic Category and Research Objectives

As noted in the previous section, studies addressing the generic “Benthic habitat” category constituted the largest share of the reviewed literature and involved the broadest range of sensors. Multispectral satellite platforms dominated this category, with WorldView-2 (21 studies), Landsat-8 (17) and Sentinel-2 (16) being the most frequently used sensors. Seagrass constituted the second most intensively studied benthic class and was predominantly mapped using Landsat-8 and Sentinel-2 (8 studies each), followed by WorldView-2 (7). Coral-focused studies were also well represented, particularly those employing Landsat-8 (10), WorldView-2 (7), and DJI-based UAV imagery (5), indicating the use of sensors with different spatial resolution. In contrast, algae-related studies were comparatively scarce. With the exception of WorldView-2, which was used in two studies [20,144], all other sensors appeared only once in this category, including the monochrome and multispectral Grasshopper systems [100], high-resolution satellites as Ikonos [33], GeoEye-1 [145] and SPOT-7 [110], and a multi-sensor combination involving Sentinel-2, Dove, DJI UAV, and ASD FieldSpec [47]. These patterns are synthesised in Figure 11, presented as a heatmap showing the frequency of sensor use across benthic categories, with colour intensity ranging from dark red (highest frequency) to white (no occurrences). The heatmap highlights the strong concentration of multispectral satellite sensors in broadly defined benthic habitat classes.

When analyzing sensor use across specific taxonomic groups, the most broadly applied sensors were WorldView-2 and Landsat-8, covering eight and seven different major taxonomic groups (phyla or divisions), respectively. However, certain groups showed clear associations with specific sensor types, likely driven by their spatial scale, habitat complexity, or detectability. For example, Annelida were consistently studied using high-resolution hyperspectral sensors such as ASD FieldSpec and Hyspex VNIR-1600, which are suitable for capturing fine-scale benthic features. Porifera, in contrast, were primarily mapped using Landsat-8, a multispectral satellite sensor with moderate spatial resolution appropriate for broader habitat-level detection. Studies on mollusks demonstrated the use of a wide array of platforms, including Landsat-5, -7, and -8, RapidEye, SPOT-4, SPOT-7, and WorldView-2 and -3.

Focusing on sensor usage within individual research objectives, mapping applications were primarily supported by satellite-based systems, with WorldView-2 (26 studies), Landsat-8 (24), and Sentinel-2 (17) being the most frequently employed sensors. Smaller commercial satellite platforms, such as QuickBird (8), Dove (6) and Pleiades-HR (3) [78,105,107], were also commonly employed for mapping purposes. Among airborne sensors, HySpex VNIR-1600 (5), CASI (4), AVIRIS (4), and AisaEAGLE-1K-OGS and CAO (Carnegie Airborne Observatory-2) [49,150] (2 each) contributed additional hyperspectral capabilities. The ASD FieldSpec spectrometer (8) was primarily used in support of mapping campaigns. UAVs, particularly DJI UAVs (12) and the Micasense RedEdge-M camera on DJI (2), further reflect the growing role of low-altitude platforms in shallow-water habitat mapping.

Monitoring was the second most frequently cited objective; however, only a limited number of remote sensing systems were repeatedly employed for this purpose. Sentinel-2 [34,120,151] and sensors from the Landsat program (specifically Landsat-4, -5, -7, and -8) appeared in two to three studies each [152,153,154], with all of them corresponding to satellite-based platforms.

Beyond the predominant applications of mapping and monitoring, several sensors demonstrated notable versatility by supporting a broader range of objectives in shallow-water remote sensing. WorldView-2 stands out as the most multi-purpose sensor, also being used in studies focused on all remaining main objectives, reflecting its high spatial resolution and consistent availability. Sentinel-2 had the second broadest thematic coverage, with documented applications in all objectives except conservation, highlighting its growing relevance as a free, open-access platform supporting mapping, monitoring, ecological knowledge, restoration, and management applications. Next was the Landsat-8, which, beyond its extensive use for mapping, was also employed in studies related to conservation (2), identification (1), monitoring (3), and restoration (1). This likely resulted from its moderate spatial resolution, high temporal frequency, and long-term data continuity.

4. Discussion

4.1. Key Technological Trends

One of the clearest technological trends identified between 2014 and 2023 is the consolidation of open access satellite missions as the backbone of shallow-water ORS. The widespread use of Landsat-8 [19,28,39,41,61,62,82,84,99,115,119,137,138,146,152,153,154,155,156,157,158,159,160,161,162,163,164,165,166,167,168,169,170,171,172] and Sentinel-2 [34,37,42,45,46,47,56,71,81,115,116,117,118,119,120,121,122,123,124,125,137,151,161,166,167,173,174,175,176] highlights their central role in benthic habitat monitoring at broad spatial scales, supported by global coverage, systematic acquisition, and free-data policies. These characteristics facilitate long-term monitoring and cross-regional comparisons.

Commercial very-high-resolution sensors, particularly those developed by Maxar Technologies [17,18,20,23,32,40,44,51,52,53,54,63,64,65,66,67,85,99,115,141,143,144,145,148,149,166,168,169,175,177,178,179,180,181,182,183,184,185,186,187,188,189,190,191,192,193,194,195,196,197,198,199,200], remain important for habitat-scale applications requiring detailed spatial information. Their use supports the identification of fine-scale ecological patterns, although acquisition costs continue to limit broader adoption.

UAV applications expanded rapidly after approximately 2018 [16,29,35,38,43,47,50,55,68,69,71,88,89,90,91,92,93,94,95,96,97,103,107,175], driven by lower platform costs, lightweight multispectral sensors, and advances in classification methods. UAVs provide centimetric-resolution observations suitable for shallow and spatially constrained environments, but their operational range and dependence on environmental and regulatory conditions remain limiting factors.

Microsatellite constellations, particularly Planet Labs systems [47,57,58,59,60,126,127,166,167,196,197], reflect an increasing emphasis on temporal resolution and near-continuous monitoring. Their high revisit frequency complements traditional satellite missions by improving temporal coverage despite more limited spectral capabilities.

Airborne hyperspectral sensors [21,24,26,27,28,34,36,42,49,75,76,77,78,79,80,83,84,98,102,114,139,147,150,166,190,201,202,203,204,205], although less frequently applied, remain among the most powerful tools for detailed benthic discrimination. Their high spectral resolution allows for improved detection of subtle differences in benthic composition. However, their operational complexity and high cost limit broader adoption. Deployment often requires dedicated aircraft, expert personnel, and sophisticated sensor integration, along with extensive post-processing workflows. As a result, their use is generally restricted to government agencies or academic research institutions, such as those linked to NASA or university-led observatories, and access typically depends on collaborative frameworks or competitive funding schemes. Consequently, while hyperspectral approaches demonstrate strong analytical potential, their scalability and operational integration into routine monitoring remain limited.

A further notable trend is the increasing integration of multiple platforms within individual studies [34,47,71,78,166,167]. Hybrid approaches combining satellite observations with UAV-based data, in situ surveys, and field spectroscopy are progressively used for calibration, validation, and multi-scale analysis. Typically, satellite imagery provides regional-scale habitat mapping and temporal coverage, UAV observations contribute centimetric spatial detail, field spectroscopy supports spectral library development and calibration, and in situ observations are used for validation and ecological interpretation. An example is provided by the study of eelgrass mapping in Elkhorn Slough, California, which combined Sentinel-2 imagery, Google Earth products, airborne hyperspectral PRISM observations, and bio-optical modelling to improve habitat detection under turbid water conditions [34]. This convergence reflects a methodological maturation of ORS, shifting from single-sensor frameworks toward integrative strategies that leverage complementary strengths in spatial, temporal, and spectral resolution.

4.2. Research Priorities and Taxonomic Focus

Research patterns in the reviewed literature reveal a heterogeneous taxonomic focus, with variations appearing to be driven mainly by technological developments, data availability, and research interests rather than by temporal changes in the ecological targets investigated.

A substantial proportion of studies were grouped under the integrative category “benthic habitat” (97 cases; 54%), encompassing classifications that combined biological and physical components such as seagrasses, macroalgae, corals, sand, rock, and rubble. This pattern reflects a major strength of ORS: the capacity to integrate multiple benthic elements within heterogeneous coastal environments. However, species-level discrimination often remains limited by sensor resolution, water column effects, and habitat complexity.

Within this ecosystem perspective, coral reef ecosystems remain a major focus of ORS applications, particularly in relation to climate change. Coral reefs are among the most biodiverse and ecologically valuable marine systems, yet they are increasingly threatened by rising sea surface temperatures, which drive widespread coral bleaching and associated mortality [206]. A major scientific breakthrough over the last decade has been the transition from opportunistic, site-specific mapping toward operational, multi-temporal monitoring of reef condition and thermal stress at regional to global scales [5,207]. Time-series imagery from platforms such as Sentinel-2 and Landsat is now routinely used to detect and quantify bleaching events, assess spatial patterns of thermal stress, characterize habitat condition, and support reef resilience assessments [58,152,167,207]. These advances have enabled the development of operational monitoring initiatives such as NOAA’s Coral Reef Watch (CRW), which provides global bleaching alert systems based on sea surface temperature anomalies, and the Allen Coral Atlas [208], which delivers standardized coral habitat mapping products and bottom reflectance analyses at unprecedented spatial scales. Collectively, these developments illustrate how ORS has evolved from a primarily descriptive mapping tool into an operational framework supporting climate adaptation, ecological forecasting, and large-scale reef monitoring.

Seagrass ecosystems have also received considerable attention due to their ecological and socio-economic importance. Species-level mapping is comparatively more common, including applications involving Posidonia oceanica [81,93,104,118,149,164,174,175,203,209] and Thalassia testudinum [31,114,170,188,191,201,202]. This likely reflects the relatively homogeneous structure of many seagrass meadows compared with the taxonomic richness and structural complexity of coral reefs. Consequently, coral applications often remain at broader functional levels, whereas species-level discrimination is more feasible in homogeneous systems or in approaches based on close-range photogrammetry, which were outside the scope of this review.

Microphytobenthos, i.e., communities of microscopic photosynthetic organisms inhabiting surface sediments, appeared in several integrative assessments [22,36,56,75,82]. Although still underrepresented, its increasing inclusion reflects its ecological importance and potential suitability for hyperspectral and UAV-based applications, particularly in intertidal environments [22].

Despite these advances, ORS applications remain uneven across taxa. Groups such as mollusks, echinoderms, and sponges are rarely the primary focus of mapping efforts [17,24,38,68,141], mainly because their spectral ambiguity and fragmented spatial distribution limit detection. Their limited representation therefore appears to reflect methodological constraints rather than ecological relevance.

Overall, current ORS applications tend to prioritize benthic systems combining ecological importance, management relevance, and favourable spectral characteristics. While this trend is operationally justified, it may also introduce biases in the representation of benthic biodiversity.

4.3. Geographic and Biogeographic Biases

The spatial distribution of studies reveals a strong concentration in tropical and subtropical regions, with approximately three-quarters of research conducted between 23°N and 23°S. Southeast Asia, the Caribbean, and parts of Oceania account for 63% of case studies, reflecting the global importance of coral reefs and associated shallow-water habitats in these regions. Despite this concentration, some reef-bearing regions appear comparatively underrepresented relative to their ecological relevance, particularly parts of East Africa, the Western Indian Ocean, and several tropical island systems. However, a quantitative assessment of study density relative to reef extent was beyond the scope of this review.

The predominance of low-latitude systems is partly explained by their exceptional biodiversity and conservation relevance. However, it also introduces methodological biases because optical properties, benthic composition, water clarity, and environmental variability differ substantially among tropical, temperate, and polar environments. Consequently, workflows and correction approaches developed for clear tropical waters may not perform equally well in turbid, seasonally dynamic, or high-latitude systems [210].

Temperate ecosystems, including kelp forests, mixed sediment habitats, and soft-bottom systems, remain comparatively underrepresented despite their ecological importance [22,24,29,47,55,68,69,75,77,79,80,82,93,94,96,101,108,143,144,164,192]. Polar and subpolar coastal regions are also less represented due to logistical constraints, extreme illumination regimes, cloud cover, and seasonal ice [62,95,97,100,110,165,177,205]. Limited methodological adaptation to these environments likely reinforces existing geographic biases.

These patterns are consistent with the distribution of scientific infrastructure and remote sensing capacity. Countries such as the United States, Australia, Indonesia, and France contribute disproportionately to both authorship and study locations, whereas several biodiversity-rich regions remain less represented. This imbalance likely reflects differences in funding, technological access, and institutional capacity.

Overall, current ORS frameworks have been developed mainly within a limited set of environmental and institutional contexts, which may restrict their transferability to underrepresented ecosystems.

4.4. Methodological Challenges in ORS of Benthic Ecosystems

Despite the rapid expansion of ORS applications in benthic ecology, several methodological limitations still constrain the robustness, comparability, and operational transferability of derived products. These challenges reflect both technical and conceptual difficulties associated with mapping biologically diverse and optically dynamic shallow-water environments. In particular, the review reveals substantial heterogeneity in preprocessing workflows, reporting practices, and validation strategies.

One of the most critical issues identified is the variability in atmospheric correction approaches. This step is fundamental for removing atmospheric scattering, aerosol effects, and adjacency contamination, yet its implementation remains inconsistent across studies. For multispectral satellite missions, tools such as Sen2Cor [56,71,117,118,119,120,175] and ACOLITE [34,81,160,164,169] are increasingly adopted, especially for Sentinel-2 imagery, while simplified empirical approaches, particularly Dark Object Subtraction (DOS), remain widespread due to their operational simplicity and low computational cost. However, these empirical methods may reduce cross-sensor comparability, particularly in multi-temporal or multi-platform analyses. In contrast, physically based radiative transfer models such as FLAASH [17,23,31,51,53,63,75,82,106,115,119,143,159,166,175,183,184,189,195,198,199,205], ATCOR [24,62,76,77,80,85,116,164,169,177,190], and the 6S-based atmospheric correction framework [20,57,59,61,137,148,149,184,190] offer greater theoretical robustness but are less frequently used due to their higher data requirements and technical complexity. For non-specialist users, these differences imply that habitat maps generated from similar imagery may not always be directly comparable when different preprocessing approaches are applied. This may affect the consistency of long-term monitoring programmes and limit the transferability of mapping products intended for environmental management and conservation applications.

The relatively limited use of aquatic-specific processors represents another important constraint. Tools such as SeaDAS [73,165], C2RCC [45,74,174], and POLYMER [45] have been specifically developed to address the optical complexity of coastal and shallow-water environments, including high turbidity, adjacency effects, and sunglint contamination. Nevertheless, their adoption remains comparatively low in benthic habitat mapping studies. This underuse may reflect limited familiarity within the benthic ecology community, as well as challenges related to parameterization and validation in optically complex waters. Water column correction also shows marked variability. The empirical depth-invariant approach introduced by Lyzenga remains the most widely applied method [40,41,42,51,54,59,71,109,111,114,117,118,121,122,123,124,126,127,137,149,151,154,157,162,163,164,170,171,176,179,180,194,199,211], largely due to its practicality, low data requirements, and effectiveness in shallow, optically clear environments. However, bio-optical and physics-based inversion approaches, such as those developed by Lee and collaborators [45,78,83,190], remain underutilized despite their stronger theoretical basis and potential transferability across environmental conditions. The omission of water column correction in some studies may be justified when objectives focus on relative classification or when depth variability is limited. Nonetheless, inconsistent application and reporting of these procedures reduce reproducibility and hinder cross-study comparisons.

Surface reflectance distortions caused by sunglint represent another frequently overlooked source of uncertainty. Although empirical correction using near-infrared bands is well established and operationally accessible [18,31,32,59,66,71,85,121,122,123,149,153,161,176,177,179,180,190,194,195,205], its application is inconsistently documented. This omission is particularly relevant in tropical and subtropical environments, where sunglint can obscure benthic spectral signals.

Beyond correction algorithms, the review highlights a major gap in the reporting of preprocessing levels and input products. Among the cases where this information was reported, Level 1 products were more frequently used than atmospherically corrected Level 2 datasets, suggesting that many workflows rely on user-defined correction procedures. Sentinel-2 Level 1C imagery was particularly prominent [42,45,56,71,115,116,117,118,119,120,121,122,123,124,125]. However, for approximately three-quarters of the satellite sensors reviewed, the processing level of the input product was not explicitly stated. This lack of detail complicates the interpretation of methodological differences across studies and hampers efforts to develop standardized workflows.

Validation practices represent another major methodological bottleneck. Accuracy assessment procedures show substantial heterogeneity across the reviewed literature. Differences in study design, data sources, and objectives prevented their recategorization into a standardized framework, limiting quantitative comparison across studies. Examples of validation approaches reported in the literature include independent field surveys based on diving observations, underwater imagery and video transects, in situ sampling, and comparisons with existing habitat maps [44,57,97,123,138,147,161,164,194]. These datasets are commonly used to construct confusion matrices and derive accuracy metrics.The limited availability of reference datasets for validation is frequently associated with logistical and financial constraints, particularly in remote or resource-limited regions. This limitation has direct implications for the credibility and operational uptake of ORS products in conservation and management contexts. Without robust validation, uncertainty estimates remain difficult to interpret, reducing confidence in habitat maps used for decision-making.

Overall, the methodological heterogeneity identified in this review suggests that benthic ORS remains in a transitional phase between exploratory and operational applications. Greater standardization in preprocessing workflows, transparent reporting of input products and correction procedures, and improved validation protocols will be essential to enhance comparability, reproducibility, and long-term monitoring capacity.

4.5. Policy and Conservation Implications

ORS provides a scalable approach for supporting long-term ecological monitoring at regional and national scales, particularly within marine protected areas (MPAs), coastal observatories, and environmental assessment frameworks. In Europe, these applications are closely linked to policy instruments such as the Marine Strategy Framework Directive and Natura 2000, which require spatially explicit information on habitat extent and condition. Similar needs are emerging globally, reinforcing the relevance of satellite-based monitoring for environmental governance.

The widespread use of ORS for mapping and monitoring applications [16,23,24,32,36,39,40,41,42,44,54,56,59,61,62,63,74,84,90,91,101,117,118,137,141,143,146,148,166,171,177,185] reflects its alignment with major conservation initiatives, including the Convention on Biological Diversity, the Sustainable Development Goals (particularly SDG 14), and the Kunming–Montreal Global Biodiversity Framework. In this context, ORS supports the generation of spatial indicators, ecosystem condition assessments, and long-term monitoring of environmental change.

Global conservation targets, including the commitment to protect at least 30% of marine and terrestrial ecosystems by 2030, further emphasize the importance of operational monitoring systems. Satellite-derived habitat mapping, change detection, and multi-temporal analyses increasingly support conservation planning, protected area assessment, and ecosystem resilience evaluation. However, the integration of ORS into active management remains uneven. Objectives related to restoration and management remain comparatively underrepresented in the reviewed literature [19,27,65,70,81,93,99,109,122,142,156,181,182,183], suggesting that ORS applications are still concentrated on baseline mapping and monitoring. Although the growing availability of high-resolution and multi-temporal datasets offers considerable potential for restoration assessment, degradation detection, and ecosystem recovery monitoring, these applications remain inconsistently developed across regions and ecosystem types.

Finally, effective implementation requires standardized and interoperable classification systems. Frameworks such as the European Nature Information System (EUNIS) [212] and initiatives such as the European Marine Observation and Data Network (EMODnet) [213] provide pathways for integrating remote sensing products into operational monitoring and reporting. Strengthening compatibility between technological capabilities, ecological frameworks, and policy requirements will be essential to maximize the contribution of ORS to conservation and sustainable coastal management.

4.6. Future Research Directions and Recommendations

The evolution of ORS in benthic ecology will continue to be driven by technological innovation, expanding sensor availability, and increasing conservation demands. Future research should therefore aim not only to improve technical performance, but also to address the methodological, ecological, and geographic gaps identified in this review.

One of the most promising directions is the wider adoption of hyperspectral imaging [214]. These systems offer exceptional spectral resolution and enable the detection of subtle differences among benthic substrates and biological assemblages, particularly in heterogeneous and optically complex environments such as coral reef–seagrass mosaics, macroalgal forests, and turbid coastal ecosystems [4,215]. Although airborne hyperspectral sensors remain among the most powerful tools for detailed benthic discrimination (Section 4.1), their operational complexity and high cost have historically limited their use to institutions with access to dedicated aircraft and specialized expertise. However, the landscape is changing rapidly with the emergence of new spaceborne and UAV-based platforms. The launch of high-resolution hyperspectral satellites is expected to broaden access to this technology. The recent deployment of Planet Labs’ Tanager-1 satellite, developed through the Carbon Mapper partnership [216,217], represents a major step toward operational spaceborne monitoring. With a spectral range of 400–2500 nm and a spatial resolution of 30 m, Tanager-1 opens new opportunities for global-scale benthic mapping and cross-platform validation [218]. Even more promising for fine-scale applications is the emergence of commercial constellations such as Pixxel’s Firefly [219]. With six operational satellites already in orbit, Firefly delivers the world’s highest-resolution hyperspectral imagery at 5 m ground sampling distance across more than 135 bands (450–900 nm) and a 24-h revisit frequency. This combination of spatial, spectral, and temporal capabilities may enable monitoring of dynamic coastal ecosystems at unprecedented scales. Complementing these commercial hyperspectral initiatives, NASA’s Eagle (Explorer for Artemis Geology Lunar and Earth) mission, formerly developed under the Surface Biology and Geology (SBG) concept, represents an important future hyperspectral Earth observation program. Scheduled for launch later this decade, Eagle is expected to provide high-spectral-resolution observations with open access data products, offering considerable potential for large-scale benthic biodiversity monitoring and cross-platform applications.

At finer spatial scales, hyperspectral technologies are also expanding rapidly across emerging platforms. The growing availability of lightweight, compact sensors deployed on unmanned aerial vehicles (UAVs) is bridging the critical gap between local ecological surveys and regional satellite observations. Instruments such as the Headwall Nano-Hyperspec, Resonon Pika L, Specim FX and the UHD185 developed by Cubert GmbH [220] are increasingly accessible, enabling centimeter mapping of benthic communities. This trend is expected to facilitate multi-scale approaches, allowing researchers to compare and refine satellite-derived products using high-resolution local observations and to monitor small-scale features such as restoration sites, invasive species patches, or fragmented habitats that remain unresolved by spaceborne sensors. However, UAV-based hyperspectral observations should not be considered direct in situ validation, as independent underwater surveys, field observations, imagery, or sampling remain necessary to provide ground-reference information in submerged environments.

In parallel, UAV systems themselves are expected to continue expanding in importance as standalone platforms. Their flexibility, relatively low cost, and ultra-high spatial resolution make them particularly well suited for monitoring shallow and spatially constrained environments. Future research should prioritize the use of UAV-based mapping in poorly studied regions, especially where satellite observations are limited by persistent cloud cover, turbidity, or inadequate spatial resolution.

Beyond optical technologies alone, the integration of ORS with complementary sensing systems represents a critical frontier. Acoustic mapping techniques (e.g., side-scan sonar and multibeam echosounders) [221] and bathymetric LiDAR systems [222] provide structural, geomorphological, and depth information that optical sensors cannot directly retrieve. Multimodal data fusion is particularly valuable in optically complex or deeper environments, as well as in structurally heterogeneous habitats. These approaches can enhance habitat discrimination, improve ecological interpretation, and increase the robustness of classification outputs.

Methodological improvements in preprocessing remain a fundamental priority. Advances in atmospheric, water column, and sunglint correction are essential to reduce optical distortions and ensure reliable benthic mapping. While classification approaches continue to evolve, the robustness of mapping outputs depends on the quality and consistency of preprocessing workflows. In this context, field validation remains indispensable, although the development of scalable remote workflows will be necessary to support regional and global monitoring initiatives. A structured and transparent methodological pipeline, such as the one illustrated in Figure 12, can help guide the sequential decision-making process from study design to final outputs. The workflow begins with defining the study objectives and characterizing the environmental conditions of the study area, which subsequently determine platform and sensor selection. Preprocessing steps, including atmospheric and water column corrections, are then adapted according to habitat characteristics and image properties. Classification and validation stages follow, supported by auxiliary information such as bathymetry, underwater observations, spectral measurements, and reference datasets. Finally, the framework links these methodological decisions with the generation of thematic products intended for monitoring, conservation, and management applications.

Building on these developments, artificial intelligence (AI) and machine learning represent an important emerging frontier in benthic remote sensing [10,223]. These approaches offer strong potential for automated data analysis, anomaly detection, and large-scale pattern recognition, particularly when integrated with multi-source environmental data, historical imagery, and sensor metadata. Although classification algorithms were not systematically assessed in this review, their rapid development is closely linked to these advances and will likely play an increasing role in future ORS applications. Their effective implementation will depend on the availability of high-quality annotated datasets, improved model transparency, and closer integration with ecological knowledge, especially in the context of operational monitoring and adaptive management frameworks.

Beyond these technological and methodological advances, a critical future priority is addressing the pronounced geographic imbalance identified in this review. Research efforts remain concentrated in a limited number of regions, while many biodiversity-rich and vulnerable coastal ecosystems remain understudied. Improving global equity in ORS research will require capacity building, open access data, technology transfer, and sustained international collaboration. Such efforts are essential not only to expand scientific knowledge but also to support conservation and management in regions where monitoring needs are most urgent.

Finally, effective ORS applications depend not only on technological capacity but also on robust ecological understanding of the biological systems being mapped. Future studies should integrate biodiversity baselines, species assemblages, depth distributions, functional traits, and seasonal dynamics. Aligning technological advances with ecological and conservation priorities will ultimately determine the long-term impact of ORS in marine biodiversity science and ecosystem-based management.