Sub-Antarctic Auckland Islands Seaﬂoor Mapping Investigations Using Legacy Data

: This paper demonstrates the richness of data collected for nautical charting and considers ways in which chart data can support scientiﬁc research, through a case study of two modern navigation surveys undertaken in the Auckland Islands. While legacy charts have coarser resolution, and may synthesize different epochs together into one ﬁnal product, we examine how they may be used on their own and to complement more recent hydrographic surveys. We argue that the hydrographic and ancillary data, only a fraction of which appears on the ﬁnal chart, also has scientiﬁc value and that the hydrographic surveying principles applied during data collection are equally relevant for all seabed mapping. While the beneﬁts of full bottom coverage obtained by state of-the-art multibeam surveys are clear, there is much more to be discovered in legacy singlebeam datasets than what is displayed on the nautical chart alone.


Introduction
While everybody who works in the marine environment is likely familiar with nautical charts (charts hereafter), scientific users of these materials may not be familiar with the methods of chart production and thus may be missing important opportunities to bring additional information to their marine research. Information collected for nautical charting is a rich primary data source that is reduced for presentation on published charts. This necessary part of charting obscures information that may aid both scientific analysis of existing data and planning of additional seabed mapping activities. Also obscured are the hydrographic surveying principles that ensure charts can be used as intended, in compliance with national and international regulations.
Nautical charts are specifically designed for safe maritime navigation. They are officially issued by a nation's Hydrographic Office or a similar government organization to meet Safety of Life at Sea (SOLAS) Chapter V requirements [1,2]. Charts show least depth and highlight features of interest to the mariner, such as specific dangers (e.g., rocks), bottom type, aids to navigation, oceanographic information, coastline delineation, terrestrial features, and topographic elevations. The International Hydrographic Organization (IHO) recognize the primary purpose of charts as a navigational tool, but acknowledge that the chart information held by hydrographic offices provides a useful secondary purpose as a source of detailed seabed information that can be used by many others, including coastal zone managers and oceanographers [3]. Charts are created at different scales depending on maritime use, with very large scales in ports and smaller scales along the coast and in the ocean. It is very uncommon for a single epoch of data collection to be represented on individual charts, and it is more likely they represent data compiled and improved over time. There are standard specifications for data gathered for nautical charts. Licensed mariners and most commercial operators are legally required to keep their charts and nautical publications up to date [4].
In contrast to charts, seafloor terrain and habitat maps are created to represent the entirety of the seafloor at the highest resolution possible, are not subject to legal or standardized requirements for data collection, processing, or presentation, and are intended for a variety of uses. Seafloor maps can identify areas characterized by sediment type, show locations suited to particular species or highlight areas requiring protection, demonstrate benefits to areas that are receiving protection, and identify change over time [5][6][7]. As a result, these maps are characterized by much more variability than charts. Knowledge of the epoch represented by map data can be crucial for studies considering oceanographic and geologic processes from tidal cycles to millennia. Similarly, data resolution and measurement uncertainties vary and may affect how the map can or should be used. Derived geomorphic attributes, such as depth, slope, and rugosity, which aid map creation and use [8], are also subject to these limitations. Seafloor map design and scale depends on the scientific objectives of the user [9], for example, large scale descriptive geomorphic modeling [10,11] or higher-resolution benthic analysis [12].
Today, all data used to create a chart is collected and archived digitally. As we look backward in time, different components of a project will have been collected and stored in discrete packages [13]. For example, singlebeam echosounder traces recorded on paper are annotated with handwritten or digitally stamped fix numbers. Fix numbers allowed the user to find the corresponding position, logged in a fieldbook or computer data file. Working with such legacy data today may require scanning the paper records and geolocation of the data, possibly involving interpolation between fixes. Legacy data are not necessarily uniformly sampled. For example, the repetition rate of positioning from sextant and theodolite observations proceeds at a different rate than that of echosounding. Raw ancillary data, such as grab sampling, may only be preserved as written field observations while today the digital record, including photographs, would be archived together with other data.
This case study in the sub-Antarctic Auckland Islands, New Zealand ( Figure 1) includes both recent and legacy hydrographic data collected primarily for charting. Collecting and processing high-quality, high-resolution bathymetric data used to generate marine habitat maps requires expert knowledge and is time consuming, thus it is expensive. The isolation and inhospitable climate of the Auckland Islands exacerbates these issues, but the area is of interest for environment and climate research [14][15][16][17]. Our interest in the data is scientific-seabed morphology in the eastern fjords and shelf records past glacier activity on the islands [18].
Our case study may be particularly useful to researchers who do not have access to state-of-the-art equipment, such as multibeam echosounders (multibeam hereafter), or as a reconnaissance tool for those intending to work in areas that have not been recently surveyed due to factors, such as isolation and associated field costs. It is timely as governments and hydrographic offices worldwide consider the high cost of gathering data for their charts and look for ways to ensure these high-quality data sources are of maximum benefit to secondary users [19].

Background
In New Zealand, charting responsibility is held by Land Information New Zealand (LINZ), who contract the collection of the underlying hydrographic data to commercial surveyors. A risk-based approach is used to prioritize survey operations, which are shared in the long-term national hydrographic survey plan, HYPLAN [1].
Hydrographic data collected in New Zealand waters for the purposes of updating nautical charts is held by LINZ and largely accessible online via the 'LINZ Data Service' (https://data.linz.govt.nz/). Up-to-date georeferenced navigational charts as well as their individual components, such as sounding points and contours, may be downloaded. Metadata includes elements, such as the date and design scale of use, and points users to other documentation, which will allow appropriate and informed use of the data. In the case of a chart, these will provide detailed information on the collection and processing methods used to generate the chart (e.g., hydrographic standards and specifications for nautical charts and publications) as well as how to interpret the symbols and abbreviations used. LINZ direct mariners to NP5011 (INT1) Symbols and Abbreviations used on Admiralty Charts [20]. All data is licensed for reuse under Creative Commons BY 4.0 NZ [21].
Older data is often identified through the shapefile of an area, allowing the user to approach LINZ to request copies of items, such as scanned hand-drawn airsheets containing depths and contours from singlebeam echosounder (singlebeam hereafter) tracks, field reports, and multibeam surfaces. Prior to 1996, the Royal New Zealand Navy (RNZN) had charting responsibility for New Zealand waters. Raw data from that era is not held by LINZ [22].
Users should be aware of the specific attributes of hydrographic charts, which may differ from maps created for other purposes.

1.
Charted depths are referenced to Chart Datum (CD) (approximately Lowest Astronomical Tide (LAT)) while heights are above Mean High Water Springs (MHWS) [23,24]. Maps created for other purposes may use Mean Sea Level (MSL), ellipsoidal heights (when using Global Navigation Satellite System (GNSS)), or a self-generated reference, and, in our experience, published seafloor maps often fail to specify any datum at all. Offsets between datums will be large in areas with large tidal ranges. Depths on products prior to chart publication-such as fairsheets-are usually on a Sounding Datum (SD) derived by the field surveyor. It may or may not be the same as CD [23,24]. 2.
Depths indicated on charts are shoal biased. Processing of depth data preserves shallow areas, with very shallow points designated to ensure they are maintained through any processing, binning, surface, or contour creation [13,25,26].

3.
The precision of depths indicated on charts is standardized. In New Zealand, LINZ requires one decimal place display for depths of 0.1-30.9m (e.g., 5.68 becomes 5.6), and integer values for depths greater than 31m. Drying heights are rounded up to the nearest decimeter [23,24]. Depth data collected for charts may be manually cleaned, filtered, or processed with algorithms, such as the Combined Uncertainty Bathymetric Estimator (CUBE) [23][24][25][26][27].

4.
Because of the legal and risk management implications of charted data, all field measurements are well calibrated with horizontal and vertical uncertainties quantified and field checks undertaken. Hydrographers collecting data for nautical charts will meet IHO s-44 standards as a minimum [28]. In New Zealand, LINZ has refined these in their Contract Specifications for Hydrographic Surveys (HYSPEC) [23,24], which provides a useful guide for users of New Zealand chart data as readers will see every element of data collection laid out.
Datum, processing and calibrations, checks, and uncertainty information are always recorded in chart metadata and reports [13,23,24,26,29], though this may not exist for older legacy data [30]. This is essential information for anyone trying to learn as much from a chart as possible, or synthesize chart data with their own.
While in the field, hydrographers collect ancillary information that will aid the mariner. Examples of this include bottom samples taken to provide information for safe anchoring, tide, and current observations. Here too, chart data is biased toward the concerns of the mariner, for example, bottom samples tend to be collected at potential anchorages. In our case study on the Auckland Islands, the ancillary data includes magnetic anomalies and areas of particularly dense kelp [26,29,31]. Grain size classifications from bottom samples can be used for initial benthic mapping and tide and current information can extend a habitat map to consider the ocean conditions associated with a particular habitat type. Daily logs, weather reports, and final report observations may also be helpful for planning new fieldwork, especially in remote locations, such as our case study.
The Auckland Islands (Maungahuka in the Māori language) are an isolated island group in the Southern Ocean. At around 51 • S 166 • E they are about 400 km south of mainland New Zealand. They are part of the New Zealand Sub-Antarctic Islands United Nations Educational Scientific and Cultural Organization (UNESCO) World Heritage Site [32]. The islands show terrestrial and submarine evidence of Quaternary glaciation [14][15][16][17][18]. Interest in the island group is growing, with increasingly frequent visits by cruise vessels, conservation, fishing, and military patrol ships. Prior to post-2015, nautical charts around the Auckland Islands were compiled from depth information collected between 1840 and 1991, the latter undertaken with singlebeam by the RNZN along the eastern shelf and into some eastern inlets [29,33]. Other inlets had no depth information. In recognition of the growing vessel traffic, conservation risks and other concerns LINZ contracted hydrographic survey firm, iX Survey (now iXblue), to undertake a multibeam survey of the entire eastern side of the island in 2015. This provided near complete bathymetric coverage of an area 660 km 2 (c.f. terrestrial island area of 568 km 2 ) and was the first time some inlets had been charted [26,31].

Materials and Methods
We analyzed and compared legacy 1991 singlebeam (referred to here as pre-2015) and recently-collected multibeam (referred to here as post-2015) bathymetric data, in combination with ancillary chart information in an investigation of the seafloor of the Auckland Islands (Table 1). The 'Chart Data Collection Date' is obtained from the Source Data diagram on the charts. Fairsheets and sheet images are the official bathymetric product created by the hydrographer. They show selected soundings in a grid-like fashion over the survey area, along with contours and points of interest. Until recently, charts were always created from these sheets and spatially uniform depth uncertainty could be reported. Now, the high-density data collected by modern multibeam is characterized by spatially variable measurement uncertainty, and statistically derived surfaces are more commonly used [34].  Multibeam sounders can also collect backscatter data and this may be used for habitat analysis [35,36]. Collection, but no processing, was specified by LINZ in post-2015 and the data currently only exist as raw files [26]. As minimal specifications existed during its collection, this data is likely to face similar issues to those experienced by Dolan et al. [19]. Those authors found large backscatter variations when contracted hydrographers focused solely on bathymetry collection on the Norwegian shelf. Backscatter is not considered here. Starting in 2016, backscatter requirements have been instituted in New Zealand [23].
Reports HI 158 and HS42 were consulted and information that would aid either analysis of the existing survey data, or support those planning to visit the area in the future, were compiled.
The pre-2015 fairsheet 2862-25 was sent with a group of .tiff files covering the entire RNZN survey area, all of which required georeferencing. Processed multibeam data from the post-2015 Auckland Islands survey was provided by LINZ in CARIS HIPS and SIPS (Hydrographic Information Processing System and Sonar Information Processing System) proprietary CSAR (Caris Spatial Archive) format at 1m grid note spacing. This was exported as a raster GeoTIFF.

5.
Visual analysis: Georeferenced charts, multibeam geotiff, and fairsheets were all imported for initial examination of data and display attributes. This included identification of changes between charted epochs, comparison of differences between the chart products and other depth data, and consideration of the vertical datums of each. 6.
Bathymetric analysis: Here, we focus on Norman and Hanfield Inlets, an area of geologic interest [14,18] covered by chart data from both epochs. Depths from fairsheet 2862-25 were digitized manually in ArcMap at the centroid of each value posting. Contours were digitized as polylines following the fairsheet contour. Location errors are estimated to be ±3.5 m and ±2.5 m, respectively, using the line widths and resolution of map elements. Both were processed with the ArcMap Spline with Barriers function at 3 m (the sonar footprint given at a depth of 10 m in Report HI 158) to create a depth raster. Slope (for each cell), ruggedness (Vector Ruggedness Measure (VRM) in neighborhood of 5), and aspect (downslope direction of maximum date of change in each cell) were calculated using the Benthic Terrain Modeler (BTM) v.3 [37] for both epochs. 7.
Ancillary data: Bottom sample and other chart data were digitized from the two versions of NZ2862 with the modern sample records checked against those in Report HS42. Sample characteristics from the charts and reports were classified for rendering as a habitat map using the following: Rock, gravel, pebbles, coral, shells, broken shells, coarse sand, sand, fine sand, and mud, with samples described using multiple classes given an average value. The resulting data were gridded using an Inverse Distance Weighting (IDW) function and a 100 m cell size.

Results
The 1991 and 2015 Reports of Survey on the singlebeam and multibeam surveys, respectively, contain descriptive and technical components of use to scientists. The key elements of use to support analysis or planning for future voyages are described in Table 2. Planning considered this would be extreme, but conditions were generally favorable with only 3 days of 73 lost to weather downtime. Calmer offshore in the mornings.

Bathymetry
Thick "Bull Kelp" growing "on all shoals of 20 m or less" caused echosounder multipath and access difficulties.
Surveyors consulted high-resolution satellite images as part of their field planning. Kelp (Durvillaea Antarctica) " . . . thick and often impenetrable" up to 100 m off coast prevented measurements. Large swell created dangerous conditions beside the coast in exposed areas.

Tides
High water anomaly referred to in NP51 NZ Pilot observed.
Surveyed through kelp at mid-tide when patches were "tow(ed) under".

Other observations
Safety considerations -all parties equipped for 3 days solo in field no matter how short task duration. Suitable landing sites listed.
Many coastline photos. Methods for working in kelp areas discussed.
To compare epochs, we must be certain about the vertical datum used at each stage of the charting process. In our case, we consulted with LINZ and confirmed the SD in post-2015 was equal to CD [38]. The same should be true for the pre-2015 data as indicated in the report ( Table 2). That is, that the pre-2015 control locations are stable and recoverable. Without this information in the post-2015 report we would need some other means to co-reference the datasets, perhaps by selecting several permanent seabed features (e.g., large rocky areas) and holding the depths of these fixed. Additional information would be required in seismically active areas where there may be relative motion over short distances.

Visual Analysis
The most recent version of chart NZ2862 "Plans in the Auckland Islands" is shown in Figure 2. Island is reported to be highly magnetic", while 2017 comments "A local magnetic anomaly is reported to exist in the area indicated on this chart". This is likely related to the volcanic origins of the island group [39] (Figure 2). Interestingly, the post-2015 chart (Figures 2 and 3b) includes far less detail than its predecessor. The pre-2015 fairsheet and post-2015 standard sheet both contain more information than what is displayed on the charts. The shape of the shoreline is more detailed in 2015. Pre-2015 coastline was taken from aerial photo enlargement and checked with optical instruments. Post-2015 coastline was delineated from high-resolution satellite imagery and checked with GNSS observations (Table 2). Numerous kelp beds are shown on the pre-2015 fairsheets (Figure 6d), fewer are shown on the post-2015 standard sheets, and even fewer appear on the charts (Figures 3a and 6d). Instead, kelp bed locations on the chart are more often shown as shallow areas and reef symbols. The digital rendering of depths on the post-2015 sheet image is less orderly than the hand-drafted pre-2015 fairsheet (Figure 3d). The fairsheet from pre-2015 contains information about shoreline features, composition of beaches, and annotated notes while the standard sheet from post-2015 has fewer observational notes.

Bathymetric Analysis
Derived products, such as shaded relief images, and maps of bed slope and other quantities can be generated using the legacy fairsheet data (Figure 3c). The 1991 fairsheet depths, spaced~75 m apart, were interpolated to a 3 m grid for this purpose. After noting the use of singlebeam at the heads of some bays mentioned in the post-2015 survey (Table 2), care was taken to confirm that this did not apply in either Norman or Hanfield Inlets. Thus, the post-2015 shaded relief surface (Figure 3f) does not have any additional equipment varying uncertainties or resolutions that should to be taken into account. The different error characteristics of the pre-and post-2015 data must be examined in their own right before derived products are compared quantitatively, for example, differencing surfaces for change detection; see, for example, [30,40]. Nevertheless, qualitative comparison of the two data sets (Figure 3e,f) reveals some interesting differences-and motivations to return to the legacy fairsheet data.
Finer resolution hydrographic data does not necessarily yield more scientifically useful data in the final nautical chart. For example, the older chart (Figure 3a) indicates more detail on the Norman Inlet sill. The features are relatively smoother on the more recent chart (Figure 3b) and it barely shows remnants of retreat moraines along the shorelines, inland of the main sill. The sills are evidence of past glacier extent and these features likely impounded terrestrial lakes when the sea level was lower than at present [15,18]. Consulting only the more recent chart when planning new fieldwork or inferring details of past glaciation could thus be misleading.
Resolution  (Figure 3f) is 54.7 m. There has been little change in the underlying data, but the chart has changed considerably.
All the concerns discussed above are apparent in the difference map created by subtracting the surface created by digitized pre-2015 fairsheet from the post-2015 surface (Figure 4). A grid-like pattern of alternating positive and negative differences in flat areas is due to different grid spacings of two datasets. Echo sounding records the shoalest point within the beam footprint. Multibeam sounding footprints are smaller than singlebeam, thus they more precisely represent the seabed within each footprint along the swath. With the use of GNSS positioning, the horizontal location of this shoalest point is similarly more precise [13]. This effect is evident in the negative differences along the fiord walls ( Figure 4a). The post-2015 multibeam survey better represents their steep slopes than does the pre-2015 survey. The positive differences along the coastline are due to the wider multibeam swath, which measures much further beyond the nadir region of the vessel. A blunder involving vertical datum would generate a systematic offset between the two datasets. Following our suggestion above, we compared depths on two prominent seabed features (blue circles in Figures 3e and 4), rocky areas that we assume to be persistent across epochs. We focused on the difference in the shoalest depth on each feature because this is a specific focus of data collection for charting. Rather than an offset, we found +1.2 m and −1.1 m differences (Figure 4c,d ). The legacy data are too coarse to allow the distributions of the differences to be compared. It is also important to recognize that due to equipment limitations, the pre-2015 survey may not have actually found the shoalest point despite careful investigation of the area. Examining larger or more features might clarify the source of the offset, but our interpretation is that it is due to better representation of fiord wall slopes in the post-2015 survey.
Morphologic attributes of the sea floor are more well-resolved by the higher resolution multibeam data, but this may be more important for some attributes than others ( Figure 5). In our case study, the slope result is similar for both (Figure 5a,b), while some details of aspect and ruggedness appear to be missing in the pre-2015 data set. The overall sense of relatively rougher and smoother areas is the same in both data sets, while the higher-resolution multibeam surveys captured more detail than the older singlebeam surveys.

Ancillary Information
Bottom sample classifications were interpolated to make sediment type maps for easy visualization ( Figure 6). The gradients do not represent transitions in sediment type, rather, changes in the color maps represent changes in knowledge of the sediment type. Our post-2015 map incorporates both preand post-2015 data. The main difference between the two is additional spatial detail and an improved knowledge of the seabed materials in Carnley Harbor and offshore, thanks to samples collected in 2015. In most cases, the additional samples confirm the earlier classifications. This is encouraging as common practice is for sediment size to be determined by visual inspection only and thus be somewhat subjective. The heads of all inlets surveyed are mud or fine sand and the sills where samples have been collected tend to be characterized by coarser sediments. The "R", rocky seabed, identification may represent either a seabed sample or denote the location of a shoal feature where the mariner is advised to take care.
It should not be assumed that the more recent chart represents the more complete or correct ancillary data. While samples add spatial detail in Norman and Hanfield Inlets (Figure 5d,e), some ancillary information about rocks ("R" in Figure 3a,c ), shoreline conditions, and kelp beds (light green diamonds in Figure 6d) has been removed from the post-2015 charts. This change could represent improved knowledge, a change in cartographic standards, or a change in the physical environment. In one example, an "R" near the head of Hanfield Inlet shown in the pre-2015 chart and fairsheet is not present in the post-2015 versions. Considering the physical setting-proximity to steep cliffs and a waterfall (see also Figure 6 in [18])-rockfall debris is likely to accumulate along the shoreline. We conclude that omitting the ancillary information is a cartographic choice. All available charts and ancillary data should be considered when using and interpreting both new and legacy data sets.

Discussion
In some respects, the bathymetric elements of this case study are a simple demonstration of the ability of modern equipment to improve data capture. However, we also demonstrate that older legacy data is useful, especially when paired with the other ancillary information available on charts. The closing paragraph of the iX Survey report expresses the hope of the hydrographer that the new charts will enable future research and conservation activities to happen safely in the Auckland Islands area [26].
The use of existing data during reconnaissance is expected. Indeed, legacy data supports campaign length calculations in terms of swath coverage, and allows speed/coverage/density tradeoffs to be considered as realistically as possible. We suggest here that other uses are also valuable. While legacy data products may appear at first glance to be sparser than more modern data, when examined comprehensively (such as on a fieldsheet) they may provide a more detailed view of the seabed than anticipated.
Care must be taken to understand the limitations inherent in nautical charts. Comparison between charted data and hydrographic fieldsheets or surfaces in Figure 3 demonstrates the wealth of information not presented to the target audience of nautical charts (the mariner concerned with safe navigation). A nautical chart has an uneven spacing of contours-typically at 5, 10, 20, 30, 50, and 100 m-such that features of interest to geomorphic and habitat studies and campaign planning may be obscured. The reduction in chart data discussed in Section 4.1 seen in the number of soundings labelled, terrestrial contours, and smoothing of the bathymetric contours serves a purpose-the nautical cartographer must be confident that what is displayed is all that is needed by the mariner. In the interest of making our case, we have given several examples of interesting features that might not be recognized using only the most recent nautical charts in our study area. However, other features would not have been anticipated at all, for example, the shoaling and partial sills identified in Deep, Chambres, Granger, Waterfall Inlets, and Carnley Harbour using the post-2015 surface [18].
Derived bathymetric products all reflect the resolution of the underlying data, although we found that general characteristics were similar across the pre-and post-2015 data sets. Depending on the intended application of derived products, high-cost, higher resolution data may not substantially change the result of any given analysis. From the point of view of planning, areas characterized by shorter or longer wavelength variability and thus of interest for particular types of field campaigns would be similarly identified using either data set.
The analysis of seabed sediment type is useful, but limited by several factors. Firstly, the samples are described qualitatively by (well-meaning) hydrographers. In our case study, "gravel" was identified several times in the pre-2015 survey, while the term "pebble" was used in the post-2015 survey. Both terms can be used for grain sizes ranging from 4 to 64 mm, but such details are not explained in the survey reports. Survey specifications requiring photographs of the seabed, sieving, sample retention, or coding aligned with the SEDCLASS, the U.S. Geological Survey ArcMap Sediment Classification Tool [41], would address this issue. Additionally, the locations of seabed samples are biased toward anchorages in our pre-2015 case [29], unless a specific request has been made, for example, along the perimeter of the survey area, in our post-2015 example [26]. There is neither the sampling density nor a protocol that might be expected for a full geological mapping survey.
Other information on a chart may support geological and biological investigations. For example, kelp beds (Figure 6d) only grow on rocky or stony substrates. This information may go unrecognized if only the nautical chart is consulted. By utilizing the fairsheet data along with plotting other rock symbols (submerged, awash, drying), users of legacy chart data (and potentially other products, such as high-resolution satellite images) may find that much of the hard substrate in the survey area can be said to be 'known' to some extent, so efforts can focus in the softer, smoother sediment. As multispectral backscatter research indicates it is of most use in these finer sediment areas [42], this initial reconnaissance can be used as a quick tool to focus modern technologies to places they will be most effective.
Should chart data be synthesized in our seafloor maps? We have demonstrated some benefits of this additional data, but it is important to consult the hydrographic survey reports and consider the challenges we have also identified, including shoal bias, truncated precision, and clarity regarding SD or CD referenced depths. In [18], we used tidal and geodetic observations to ascertain the CD to MSL offset for the area to link our map with terrestrial observations. Once datum issues are resolved it is very useful that the data are collected to recognized standards, calibrated thoroughly, and the means to analyze the measurement uncertainty exists. Previous calculations of the vertical uncertainty associated with the data in the Auckland Islands puts the post-2015 multibeam depths at ±0.4 m to ±1.6 m @ 95% Cl (using MB-1 specs [18,24]) and the older data at ±0.5 m to ±3.3 m at 95% Cl (assuming IHO Order 1 inshore and Order 3 offshore) [18,28]. Calibrations and uncertainty considerations are the strength of hydrographic measurements, and while older equipment may have meant researchers undertaking seafloor mapping fieldwork did not see the effects of poorly calibrated data, this is no longer the case with the high-precision rapid data collecting gear used today. Lecours et al. and Lucieer et al. [43][44][45] demonstrate the effect of motion sensor calibration issues on data derivatives and advocate for more consideration of this by researchersusing digital terrains in seafloor mapping. Synthesis of hydrographic calibration, checks, metadata, and reporting methods from nautical charting into general seafloor mapping would be of great benefit to the marine scientific community today and for users of our data in the future. This is recognized in [46]. Quantitative change detection via differencing datasets from different epochs is a primary motivation of work, like ours [30,42,47]. In this case study, differences appear to be associated with methods of data collection and processing rather than environmental change. The mean difference we calculated, -0.6m (Figure 4b) lies within the range of total propagated vertical uncertainty, ±0.6 m to ±3.7 m at 95% Cl. Another approach to detect change and investigate measurement uncertainties would be to down-sample the raw post-2015 multibeam data at the locations of the pre-2015 singlebeam depth points and compare the two. This approach could be particularly useful with more widely spaced legacy data.
The samples on the chart could be integrated into seafloor mapping classifications. Interpretation challenges may arise when details are removed in a chart update without explanation. In our case study, does the removal of the "R" indicating a rocky seabed mean the annotation on the pre-2015 chart was incorrect, or that rock that was there in pre-2015 is no longer present in post-2015, indicating change over time in this environment? In our experience conducting nautical charting surveys, one rule-of-thumb was to classify a site with three or more unsuccessful grabs as "R". Thus, even if the designation is removed on cartographic grounds (such as relevance to the intended user), there may be other reasons to exercise caution. In the case study presented here, we examined the surrounding terrestrial features and concluded that rockfall was likely in at least one location where an "R" had been removed. If such seabed features were important to our study, we would have a clear reason to visit such sites, but only because we compared the new and old charts and sheets.
We recognize that in New Zealand, we benefit from an all-of-government open data environment that enabled us to free and open access of the data used in this case study. Other studies incorporating legacy data have obtained it from, for example, British Admiralty Charts [40], United Stations National Ocean Service (NOS) and the National Oceanic and Atmospheric Administration (NOAA) [30,48], and the Hydrographic Service of the Royal Netherlands Navy [47]. While we did not have access to the raw data from 1991 through LINZ, researchers in other countries may have access to scans of archived analog recordings, for example, the NOAA Marine Geophysical Trackline Data documented by the National Geophysical Data Centre [48,49]. An additional benefit of access to analogue records is that signal intensity could be analyzed for applications, like acoustic ground discrimination for habitat classification, and could be applied to the legacy data. [7].

Conclusions
This short case study reveals the important primary and ancillary data that exist behind the nautical charts commonly consulted for seabed mapping voyage planning or during data analysis. Charts and ancillary data from surveys in pre-2015 and post-2015 are shown to provide useful seafloor information in the Auckland Islands. The addition of the primary data (fairsheets and surfaces) greatly improves the bathymetric analyses able to be undertaken. Improvements can be made to seabed sampling specifications for nautical charting to support a wider variety of users and those users would benefit from understanding the existing charting specifications so that they can correctly use existing data, or indeed ensure their data is useful to others in the future.