Coastal Mapping. Detection of coastal morphology, tidal flats, and shallow water bathymetry using high-resolution satellite imagery covering sites in West and North East Greenland. Technical Report No. 112

1 Mapping the Coastal Zone in North East Greenland Detection of coastal morphology, tidal flats, and shallow water bathymetry using high-resolution satellite imagery covering sites in West and North East Greenland Mikkel Høegh Bojesen, Rasmus Eskerod Borgstrøm, Kirsty Langley, Eva Mätzler, Alexandra Messerli, Mikkel Lydholm Rasmussen, and Karl Brix Zinglersen Technical Report no. 112 Greenland Institute of Natural Resources In collaboration with Asiaq Greenland Survey and DHI GRAS A/S 2 Title: Coastal Mapping. Detection of coastal morphology, tidal flats, and shallow water bathymetry using high-resolution satellite imagery covering sites in West and North East Greenland Authors: Mikkel Høegh Bojesen, Rasmus Eskerod Borgstrøm, Kirsty Langley, Eva Mätzler, Alexandra Messerli, Mikkel Lydholm Rasmussen, and Karl Brix Zinglersen Publisher: Greenland Institute of Natural Resources Final version: 29-05-2019 Date of publication: Financial support: This project is carried out under a cooperation agreement between the Ministry of Mineral Resources & Labour and Asiaq; including a sub cooperation agreement between Asiaq, Greenland Institute of Natural Resources and DHI GRAS. Project: Project 0043: Mapping the Coastal Zone in NE Greenland ISSN: 1397-3657 ISBN: 87-91214-90-4 EAN: 9788791214905 Web: https://natur.gl/ Citation: Zinglersen KB, Garde E, Langley K, & Mätzler E (2020) RemoteID. Identification of Atlantic Walrus at haul out sites in Greenland using high-resolution satellite images. Technical report no. 112, Greenland Institute of Natural Resources, Greenland. ISBN 3 English Summary 4 Greenlandic summary – Kalaalissut Naalisarneqarnera 4 Danish summary – sammenfatning på dansk 4 1. Introduction and Background 5 1.1 Satellite based analysis 5 1.2 Field Sites 5 1.3 Fieldwork activities 7 2. Data and Methods 9 2.1 Satellite derived bathymetry 9 2.2 Intertidal zones 10 2.3 Coastal classification 11 2.4 Field survey methods 13 3. Results 14 3.1 Satellite derived bathymetry 14 3.2 Intertidal zones 16 3.3 Coastal classification 18 3.4 Validation 21 3.5 Findings and lessons learned 22 4. Conclusions & Perspectives 24 Acknowledgements 25 References 26 Appendix 1: Organization 28 Appendix 2 Data Distribution 29 Appendix 3: Decision tree for final coastal classification 30 4 English Summary This report highlights the findings of the mapping of the coastal environment on two locations in Greenland based on analysis of satellite imagery. The study includes satellite derived bathymetry, mapping of the intertidal zones, and classification of the coastal areas, which we could map successfully. The analysis involved algorithms applied to highand medium-resolution satellite imagery, validation with field measurements, and morphological analysis from coastline data. The study successfully demonstrated the feasibility of using satellite-based analysis for remote areas, but also highlighted the logistical complexities of field work conducted there. Nevertheless, the technologies hold high potential for efficient large-scale mapping in the Arctic. Greenlandic summary – Kalaalissut Naalisarneqarnera Uani nalunaarusiami Kalaallit Nunaanni sumiiffinni marlunni sinerissap qanittuani immap assiliornerit inerneri ersersinniarpagut, taakkua satellit’tikkut assilisat atorlugit nalilersorneqarneri aallaavigalugit suliaapput. Misissuinermi pineqarput immami ikkattumi immap itissusaa satellitsikkut paasissutissat aallaavigalugit suliaq, immap tinittarnerata ulittarneratalu nalunaarsorneqarneri aammalu sinerissami sumiiffiit immikkoortiterneri, iluatsilluartumik nalunaarusiorneqarsinnaasimapput. Matematikkimik suliassat naammassisarnissaannut maligassat atorlugit satellitsikkut assilisat pitsaalluinnartut naammaginartumillu pitsaassuseqartumik assilisat atorneqarput, naatsorsuinerit sumiiffinni atorsinnaassusii pineqartullu ilusai isikkuilu ineriartornerminnilu ullumimut isikkorilersimasaat misissuiffigineqarsimapput. Misissuinerit takussutissaqartilerpaat sumiiffinni tikikkuminaatsuni ungasissumiittunilu nalunaarsuissagaanni satellit’tit atorneqarsinnaalluartut, aammalu sumiiffinnut ornigulluni suliniaraanni qanoq ataqatigiissaarneqarnissaasa pisariutiginissaannut takussutissaqartitsilerpoq. Taamaakkaluartoq Issittumi nalunaarsuinissamut teknologiikkut periarfissaqarluartoq ersersinneqarpoq. Danish summary – sammenfatning på dansk I denne rapport fremhæver vi resultaterne af kortlægningen af kystmiljøet to steder i Grønland baseret på analyse af satellitbilleder. Undersøgelsen indbefatter havdybder på lavt vand beregnet fra satellitdata, kortlægning af tidevandszonen og klassificering af kystområder, som kunne kortlægges med success. Analysen bestod i algoritmer anvendt på satellitbilleder af høj og medium opløsning, validering fra målinger i felten og morfologisk analyse fra kystlinjens forløb. Undersøgelsen fremviste mulighederne for at bruge satellitbaseret analyse i fjerntliggende områder, men fremhævede også de logistiske kompleksiteter af feltarbejdet på stedet. Ikke desto mindre har teknologierne et stort potentiale for effektiv kortlægning i stor skala i Arktis. 5 1. Introduction and Background Motivation The coastline of Greenland is long and complex, both in its form and in composition. Since the predominant part of the country consists of wild nature without roads, the coastal areas are the main transportation routes. There are a multitude of skerries, archipelagos, islands and islets, which are uncharted, and these present a significant hazard and risk for accidents. The lack of navigational charts and accurate topographic maps are a particular problem in the remote areas of Greenland, such as North East Greenland. With hydrocarbon exploration expanding ever further north it is essential to fill-in crucial missing information on the coastal environment. The recent advances in satellite remote sensing and image analysis have increased the potential to obtain detailed insights, and updated information about vulnerable and remote coastal environments in the Arctic. Freely available Sentinel satellite imagery from the European Space Agency (ESA) Copernicus programme, supplemented with advanced commercial satellite imagery represents an obvious data source for updating these maps. In this report we highlight the findings of the coastal mapping projects based on satellite imagery. It includes satellite derived bathymetry, mapping of the intertidal zones, and classification of the coastal areas, which we could map successfully in two regions of Greenland. 1.1 Satellite based analysis In this study, we have demonstrated how satellite imagery can be used for mapping the coastal zones in the remote area of North East Greenland. Additionally, we developed and applied methods for deriving shallow water bathymetry, intertidal zones, and classifying different coastal types. The analysis was carried out using 1) high resolution, commercial satellite imagery of selected areas, and 2) freely available medium resolution satellite data, which has a larger spatial coverage and therefore allowed for an upscaling of the area of interest. Field measurements were carried out in order to validate the satellite-derived products. This project served as a feasibility study that successfully demonstrated the use of satellite-based methods for mapping the coastal characteristics of the study area. The challenges encountered during the project (such as sea ice coverage, turbidity, and small tidal level variability) have helped us to define the limits of the methods, thereby reducing the risks for future applications. These positive results open up the possibility for applying the newly developed methods to other coastal areas in Greenland, and thus holds an enormous potential for a cost-efficient large-scale mapping of the coastal zone in Arctic. 1.2 Field Sites Two areas of interest (AOI) - one in West Greenland and another in North East Greenland (Figure 1.2.1) were chosen for this pilot study. 6 Figure 1.2.1. Map of Greenland indicating the two AOIs used in this study, Nipisat in West Greenland and Vega Sund in North-East Greenland. West Greenland AOI – Nipisat: Nipisat is a shallow-water inlet located close to the capital of Nuuk (Figure, 1.2.2). The area is an inlet of Nuup Kangerlua (Nuuk Fjord), which has a depth range from 0.5m to 20m and a tidal range of 3-4m. This AOI was used to establish the methodological framework, data management and analysis procedures for the study. Additionally, the depth measurements acquired in Nupisat were used to calibrate and validate the satellite-derived bathymetry. The proximity of the AOI to Nuuk allowed for effective and safe collection of ground-based validation data used during the development phase of the project. Nipisat Sund was visited using a boat provided by the Greenland Institute of Natural Resources. Figure 1.2.2. Field site Nipisat Sund, a shallow-water inlet close to Nuuk North East Greenland AOI – Vega Sund: Within North East Greenland, we have focused on especially vulnerable coastal areas, for example breeding grounds of bird colonies, areas with interesting seaweed deposits or archaeological sites that are threatened due to coastal erosion. These AOI’s were selected with input from biologists and archaeologists (Fig. 1.2.3). High-resolution satellite imagery of the AOI’s were examined in conjunction with near-shore bathymetry datasets. The Vega Sund area was highlighted as the most appropriate location due to its vulnerable and valuable seaweed deposits, which would be at risk in case of an oil spill incident. Vega Sund is a shallow area between Foster Bugt bay and the inner fjords, located north of the town of Ittoqqortoormiit (Scoresbysund). Archive satellite images show that the area has been free or partially free of sea ice cover during the last years from beginning of august, deeming it a suitable AOI for satellite-derived bathymetry analysis. 7 Figure 1.2.3. The three field sites of North East Greenland Travelling and logistics for field surveys in North East Greenland are challenging, requiring expensive charters for transport and collaboration with other operators to minimize costs and enhance safety. The survey team collaborated intensely with NANOK, The North-East Greenland Company, a non-profit organization restoring cultural heritage cabins in the National Park. Together we shared logistical platforms: charter of airplanes, shared cabin accommodations, and as passengers on their boat travelling towards the study area. The locally based military units of the Mestersvig Station and the Ella Ø Station (Sirius Dog Sled Patrol) of the Joint Arctic Command, the Danish Defence, provided logistical support of driving, accommodation and boat transport. Such collaboration and kind assistance with existing logistical resources in the area highly minimized the costs and supported the operations but cannot be taken for granted in future cases. 1.3 Fieldwork activities A workshop with all project participants was held in Nuuk during Spring 2018. The goal of the workshop was to prepare field activities, and to test the preliminary near-shore bathymetry analysis. The workshop included collection of ground truthing data in a shallow tidal zone close to Nuuk. Fieldwork was conducted in North East Greenland for four weeks in July-August 2018, planned to coincide with minimal snow cover and low sea ice extent. However, abnormally high snow cover and a short melt season created unfavourable conditions for both aerial and boat logistics. Thus, as a result of the abundant sea ice cover in Vega Sund, access to Vega Sund was limited and the majority of the fieldwork was instead carried out at Ella Ø and Mestersvig. The Ella Ø Station harbour was accessed via a nearby landing strip on Trail Ø at Holms Bugt with support from the civilian and military operators. From here, Ella Ø was accessed by boat with 8 assistance from the Sirius Dog Sled Patrol. A fieldwork activity programme was subsequently established, which included setting up a GNSS base station for continuous positioning and installing two tidal gauge stations for continual tidal baseline measurements. The team covered the near shore water with an inflatable boat for depth mapping with a Garmin echo sounder, positioned using the GNSS base station. Exposed rocks and skerries were mapped, and physical depth measurements were conducted using a lead weight. On land, the team documented the coastal morphology, geological properties and vegetation cover with geolocated photos. The groundtruthing data collected here was used to validate the shallow bathymetry and coastal zones. The coastal waters around Ella Ø are deeper than those in the original AOI at Vega Sund. This resulted in a more limited near-shore bathymetry survey and tidal zone classification than intended. Despite the challenging logistical conditions, this did not have a negative impact on the overall outcome of the project. A short trip to Vega Sund was possible at the end of the field campaign thanks to logistical support from the Sirius Dog Sled Patrol. The AOI could be visited for 3-4 hours, within which echo sounds measurements and photography of the coastline were completed. 9 2. Data and Methods 2.1 Satellite derived bathymetry Advances in image analysis, mean it is now possible to estimate the depths of clear, shallow waters using optical satellite imagery. The advantage of satellite-derived bathymetry (SDB) are that it is cost effective, has a large spatial coverage of remote and dangerous regions, at a high temporal resolution. The main disadvantage of the method is that it relies on the use of optical satellite data, which has a limited water depth penetration (around one secchi depth). The SDB method is most effective in clear waters, as the penetration depth decreases in relation to increased water turbidity, low light, and the presence of ice. The SDB’s in this study were produced using Sentinel-2A & WorldView-2 satellite imagery. The high spatial resolution of Sentinel-2 and WorldView-2 imagery at 10m and 2m respectively, allows for detailed bathymetric mapping. This is of a higher spatial resolution than existing bathymetric datasets. Theoretical framework for SDB analysis There are a number of approaches that can be used to derive SDB. The background of some key methods are briefly outlined below. For this project, a DHI GRAS proprietary physical radiative transfer model (extended version of Guzinski et al. 2016; Klonowski et al, 2007; Lee et al. 1998, 1999, 2001) was applied. Other methods of note used for deriving SDB are empirical models, based on the early work of Lyzenga, 1978, refined by Stumpf, 2003. Additionally, radiative transfer-based models, such as the SAMBUCA model developed by CSIRO (Brando & Dekker, 2003; Wettle & Brando, 2006) can also be applied. The fundamental difference between the empirical models, and the radiative transfer models, is that the empirical models calculate an index that correlates with depths, but has to be transformed into water depths, while the radiative transfer methods model the water column as primarily a function of the water depth and the seabed reflectance. The model used in this study minimizes the differences between an observed satellite image and a modelled satellite image, which is created as a function of six parameters – depth, bottom type, backscattering, chlorophyll-A, gelbstoff, and the slope of the backscattering function. Through minimizing the difference between the observed and the modelled satellite images, accurate water depths can be retrieved in optically shallow waters, meaning light reflecting from the seabed is observed. Essentially, the SDB can be reduced to the following series of pseudo equations: 𝑟𝑟𝑠𝑚𝑜𝑑𝑒𝑙𝑙𝑒𝑑 = 𝑓(𝑐ℎ𝑙𝑜𝑟𝑜𝑝ℎ𝑦𝑙𝑙, 𝑔𝑒𝑙𝑏𝑠𝑡𝑜𝑓𝑓, 𝑏𝑏, 𝑏𝑥, 𝜌, 𝐻) Where 𝑟𝑟𝑠𝑚𝑜𝑑𝑒𝑙𝑙𝑒𝑑 is the modelled satellite image, 𝑏𝑏 is the backscattering, 𝑏𝑥 is the slope of the backscattering function, 𝜌 is the summed up bottom reflectance, and 𝐻 is the depth. The next step is to minimize the difference between the modelled and observed satellite images: 𝜒2=1 𝑁 [∑(𝑟𝑟𝑠𝑜𝑏𝑠𝑒𝑟𝑣𝑒𝑑 − 𝑟𝑟𝑠𝑚𝑜𝑑𝑒𝑙𝑙𝑒𝑑)] With N being the number of spectral bands, where the satellite image provides information, and 𝑟𝑟𝑠𝑜𝑏𝑠𝑒𝑟𝑣𝑒𝑑 is the observed satellite image. 16 The correlation between the survey and the satellite data in Figure 3.1.3 is high in areas where the water is clear. However, a large part of the water in the satellite image has very high sediment loads, which obscures the seabed and makes SDB impossible. As a result, the areas affected by these sediments have been removed from the dataset. The sediment cover obscured the overlap between the westernmost satellite image and the in-situ data, and therefore the SDB was not extended to the eastern reefs at the tip of Geographical Society Island. 3.2 Intertidal zones The result of the intertidal zone workflow is a raster showing land (pixel location always classified as land), water (pixel location always classified as water), and areas that have sometimes been classified as land and sometimes as water (Figure 3.2.1). These latter regions, when located along the shoreline, are interpreted to be intertidal zones. A more precise delineation of the intertidal area can be obtained by defining land and water thresholds based on low and high tide scenes. Figure 3.2.1. Vega Sund region. Upper row: Sentinel-2 scenes showing water coverage during low and high tide. Lower left: for each pixel location, the number of pixels included in the mosaic. Lower right: resulting mosaiced raster indicating areas of land, water and intertidal zones. The NE Greenland sites of Vega Sund and Ella Ø have a relatively small tidal reach of about 1m. The Vega Sund area has a number of river deltas that have created shallow deltas along the coastline and which result in narrow intertidal zones at the delta mouths, and in some inlets (Figure 3.2.1). 17 Around the island of Ella Ø, the steep sided deep fjords resulted in only a few intertidal zones being identified with the satellite imagery (Figure 3.2.2). It was clear during the fieldwork that there are narrow beaches and multiple skerries on the northern central coast of Ella Ø, however the pixel resolution of 10m of the satellite imagery likely hindered the detection of these. Figure 3.2.2 Ella Ø site. Left: Mosaiced raster indicating areas of land, water and intertidal zones. Note that sea ice can be misclassified as “land”, which can be seen in the central part of the figure. Right: Satellite image of Ella Ø around the time of fieldwork, note the patches of broken-up sea ice visible in the right section of the image (Satellite image: SnapPlanet). In contrast to NE Greenland, the test area in Nipisat Sund, Nuuk Fjord, has a much larger tidal range (3-4m). The shallow, gently sloping sound has a larger intertidal zone and numerous small islands that are visible (Figure 3.2.3). Figure 3.2.3 Nipisat Sund, Nuuk Fjord. Right: Mosaiced raster indicating areas of land, water and intertidal zones. Left: A satellite image, of Nipisat Sund around the time of fieldwork (Satellite image: SnapPlanet). 18 3.3 Coastal classification Complexity index To explore the spatial distribution of the complexity of the coastline, a multivariate grouping analysis, and a cluster and outlier analysis was carried out. The grouping analysis showed that 4 groups best represented the variation in the data, but also that the variation was considerable within each of these four groups. This made it impossible to distinguish the four groups statistically at the varying distances (0,1km; 1km and 10km). In the cluster and outlier analysis we found that the spatial autocorrelation was significant, meaning that the areas of high and low complexity are significantly clustered. We used the Anselin Local Moran’s I method to further investigate the location of the clustering. In cluster and outlier analysis one very important aspect is to define neighbouring points, since these will be used in the statistical analysis comparing neighbouring observations. For this analysis, neighbourhood distances were calculated to be 2750m using Global Moran’s I. The resulting Figure 3.3.1, b, shows clusters and outliers. High value clusters means that high values are surrounded by other high values, relative to each other (pink dots) and low value outliers, where low values are surrounded by high values (dark blue dots). Figure 3.3.1 a & b. Complexity Index measurements (a) and cluster and outlier analysis (b) in Vega Sund. In effect Figure 3.3.1, a, shows the absolute complexity at a defined scale (1km) whereas Figure 3.3.1, b, shows the complexity of each observation relative to its neighbours. In combination these two figures allow us to compare the significance of complexity in the case oil spill emergency response planning, at a regional level. a b 19 The complexity index analysis is scale dependant, and therefore the scale selection for the analysis depends on the level of detail needed. For example, in the case of an oil spill emergency response, information on the size and spread of the oil spill would be important factors in selecting the correct scale. It is equally important to determine how much variation exists along the coastline in order to set an appropriate scale for analysis. Archipelagos As seen below in Figure 3.3.2, large parts of Vega Sund are regarded as an archipelago. This classification is not excluded, since important information about coastline shape, i.e. complexity index, and coastline content, i.e. surface type information from the supervised classification, can also feed into the oil spill information system. Figure 3.3.2. Archipelagos defined as group of islands where the ratio of the area of the water to the area of the land is as high as 9:1. 20 Supervised classification Ella Ø has a complex coastline with a variety of coastal types. As seen in Figure 3.3.3 a rocky (and steep) coast dominates the southern side of Ella Ø, whereas much more diverse coastal surfaces are found on the western side of the island. Here a combination of sand, fluvial flood plains, unconsolidated rock and other geology/ vegetation are found. This results in a complex and diverse environment, which is very difficult to clean up in an oil spill response scenario. Figure 3.3.3. Final classification of coastal composition at Ella Ø. 21 3.4 Validation Validation of the supervised classification is important in order to assess the accuracy, plausibility and robustness of the classification. In order to assess the classification, verification photos were taken at the field site, as described in section 2.4. Three example validation photos are shown in Figure 3.4.1. Figure 3.4.1. Validation of classification, using verification photos. A, shows semi vegetated area, B shows an example of a rocky coast and C, shows an unconsolidated rock beach. Photo A in Figure 3.4.1 shows a vegetated area with small areas of clear standing water. According to the classification this location is classified as water. Despite also being vegetated this area is clearly influenced by water. Photo B shows how the rocks directly meet the sea, forming a steep rocky coast, which is also depicted on the classified map. One challenge when making a supervised classification is that even within each 10m pixel more surface classes are represent. In effect each pixel on the map depicts which of the possible classes is the most abundant within that area of 100m2. Photo C is taken in an area characterised by a loose gravel surface with a certain degree of moisture due to the vegetation. The area is classified as fluvial flood plain, sand, rock and snow. This loose structure of the surface could also be termed as unconsolidated rock. This example demonstrates another challenge with a supervised classification. One has to define classes for classification detailed enough to represent the degree of variation in the environment in question, and at the same time define classes broad enough to end up with an applicable classification for managerial purposes. Thus, the question is how to define robust and applicable class boundaries. 22 Despite the uncertainties discussed above, the classification generally fits well with the validation photos and a visual inspection of the VHR imagery. For a further quantitative validation to be made, a reference dataset is needed. Unfortunately, such a reference dataset is not available for this study. A more widespread validation campaign was planned in Vega Sund, however, due to poor weather conditions, only a small fraction of this was carried out. In future studies the planning of validation sampling should take into account sampling area size and spatial distribution relative to the size of the AOI since this will influence the strength of validation data, and overall end product. 3.5 Findings and lessons learned Many valuable experiences have been gained throughout the entire project; from the ground truthing fieldwork, to satellite data acquisition and data analysis. We hope that these experiences will enable us to adapt our methods for future projects. The main findings can be listed as follows: Product Nipisa, W Greenland NE Greenland satellite derived near-shore bathymetry map up to 8 m depth X X intertidal zone map X X coastal complexity X eight coastal surface classes X robust validation setup (see report) X X Details of some of the challenges encountered in different areas of the project are outlined below. Challenges in classification In the classification of the coastal zone we paid attention to a variety of surface material classes. In doing so we encountered some challenges that can be accounted for and avoided in future work. The surfaces we mapped are to some degree influenced by water and other biophysical factors. Since we are working with optical data, the spectral signal variation will vary across the study area. One issue of concern is whether the training areas and validation areas are representative for the variation across the study area, since this will influence the accuracy of the final classification. Challenges with SDB In the production of SDB for the study area, the challenges relating to the complicated geography of Greenland became apparent. The high sediment load from melting glaciers obscures the seabed in many areas. This made it very difficult to derive SDB in these regions, as well as limiting the potential extent of the Greenlandic coastline that can be mapped with this method. Another challenge tied into the complex geography is the rapid changes in water depth, caused by the rocky coastline in parts of Greenland. This was highlighted at the Ella Ø study site. In those areas, the 23 10m spatial resolution of Sentinel-2 data may not be sufficient to accurately represent the bathymetry, and therefore a large scatter is likely to be observed in validation dataset. Challenges with field work operations In order to assess and validate the satellite data for bathymetric mapping, intertidal area and terrain classification, in-situ data collection is still currently preferred. The fieldwork carried out in North East Greenland was both expensive and time-consuming. Like all fieldwork in Greenland, it was at times somewhat unpredictable due to difficult weather conditions, localised sea ice conditions and thus lead to some challenges with logistics of the data collection. Whilst the survey team adapted to the dynamic environmental conditions and was efficient in data collection when conditions allowed, nevertheless only a fraction of the original planned data was collected. This highlights the need for a more flexible data acquisition plan for both the ground-based data and satellite data over the study area. For the in-situ data collection, it is recommended to use high quality instruments, such as a multi beam echo sounder in conjunction with GNSS GPS systems to allow for more complete mapping of the seabed and accurate geographic positioning. 24 4. Conclusions & Perspectives As demonstrated during this project, satellite derived information holds an enormous potential for obtaining geo-information for remote and challenging regions, such as North East Greenland. By taking advantage of freely, and/or commercially available satellite imagery, advanced image analysis and local knowledge it is possible to provide a detailed characterisation of the coastal zone in arctic waters. The demonstrated methods have the advantage of being cost-efficient, objective and without the risks associated with traditional survey methods. The derived map products have all been developed with satellite-based information and are produced with generic and transferable image analysis methods. This facilitates a larger upscaling of the maps to extend the coverage over large and poorly mapped regions of the arctic. The developed methods and the derived results of the project will be included in the production of Oil Spill Sensitivity Atlases in the Arctic. An important element of the existing Oil Spill Sensitivity Atlases for coastal Greenland, focuses on the analysis of the oil spill resistance of the coast. Traditional analysis is based on manual interpretation of the general coastal morphology and geology from available topographic maps, available aerial photography and low-resolution satellite images, and how these factors determine how oil from a spill will be absorbed by the coastal materials or washed off. In this regard, the information of interest is the predominantly the slope of the coast’s terrain, straight coastlines or small beaches, general geology of rock, boulders, pebbles, sand or sediments. The classification procedures outlined in this study, using medium resolution Sentinel-2 (10m) and WorldView-2, 2 m resolution satellite images, provides the possibility to apply quantitative and automated analysis methods on large image datasets with increased temporal resolution. Additionally, this allows for automated time-series analysis, thereby reducing the biases introduced by traditional manual digitisation methods. We expect the coastline complexity index, the surface classification and tidal zone products of the analyses to be useful in the preparation of future Oil Spill Sensitivity Atlases. The complexity index can detect straight or contoured coastlines and identify the risk of concentrations of oil in pocket beaches or other complex morphologies. The surface classification can play a vital role in determining the types of geology types that absorb or repel oil. The tidal zone mapping and SDB would be new features of an Oil Spill Sensitivity Atlas in Greenland. These products have not previously been investigated within the atlas framework. By applying temporal analysis for the tidal zone mapping and novel algorithms for shallow water bathymetry mapping, new critical areas of important marine or tidal habitats that are highly sensitive to oil spills are identified. These highly sensitive areas would otherwise require costly and detailed fieldwork in the remote areas in order to ascertain the same level of detail. To support the infrastructure of operations in the area, the SDB can be utilised for designating possible natural harbours and bays to capture oil residues. This new information would be beneficial to the atlas, as oil spill materials are often saturated into the tidal soils, depending on the geology. Consequently, we recommend the implementation of high-resolution satellite image analysis in the mapping of oil spill sensitive areas and spill removal infrastructure. 25 Acknowledgements The work carried out in this project has been financed via the Greenland Minerals Authority. The project partners are thankful for the work opportunities that has been made possible by this funding. The project partners also want to express their gratitude towards a number of institutions who have provided assistance and knowledge sharing, which has been of great value to the outcome of this project. These institutions are: ● Nordøstgrønlandsk Kompagni Nanok ● Joint Arctic Command Head Quarter Nuuk ● Joint Arctic Command Station Daneborg and Ella Ø (Sirius Dog Sled Patrol) ● Joint Arctic Command Station Mestersvig ● Royal Danish Navy, 1. Squadron HDMS Hvidbjørnen (F360) A special thank goes to Nanok Kompagniet, which provided us residence at Ella Ø, and helped us with invaluable advice during the field work and blessed us with great stories from their adventures as Sirius patrols. Also, a big thank you to Joint Arctic Command and Sirius, who helped us with all kinds of logistics. Thanks to you we succeeded with our project.