Glacier area changes in the Arctic and high latitudes using satellite remote sensing

ABSTRACT Glaciers have been retreating over the last century as a result of climate change, particularly in the Arctic, causing sea levels to rise, affecting coastal communities and potentially changing global weather and climate patterns. In this study, we mapped 2203 glaciers in Novaya Zemlya (Russian Arctic), Penny Ice Cap (Baffin Island), Disko Island (Qeqertarsuaq, Greenland) and part of Kenai (Alaska), using Object-Based Image Analysis (OBIA) applied to multispectral Landsat satellite imagery in Google Earth Engine (GEE) to quantify the glacier area changes over three decades. Between 1985–89 and 2019–21, the results show that the overall glacier area loss in Novaya Zemlya is 1319 ± 419 km2 (5.7% of area), 452 ± 227 km2 (6.6%) for Penny Ice Cap, 457 ± 168 km2 (23.6%) in Disko Island and 196 ± 84 km2 (25.7%) in Kenai. A total of seventy-three glaciers have disappeared completely, including sixty-nine on Disko Island, three in Novaya Zemlya and one in Kenai.


Introduction
The cryosphere is a key component of the global climate system, and as temperatures continue to rise, the cryosphere continues to shrink, leading to a number of significant impacts (IPCC, 2021).In particular, melting glaciers are a major concern, as they contribute to sea level rise, coastal erosion, hazards such as avalanches and glacial floods in mountainous regions and the disruption of essential fresh water supplies (Ding et al., 2021).
Glaciers in the Arctic region play a vital role in the global climate system by reflecting sunlight and trapping cold air, which helps to regulate global temperatures (Previdi et al., 2021).Over the last few decades, the Arctic is warming faster than any other region in the world, a phenomenon known as Arctic amplification (e.g.Dai et al., 2019).Positive feedback mechanisms, such as ice-albedo feedback, where sea ice melts and more sunlight is absorbed by the darker ocean, lead to further warming (Screen & Simmonds, 2010).As a result of these processes, studies have reported that the Arctic is warming at least twice as rapidly as anywhere else in the world (Schädel et al., 2018;You et al., 2021), with a recent study estimating that the Arctic has been warming nearly four times faster than the rest of the world (Rantanen et al., 2022).
In the Arctic, glaciers, ice caps, and the Greenland Ice Sheet have all retreated in the past 100 years and are starting to retreat faster since 2000, causing changes in the surface albedo and global sea level rise (Moon et al., 2019).Glaciers, snow cover, and sea ice are expected to decrease across this century, which means that more incoming solar radiation will be absorbed by open water and land, leading to increased melting and heating across the Arctic (IPCC, 2019).
The Global Land Ice Measurements from Space (GLIMS) initiative provides a global database of glacier outlines, mostly derived from satellite imagery (Raup et al., 2007).Glacier outlines, especially outlines of the same glaciers mapped over time, are an important dataset and are necessary for assessing the impact of climate change.However, mapped outlines are only available at a single point in time for most of the World's glaciers, which limits their use for understanding the long-term impacts of climate change.
Given the importance of glaciers to the Arctic region, multi-temporal glacier outlines are necessary to provide a clear understanding of the climate change impacts on Arctic glaciers.In this study, we used a satellite remote sensing approach in Google Earth Engine (GEE) to create multi-temporal outlines of glaciers in different Arctic regions, providing a valuable dataset that can be used to rapidly assess the impacts of climate change, to estimate glacier mass changes, and to predict glacier contributions to sea level rise (e.g.Hock et al., 2019;Zemp et al., 2019).

Study area
This study focused on four different regions within the Arctic and at high latitudes: Novaya Zemlya (Russian Arctic), Penny Ice Cap (Baffin Island), Disko Island (Greenland), and the Kenai Peninsula (Alaska) (Figure 1).A total of 2203 glaciers were selected for mapping using the Randolph Glacier Inventory (RGI) version 6.0 (RGI Consortium, 2017) Novaya Zemlya is located north of Russia's mainland, between the Barents and Kara Seas (Figure 1).Penny Ice Cap is located on Baffin Island in the Canadian Arctic Archipelago, which is the largest land ice region outside of the Greenland and Antarctic ice sheets (Figure 1), while Disko Island (Greenlandic: Qeqertarsuaq) is the largest island in Greenland, located off the west coast.The Kenai Peninsula (hereafter, Kenai) is located in south-central Alaska, between the Cook Inlet and the Gulf of Alaska (Figure 1).The four study regions were selected because they contain glaciers that are spread widely across the Arctic and high latitudes, and therefore can provide useful insights into how glaciers are being impacted by rising atmospheric and sea surface temperatures in the pan-Arctic.

Data sources
NASA's Landsat satellite programme has been collecting imagery of the Earth since 1972 and is the longestrunning earth observation satellite programme (Masek et al., 2020).Landsat satellites collect data in the visible to infrared part of the electromagnetic spectrum to monitor different features of the Earth such as land cover, vegetation, surface temperature, and more.The main sensors of Landsat satellites are Multi-Spectral Scanner (MSS), Thematic Mapper (TM), Enhanced Thematic Mapper Plus (ETM+), Operational Land Imager (OLI), and Thermal Infrared Sensor (TIRS) (Hemati et al., 2021).This study uses Landsat 5 (TM), Landsat 7 (ETM+), and Landsat 8 (OLI/TIRS) for glacier mapping.
Because of the large swath width, the multispectral capabilities, and the long time span of data collection, Landsat data have proven to be an effective asset for glacier mapping and for creating multi-temporal outlines of glaciers (e.g.Mölg et al., 2018;Nuth et al., 2013).The images for this study were carefully selected in GEE during the boreal summer-time period (July-September) with the least amount of cloud cover and snow patches, across three time periods depending on image availability: 1985-1989, 2000-2002, and 2019-2021.To map glaciers, we use Collection 1 Level-2 Tier 1 surface reflectance products (Table 1).These images are orthorectified and atmospherically corrected.In particular, Tier 1 images have the best quality, and are considered suitable for time-series analysis (Masek et al., 2020).The United States Geological Survey (USGS) also provides images in Tier 2; however, Tier 2 images have issues with geometric correction but may still be usable (Hemati et al., 2021).

Glacier mapping: object-based image analysis
Object-based classification uses both spectral and spatial information such as size, shape, texture, and context from the surrounding pixels for image classification (Blaschke, 2010).Because pixel-based classification relies entirely on the spectral information contained within each pixel, it can result in 'salt and pepper' noise in the final classification (Ma et al., 2019).In this study, we used Object-based Image Analysis (OBIA) approach in Google Earth Engine to map glacier changes.Google Earth Engine is a cloud-based platform for Earth Observation data processing and scientific research that allows users to access, analyse, and visualise a multi-petabyte catalogue of data, making GEE one of the most powerful tools available for remote sensing analysis (Gorelick et al., 2017).Below, we provide a brief summary of the approach used as the full details of the methodology are provided in Ali et al. (2023).
In the OBIA approach, segmentation is an important step that groups similar pixels into a cluster or image objects.In this study, we use Simple Non-Iterative Clustering (SNIC), an improved version of simple linear clustering, to segment the Landsat image (Achanta & Süsstrunk, 2017).Based on the segmented image, a Random Forest algorithm with ten trees was trained on manually selected samples of 'glacier' and 'non-glacier' classes throughout the scene (e.g.Nery et al., 2016;Praticò et al., 2021).To train the classifier, we used approximately 100 'glacier' and 100 'non-glacier' samples per image.To create glacier outlines, the classified image was converted from raster to vector, and outlines were exported from GEE to ArcMap 10.5.1.Each glacier outline was carefully examined and corrected where required for the area change analysis.The RGI 6.0 internal boundaries were used to divide the connected glaciers to enable the computation of area change for each glacier.

Map description
The main map contains more than two thousand glacier outlines that were produced using our approach.This breaks down to 480 glacier outlines for Novaya Zemlya, 523 for Penny Ice Cap, 748 for Disko Island and 452 for Kenai.The main map was produced using ESRI ArcGIS Pro in A1 size (594 mm x 841 mm).Because each area is located in a different part of the Arctic, a specific Universal Transverse Mercator (UTM) projection was set for each study region (Table 2).To provide better visualisation and topographic context of the mapped glacier outlines, we used the built-in base maps of ArcGIS Pro: a World Ocean base map for the main layouts, and for the inset map a hill shaded World Topographic map.
The main map shows the area change of each glacier in percentage of glacier area from 1985-89 to 2019-21, (A) shows Novaya Zemlya, (B) Penny Ice Cap, (C) Disko Island, and (D) Kenai.The light-yellow colours show glaciers that have lost the least amount of area in percentage (0-0.9%)while dark red represents the glaciers that have lost more than 30% of their area.The colours of this map were carefully selected to be colour blind safe using the COLORBREWER 2.0 website 'https://colorbrewer2.org/'(e.g.Crameri et al., 2020).
In this study, we created a new dataset of glacier outlines at three time periods: 1985-89, 2000-02, and 2019-21 (Figure 2), that shows the area change of each glacier in both percentage and km 2 in two time periods (1985-89 to 2000-02 and 2000-02 to 2019-21).This long-term record of glacier change is essential for understanding the impact of climate change on glaciers in these areas over time.This dataset also provides the specific date of when the image was captured by the satellite, as well as the area change analysis which was carried out on the manually corrected outlines.
Figure 3 shows the rate of area loss of glaciers in km 2 / year from 1985-89 to 2000-02 and 2000-02 to 2019-21, revealing that the area loss was higher from 2000- 02 to 2019-21, compared to 1985-89 to 2000-02.The rate of area loss of Novaya Zemlya increased by 4.9 km 2 /year (from 36.1 km 2 /year to 41 km 2 /year) between 2000-01 and 2019-21, while Penny Ice Cap selected glaciers loss increased by 3.2 km 2 (from 10.7 km 2 /year to 13.9 km 2 /year), Disko Island glacier area loss increased by 6.9 km 2 (from 10 km 2 /year to 16.9 km 2 /year), and Kenai selected glacier area loss increased by 4.3 km 2 (from 3.7 km 2 /year to 8 km 2 /year).

Conclusion
Based on Landsat data and utilising an OBIA approach in GEE, we created outlines of more than two-thousand glaciers from four different regions of the Arctic: Novaya Zemlya, Penny Ice Cap, Disko Island, and Kenai for three different time periods: 1985-89, 2000-02, and 2019-21.These multi-temporal outlines provide the total area change of each glacier within the study area in km 2 and percentage for three time periods, which is an important data set for rapidly quantifying the impacts of climate change on glacier loss in the Arctic and high latitudes.
The final map enables the glacier outlines to be viewed in greater detail, providing valuable visual insights about the state of glaciers in each of four study regions.The data allowed the glacier changes in each location to be quantified and show that in the timeframes of this study outlined in Table 1.
Between 1985-89 and 2019-21, the results show a clear reduction in the total glacier area in all four regions and demonstrated that all glaciers in the four areas are responding to climate change.However, glacier area loss was greater in 2000-02 to 2019-21, compared to 1985-89 to 2000-02.Seventy-three glaciers have completely retreated, including sixty-nine on Disko Island, three on Novaya Zemlya, and one on Kenai.
It is important to conduct comprehensive regional analysis on glacier changes in the Arctic and high latitudes in order to comprehend the decadal changes and the likely direction of glaciers in these regions as the world continues to warm.Because the melting of these glaciers may have a significant impact on global sea level rise, accurate measurements and mapping large extent changes, as well as the precise tools that can facilitate rapid mapping of glaciers are necessary    at the regional scale.Platforms such as Google Earth Engine, combined with the extensive Landsat archive and approaches such as Object-Based Image Analysis, help provide these valuable tools.

Software
The initial outlines of glaciers were created using Google Earth Engine, and the main map was created using ArcGIS pro version 2.7.0.

Figure 1 .
Figure 1.The four study areas are, (A) Novaya Zemlya; (B) Penny Ice Cap; (C) Disko Island; and (D) Kenai.The red polygons in the inset images are the selected glaciers from Randolph Glacier Inventory (RGI).

Figure 2 .
Figure 2. A close-up example of the area changes of glaciers mapped in Disko Island in 1985, 2001, and 2019.Panel A shows black polygons derived from Landsat 5 in 1985, panel B shows yellow polygons derived from Landsat 7 in 2001 and panel C show red polygon derived from Landsat 8 in 2019.D shows a composite of the mapping done in three time periods, each colour represents a separate year and clearly illustrates the amount of area change at this location.This map uses WGS1984 UTM 22N projection.

Table 1 .
Details of images that are used in this study for each location.

Table 2 .
Shows the UTM projection information used for each study area.