3D VISUALIZATION AND ANALYSIS OF PATAGONIAN GLACIERS CHANGES USING EARTH OBSERVATION DATA

Introduction

Glaciers, defined as “perennial mass of ice, and possibly firn and snow, originating on the land surface by the recrystallization of snow or other forms of solid precipitation and showing evidence of past or present flow” (Cogley et al., 2010, p. 45) are highly valuable as they provided ecosystem services which include water provision, sediment, nutrient inputs, flow regulation, flood mitigation, and biodiversity conservation (Rivera et al., 2023). As complex natural ecosystems, they are sensitive to climate change, reducing or increasing their extent in response to temperature and precipitation variation, making them crucial climate change indicators (Manquehual-Cheuque & Somos-Valenzuela, 2021). Due to these factors, glaciers are being studied more, with recent studies revealing that most glaciers worldwide are disappearing at an exponential rate due to warming temperatures, contributing to world sea level rise by 27 millimeters since 1961 (Bates, 2020).

Despite the increase in glacier analyses, glaciers in mountain regions are still poorly known due to difficult access and challenging environmental conditions (Carrasco-Escaff et al., 2023). This is the case for the Patagonian Andes, which is the largest glacierized area in South America with over 20000 km² of glaciers, concentrated mainly in the Northern Patagonian Icefield (NPI) and Southern Patagonian Icefield (SPI) (Pellicciotti et al., 2014). Studying this region is essential not only as most of the glaciers have been shrinking in recent decades, contributing a significant proportion of meltwater to the sea level rise in the 20th century (Rignot et al., 2003), but also because some of them are evidencing anomalous advancing (Rivera et al., 2012).

In this context, remote sensing technologies and Earth Observation (EO) data have been widely used since the 1990s to monitor global environmental dynamics and glacier changes in hard-to-reach areas (Yu et al., 2023). This includes EO satellite programs such as Landsat, Sentinel, and the European Centre for Medium-Range Weather Forecasts (ECMWF), with the ERA5-Land dataset for climate analyses. These technologies have introduced advantages such as global coverage, multi-spectral data, and high spatial-temporal resolution (Genzano et al., 2020). However, due to the increasing volume of EO data, known as “big data,” it is not feasible to access, collect, and analyze data using traditional methods (Di Tullio et al., 2018). Therefore, cloud computing platforms, such as Google Earth Engine (Google Earth Engine), have gained popularity as efficient ways for storing, accessing, and analyzing petabytes of EO data, offering free access to fast computations via the internet (Amani et al., 2020).

Although Google Earth Engine provides a solid framework for glacier analyses and visualization in 2D, having a 3D perspective is fundamental as glacier processes are highly determined by their altitude (Manquehual-Cheuque & Somos-Valenzuela, 2021). Currently, with the rise of Web Graphics Library (WebGL) and JavaScript (JS) libraries like CesiumJS, which enable the creation of 3D online environments (Schanche, 2020), 3D applications can be developed to enhance data analysis and visualization, providing deeper insights into environmental changes (Van Ackere et al., 2016).

Objectives

This diploma thesis aims to address the lack of studies and 3D visualizations of NPI and SPI glaciers by developing a Google Earth Engine web application for the creation, visualization, and export of remote-sensing and time-series products that permit the quantitative and qualitative estimation of glacier area, LST, and air temperature changes of the mentioned icefields, in the summer periods between 2018 and 2023, using Landsat 8-9, Sentinel-2, and ERA5-Land imagery. Moreover, it aims to provide a 3D web application that allows users to add, visualize, interact, and compare the remote-sensing and time-series products obtained from the Google Earth Engine web application. The specific goals of the study are:

• Conduct a time-series analysis to estimate changes and identify trends in glacier area, LST, and air temperature.
• Design and publish an interactive, user-friendly Google Earth Engine application where users can create, visualize, and export remote sensing and time-series products for the NPI and SPI glaciers and custom Areas of Interest (AOI).
• Develop and publish a 3D web application using CesiumJS that allows users to dynamically visualize and compare the remote sensing and time-series products derived from the Google Earth Engine web application using widgets such as swipe, sidebar, and display controls.
• Evaluate the user’s reception of the Google Earth Engine and 3D web applications by implementing an online questionnaire.

The results of this work will be open-source, ready-to-use applications that scientific and general communities can use and customize. These applications will allow users to visualize glaciers interactively and innovatively, and generate EO outputs that can be used in user-specific analyses. Such accessibility and customizability will significantly enhance understanding of the Patagonian glacier processes and promote interest in these natural features.

Patagonia

The study area for this research is the Patagonian Andes, considered the largest glacierized area in South America, with a total glacier extent of 20000 km², spread over 46°S and 52°S latitudes (Carrasco-Escaff et al., 2023). Specifically, this research focuses on the NPI and SPI (Figure 2), where most glaciers are concentrated, with an extent of 4200 km² and 13000 km², respectively (Pelliccioti et al., 2014). For these icefields, based on the RGI 7.0 inventory, glaciers with an area greater or equal to 10 km² were filtered and selected, resulting in 27 NPI and 56 SPI glaciers.

For the NPI, the biggest glaciers selected were “San Quintin” (Figure 3A) and San Rafael, with areas of 815.29 km² and 684.55 km², respectively. Meanwhile, for the SPI, the largest were “PIO XI” and “Viedma” (Figure 3B), with extents of 1233.08 km² and 896.36 km². Regarding the glaciers' median height, it goes from 690 m to 2040 m. The NPI and SPI are constituted mainly by calving glaciers that terminate into fjords or lakes (Aniya et al., 1997), which are sensitive to temperature, wind speed, precipitation, and altitude variations. Consequently, climatic changes have a direct impact on their area, volume, and glaciological processes (Carrasco-Escaff et al., 2023 ; Florath et al., 2021; Sáez et al., 2019). In the last few years, NPI and SPI glaciers have been experiencing a dramatic glacier retreat, area, and volume loss due to climate variability (Bates, 2020), leading to increased scientific studies that aim to understand their status and ongoing climatic dynamics. However, glaciers in NPI and SPI react differently to the same climate perturbation (Rivera et al., 2012 ; Sagredo & Lowell, 2021), which makes it difficult to develop regional studies.

Methodology

The workflow developed in this thesis (Figure 4) starts with retrieving the NPI and SPI glacier outlines from RGI 7.0. The data was filtered and exported as shapefiles using ArcGIS Pro. Then, in GEE, the feature collection for the glacier outlines, and the image collections for Landsat 8,9, Sentinel-2, and ERA5-Land imagery were created. Furthermore, inspired by Zhang et al. (2021), Turpo Cayo et al. (2022), and Arif et al. (2021) methodologies, glacier area estimations, median temperature composites, and time series charts for each summer period were generated by the combination of several processing functions. After the time series analysis, a GUI was designed and launched with three principal panels (Main, Map, and Charts) and different widgets. With this GEE web application, the summer median temperature composites and glacier area delimitation images were exported in TIFF format. Then, in ArcGIS Pro, the images were processed to have comparable symbology and be uploaded to Cesium ion. The Comma-separated Values (CSV) files exported from the GEE application were used to create line charts within Microsoft Excel by fitting a linear regression to the values, estimating R², RMSD, and RRMSD, and then uploading them to GitHub. With the required data hosted in the cloud, the 3D CesiumJS web application was developed using Visual Studio Code. Finally, user evaluation for both web applications was performed by collecting feedback from an online ArcGIS Survey123 form.

Results

This chapter presents the results of the study, including the outputs from the time-series analysis, the GEE and 3D CesiumJS web applications, and the findings from the ArcGIS Survey123 form.

Time-Series Analysis Ouputs

With the time-series analysis performed within the GEE cloud, 1245 summer median composites and 249 time-series charts, equivalent to the 83 analyzed glaciers, were generated through the GEE web application GUI. This data can be accessed through the application itself and the “PatagonianGlacier” GitHub repository, which includes the pre-processed and processed datasets for the summer median composites and time-series charts. The repository is available online at https://github.com/felipecamachoh/PatagonianGlaciers.

With the processed time-series charts, fitted a linear regression and having calculated R², RMSD, and RRMSD, it was compared the rate of change in glacier area, LST, and air temperature across all studied glaciers, identifying those with the most significant changes. Additionally, with the RGI 7.0 glacier area values, the year of measurement, and the glacier area for the summer period of 2022/23, the total area loss between this period of time was estimated as a complementary output. The complete consolidated tables can be accessed at https://github.com/felipecamachoh/PatagonianGlaciers/tree/main/RateChanges_Tables.

Google Earth Engine Web Application

After designing the GUI for the GEE code implementation, using widgets such as panels, buttons, and charts, the “Patagonian Glaciers Monitoring Application” GEE web application was launched and publicly shared. It is accessible online at https://felipecamachoh.users.earthengine.app/view/patagonian-glaciers-monitoring. The layout for the web application (Figure 5 & Figure 6) consists of three panels. The Main panel provides the user with general information, interactive buttons, and selectors for running the time series analysis, and creating the summer median composites and time-series charts. The Map panel provides an interactive map element for visualizing the Patagonian glaciers and the remote sensing outputs. Lastly, the time series chart panel displays the time series charts.

3D CesiumJS Web Application

The result of the CesiumJS web application development methodology, is the “3D Patagonian Glaciers Viewer” web application (Figure 7, 8 and 9), available at https://felipecamachoh.github.io/PatagonianGlaciers/3DMonitoringApp.html. This application allows the user to visualize and compare the remote sensing outputs from the GEE web application fully in 3D and to access the time-series charts and rate of change for the glacier area, LST, and air temperature variables.

Survey Review

From the ArcGIS Survey123 form created to evaluate the user experience with the GEE and CesiumJS web applications, available at https://survey123.arcgis.com/share/f81642724caf45bc8053c291a1d440a5, 12 responses were received in total, with 11 participants aged between 25 and 34 years old and one aged over 45. With respect to users' experience with the “Patagonian Glacier Monitoring Application” in terms of interface attractiveness, intuitiveness, and general performance, it was found that most of the users were highly satisfied with the interface, as seven rated it as excellent, and five as good. However, for performance, although most rated it as good or excellent, two users rated it as fair (Figure 10).

Concerning the user's experience with the “3D Patagonian Glaciers Viewer” Cesium JS application, the ratings for the user interface and performance were better than those of the “Patagonian Glaciers Monitoring Application”. The participants rated the interface and performance as either good or excellent, with the interface receiving excellent responses from 75% of users and performance garnering more than 50% excellent ratings. Moreover, compared with the GEE web application that received two bug reports, none of the participants evidenced unexpected behaviors in the 3D CesiumJS web application. Finally, all the contestants agreed that the 3D perspective provided by the “3D Patagonian Glaciers Viewer” application facilitated the understanding of glacier area and temperature changes in the Patagonian glaciers (Figure 11).

Conclusions

Glaciers are considered among the most important natural features as they provide ecosystem services such as water provision, flood mitigation, and biodiversity conservation. Moreover, as complex ecosystems, they are accurate indicators of climate variability. Glaciers worldwide have shown significant area retreats and temperature increases in the last few years, which has motivated research on ongoing glacier dynamics. Despite the increase in glaciology research, glacierized regions such as the Patagonian Andes are still poorly known due to harsh access conditions and disinterest from international scientific communities. In particular, the Patagonian Andes is a critical region, as it is the largest glacierized area in South America, with over 20000 km², distributed mainly between the NPI and SPI. In this context, EO data and techniques, including satellite imagery, cloud computing platforms, and 3D visualization methods, have great potential to overcome the lack of studies and promote interest in these natural ecosystems for both scientific and non-scientific communities.

Motivated by the mentioned context, this diploma thesis focused on developing a GEE web application that allows the creation, visualization, and export of remote-sensing and time series products for the Patagonian region. Moreover, using this application, the goal was to conduct a time-series analysis of glacier area, LST, and air temperature changes for the NPI and SPI glaciers between the 2018 and 2023 summer periods, using Landsat 8-9, Sentinel-2, and ERA5-Land Imagery. Furthermore, the thesis aimed to develop a 3D web application to visualize, interact, and compare the remote sensing and time-series products obtained from the GEE web application. Lastly, it evaluated the user experience with both web applications.

The resulting GEE web application proved to serve its purpose, as it allows users to generate time-series outputs and median summer composites for 83 Patagonian glaciers through an attractive and user-friendly GUI. From this application, 1245 summer composites and 249 time-series charts were generated to evaluate the rate of change of glacier area, LST, and air temperature for the mentioned glaciers, evidencing a significant glacier area retreat and temperature increases overall while also providing detailed information for each glacier independently.

Regarding the 3D web application developed using CesiumJS, it provides a fully 3D interactive and high-resolution experience where users can use the swipe and sidebar widgets to visualize, compare, and assess the ongoing Patagonian glacier changes. Moreover, it demonstrated the potential and advantages of a 3D perspective over the classic 2D web approach to generate impact and grab the general population's attention in scientific discussions.

In conclusion, the time-series methodology and web applications proposed and developed in this study are a major step toward improving the glaciology knowledge of the Patagonian region. They evidenced the hidden potential of GEE and 3D visualizations for scientific dissemination and the importance of GUI design when creating attractive web experiences. Furthermore, as the source code for both GEE and CesiumJS web applications is publicly available, the users will be able to extend their functionalities by introducing new data, adding widgets, or changing the AOI. It is expected that with the outputs generated from this study, scientific and non-scientific communities will gain awareness of the importance of the glaciers and their ongoing changes. The next steps will be to validate the accuracy of the time-series outputs and to continue updating the application considering the upcoming summer periods.

Elements

i = 0;

while (!deck.isInOrder()) {
    print 'Iteration ' + i;
    deck.shuffle();
    i++;
}

print 'It took ' + i + ' iterations to sort the deck.';