Thesis title: Earth Observation big data exploitation for water reservoirs and glaciers continuous monitoring
Surface freshwater, available in the form of lakes, rivers, reservoirs, wetlands, snow, and glaciers, is a key resource for ecosystems, climate, and human activities like agriculture and industry. Climate change threatens these resources through rising temperatures, altered precipitation, and glacier melt, impacting freshwater availability, quality, and distribution worldwide.
Therefore, it is crucial to develop effective methodologies based on up-to-date technologies to homogeneously monitor surface freshwater on a large scale.
The aim of this thesis is to monitor inland water and glacier levels using satellite altimetry.
Firstly, we proposed an automatic, reliable and worldwide operational workflow based on GEDI (Global Ecosystem Dynamics Investigation), the NASA (National Aeronautics and Space Administration) LiDAR ( Light Detection and Ranging) altimeter hosted on the ISS (International Space Station), for the large scale monitoring of inland water surface levels, benefiting from the availability of the whole time-series of GEDI data within the archive of the Google Earth Engine (GEE) platform.
Leveraging the extensive computational capabilities of GEE, we were able to analyse millions of footprints and to efficiently and reliably employ GEDI as a remote hydrometer.
Our workflow is based on a rigorous spatio-temporal outliers rejection procedure and on the spatial aggregation of the remaining high-quality footprints to estimate a per-epoch median water level and its precision for the considered lake surface.
We carried out a comprehensive assessment by comparing the GEDI retrieved water level time series with in situ gauge data for 11 lakes of variable extent (from several tens to several thousands km2) across three continents.
The developed workflow achieved a homogeneous intrinsic precision of GEDI water levels of 14 cm, and an overall accuracy of 35 cm compared to reference gauge stations. This was accompanied by a strong overall correlation of 0.76 and a slight overestimation bias (6 cm), which is negligible w.r.t. the overall accuracy.
Our GEDI-based workflow can be easily applied to provide reliable inland water level time series for any lake with available GEDI data, offering higher temporal resolution than other altimeters. This lays the foundations for using GEDI for large scale water cycle monitoring, particularly in remote areas where installing hydrometric gauges is not feasible.
Secondly, we carried out a preliminary assessment of SWOT (Surface Water and Ocean Topography), a mission launched in December 2022 with the aim to address the crucial environmental goal of water monitoring to support preparedness for extreme events and facilitate adaptation to climate change on global and local scales. The first assessment was carried out on Lake product Level 2 version 1.1, also known as "L2_HR_LakeSP". The analysis covered six diverse lakes across three continents, revealing an average median bias of 0.08 m of SWOT data, after the removal of outliers, with respect to the considered reference data acquired from various sources (Hydroweb, DAHITI, for North America from Gauge measurements). We found an overall precision of 0.22 m, with an average correlation of 68% between SWOT and reference time series. However, the accuracy varied in the considered six lakes, with biases up to several decimeters in some cases. These discrepancies may stem from residual inconsistencies between {the} vertical reference frame of SWOT and the considered reference data.
In summary, this initial analysis of the "L2_HR_LakeSP" product, Version 1.1, demonstrated the promising potential of SWOT for monitoring seasonal variations in water levels. Nevertheless, significant anomalies were found in delineating the water bodies, particularly in higher latitudes, suggesting potential difficulties for the sensor in accurately finding the pixels that capture the water's surface elevation in those regions. Additionally, a noticeable reduction in accuracy was observed towards the end of the monitoring period. These preliminary findings indicate some issues that should be addressed in future investigations on the quality and potential of SWOT's lake products.
Following the release of the validated SWOT v2.0 product, we extended our analysis to 3 additional lakes. The automated data acquisition process was adjusted to account for the presence of outliers by implementing an appropriate detection and removal workflow. The time series extracted from SWOT data for the final comparison with the in situ gauge measurements included 70 observations for Lake Bodensee, 46 for Lake Garda, and 111 for Lake Léman.
The comparison demonstrated SWOT's capability to detect water level variations with a 92% correlation and an average precision of approximately 0.06 meters, a significantly lower value than the previous version. However, a residual bias of around 0.42 meters compared to hydrometric data was observed, the cause of which remains unclear. One potential explanation for this discrepancy may lie in differences between the height reference frame used by SWOT and those adopted by the hydrometric stations used as reference. Further investigations are required to resolve these inconsistencies in future applications.
Overall, the SWOT mission shows considerable potential for acquiring valuable inland water level data, particularly in situations where in situ measurements are not feasible. This makes SWOT a promising alternative for monitoring inland water reservoirs and, more broadly, for managing water resources.
Finally, we presented a novel approach for analyzing and retrieving glacier surface levels using GEDI altimetry data.
The proposed method was entirely implemented within GEE, and applied to three glaciers in the Alps across nine GEDI acquisitions, each one compared to the corresponding reference Digital Surface Models (DSMs). The glacier profiles along the GEDI tracks revealed the valuable information GEDI provides for glacier surface elevation, showing an overall correlation of 0.99 and a low mean difference (0.04 meters), with an average of 135 GEDI footprints analyzed per glacier.
Furthermore, the findings suggest that GEDI can detect seasonal effects on glaciers. Acquisitions conducted before the melting season tend to show higher elevations compared to the reference DSM acquisition date, while acquisitions after the melting season show lower elevations as the snowpack diminishes.
To sum up, this thesis provides a comprehensive framework for leveraging satellite altimetry data, particularly from GEDI and SWOT, to monitor inland water and glacier surface levels at a global scale. The workflows and methodologies developed here demonstrate promising accuracy and reliability, establishing satellite-based remote sensing as a viable tool for water resource management and climate change adaptation. These findings pave the way for future advancements in remote hydrology, offering essential insights for areas where traditional measurements are challenging, and contributing to a more sustainable and resilient approach to managing freshwater resources.