Inferring elevation variation of lakes and reservoirs from areal extents: Calibrating with altimeter and in situ data

Abstract

Water security is a major national security issue that can impact food and energy production and sometimes political stability in countries all over the globe. Monitoring water supplies is critical for identifying potential crises before they begin. As a means to estimate lake and reservoir storage for sites without reliable in situ water stage data, this study defines correlations between satellite-based reservoir elevation and aerial extents. Water body levels from altimeter water levels (i.e. TOPEX/Poseidon, Jason series) were correlated with areal extents observed in historic multispectral (i.e. MODIS and Landsat TM/ETM + /OLI) imagery for 18 sites globally. Water levels measured using in situ observations were used to validate the relationships between water level, altimetry data, and surface area for six of the sites. Altimeters were generally more accurate (RMSE: 0.40–0.49m) for estimating lake elevations in Iraq and Afghanistan than the modeled elevation data using multispectral sensor areal extents: Landsat (RMSE: 0.25–1.5m) and MODIS (RMSE 0.53–3.0m). Correlations between altimeter data and Landsat imagery processed with Google Earth Engine confirmed similar relationships exists for a broader range of lakes without reported in situ data across the globe (RMSE: 0.24–1.6m). Both sensor types have limitations. Altimeters have fixed orbits, which limits geographical coverage, though multispectral sensors are sensitive to atmospheric conditions. Thus, while altimetry is still preferred to an aerial extent model, lake surface area derived with Google Earth Engine can be used as a reasonable proxy for lake storage, expanding the number of observable lakes beyond the current constellation of altimeters and in situ gauges.