Advancing Alpine lake monitoring and modelling through calibration, validation, and dissemination of high-resolution thermal remote sensing products

Lake thermal dynamics provide critical insights into regional and global climate change, and play a regulatory role in lake biogeochemical cycles. In situ measurements, remote sensing, and hydrodynamic modelling are key sources for monitoring lake temperature. In situ data are essential for calibration and validation (cal/val) of satellite products and numerical models, but are often scarce or irregular for many lakes. Additionally, data assimilation of lake surface water temperature (LSWT) products can improve numerical models. Satellite thermal imagery has been widely used for LSWT monitoring at regional and global scales. However, current operational LSWT services are limited to 1 km resolution, thereby excluding small lakes. High-resolution, high-revisit Earth observation missions, such as ECOSTRESS, LSTM, TRISHNA, and SBG, extend LSWT services to smaller lakes, but require dedicated cal/val efforts due to their unique radiometric and geometric properties. Collecting reliable skin temperature and ancillary datasets across diverse lakes and optimizing LSWT retrieval algorithms is thus urgently needed.

Our research, within the ESA-funded TRISHNA – Science and Electronics Contribution (T-SEC) project, focuses on validating and improving high-resolution LSWT products, and openly publishing final products for lakes in the Alpine region. We operate automated reference stations in four Swiss lakes: Lake Geneva, Lake Aegeri, Lago Bianco, and Greifensee. These lakes comprise a variety of morphological, bio-physical, and meteorological features, and are located along an elevation gradient in pre-, sub-, and high-alpine environments. Skin, sky, and bulk temperatures, as well as meteorological data are available for all sites. We evaluated Landsat 7/8/9 LSWT products from USGS Collection-2 Level-2 data and the single-band Acolite-TACT algorithm. Our matchup comparisons yielded a Mean Absolute Error (MAE) of < 1.2 °C, a Mean Bias Error (MBE) < 0.1 °C and a correlation coefficient (R²) of > 0.94. However, official Level-2 ECOSTRESS data showed weaker performance (MAE > 2.4 °C, MBE < -2 °C, and R² < 0.85), highlighting the need for further cal/val and algorithm refinements, particularly for emissivity corrections.

Landsat validated algorithms are used for operational monitoring via AlpLakes web platform (www.alplakes.eawag.ch), which integrates satellite data, in situ measurements, and hydrodynamic models. AlpLakes’ scalable design enables rapid integration of new lakes and products. For example, we aim to disseminate our tools across lakes in the Alpine region under the EU Interreg AlpineSpace project DiMark (https://www.alpine-space.eu/project/dimark/). This pipeline will also facilitate the adoption of upcoming missions and timely dissemination of validated products. Ultimately, our research and datasets will support lake monitoring and modelling activities in Switzerland and beyond. Moreover, integrating satellite data, hydrodynamic models, and in situ measurements (e.g., assimilating LSWT products into existing models) will enhance understanding of short-term events and long-term trends in lakes, fostering interdisciplinary research and providing deeper insights into underlying bio-physical processes.