Reconstruction of the Hydro-Thermal Behavior of Regulated River Networks of the Columbia River Basin Using Satellite Remote Sensing and Data-Driven Techniques

Abstract

The use of satellite-based thermal infrared remote sensing has facilitated the assessment of surface water temperature on a large scale. However, the inherent limitations of this remote sensing technique make it difficult to assess rivers unless ambient conditions are cloud-free, devoid of steep terrain and the rivers are at least 60 m wide. To address these challenges that limit the spatiotemporal continuity of satellite-based hydro-thermal data, we harnessed the extensive coverage from the Landsat missions' thermal infrared sensors and data-driven techniques to estimate surface water temperature of rivers. Out of the tested data-driven techniques, we selected the Random Forest Regressor as our prime non-linear approach for estimation of surface water temperature in rivers. Using the selected technique, proposed as THORR (Thermal History of Regulated Rivers), we successfully reconstructed a multi-decadal, continuous spatiotemporal surface water temperature record for regulated rivers in the Columbia River Basin. Using 42 years of data, the surface water temperature could be predicted on average with 0.71° C of absolute error regardless of the dam's potential thermal influence in the downstream reaches. The reconstructed hydro-thermal behavior generated from THORR revealed a long-term downstream warming trend along the Columbia River. The open-source THORR tool can be extended to any river system around the world that is not gauged with in-situ temperature measurements for the reconstruction of hydro-thermal behavior.