Spatial heterogeneity in geothermally-influenced lakes derived from atmospherically corrected Landsat thermal imagery and three-dimensional hydrodynamic modelling

• Effect of atmospheric profiles on precision of Landsat water temperature retrieval.
• Surface water temperatures validated using a high-frequency sensor data.
• Landsat data yield accurate lake surface water temperature RMSE 0.48 K.
• Validation of hydrodynamic lake models.

Atmospheric correction of Landsat 7 thermal data was carried out for the purpose of retrieval of lake skin water temperature in Rotorua lakes, and Lake Taupo, North Island, New Zealand. The effect of the atmosphere was modelled using four sources of atmospheric profile data as input to the MODerate resolution atmospheric TRANsmission (MODTRAN) radiative transfer model. The retrieved skin water temperatures were validated using a high-frequency temperature sensor deployed from a monitoring buoy at the water surface of Lake Rotorua. The most accurate atmospheric correction method was with Moderate Resolution Imaging Spectroradiometer (MODIS) atmospheric profile data (root-mean-square-error, RMSE, 0.48 K), followed by radiosonde (0.52 K), Atmospheric Infrared Sounder (AIRS) Level 3 (0.54 K), and the NASA atmospheric correction parameter calculator (0.94 K). Retrieved water temperature was used for assessing spatial heterogeneity and accuracy of surface water temperature simulated with a three-dimensional (3-D) hydrodynamic model of Lake Rotoehu, located approximately 20 km east of Lake Rotorua. This comparison indicated that the model was suitable for reproducing the dominant horizontal variations in surface water temperature in the lake. This study demonstrated the potential of accurate satellite-based thermal monitoring to validate temperature outputs from 3-D hydrodynamic model simulations. It also provided atmospheric correction options for local and global applications of Landsat thermal data.