Determination of the bulk temperature from NOAA/AVHRR satellite data in a midlatitude lake

Abstract

The spatial and temporal variability of the bulk temperature gives important insights into biological and hydrodynamic processes. However, standard algorithms for satellite data only provide information of the surface temperature. The comparison of current and new split-window coefficients applied on NOAA-14/AVHRR brightness temperatures of Lake Constance showed that a regional adaption was most promising. To derive the bulk temperature information, a priori progression from a weather station was included into the AVHRR analysis. Among the weather is data, the mean temperature of the three preceding days and the day of the year were the most relevant additional information. By a multiple regression approach the bulk temperature in the upper 4 m of Lake Constance could be determined with an accuracy of ±1.20 °C. The training of a neural network improved the determination of the bulk temperature to ±1.04 °C. An extended field campaign demonstrated that the algorithm is also applicable to other sensors with the same spectral band settings (in this case NOAA-16/AVHRR) with an acceptable error and that it is equally accurate over the entire lake.