Water temperature influences physical and chemical parameters of rivers and streams and is an important parameter for water quality. It is a crucial factor for the existence and the growth of animal and plant species in the river ecosystem. The aim of the research project ”Remote sensing of water surface temperature” at the Federal Institute of Hydrology (BfG), Germany, is to supplement point measurements of water temperature with remote sensing methodology. The research area investigated here is the Upper and Middle Rhine River, where continuous measurements of water temperature are already available for several water quality monitoring stations. Satellite imagery is used to complement these point measurements and to generate longitudinal temperature profiles for a better systematic understanding of the changes in river temperature along its course.
Several products for sea surface temperature derived from radiances in the thermal infrared are available, but for water temperature from rivers less research has been carried out. Problems arise from the characteristics of the river valley and morphology and the proximity to the riverbank. Depending on the river width, a certain spatial resolution of the satellite images is necessary to allow for an accurate identification of the river surface and the calculation of water temperature. The bands from the Landsat ETM+ sensor in the thermal infrared region offer a possibility to extract the river surface temperatures (RST) of a sufficiently wide river such as the Rhine. Additionally, problems such as cloud cover, shadowing effects, georeferencing errors, different emissivity of water and land, scattering of thermal radiation, adjacency and mixed pixel effects had to be accounted for and their effects on the radiance temperatures will be discussed.
For this purpose, several temperature data sets derived from radiance and in situ measurements were com- pared. The observed radiance temperatures are strongly influenced by the atmosphere. Without atmospheric correction, the absolute mean difference between RST and in situ measurements was 1.1°C with a standard devi- ation of 1.3°C. Thus, a correction of atmospheric influences on radiances measured at the top of the atmosphere was necessary and two different methods for atmospheric correction (ATCOR2 and the Atmospheric Correction Parameter Calculator) were applied. The correction results showed that for both methods, the correct choice of atmospheric profiles is very important. With the calculator, an absolute mean difference of 0.8 ± 1.0°C and with the selected overall best scenes, an absolute mean difference of 0.5 ± 0.7°C was achieved.
The selected corrected RST can be used to interpolate between in situ measurements available only for a limited number of points along the river course and longitudinal example profiles of the surface water temperature in the Upper and Middle Rhine could be calculated for different seasons. On the basis of these profiles, the increasing temperature gradient along the Upper Rhine could be identified and the possibility to detect heat or cooling discharge from tributaries and other sources is evaluated.