Improving the accuracy of time-lapse thermal infrared imaging for hydrologic applications

Abstract

In recent years, thermal infrared (TIR) cameras have improved in resolution and accuracy while their cost has declined. By deploying a ground-based TIR camera to collect time-lapse images, it is now possible to acquire high-resolution stream temperatures through both space and time. However, while ground-based TIR is useful for qualitatively identifying stream temperature differences, acquisition of absolute stream temperatures remains difficult due to interference from reflected radiation. Therefore, improved correction approaches are still needed to extract absolute stream temperatures from ground-based, time-lapse TIR imagery. Using >1100 TIR images acquired every 10 minutes during two field seasons, we assess two methods for correcting time-lapse, ground-based TIR stream temperature data: (1) an analytical method derived from the literature that corrects for atmospheric transmissivity, reflected temperatures and water surface emissivity, which did not improve TIR temperature accuracy, and (2) an empirical approach that uses an offset correction created from in-stream control point temperatures, which reduced the mean absolute temperature difference between the TIR and in-situ stream temperatures. Examination of the analytical method revealed its sensitivity to reflected temperatures from the surrounding environment, a problem that is particularly pronounced in ground-based TIR imagery due to the lower stream emissivity at more oblique viewing angles. Since reflected temperatures and stream surface emissivity can be difficult to quantify and are misrepresented in previous hydrologic literature, the empirical correction approach offers an alternative method for extracting absolute stream temperatures from ground-based TIR imagery affected by these factors.