Parameter Uncertainty of a Hydrologic Model Calibrated with Remotely Sensed Evapotranspiration and Soil Moisture

Remotely sensed (RS) observations are becoming prevalent for hydrological model calibration in sparsely monitored regions. In this study, the parameter uncertainty associated with a hydrological model calibrated with RS evapotranspiration (ET) and soil moisture (SM) is investigated in detail using a Markov chain Monte Carlo (MCMC) approach. The daily Commonwealth Scientific and Industrial Research Organization (CSIRO) Moderate Resolution Imaging Spectrometer (MODIS) ReScaled potential ET (CMRSET) and SM retrievals from the Advanced Microwave Scanning Radiometer-Earth Observing System (AMSR-E) are used to calibrate a simplified Australian Water Resource Assessment Landscape (AWRA-L) model at 10 small catchments in Eastern Australia. The study inspects the changes in parameter uncertainty with respect to different RS observations and catchment rainfall conditions and the impact of parameter uncertainty on model predictions. Results suggest that uncertainty in posterior parameter distributions increases from high- to low-rainfall catchments due to the intricate nonlinear relationship between rainfall and runoff in low-yielding catchments. Uncertainty is narrower for ET calibrations than SM calibrations, representing higher uncertainty associated with SM data processing. The study concluded that quantification of parameter uncertainty alone is not enough to provide satisfactory prediction uncertainty.

Automated summary

In this study, parameter uncertainty associated with a hydrological model calibrated with remotely sensed evapotranspiration and soil moisture data was examined using a MCMC approach. Results showed that uncertainty was higher for soil moisture calibration compared to evapotranspiration, and increased in low-rainfall catchments.