What is the impact of calibration procedure on improving hydrological simulations with reanalysis and satellite-based evaporation products?

Dembele, M.; Ceperley, N. C.; Zwart, S.; Mariethoz, G.; Schaefli, B. (2019). What is the impact of calibration procedure on improving hydrological simulations with reanalysis and satellite-based evaporation products? In: AGU Fall Meeting.

The need to improve the predictive skill of hydrological models has led to the use of diverse reanalysis and satellite remote sensing data sources for model parameter estimation. Evaporation datasets are among the most used products in multivariate (i.e. multiple variables) model calibration. However, satellite-based and reanalysis evaporation data can be integrated in the calibration procedure in various ways depending on model structure, computational requirements, and other factors. Consequently, low model performances might be attributed to inadequate calibration strategies rather than to a low reliability of the used evaporation products, which also vary in performances across regions. <P />This study evaluates the performance of 12 different global evaporation products (MODIS, SSEBop, ALEXI, CMRSET, SEBS, GLEAM v3.2a/v3.3a/v3.2b/v3.3b, ERA5, MERRA-2 and JRA55) in improving the spatiotemporal variability of multiple hydrological state and flux variables (i.e. streamflow, evaporation, soil moisture and terrestrial water storage) simultaneously, when combined with streamflow data. Four different calibration strategies that differently harness the information content of the evaporation products are tested: (i) basin average, (ii) temporal pixel wise, (iii) spatial pixel wise, and (iv) bias insensitive pixel wise. A set of 48 combinations of evaporation products and calibrations procedures are tested and compared to the benchmark model calibrated only on streamflow. Independent datasets of soil moisture (ESA CCI), terrestrial water storage (GRACE), and land surface temperature (MODIS), and in-situ data of eddy-covariance towers are used for model evaluation. The methodology is tested with the fully distributed mesoscale Hydrologic Model (mHM) over ten years (2003-2012) in the predominantly semi-arid Volta River basin in West Africa. <P />Results highlight which combinations of satellite products and calibration procedures are best and worst for reproducing hydrological processes. Most combinations perform better than the benchmark. Finally, we found the spatial pattern of satellite-based evaporation products to be the most relevant feature, which can be used without absolute values to improve the predictive skill of hydrological models.

Interest & Impact

Downloads

0 since deposited on 06 May 2020
0 in the past 12 months

Citations

5 Citations in Web of Science ®
6 Citations in Scopus

Search

in Google Scholar™

Services

Actions (login required)

Edit item

Actions (login required)

Edit item