Implementation of surface soil moisture data assimilation with watershed scale distributed hydrological model

Abstract

This paper aims to investigate how surface soil moisture data assimilation affects each hydrologic process and how spatially varying inputs affect the potential capability of surface soil moisture assimilation at the watershed scale. The Ensemble Kalman Filter (EnKF) is coupled with a watershed scale, semi-distributed hydrologic model, the Soil and Water Assessment Tool (SWAT), to assimilate surface (5 cm) soil moisture. By intentionally setting inaccurate precipitation with open loop and EnKF scenarios in a synthetic experiment, the capability of surface soil moisture assimilation to compensate for the precipitation errors were examined. Results show that daily assimilation of surface soil moisture for each HRU improves model predictions especially reducing errors in surface and profile soil moisture estimation. Almost all hydrological processes associated with soil moisture are also improved with decreased Root Mean Square Error (RMSE) values through the EnKF scenario. The EnKF does not produce as much a significant improvement in streamflow predictions as compared to soil moisture estimates in the presence of large precipitation errors and the limitations of the infiltration-runoff model mechanism. Distributed errors of the soil water content also show the benefit of surface soil moisture assimilation and the influences of spatially varying inputs such as soil and landuse types. Thus, soil moisture update through data assimilation can be a supplementary way to overcome the errors created by inaccurate rainfall. Even though this synthetic study shows the potential of remotely sensed surface soil moisture measurements for applications of watershed scale water resources management, future studies are necessary that focus on the use of real-time observational data.

Application of Remote Sensing Data to Constrain Operational Rainfall-Driven Flood Forecasting: A Review

“…Draper et al [197] found that the assimilation of ASCAT data improved the streamflow prediction to some extent, but the improvement may mainly result from the correction of large bias from precipitation, which was suggested to be addressed through bias-aware data assimilation approaches. Han et al [185] found that the improvements in streamflow were much weaker than in soil moisture, and not consistent in all sub-areas. The study by Matgen et al [184] illustrated that introducing soil moisture information brings limited or no extra improvement in streamflow prediction if the model is well calibrated by streamflow gauges.…”

“…Beside a couple of tests using synthetic RS-SM data [104,185,196], ASCAT [102,103,182,184,186,191,192,197], AMSR-E [103,180,182,192] and SMOS [102,103,182,183,198] soil moisture products were widely implemented during this period. Simultaneously assimilating multi-source RS-SM products has also been tested recently [102,103,182,192].…”

“…The choice of the data assimilation algorithm turned to be more consistent in this period. The majority of studies chose the ensemble Kalman filter (EnKF) [102][103][104]180,182,183,185,191,196,198], while just a few used nudging [186], hard updating [192] and EKF [197]. There were also few particle filter (PF) applications.…”

“…With the development of flood forecasting and data assimilation techniques, there is an increasing need to adopt RS-SM assimilation approaches for semi-distributed models. Quite a few attempts have been carried out for semi-distributed model updating [103,179,183,185,193,196], among which the challenges moving from lumped to semi-distributed models have also been explored [103,183].…”

Identification of hydrological model parameter variation using ensemble Kalman filter

“…The DA method has been widely applied in hydrology for soil moisture estimation (Han et al, 2012;Kumar et al, 2012;Yan et al, 2015) and flood forecasting (Y. Liu et al, 2012;Abaza et al, 2014).…”