|Chen, Fan -|
Submitted to: Watershed Management Conference Proceedings
Publication Type: Abstract Only
Publication Acceptance Date: January 14, 2010
Publication Date: August 1, 2010
Citation: Chen, F., Crow, W.T. 2010. Potential of improving flow modeling with SWAT using remotely sensed soil moisture retrievals [abstract]. Watershed Management Conference Proceedings. 2010 CDROM. Technical Abstract: Antecedent soil moisture plays a vital role in determining the rainfall-runoff partitioning and is hence important to be properly quantified in any watershed modeling efforts. Although such information is rarely available in most study area, remote sensing provides the most viable solution to characterize the antecedent soil moisture condition. The objective of this study is to assess the potential of utilizing remotely sensed soil moisture retrievals to update the soil moisture state of a watershed hydrological model in order to improve the runoff prediction. A case study is performed at the Cobb Creek watershed in southwestern Oklahoma with the semi-distributed, physically based model Soil and Water Assessment Tool (SWAT). Ground-observed and SWAT-modeled storm-scale runoff ratio (RRo and RRm, respectively) were compared to the respective pre-storm soil moisture ('o, 'm) states at 5, 25 and 45 cm depths. RRo was found to significantly correlate to pre-storm 'o at all three depths at the 5% level. With uncalibrated parameter values, simulated stream flow was unacceptable and RRm only correlated to 'm at the 45 cm depth. After calibrating for stream flow, correlation coefficient increased at all three depths. However, error in soil moisture still account for up to 36% of the error in runoff ratio. On the other hand, error in soil moisture is significantly correlated to that of runoff ratio prediction when near-surface soil moisture retrieval ('r) by the advanced microwave sensing radiometer (AMSR-E) is substituted for ground-based observation. Such correlation is considerably stronger with the uncalibrated SWAT model. The results suggest the potential to improve the storm-scale runoff prediction in SWAT by correcting/updating the soil water states using AMSR-E. For data-poor or ungaged watersheds, remotely sensed soil moisture products may offer a solution to improve the runoff prediction without high-quality forcing data (e.g. rainfall) and a priori knowledge of proper parameter values. An ongoing research is underway to assimilate the AMER-E soil moisture into the SWAT model with a Kalman filter and to assess the potential of assimilating other remotely sensed soil moisture products such as the ALEXI soil moisture fields.