Retrieving Soil Moisture States Using Streamflow Assimilation
Christoph Rudiger, Jeffrey P. Walker, Jetse D. Kalma, Garry R. Willgoose, and Paul R. Houser
It has been shown that soil moisture has an important impact on seasonal to interannual climate prediction through evapotranspiration controls, especially in heavily forested areas like the Amazon. Hence, it is important that the land surface component of climate models have an accurate initialisation of soil moisture. While remote sensing of soil moisture holds much promise for near-surface soil moisture measurement, and root zone soil moisture retrieval when assimilated into a land surface model, its application to such heavily forested areas is limited. This is due to the masking effect of dense vegetation canopies on remote sensing signals. However, soil moisture also has a strong impact on streamflow, through its control on baseflow and partitioning of rainfall into infiltration and runoff. Thus the use of streamflow data to constrain model predicted soil moisture is a potentially viable alternative to near-surface soil moisture remote sensing. This research demonstrates this potential using a synthetic twin-experiment. The study is based on typical conditions for both a semiarid and humid environment, using the catchment-based land surface model used by NSIPP (NASA Seasonal to Interannual Climate Prediction Project). First we produce a "truth" dataset which provides the streamflow observations and soil moisture validation data. Second, we make an "openloop" simulation where only the initial soil moisture states have been degraded to represent our lack of knowledge on soil moisture. We then assimilate streamflow observations from the truth run into the degraded simulation, in order to retrieve back the true initial soil moisture states. The results shown from this demonstration are for single subcatchments of much larger catchments, so that runoff routing could be ignored. Future research will include larger nested catchments interconnected via a routing model.