Root Zone Soil Moisture Retrieval Using Streamflow and Surface Moisture Data Assimilation in Nested Catchments

Correct knowledge of soil moisture is important for improving the prediction of coupled land surface atmosphere interactions. This is due to the control that soil moisture exerts on the latent and sensible heat flux transfer between the land surface and atmosphere. Because of this strong dependence on moisture availability, improved atmospheric prediction requires correct initialisation of soil moisture states within the hydrological model. Satellite remote sensing and ground point measurements present two techniques for obtaining soil moisture observations. While point measurements allow for the collection of high resolution data through the soil profile, it is limited to a local or at most regional scale due to instrument and logistical constraints. On the other hand, satellite remote sensing is limited to the top few centimetres but yields good spatial information over large areas. However surface soil moisture remote sensing is limited to regions of low-to-moderate vegetation cover, as dense vegetation masks the soil moisture signal. This makes remote sensing of surface soil moisture impossible for heavily vegetated regions such as the Amazon or south-east Asia; regions which have been shown to have the most potential for improved predictability of precipitation when knowledge of soil moisture values is improved. Hence, an alternate approach for soil moisture information is required. Recent work by the authors has shown the potential for assimilating streamflow measurements to retrieve soil moisture in a small single catchment. In those studies a variational-type data assimilation approach was used to account for the fact that observed streamflow is the result of rainfall and soil moisture conditions at some time in the past. While this approach was able to retrieve the root zone soil moisture well, the surface soil moisture was not well retrieved. It is thus suggested to combine the two approaches to utilise their individual strengths, particularly in regions of mixed vegetation conditions. In this paper we report on use of the variational data assimilation approach to assimilate streamflow observations and/or surface soil moisture observations This synthetic study is undertaken on three nested catchments within the Goulburn River experimental catchment in south-eastern Australia to demonstrate the approach. Three scenarios are presented: i) only streamflow observations are available for the outlet of the lowest catchment, ii) there are no streamflow observations and surface soil moisture observations are only available for the lowest catchment under the assumption that the upper and middle catchments are too densely vegetated and iii) streamflow observations are available for the lower catchment and surface soil moisture observations for the middle catchment. This synthetic study identifies the potential of using different observations, where and when available, for the retrieval of soil moisture initial states. Results are shown for soil moisture and runoff retrieval and the subsequent changes in surface heat fluxes. The assessment is based on a comparison between assimilated, truth and non-assimilated (control) simulations. It was found that the assimilation of streamflow has a significant improvement in the retrieval of profile and root zone soil moisture in all three catchments, but displays limitations in retrieving the surface soil moisture state. In contrast, the assimilation of surface soil moisture in the lower catchment alone does not have any effect on the upstream catchments, as there is no feedback between the downstream and upstream soil moisture and respective runoff. Finally, the joint assimilation of both streamflow and surface soil moisture observations leads to a further improvement from the streamflow assimilation alone.