Enhanced Soil Moisture Initialization Using Blended Soil Moisture Product and Regional Optimization of LSM-RTM Coupled Land Data Assimilation System.

Prediction of soil moisture dynamics is high priority research challenge because of the complex land-atmosphere interaction processes. Soil moisture (SM) plays a decisive role in governing water and energy balance of the terrestrial system. An accurate SM estimate is imperative for hydrological and weather prediction models. Though SM estimates are available from microwave remote sensing and land surface model (LSM) simulations, it is affected by uncertainties from several sources during estimation. Past studies have generally focused on land data assimilation (DA) for improving LSM predictions by assimilating soil moisture from single satellite sensor. This approach is limited by the large time gap between two consequent soil moisture observations due to satellite repeat cycle of more than three days at the equator. To overcome this, in the present study, we have performed DA using ensemble products from the soil moisture operational product system (SMOPS) blended soil moisture retrievals from different satellite sensors into Noah LSM. Before the assimilation period, the Noah LSM is initialized by cycling through seven multiple loops from 2008 to 2010 forcing with Global data assimilation system (GDAS) data over the Indian subcontinent. We assimilated SMOPS into Noah LSM for a period of two years from 2010 to 2011 using Ensemble Kalman Filter within NASA's land information system (LIS) framework. Results show that DA has improved Noah LSM prediction with a high correlation of 0.96 and low root mean square difference of 0.0303 m3/m3 (figure 1a). Further, this study has also investigated the notion of assimilating microwave brightness temperature (Tb) as a proxy for SM estimates owing to the close proximity of Tb and SM. Preliminary sensitivity analysis show a strong need for regional parameterization of radiative transfer models (RTMs) to improve Tb simulation. Towards this goal, we have optimized the forward RTM using swarm optimization technique for direct Tb assimilation. The results indicate an improvement in Tb simulations based on the multi polarization difference index approach with a correlation of 0.81 (figure 1b (e)) and bias of < 5 K with respect to the SMOS Tb.