ECMWF EDA-SEKF Soil Moisture Analysis for Numerical Weather Prediction

The current European Centre for Medium-Range Weather Forecasts (ECMWF) soil analysis is part of a weakly coupled land-atmosphere data assimilation system. Form 2019, the ECMWF Ensemble Data Assimilation (EDA) spread will be used in the simplified Extended Kalman Filter (SEKF) soil analysis to compute the Jacobians. It will replace the current SEKF approach where the Jacobians are computed in finite differences using a set of one control and one perturbed coupled model trajectory per analysed variable. In addition the Soil Moisture and Ocean Salinity (SMOS) neural network soil moisture product will be assimilated along with the ASCAT level 2 surface soil moisture product. To this end a new SMOS neural network processor was recently implemented at ECMWF using SMOS level 1 brightness temperature and trained on ECMWF soil moisture.

This paper presents the new EDA-SEKF ECMWF soil analysis and it shows its impact on Numerical Weather Prediction. The EDA Jacobian approach in the SEKF enhances the coupling between land and atmospheric assimilation systems by ensuring more dynamic Jacobian estimates than in current finite difference approach. It is also highly interesting for operational perpectives since it significantly reduces the SEKF computing cost. The impact of using SMOS neural network and the EDA Jacobians is relatively neutral on medium range weather forecasts, with however a small but significant improvement in two meter temperature forecasts in the short range in the northern hemisphere and slight improvements in soil moisture.

These recent developments both in land surface satellite data assimilation and land-atmosphere coupling are relevant for both NWP and flood forecasting activities at ECMWF. In the near future the impact of assimilation of soil moisture satellite products will be evaluated for consistent NWP and flood forecasting applications with a particular focus on extreme events.