Alexander Loew1, Mike Schwank2 Combining precipitation and soil moisture observations: a way for improved estimates of land surface water fluxes? Alexander Loew1, Mike Schwank2 1KlimaCampus Hamburg / Max-Planck-Institute for Meteorology (alexander.loew@zmaw.de), 2Swiss Federal Institute for Forest, Snow and Landscape Research (WSL) Motivation The accuracy of land surface model simulations is dependent on the reliability of model input data like, for example, meteorological information or land cover and soil information. Uncertainties of simulations of soil water fluxes are directly related to the accuracy of available precipitation data. As precipitation is characterized by small temporal and spatial correlation lengths, the uncertainties in precipitation data increase with decreasing density of available precipitation gauges. As soil moisture is directly linked to precipitation dynamics, its variation can be used as a proxy for precipitation variability. The present study investigates the potential of using surface soil moisture information based on microwave remote sensing observations to partly compensate for uncertainties in precipitation data (Loew et al., 2009) Data Figure 3: ELBARA L-band (1.4 GHz) microwave radiometer (top); measured brightness temperatures and soil moisture (right). Polarization ratio PR is used as a proxy for the surface soil moisture dynamics. Sensitivity studies revealed that the polarization ration is a suitable soil moisture proxy for the data used in the present study. However, more elaborated soil moisture retrieval algorithms, based on microwave emission models, could be used as well, but did not show superior performance in the present study. Figure 1: Network of meteorological stations used in the study; red: location of microwave radiometer, green: locations of various precipitation stations, listed in Table I with their coordinates and distances to the location of the microwave radiometer Microwave emission model Figure 2: Impact of precipitation variability on soil moisture variability; soil water dynamics based on antecedent precipitation index (API). Water loss due to evapotranspiration and runoff is modelled following the approach of Crow et al. (2007) Methodology A simple soil water balance model (API model) is combined with surface soil moisture observations derived from passive microwave observations. The model is forced using observed precipitation from various meteorological stations. As shown in Figure 2, using precipitation data from different stations results in highly variable model predictions of soil water content [1,2]. Uncertainties of precipitation measurements are therefore mimiced in the study by using observed precipitation from stations beeing located a couple of kilometers apart from the location of the microwave radiometer. The uncertainties in the precipitation data are therefore known in the present study. Uncertainties in the precipitation observations are reduced by assimilating a surface soil moisture proxy from microwave remote sensing data into the API model [3]. Neither the model error Q or observational error S are known in that case. The assimilation procedure solves itteratively for S and Q by analyzing the assimilation filter innovations u. In case they are zero mean and serially uncorrelated, convergence is achieved [4]. Model Analysis Diagnostics Y N observations Itterate S and Q ... ... until u zero mean and serially uncorrelated Soil water model Validation Assimilation framework L-band Microwave observations Rainfall data Data Models In situ soil moisture S1 S2 S3 S0 Increments D Soil moisture q1 1 2 3 4 t q2 q3 q(R0) vs. qinsitu R0-Rk vs. Dk 5 qinsitu Abbreviations Model T: model error covariance Q: model error API: antecedent precipitation index Observations q: soil moisture observations S: observation error Analysis K: Kalman gain u: normalized filter innovations D: filter increment Results The assimlation filter adds or removes water from the model. These filter increments provide a diagnostic tool to evaluate whether the amount of water added or removed by the filter is in agreement with actual uncertainties of the precipitation data [5]. The Figure shows an example of the comparison between precipitation error against the filter increments Δ. A strong correlation between the actual precipitation error and the filter increments indicates, that the method is able to partly compensate for the uncertainties in precipitation data by using additional information on soil moisture dynamics. All stations show a positive correlation between Δ and the rainfall error, which indicates that the model simulations benefit from the integration of the microwave data. The analysis of the relationship between simulated and observed soil moisture shows an improvement of the simulated soil water dynamics for all the stations in the assimilation case. Conclusions Assimilation of remote sensing based surface soil moisture information helps to compensate for uncertainties in precipitation data First study showing these results using L-band radiometer data at the local scale Potential applications for combining satellite soil moisture data (e.g. SMOS) with satellite precipitation data (e.g. Mega-Tropiques) References Loew, A.; Schwank, M. & Schlenz, F.: Assimilation of an L-band microwave soil moisture proxy to compensate for uncertainties in precipitation data. IEEE Transactions on Geosciences Remote Sensing, 2009, 47, 2606 -2616. Crow, W. & Bolten, J.: Estimating precipitation errors using spaceborne surface soil moisture retrievals. Geophyical Research Letters, 2007, 34, L08403.