Soil Moisture Analysis – Results from E-LDAS

1. Soil Moisture Analysis – Results from E-LDAS M. Drusch1, H. Wilker2, G.Seuffert3, P. Vitrbo1 1 ECMWF, Reading, UK, 2 Meteorological Institute Bonn University, Bonn, Germany, 3 BMVBW, Bonn, Germany Atm. initial conditions + dynamics forcing from ECMWF reanalysis (ERA40) [mm / day] [mm / day] precipitation runoff evaporation Single-column model of the ECMWF NWP model + microwave emissivity model soil moisture analysis increments Observation of precipitation + radiation empirical corrections for low incoming solar radiation F1,F2 sequential, intermittent assimilation Nudging OI Introduction Motivation standard deviations of forecast and analysis error forecast increment analysis increment σ observations observations observations dam2 12 UTC – 6 UTC May 2002 an 12 UTC – fc 12 UTC May 2002 dam2 [mm / day] Increments (daily) First guess: T2m,RH2m,Tb ρ A well posed analysis is a better estimate of the true state than either the background information or the observation data sets available. Analyses can be used as: - initial state for a numerical weather forecast - reference against which to quality check other observations - pseudo observation for e.g. satellite retrieval algorithm development The ratio between analyses increments and forecast increments is one indicator of the quality of the forecast system. In case of 500 hPa geopotential height the analysis increments are one order of magnitude smaller compared to the forecast increments (Fig. 2, upper panel). The increments are comparable for root zone soil moisture (Fig. 2, lower panel). Consequently, analysis increments are a sizeable part of the water budget (Fig. 3). correlation of forecast error Observations: T2m,RH2m,Tb Soil moisture analysis scheme OI or Extended Kalman Filter analysis analysis analysis Soil moisture Background error OI / Tiled Surface forecast 06 12 18 forecast increment analysis increment (mm/6h)2 (mm/6h)2 Fig. 3: Water budget components for the Baltic Sea Area calculated from the ERA 40 data set. Fig. 2: Mean square increments for 500 hPa geopotential height (top pannel) and root zone soil moisture (lower pannel) for May 2002. Fig. 1: Schematic view of a sequential, intermittent assimilation system (e.g. ECMWF’s atmospheric 4DVar). Current operational DA systems Summary I • Optimal Interpolation is a simplification of BLUE: • The gain K is calculated by solving the system of linear equations • minimizing the variance of the resulting analysis error (Douville et al. 2000). • Soil moisture observations have not been available. Two meter temperature • and relative humidity are used as proxy observations for root zone soil moisture. • The Optimal Interpolation scheme is only locally optimal (at best). • The analysis will produce optimal surface fluxes rather than optimal soil moisture. • Soil moisture increments are comparable with the forecast increments and must be • regarded as a sizeable contribution to the water budget. • The state of the soil moisture analysis is comparable to the state of the atmospheric • analysis (dynamics) 20 years ago. a) b) Future analysis systems • Analyzed 2 meter temperature and humidity are being assimilated. • Statistics of background errors were derived from Monte Carlo Experiments. New Observations Observation Operators c) Advanced DA systems • radiative transfer model / • transfer function • radiative transfer model & • land surface model • routing scheme • microwave brightness • temperature • microwave backscatter • coefficient • infra red heating rate • streamflow • Extended Kalman Filter • Ensemble Kalman Filter forecast error correlation • Both, improvements in DA and the physical model, are needed to reduce • analysis increments. OI is significantly better compared to Nudging • Techniques (Fig.4) -0.82 -0.92 -0.90 0.83 0.93 0.91 -0.99 value Fig. 4: Mean soil moisture increments for July 1988. E-LDAS single-column experiment The Extended Kalman Filter • The standard KF is the optimal sequential data assimilation • method for linear dynamics and measurement processes with • Gaussian error statistics. • The EKF is a variant that can be used for nonlinear problems. • The EKF works sequentially, applying in turn a forecast step • and an update step. • Propagating the background error covariance matrix is • computationally very demanding (and the limiting factor in • NWP applications). • In land surface applications, horizontal correlations are being • neglected, which reduces the costs considerably. Update at Ti: observations observation [after Seuffert et al. 2003] EKF Model: ECMWF IFS ‚single column‘ Observation: Land Surface Microwave Emission Model Operator Assimilation scheme: Extended Kalman Filter Daten: SGP97, LW02, 15.6.-15.7.1997 Ti-24 Ti Ti+24 Propagation Ti-24 to Ti: Propagation Ti to Ti+24: [after Reichle et al. 2002] Results Summary II errors: σQ1 = 0.01 m3 / m3 σQ2 = 0.01 m3 / m3 σQ3 = 0.0015 m3 / m3 σT2m= 2 K σrH = 10 % σTb = 2 K Observation bias correction! errors: σQ = 0.005 m3 / m3 σT2m= 2 K σrH = 10 % σTb = 2 K Observation bias correction! errors: σQ = 0.005 m3 / m3 σT2m= 2 K σrH = 10 % σTb = 2 K Reference Run ‘Bias-free’ Run • Field experiments are necessary to develop and evaluate analysis • systems. • The EKF is a computationally affordable method, which yield • promising results in the SCM experiment. • The assimilation of screen level parameters and brightness • temperatures improved the modelled surface fluxes. • A realistic set of background errors is needed to obtain reliable root • zone soil moisture. • Assimilation of brightness temperatures does not prevent the • assimilation system from abusing root zone soil moisture to correct • surface fluxes. • Assimilation of brightness temperature is an effective way to correct • for errors in the precipitation forcing. • ‘Observation’ bias correction is a fundamental problem. It is • probably as important as the correct choice of observation and • model errors. ‘Dry’ Run Best set up - surface sm : KTB root zone sm: KTB evaporative fraction: KTB Best set up - surface sm : KTB root zone sm: CTRL / KTB evaporative fraction: KTB Best set up - surface sm : KTRB root zone sm: CTRL / KTB evaporative fraction: KTRB Douville, H., P. Viterbo, J.-F. Mahfouf, and A.C.M. Beljaars, 2000: Evaluation of the Optimum Interpolation and Nudging Techniques for soil moisture analysis using FIFE data, Mon. Wea. Rev., 128, 1733-1756 Reichle, R., J.P. Walker, R.D. Koster, and P. Houser, 2002: Extended versus Ensemble Kalman Filtering for Land Data Assimilation, J. Hydromet., 3, 728-740 Seuffert, G., H. Wilker, P.Viterbo, M. Drusch, and J.-F. Mahfouf, 2004: On the usage of screen level parameters and microwave brightness temperature for soil moisture analysis, J. Hydromet., 5, 516-531