1
Use of satellite soil moisture data for NWP at the Italian Air Force Meteorological Centre Paride Ferrante, Francesca Marcucci , Valerio Cardinali, Lucio Torrisi [email protected], [email protected], [email protected], [email protected] COMET - Italian Air Force Operational Centre for Meteorology KENDA (Kilometre-scale ensemble data assimilation) EnKF DA LETKF Formulation (Hunt et al,2007) Model and sampling errors are taken into account using: Analysis Ensemble Mean Analysis Ensemble Perturb. 2 1 a 1 a ~ 1 W )) ( ( ~ w a b bT a P m x H y R Y P ) ) ( ) ( ( ),...., ) ( ) ( ( Y ) 1 ( P ~ 1 b 1 1 a b b m b b b bT x H x H x H x H Y R Y I m a b a b W X w X a b a X x x 6-hourly assimilation cycle 40 ensemble members + deterministic run with 0.09° (~10Km) grid spacing (COSMO model), 45 vertical levels (T,u,v,pseudoRH,ps) set of control variables Observations: RAOB (also 4D), PILOT, SYNOP, SHIP, BUOY, Wind Profilers, AMDAR-ACAR-AIREP, MSG3-MET8 AMV, MetopA-B scatt. winds, NOAA/MetopA-B AMSUA/MHS and NPP ATMS radiances+ Land SAF snow mask, IFS SST analysis once a day = 0.95 σ a 2 = variance an. pert. “Relaxation-to-Prior Spread” Multiplicative Inflaction according to Whitaker et al (2010) Additive noise from scaled ECMWF EPS pertubations Lateral Boundary conditions from the most recent IFS deterministic run perturbed using ECMWF EPS Climatological Perturbed SST Adaptive selection radius using a fixed number of effective observations (sum of obs weights) H-SAF ASCAT soil moisture quality control. Data are rejected if: snow: the analysed snow amount is greater than 0.05 kg/m^2 (not active) Sea point (check land sea mask) frost: the 2m Temperature analysis is below 275.15 K (not active) wetlands: the inundation and wetland amount has a value greater than 15% mountains: the topographic complexity has a value greater than 20% ASCAT estimated error: greater than 8% (ECMWF value, UKMO uses 7%) This check rejects ASCAT data from regions with dense vegetation and sand dunes Soil type =1 or 2 (ice and rock) “processing flag” 0 (quality of retrieval) Ens.mean Observation Increments > 2.5 ( estimated from 1 year statistics for each soyl type) Observation error: e_o= 2 x BUFR estimated error (suggested by P. De Rosnay ECMWF ) SOIL MOISTURE ASSIMILATION: VERIFICATION RESULTS (parallel test suite from 22 jun 2016 to 23 jul 2016) Verification results with respect SYNOP and TEMP observations Synop-2m dew point temperature: A little improvement of rmse and bias is observed No impact for other variables (not shown) TEMP wind vector: A little improvement of rmse is observed No impact for other variables (not shown) SOIL MOISTURE ASSIMILATION: PRE - PROCESSING OF ASCAT HSAF SOIL MOISTURE DATA ASCAT soil moisture Data provided by EUMETSAT within the H-SAF project, one of the 8 EUMETSAT SAFs, lead by the Italian Air Force Met Service frequency: 5.3 GHz (microwave C-band) VV polarization Able to provide a triplet of backscattering coefficients for each swath 25 km resolution From backscattering coefficient measurements it is possible to retrieve the soil moisture content in the first 2 cm below the soil by mean of microwave technique thanks to the high sensitivity of microwaves to the water content in the soil surface layer (for microwave frequencies in the C-band (< 10 GHz) the addition of liquid water to the soil strongly increases the soil dielectric constant, and so the backscattering coefficients) H-SAF ASCAT derived Soil Moisture: degree of saturation (%) in the first 2 cm COSMO TERRA_ML model soil moisture: liquid water content (m H2O) in the various model layers To compare observed and model values the model values are transformed (to have quantities independent from the thickness of the layers) in volumetric water content (m^3/m^3) in the first 2 cm, using CDF matching method (ECMWF approach) To scale the ASCAT derived soil moisture to the model climatology so that the cumulative distribution functions (CDF) of satellite and model soil moisture match (performed for each soil type separately). • 1 year time series of ASCAT and model SM data (january 2015 - january 2016) local regression analysis global regression analysis CDF Matching Method: Rescalation of ASCAT observation values to the model values LETKF 10 km 45 v.l. b slope, a intercept TEST 1 :soil moisture observations influence ONLY the low level atmospheric variables : l_soil_ana = false , horizontal localization (100 km) , vertical localization (10 lower levels) A little improvement is noticed compared to the reference case (no use of ASCAT soil moisture data) lowatm (TEST1) full wso (TEST2) noLowatm (TEST3) Verification results with respect SYNOP observations TEST 2 : soil moisture observations influence BOTH the low level atm + soil variables (100km, 10 lower levels) TEST 3 : soil moisture observations influence only soil variables BIAS RMSE Small improvement in TD and CCT bias for TEST2 Small improvement in TD and DD rmse for TEST2
