Data assimilation experiments for AMMA, using radiosonde observations and satellite observations over land F. Rabier, C. Faccani, N. Fourrié, F. Karbou, J-P Lafore, P. Moll, M. Nuret, J-L Redelsperger Météo-France and CNRS, Toulouse, France A. Agusti-Panareda ECMWF, Reading F. Hdidou Direction de la Météorologie Nationale, Morocco O. Bock IGN, France

AMMA: The African Monsoon Multidisciplinary Analysis Better understand the mechanisms of the African monsoon and prevent dramatic situations (Redelsperger et al, 2006) Enhanced observations over West Africa in 2006 In particular, major effort to enhance the radiosonde network (Parker et al, 2008)

Impact of using the AMMA radiosonde dataset New radiosonde stations Enhanced time sampling AMMA database: additional data which were not received in real time + enhanced vertical resolution Bias correction for RH developed at ECMWF (Agusti-Panareda et al) Data impact studies With various datasets, With and without RH bias correction Number of soundings provided on GTS in 2006 and 2005 Period: 15 July- 15 September, 0 and 12 UTC

Impact on mean TCWV CNTR: data from GTS AMMA: from the AMMA database AMMABC: AMMA + bias correction PreAMMA: with a 2005 network NOAMMA: No Radiosonde data

Validation of Total Column Water Vapour analyses: Comparison with GPS data at Tombouctou CNTR: data from GTS AMMA: from the AMMA database AMMABC: AMMA + bias correction PreAMMA: with a 2005 network NOAMMA: No Radiosonde data GPS: Observations Very poor performance of NO AMMA Best performance of AMMABC NO AMMA AMMABC Observations

Impact on quantitative prediction of precipitation over Africa Higher scores for AMMABC Lowest scores for NO AMMA CNTR: data from GTS AMMA: from the AMMA database AMMABC: AMMA + bias correction PreAMMA: with a 2005 network NOAMMA: No Radiosonde data

Downstream impact Impact on geopotential at 500hPa, averaged over 45 days 48hr forecasts: AMMABC vs PREAMMA

3 day range: AMMABC vs PREAMMA

Fit to European Radiosondes Scores at the 3 day range, PREAMMA versus AMMABC

11 Assimilating low-level humidity observations over land Assimilation of MW observations over land New methods for estimating the land surface emissivity (Karbou et al. 2006) operational at Météo-France since July Karbou et al, 2009 Microwave observations over land High emissivity (~1.0) Only channels that are the least sensitive to the surface are currently assimilated Remaining large uncertainties on land emissivity and skin temperature Top of Atmosphere Energy source Surface (emissivity, temperature) (1) Upwelling radiation (2) Downwelling radiation (3) Surface emission Signal attenuated by the atmosphere

12 Impact of assimilating low-level humidity observations over land on the African Monsoon during AMMA Density of assimilated AMSU-B Ch5 during August 2006 ControlExperiment Improved emissivity parametrisation Better simulation by the Radiative Transfer Model of the low-level peaking channels Possibility to assimilate more channels Experiments performed during AMMA in 2006

Assimilation of humidity observations over land Assimilation of AMSU-B Ch2 (150 GHz) & Ch 5 (183±7 GHz) over land, 45 days TCWV (EXP) - TCWV (CTL) TCWV (CTL) Karbou et al, 2009

Humidity bias correction (from ECMWF) over the AMMA region is beneficial Significant positive impact of additional AMMA RS data on the humidity analysis and on precipitation over Africa Positive downstream impact over Europe Using more satellite data over land also has a large positive impact in the Tropics Results in a AMMA special issue Weather and Forecasting Summary of AMMA results

Radiosonde RH Bias correction Well-documented dry bias for Vaisala sonde types (e.g. Wang et al., 2002, Nuret et al., 2008). Motivation: In West Africa many radiosondes are located within a region of strong low-level moisture gradient and there is lack of ppn in the short-range forecast over Sahel. Can be used in data impact studies of enhanced AMMA radiosonde network, AMMA reanalysis experiment and water budget studies within the AMMA project. Based on the ECMWF operational RS bias correction implemented in CY32r3. Main differences between AMMA and OPER. RS RH bias correction: Takes into account the dependence of bias on the observed RH values, which is very important in the Sahel because of its pronounced seasonal cycle. Agusti-Panareda et al

Radiosonde (RS) RH Bias correction: RESULTS Comparison with GPS TCWV RS-GPS: BIAS Olivier Bock UNCORRECTED RS CORRECTED RS Agusti-Panareda et al

Impact of radiosonde bias correction: RESULTS Mean total daily PPN FC (T+42-T+18) [mm/day] 1 to 31 Aug 2006, 12 UTC RSBIAS CORRECTION RSBIAS CORRECTION – CNTRL CNTRL OBS: RFE 2.0 (NOAA CPC) Agusti-Panareda et al

The ECMWF AMMA reanalysis Anna Agustí-Panareda, Carla Cardinali, Jean-Philippe Lafore, … Period: 1 May – 30 September 2006 Resolution: T511 (~40 km), L91 Extra data used: sonde profiles of wind, temperature and humidity extracted from the AMMA database IFS cycle with improved physics: CY32r3 (Bechtold et al., ECMWF Newsletter No. 114, Winter 2007/08, pp ) AMMA radiosonde humidity bias correction (Agustí-Panareda et al. 2008, submitted to Q.J.R.Meteorol.Soc) Agusti-Panareda et al

DFS= degrees of freedom for signal DFS =Tr (δH(x a )/ δy) Calculated for each station, averaged 1-15 August 2006 Large impact of additional AMMA data Faccani et al More influence

Assimilation of humidity observations over land Objective scores %radiosondes geopotential, 24-hr fcst, 34 cases, CTL --- BIAS/TEMP __ EQM/TEMP EXP --- BIAS/TEMP __ EQM/TEMP PRESSURE (hPa)

OPER EXPERIMENT Correlations between observations and RTTOV simulations AMSU-A ch4, August 2006 First step towards the assimilation of surface sensitive observations over land The effect of land surface emissivity

Number of assimilated ch7 AMSU-A data (Temperature 10 km), August 2006 OPER EXPERIMENT Change of land emissivity A land emissivity parameterisation at Météo-France

Time series of global correlations between observations and RTTOV simulations over land : AMSU-B ch2 (150 GHz), August 2006 Impact of emissivity on simulations Simulations from CTL Simulations from EXP

Similar results obtained at ECMWF Monthly averaged RR better with bias correction Faccani et al, 2009 Impact on monthly mean precipitation over Africa Very poor performance of NO AMMA Best performance of AMMABC AMMABC: AMMA + bias correction PreAMMA: with a 2005 network NOAMMA: No Radiosonde data CPC: Observations