UPDATES IN THE ASSIMILATION OF GEOSTATIONARY RADIANCES AT ECMWF Carole Peubey, Tony McNally, Jean-Noël Thépaut, Sakari Uppala and Dick Dee ECMWF, UK Abstract Currently, ECMWF assimilates clear sky radiances (CSRs) from geostationary platforms into its 4D-Var analysis system (Courtier et al, 1994), alongside other satellite and conventional data. So far, CSRs from the water vapour channels only are assimilated, which mainly generates increments in upper tropospheric relative humidity (RH). Since 2006, a number of changes have occurred in both the network of CSRs used at ECMWF and the way these are assimilated. This note presents the latest updates of the geostationary CSR assimilation in the operational system. An overview of the monitoring of EUMETSAT reprocessed CSRs in the framework of the ERA-interim project is also shown, together with the plans for the future use of geostationary CSRs at ECMWF. RECENT UPDATES IN GEOSTATIONARY NETWORK IN ECMWF Several changes have occurred in the ECMWF geostationary network since 1 January 2006. GOES- 10 (135 W), MET-8 (3.4 E) and MET-5(63 E) stopped their operational missions. GOES-10 was relocated from 135 W to 60W, replaced by GOES-11 over the west Pacific, while MET-7 was moved from 0 to 57 E assuring the continuity of MET-5 operational service over the Indian Ocean. CSRs data from two new satellites have been added to the ECMWF assimilation system: MET-9 (0 ) and MTSAT-1R (140 E, monitoring only). New CSR datasets have been evaluated. MTSAT CSR data are currently passively monitored. Assimilation tests show that they have a neutral to positive impact on forecast scores, leading to a slight decrease in the forecast error in the 200 hpa RH in the Northern Hemisphere (Fig.1). MTSAT WV channel will be actively assimilated in the near future. Comparisons between MET-8 and MET-9 data show that the two CSR datasets have equivalent levels of noise (Fig.2). There is an offset of around 0.6 K in the 7.3 µm channel, which disappears after the bias correction is applied (see below). Figure 1: Difference of root mean square forecast error between assimilation experiments with and without MTSAT WV-channel.
Figure 2: Comparison between MET-8 (blue) and MET-9 (red) WV CSRs ADAPTATIVE BIAS CORRECTION SCHEME From September 2006, a new variational bias correction scheme (VARBC, Dee 2004) has been applied to all radiances, which consists of an online correction calculated during the minimization of the 4D-Var cost function (example shown on Fig.3). For the CSRs, the correction is based on a constant offset and a set of predictors which are the 100-300hPa and 200-50hPa thicknesses and the total water column. Figure 3: Example of bias correction for GOES-11 CSRs.
The main advantage of VARBC is its capacity to update the bias correction at each assimilation cycle. For instance, Fig.4 shows how a calibration problem in MET-5 data can successfully be corrected by VARBC, while the previous static corection scheme did not provide an adequate response. Figure 4: Example of responses to a calibration problem in MET-5 data, with a static bias correction scheme (left) and with VARBC (right, courtesy Thomas Auligne). QUALITY CONTROL After preliminary checks, CSRs are compared to the model first-guess brightness temperature. This is calculated with RTTOV-8 from model fields horizontally interpolated to the observation locations. There are two criteria to reject poor CSR observations in the WV channel, which mainly occur through cloud contamination: to use the % of clear pixels provided with each observation as an indicator of possible cloud contamination. This quality indicator needs to be carefully tuned for each dataset, as different data providers use different methods to assign cloud flag to observations, as shown on Fig.5, more particularly in the case of MTSAT and METEOSAT CSRs. to reject data for which the model departure in the window channel is outside a chosen range. The window channel data need thus to be bias corrected. Different tests have been performed to find the quality control which optimizes the use of geostationary CSR data in the ECMWF assimilation system. The current setting is shown on Table.1. The difference of treatment between satellites reflects the different method of cloud clearing used in the generation of the CSR product. satellites METEOSAT GOES MTSAT Clear pixel test (%) Sea 70 2 3 Land 4 5 6 Window channel Sea 7 8 9 test Land 10 11 12 Table 1: The current setting of the quality control criteria for WV CSRs
Figure 5: The number of observations as a function of the percentage of clear pixels MONITORING OF REPROCESSED EUMETSAT CSRs The ERA interim project allows the quality of reprocessed METEOSAT CSRs from 1989 onwards to be assessed. There is a good consistency between the calibration of the different geostationary satellites in the WV and window channels. First-guess departures are around +4K in the WV channel and -2K in the window channel. By comparison, HIRS first-guess departures in the corresponding channels are smaller but the signs of the departures are consistent with METEOSAT ones. Reprocessed data are more affected by clouds than are current operational data, as attested by scatterplots of first-guess departures versus observations (Fig.6, left panel). If the data were to be assimilated, this problem could be circumvented by applying a strong selection criterion on the number of clear pixels (Fig.6, right panel), but this would greatly reduce the number of assimilated data. Despite this difficulty, experiments suggest a slight positive impact of assimilating MET-3 WV channel on forecast scores above Europe (Fig.7). Figure 6: Observations-minus-First-guess versus observation scatterplots for MET-3 (in K).
Figure 7: Difference of root mean square error forecasts between assimilation experiments with MET-3 WV channel and without MET-3 WV-channel. FUTURE USE OF GEOSTATIONARY RADIANCES Currently, only clear sky infrared radiances are assimilated at ECMWF. Cloudy data are hence discarded, although these represent a large amount of observations and are situated in meteorologically active areas (McNally, 2002). In the future, it is hoped to assimilate cloudy radiances from all infrared sounders currently used in ECMWF. The main difficulty is the non-linearity of the observation operator for clouds. Different strategies can be used. One could be to start working with the most linear cases, corresponding to observations which are already well simulated by the model first-guess but other strategies are currently investigated. REFERENCES Courtier, P, J-N Thépaut and A. Hollingsworth, 1994: A strategy for operational implementation of 4D- Var, using an incremental approach, Q. J. R. Meteorol. Soc., 120, pp 1367-1387 Dee, D. P., 2004: Variational bias correction of radiance data in the ECMWF system, Proceedings of the ECMWF workshop on assimilation of high spectral resolution sounders in NWP. Reading, UK, 28 June - 1 July Harris, B.A. and G. Kelly, 2001: A satellite radiance-bias correction scheme for data assimilation, Q. J. R. Meteorol. Soc., 127, pp 1453-1468 McNally, A.P., 2002: A note on the occurrence of cloud in meteorologically sensitive areas and the implications for advanced infrared sounders, Q. J. R. Meteorol. Soc., 128, pp 2551-2556