Verification of NWP Model Analyses and Radiosonde Humidity Data with GPS Precipitable Water Vapor Estimates during AMMA

Abstract

Abstract This paper assesses the performance of the European Centre for Medium-Range Weather Forecasts-Integrated Forecast System (ECMWF-IFS) operational analysis and NCEP–NCAR reanalyses I and II over West Africa, using precipitable water vapor (PWV) retrievals from a network of ground-based GPS receivers operated during the African Monsoon Multidisciplinary Analysis (AMMA). The model analyses show reasonable agreement with GPS PWV from 5-daily to monthly means. Errors increase at shorter time scales, indicating that these global NWP models have difficulty in handling the diurnal cycle and moist processes at the synoptic scale. The ECMWF-IFS analysis shows better agreement with GPS PWV than do the NCEP–NCAR reanalyses (the RMS error is smaller by a factor of 2). The model changes in ECMWF-IFS were not clearly reflected in the PWV error over the period of study (2005–08). Radiosonde humidity biases are diagnosed compared to GPS PWV. The impacts of these biases are evidenced in all three model analyses at the level of the diurnal cycle. The results point to a dry bias in the ECMWF analysis in 2006 when Vaisala RS80-A soundings were assimilated, and a diurnally varying bias when Vaisala RS92 or Modem M2K2 soundings were assimilated: dry during day and wet during night. The overall bias is offset to wetter values in NCEP–NCAR reanalysis II, but the diurnal variation of the bias is observed too. Radiosonde bias correction is necessary to reduce NWP model analysis humidity biases and improve precipitation forecast skill. The study points to a wet bias in the Vaisala RS92 data at nighttime and suggests that caution be used when establishing a bias correction scheme.