• Andersson, E., , R. Dumelow, , H. Huang, , J-N. Thépaut, , and A. Simmons, 2004: Space/terrestrial link. Outline study proposal for consideration by EUCOS management. (Available from EUCOS Secretariat, DWD Kaiserleistr, 29/35, 63067 Offenbach, Germany.).

  • Andersson, E., and Coauthors, 2005: Assimilation and modeling of the atmospheric hydrological cycle in the ECMWF forecasting system. Bull. Amer. Meteor. Soc., 86 , 387402.

    • Search Google Scholar
    • Export Citation
  • Andersson, E., , E. Hólm, , P. Bauer, , A. Beljaars, , G. A. Kelly, , A. P. McNally, , A. J. Simmons, , and J-N. Thépaut, 2007: Analysis and forecast impact of the main humidity observing systems. Quart. J. Roy. Meteor. Soc., 133 , 14731485. doi: 10.1002/qj.112.

    • Search Google Scholar
    • Export Citation
  • Auligné, T., , A. P. McNally, , and D. P. Dee, 2007: Adaptive bias correction for satellite data in a numerical weather prediction system. Quart. J. Roy. Meteor. Soc., 133 , 631642.

    • Search Google Scholar
    • Export Citation
  • Bauer, P., , G. Kelly, , and E. Andersson, 2002: SSM/I radiance assimilation at ECMWF. Proc. ECMWF/GEWEX Workshop on Humidity Analysis, Reading, United Kingdom, ECMWF, 167–175. [Available from the European Centre for Medium-Range Weather Forecasts, Shinfield Park, Reading RG2 9AX, United Kingdom.].

    • Search Google Scholar
    • Export Citation
  • Bauer, P., , P. Lopez, , A. Benedetti, , D. Salmond, , and E. Moreau, 2006a: Implementation of 1D+4D-Var assimilation of precipitation affected microwave radiances at ECMWF. Part I: 1D-Var. Quart. J. Roy. Meteor. Soc., 132 , 22772306.

    • Search Google Scholar
    • Export Citation
  • Bauer, P., , P. Lopez, , A. Benedetti, , D. Salmond, , S. Saarinen, , and M. Bonazzola, 2006b: Implementation of 1D+4D-Var assimilation of precipitation affected microwave radiances at ECMWF. Part II: 4D-Var. Quart. J. Roy. Meteor. Soc., 132 , 23072332.

    • Search Google Scholar
    • Export Citation
  • Bengtsson, L., , and K. I. Hodges, 2005: On the impact of humidity observations in numerical weather prediction. Tellus, 57A , 701708.

  • Dee, D. P., 2005: Bias and data assimilation. Quart. J. Roy. Meteor. Soc., 131 , 33233343.

  • Errico, R., , P. Bauer, , and J-F. Mahfouf, 2007: Issues regarding the assimilation of cloud and precipitation data. J. Atmos. Sci., 64 , 37853798.

    • Search Google Scholar
    • Export Citation
  • Gérard, E., , and R. Saunders, 1999: Four-dimensional variational assimilation of Special Sensor Microwave/Imager total column water vapour in the ECMWF model. Quart. J. Roy. Meteor. Soc., 125 , 30773102.

    • Search Google Scholar
    • Export Citation
  • Haseler, J., 2004: The early-delivery suite. ECMWF Tech. Memo. 454, 35 pp. [Available from European Centre for Medium-Range Weather Forecasts, Shinfield Park, Reading RG2 9AX, United Kingdom.].

  • Heckley, W. A., , G. Kelly, , and M. Tiedtke, 1990: On the use of satellite-derived heating rates for data assimilation within the tropics. Mon. Wea. Rev., 118 , 17431757.

    • Search Google Scholar
    • Export Citation
  • Kållberg, P., , P. Berrisford, , B. Hoskins, , A. Simmons, , S. Uppala, , S. Lamy-Thépaut, , and R. Hine, 2005: ERA-40 Atlas. ECMWF Re-Analysis Tech. Rep 19, ECMWF, 191 pp. [Available from European Centre for Medium-Range Weather Forecasts, Shinfield Park, Reading RG2 9AX, United Kingdom.].

    • Search Google Scholar
    • Export Citation
  • Klinker, E., , F. Rabier, , G. Kelly, , and J-F. Mahfouf, 2000: The ECMWF operational implementation of four-dimensional variational assimilation. Part III: Experimental results and diagnostics with operational configuration. Quart. J. Roy. Meteor. Soc., 126 , 11911215.

    • Search Google Scholar
    • Export Citation
  • Krishnamurti, T. N., , and H. S. Bedi, 1996: A brief review of physical initialization. Meteor. Atmos. Phys., 60 , 137142.

  • Krishnamurti, T. N., , K. Ingles, , S. Cooke, , T. Kitade, , and R. Pash, 1984: Details of low-latitude medium-range numerical weather prediction model using a global spectral model. Part 2: Effects of orography and physical initialization. J. Meteor. Soc. Japan, 62 , 613648.

    • Search Google Scholar
    • Export Citation
  • Lopez, P., , and E. Moreau, 2005: A convection scheme for data assimilation: Description and initial tests. Quart. J. Roy. Meteor. Soc., 131 , 409436.

    • Search Google Scholar
    • Export Citation
  • Lopez, P., , and P. Bauer, 2007: 1D+4D-Var assimilation of NCEP Stage IV radar and gauge hourly precipitation data. Mon. Wea. Rev., 135 , 25062524.

    • Search Google Scholar
    • Export Citation
  • Mahfouf, J-F., , and F. Rabier, 2000: The ECMWF operational implementation of four-dimensional variational assimilation. Part II: Experimental results with improved physics. Quart. J. Roy. Meteor. Soc., 126 , 11711190.

    • Search Google Scholar
    • Export Citation
  • Mahfouf, J-F., , V. Marécal, , M. Janisková, , F. Chevallier, , and J. Barkmeijer, 2001: Towards the use of satellite data for cloud and rain assimilation. Proc. ECMWF Seminar on Exploitation of the New Generation of Satellite Instruments for Numerical Weather Prediction, Reading, United Kingdom, ECMWF, 267–287. [Available from European Centre for Medium-Range Weather Forecasts, Shinfield Park, Reading RG2 9AX, United Kingdom.].

  • Mahfouf, J-F., , V. Marécal, , and P. Bauer, 2005: The assimilation of SSM/I and TMI rainfall rates in the ECMWF 4D-Var system. Quart. J. Roy. Meteor. Soc., 131 , 437458.

    • Search Google Scholar
    • Export Citation
  • Marécal, V., , and J-F. Mahfouf, 2000: Variational retrieval of temperature and humidity profiles from TRMM precipitation data. Mon. Wea. Rev., 128 , 38533866.

    • Search Google Scholar
    • Export Citation
  • Marécal, V., , and J-F. Mahfouf, 2002: Four-dimensional variational assimilation of total column water vapor in rainy areas. Mon. Wea. Rev., 130 , 4358.

    • Search Google Scholar
    • Export Citation
  • Marécal, V., , and J-F. Mahfouf, 2003: Experiments on 4D-Var assimilation of rainfall data using an incremental formulation. Quart. J. Roy. Meteor. Soc., 129 , 31373160.

    • Search Google Scholar
    • Export Citation
  • Moreau, E., , P. Lopez, , P. Bauer, , A. M. Tompkins, , M. Janisková, , and F. Chevallier, 2004: Variational retrieval of temperature and humidity profiles using rain rates versus microwave brightness temperatures. Quart. J. Roy. Meteor. Soc., 130 , 827852.

    • Search Google Scholar
    • Export Citation
  • Phalippou, L., 1996: Variational retrieval of humidity profile, wind speed and cloud liquid-water path with the SSM/I: Potential for numerical weather prediction. Quart. J. Roy. Meteor. Soc., 122 , 327355.

    • Search Google Scholar
    • Export Citation
  • Puri, K., , and M. J. Miller, 1990: The use of satellite data in the specification of convective heating for diabatic initialization and moisture adjustment in numerical weather prediction models. Mon. Wea. Rev., 118 , 6793.

    • Search Google Scholar
    • Export Citation
  • Rabier, F., , H. Järvinen, , E. Klinker, , J-F. Mahfouf, , and A. Simmons, 2000: The ECMWF operational implementation of four-dimensional variational assimilation. Part I: Experimental results with simplified physics. Quart. J. Roy. Meteor. Soc., 126 , 11431170.

    • Search Google Scholar
    • Export Citation
  • Tompkins, A. M., , and M. Janisková, 2004: A cloud scheme for data assimilation: Description and initial tests. Quart. J. Roy. Meteor. Soc., 130 , 24952518.

    • Search Google Scholar
    • Export Citation
  • Trémolet, Y., 2004: Diagnostics of linear and incremental approximations in 4D-Var. Quart. J. Roy. Meteor. Soc., 130 , 22332251.

  • Treadon, R. E., , H-L. Pan, , W-S. Wu, , Y. Lin, , W. S. Olson, , and R. J. Kuligowski, 2003: Global and regional moisture analysis at NCEP. Proc. ECMWF/GEWEX Workshop on Humidity Analysis, Reading, United Kingdom, ECMWF, 33–48. [Available from European Centre for Medium-Range Weather Forecasts, Shinfield Park, Reading RG2 9AX, United Kingdom.].

  • Tsuyuki, T., , K. Koisumi, , and Y. Ishikawa, 2003: The JMA mesoscale 4D-Var system and assimilation of precipitation and moisture data. Proc. ECMWF/GEWEX Workshop on Humidity Analysis, Reading, United Kingdom, ECMWF, 59–68. [Available from European Centre for Medium-Range Weather Forecasts, Shinfield Park, Reading RG2 9AX, United Kingdom.].

  • Uppala, S. M., and Coauthors, 2005: The ERA-40 re-analysis. Quart. J. Roy. Meteor. Soc., 131 , 29613012.

  • View in gallery

    Mean analysis difference for TCWV as a percent of the TCWV content of the BL experiment: (a) CTRL − BL, (b) (BL + RAIN) − BL, (c) (BL + CLEAR) − BL, and (d) (BL + CLEAR + RAIN) − BL.

  • View in gallery

    Zonal cross sections of the mean relative humidity analysis difference: (a) (BL + CLEAR + RAIN) − BL, (b) (BL + RAIN) − BL, (c) (BL + CLEAR) − BL, and (d) CTRL − BL.

  • View in gallery

    Mean relative analysis difference for TCWV (percent) (a) CTRL − (CTRL − RAIN) and (b) CTRL − (CTRL − CLEAR). Hatched areas indicate regions where the calculated differences reach 95% statistical significance.

  • View in gallery

    RMSE (kg m−2) of TCWV forecasts from all experiments from verification against the operational model.

  • View in gallery

    Relative RMSE (% cf. BL = 100%) of TCWV forecasts from all experiments after (a) 24 and (b) 72 h.

  • View in gallery

    RMSE (%) of relative humidity (R) forecasts from all experiments stratified (top to bottom) by level at 200, 500, 700, and 1000 hPa and (left to right) by latitude between the Southern Hemisphere, tropics, and Northern Hemisphere from verification against the operational model.

  • View in gallery

    As in Fig. 6, but for vector wind (VW; m s−1).

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 24 24 1
PDF Downloads 12 12 0

Impact of SSM/I Observations Related to Moisture, Clouds, and Precipitation on Global NWP Forecast Skill

View More View Less
  • 1 European Centre for Medium-Range Weather Forecasts, Reading, United Kingdom
© Get Permissions
Full access

Abstract

This paper presents the results from the Observing System Experiments (OSEs) with the current ECMWF data assimilation and modeling system for quantifying the impact on both analysis and forecast quality of Special Sensor Microwave Imager (SSM/I) observations sensitive to moisture and clouds as well as precipitation. SSM/I radiances have been assimilated operationally in clear-sky areas for 8 yr and in cloud- and rain-affected areas since June 2005. This paper examines experiments set up such that clear-sky and rain-affected observations were either added to a baseline with a restricted observing system configuration or withdrawn from the full system. The experiment duration was 10 weeks of which the first 14 days were excluded from the evaluation to allow the system to lose the memory of the initial conditions at day −1.

It is shown that both clear-sky and rain-affected observations account for the bulk correction of moisture in the ECMWF analysis. SSM/I data adds 1 day of forecast skill over the first 48 h when evaluated in addition to a baseline-observing system. In the tropics, the rain-affected data contributes more skill to the moisture forecast than the clear-sky data at 700 hPa and above. In the Northern and Southern Hemispheres, the effect is generally weaker and slightly in favor of clear-sky observations. A similar performance can be seen with respect to the wind vector forecast skill, which reflects the connection between the analysis of moisture and dynamics.

Corresponding author address: Graeme Kelly, European Centre for Medium-Range Weather Forecasts, Shinfield Park, Reading, Berkshire RG2 9AX, United Kingdom. Email: graeme.kelly@ecmwf.int

Abstract

This paper presents the results from the Observing System Experiments (OSEs) with the current ECMWF data assimilation and modeling system for quantifying the impact on both analysis and forecast quality of Special Sensor Microwave Imager (SSM/I) observations sensitive to moisture and clouds as well as precipitation. SSM/I radiances have been assimilated operationally in clear-sky areas for 8 yr and in cloud- and rain-affected areas since June 2005. This paper examines experiments set up such that clear-sky and rain-affected observations were either added to a baseline with a restricted observing system configuration or withdrawn from the full system. The experiment duration was 10 weeks of which the first 14 days were excluded from the evaluation to allow the system to lose the memory of the initial conditions at day −1.

It is shown that both clear-sky and rain-affected observations account for the bulk correction of moisture in the ECMWF analysis. SSM/I data adds 1 day of forecast skill over the first 48 h when evaluated in addition to a baseline-observing system. In the tropics, the rain-affected data contributes more skill to the moisture forecast than the clear-sky data at 700 hPa and above. In the Northern and Southern Hemispheres, the effect is generally weaker and slightly in favor of clear-sky observations. A similar performance can be seen with respect to the wind vector forecast skill, which reflects the connection between the analysis of moisture and dynamics.

Corresponding author address: Graeme Kelly, European Centre for Medium-Range Weather Forecasts, Shinfield Park, Reading, Berkshire RG2 9AX, United Kingdom. Email: graeme.kelly@ecmwf.int

1. Introduction

Clouds and precipitation strongly affect the global hydrology and energy cycles, mainly through the interaction of solar and infrared radiation with cloud droplets and the release of latent heat in precipitation development. The accurate representation of clouds and precipitation in numerical weather prediction (NWP) systems is crucial for daily forecasting, in particular in case of extreme weather conditions. Many catastrophic weather events are associated with extreme winds but also extreme precipitation. On climatological time scales, even droughts must be considered as weather extremes through the systematic and long-term anomalous lack of precipitation.

About 3 million satellite observations are actively assimilated per a 12-h analysis cycle in the four-dimensional variational data assimilation (4DVAR) system at the European Centre for Medium-Range Weather Forecasts (ECMWF; model cycle 30R1, June 2006) composing about 98% of the observational data volume. In general, radiance observations are only assimilated in clear skies because the forward modeling of radiance signatures of clouds and precipitation is much less accurate than in clear skies. Most crucially affected by the presence of clouds are observations that are sensitive to the mid- to lower troposphere and, in particular, to atmospheric moisture. Recent observing system experiments (Andersson et al. 2007) suggest that, over sea, the most dominant humidity observations in the analysis originate from the Special Sensor Microwave Imager (SSM/I) and the Advanced Microwave Sounding Unit (AMSU-B). Over land, radiosondes, ground-based stations, and AMSU-B observations dominate. In the upper troposphere, also infrared sounding channels from the High-Resolution Infrared Radiation Sounder (HIRS), the Atmospheric Infrared Sounder (AIRS), and radiometers on board geostationary satellites provide significant information on humidity to the atmospheric analysis.

When rainfall observations from the Tropical Rainfall Measuring Mission (TRMM) Precipitation Radar (PR) are analyzed, regional differences in precipitation amount mainly depends upon rainfall occurrence rather than rain intensity (C. Kidd 2007, personal communication; sample over all PR observations in period 1998–2006). In the tropical convergence zones, rainfall occurs at a rate of 8% or more while in areas with larger-scale precipitation systems, rainfall occurrence is of the order of 5%. These values apply only to the original spatial resolution of the PR, which is 4 km. If they are averaged to the current approximate spatial resolution of the ECMWF model, rainfall occurrence reaches about 25% and more in areas where convective precipitation regimes are dominant. This means that for data assimilation systems that screen out cloud and precipitation-affected observations a number of undesirable effects occur. First, in particular over Southern Hemisphere oceans, large areas are not well covered by observations that are sensitive to humidity so that the moisture field is mainly constrained by the model background (i.e., short-range forecast) at each analysis cycle. Second, as seen in the 40-yr ECMWF Re-Analysis (ERA-40; Uppala et al. 2005) an underestimation of atmospheric moisture by the model together with positive humidity increments from observations led to an imbalance in the hydrological budget that was accompanied by precipitation spinup. Recently, improvements in model physics and the data assimilation system have helped to substantially reduce these problems (Andersson et al. 2005).

Apart from general predictability issues with regard to clouds and precipitation, the forecasting of processes involving the hydrological cycle is generally much less accurate than that of dynamical fields. This is mainly because the moist physics parameterizations represent diabatic processes at model grid scales in a simplified way and, for example, lack an explicit description of moist convection. Also, moist processes strongly depend on less well-predicted dynamical parameters like vertical wind (Tompkins and Janisková 2004; Lopez and Moreau 2005).

Despite the above underlying problems, both the continuous availability of accurate moisture observations and the progress in physical modeling and data assimilation system development have paved the way for the use of cloud and precipitation affected observations in modern NWP systems (e.g., Mahfouf et al. 2001). While these new observations represent greater scientific challenges they should produce an improvement of NWP skill due to the lack of competing observations in cloudy areas over large parts of the globe.

Since weather system development at lower latitudes is equally dominated by convection and large-scale dynamics, the first attempts to produce better moist initial conditions for model forecasts were made through adjustments of diabatic heating using infrared sounder data (e.g., Heckley et al. 1990) and moisture initialization and heating adjustments from outgoing longwave radiation (OLR) observations (Puri and Miller 1990). The development of more complex physical initialization schemes included the use of OLR and retrieved surface rain rates from satellite data to constrain surface fluxes, moisture flux profiles, and their feedback with convection and radiation (Krishnamurti et al. 1984; Krishnamurti and Bedi 1996).

The advent of variational data assimilation schemes involving adjoint techniques initiated the employment of more complex and more consistent model physics in model initialization using cloud and precipitation observations. Apart from a number of experimental and case study–oriented methods only a few implementations in operational forecasting systems have succeeded so far (Tsuyuki et al. 2003; Treadon et al. 2003; Bauer et al. 2006a, b). This is because of a number of fundamental problems associated with the assimilation of rain-affected observations, some of which can only be pragmatically solved (Errico et al. 2007).

The ECMWF version of rainfall assimilation represents a unique step forward in that it is based on the variational assimilation of passive microwave radiance observations and that it performs the assimilation with an operational 4DVAR system on a global scale (Bauer et al. 2006a, b). Its original methodology was based on the use of near-surface rainfall rate retrievals derived from TRMM Microwave Imager (TMI) observations and was introduced by Marécal and Mahfouf (2000, 2002). The term 1D + 4DVAR refers to a one-dimensional variational data assimilation retrieval (1DVAR) of total column water vapor (TCWV) in the presence of clouds and precipitation followed by a 4DVAR assimilation of TCWV in the full system. The developments and experimental studies of Marécal and Mahfouf were instrumental for the extension of the system from derived rain rates to SSM/I radiances (Moreau et al. 2004) and for the understanding of this new observation type in an operational 4DVAR system (Marécal and Mahfouf 2003; Mahfouf et al. 2005).

The operational 1DVAR retrieval employs temperature and moisture profiles on ECMWF model levels as well as the 10-m zonal and meridional wind components in the control vector. Only the 19.35- and 22.235-GHz SSM/I channels are used as observations to avoid excessive errors associated with the modeling of frozen hydrometeors and strong particle scattering in the observation operator. Background errors are taken from the operational 4DVAR formulation while SSM/I instrument and observation operator errors are determined from the statistical analysis of spatial covariances of observation minus modeled radiance departures. More details of 1D + 4DVAR design and implementation are given by Bauer et al. (2006a, b). The system has worked successfully since its operational implementation in June 2005 and has undergone several upgrades.

Originally, clear-sky SSM/I data was also used in a one-dimensional variational retrieval of total column water vapor and near-surface wind speed in the ECMWF system (Phalippou 1996; Gérard and Saunders 1999) that was later replaced by the direct radiance assimilation in 4DVAR (Bauer et al. 2002). In a 12-h assimilation cycle, 5000 clear-sky and 10 000 cloud/rain-affected observations are used. However, clear-sky data is more heavily thinned and the seven SSM/I channels are treated as independent observations. This means that the above data samples refer to true data points with respect to the rain-affected observations and to data point multiplied by number of channels in case of the clear-sky observations.

This paper presents the first systematic performance analysis of such a system in global operational applications and is based on Observing System Experiments (OSEs) using the current ECMWF data assimilation and forecasting system. The rain assimilation system is briefly introduced in section 2 together with the description of the experimental setup. The forecast performance is presented in section 3 and conclusions are drawn in section 4.

2. Observing System Experiments

a. Data assimilation system

ECMWF employs a multi-incremental 4DVAR data assimilation system that is run in two different analysis suites twice per day. One suite runs 4DVAR over a 12-h window and produces analyses at 0000 and 1200 UTC, respectively. From these, short-range forecasts are generated that serve as model first-guess estimates of the atmospheric state for a second suite, which runs two analyses per day over 6-h windows centered at 0000 and 1200 UTC. This second suite initializes the global medium-range forecasts. Details of the ECMWF 4DVAR system can be found in Rabier et al. (2000), Mahfouf and Rabier (2000), and Klinker et al. (2000) while the latest analysis suite setup is described by Haseler (2004).

Currently, the high-resolution analyses and forecasts are produced at 25-km spatial resolution (spectral wavenumber cutoff at 799) and on 91 model levels. For the sake of efficiency, the OSEs presented in this paper were run at lower resolution [i.e., 40 km (spectral wavenumber cutoff at 511)] and on 60 model levels. In the minimization, the spatial resolution of the tangent-linear and adjoint models is increased in successive inner loops (minimization steps) from a wavenumber cutoff at 95–159 (e.g., Trémolet 2004).

A large number of observations are currently assimilated at ECMWF, comprising conventional and, for the largest part, satellite observations. Conventional observations currently assimilated in the system include radiosondes, pilots and wind profilers, synoptic observations, data from ships and buoys (moored and drifters), as well as aircraft providing reports during ascent–descent. Satellite observations include Atmospheric Motion Vectors (AMVs) derived from geostationary platform data [Meteorological Satellite-5/8 (Meteosat-5/8) and Geostationary Operational Environmental Satellite-9/10/12 (GOES-9/10/12)] and low-Earth orbiting satellites [the Moderate Resolution Imaging Spectroradiometer (MODIS) on board Terra and Aqua]; clear-sky water vapor radiances from Meteosat-5/8 and GOES-9/10/12; infrared radiances from the National Oceanic and Atmospheric Administration Satellite-17 (NOAA-17) (HIRS) and Aqua (AIRS); microwave radiances from NOAA-15 (AMSU-A), NOAA-16 (AMSU-A and AMSU-B), NOAA-17 (AMSU-B), NOAA-18 (AMSU-A, MHS), Aqua (AMSU-A); and the Defense Meteorological Satellite Program (DMSP) satellites DMSP-F13/14/15 (SSM/I); sea surface winds from scatterometers on board the Quick Scatterometer (QuikSCAT) and the European Remote Sensing Satellite-2 (ERS-2); and ozone products from NOAA-16 [the Solar Backscatter Ultraviolet (SBUV) instrument] and the Environmental Satellite (Envisat) [the Scanning Imaging Absorption Spectrometer for Atmospheric Cartography (SCIAMACHY)].

b. Experiments

During its meeting at ECMWF on 3 May 2003, the European Meteorological Network (EUMETNET) Composite Observing System (EUCOS) Scientific Advisory Team discussed the need to investigate the interdependencies between the space-based and terrestrial components of the observing system. It was suggested that such an investigation could be based on a set of carefully designed OSEs. Studies should be designed so as to provide guidance on the future development of the terrestrial-observing system in view of the increasing capabilities of the satellite-observing systems provided by the meteorological space agencies (Andersson et al. 2004). The following experiments were designed accordingly and in agreement with EUMETSAT.

All experiments were run for the period of 1 June to 15 August 2006. The initial 14-day period was excluded from the verification to allow for the system to spin up and to eliminate the effect of the operational model that initialized the first analysis. Therefore, all following results refer to the period 15 June to 15 August 2006.

The introduction of the variational bias correction (VarBC; Dee 2005; Auligné et al. 2007) allows the bias correction to adapt to a very different observational configuration used in these experiments, but allows unwanted feedbacks between observations and the climate of the model. To avoid this VarBC was only activated for the initial 14-day period and bias corrections were frozen for the remaining time. In general, the 4DVAR system is based on the assumption that the differences between observations and model-simulated parameters are not affected by biases and that their probability distributions have Gaussian shapes so that error properties can be described by error covariance statistics. Since both observations and model contain biases due to technical deficiencies as well as model inaccuracies, observation system characteristics as well as model variables are used as predictors in bias-correction schemes. The basic principle of VarBC is to include biases in the 4DVAR control vector such that the analysis derives an optimal model state that also minimizes biases. VarBC may also be affected by spinup if the observation system is drastically changed, and this is certainly the case in the OSEs presented here. The time for spinup depends on the sensor and which parameter it is sensitive to. For moisture related observations, the spinup takes a few days while for atmospheric temperature and in the mid- and higher atmosphere the spinup evolution can be much slower. To avoid such effects no bias-predictor spinup was allowed after the initial 14-day warm-up period (i.e., all bias corrections were frozen after the initial period).

The following experiments have been performed:

  1. BL: A baseline configuration that includes all conventional observations as well as AMSU-A radiance observations from NOAA-16. The latter was found to be important because the sparsity of conventional observations over the Southern Hemisphere and tropics produces very poor analyses of the mass field otherwise. This is less crucial when OSEs evaluate observations sensitive to temperature. However, moisture-related observations produce less impact on the large-scale atmospheric dynamics so that reasonable constraints must be produced by one additional satellite instrument.
  2. CTRL: A control run with the full operational observation system.
  3. BL + RAIN: Baseline + SSM/I 1D + 4DVAR rain observations from DMSP-F13/14/15. The data is thinned to one observation per 1.25° box and model time step, which produces a total number of about 50 000 rain observations per 12-h analysis window.
  4. BL + CLEAR: Baseline + SSM/I clear-sky radiance observations from DMSP-F13/14/15. The data is subsampled and thinned as for BL + RAIN so that roughly 90 000 clear-sky moisture observations (=3 DMSP satellites, observation points times number of channels) remain per 12-h analysis window.
  5. BL + RAIN + CLEAR: Baseline + SSM/I clear-sky and rain-affected observations.
  6. CTRL − RAIN: Full operational observation system, but with SSM/I 1D + 4DVAR rain observations withdrawn.
  7. CTRL − CLEAR: Full operational observation system, but with SSM/I clear-sky radiance observations withdrawn.

This setup allows the evaluation of the impact of data addition and denial on forecast skill. Given the rather well-constrained operational analysis and the rather poor baseline configuration, the denial of data will exhibit much less impact than the addition of data. However, due to the nonlinear response of the system to data addition and denial a consistent evaluation at both ends of the scale is possible. The assessment of the individual and combined contribution of SSM/I clear-sky and rain-affected observations helps understanding the degree of redundancy and complementarity between both observation types.

3. Results

a. Analysis

1) Data addition experiments

The fundamental impact of clear-sky SSM/I observations on the lower-tropospheric humidity analysis over oceans has been demonstrated by earlier OSEs with previous ECMWF model versions and the available observing systems at that time (Andersson et al. 2007). From these, Andersson et al. concluded that the SSM/I provides the strongest observational contribution to moisture below 500 hPa.

Above that level, sounder-type sensors with better sensitivity to mid- and upper-tropospheric moisture, such as the AMSU-B and the water vapor sounding channels on the geostationary imagers, take over. However, the bulk of the total atmospheric moisture is located below 500 hPa, it is therefore fair to say that the TCWV analysis over oceans is mainly driven by SSM/I observations.

Figure 1a shows the mean analysis difference in TCWV between CTRL and BL (in percent of TCWV from BL). Since SSM/I observations are only assimilated over oceans, the impact is small over large continental areas like North America and Eurasia. The SSM/I data tends to systematically moisten the largest part of the Northern Hemispheric oceans at latitudes below 50°. In the Southern Hemisphere (winter hemisphere), the effect is reversed and focused on latitudes between the equator and 30°S. Figures 1b–d demonstrate that both clear-sky and rain-affected SSM/I observations (i.e., BL + RAIN, BL + CLEAR, and BL + RAIN + CLEAR) work very similarly on the BL moisture distribution. However, BL + CLEAR seems to emphasize the moistening in northern oceans and the drying in southern oceans while the rain assimilation tends toward more moistening in the southern subtropical areas and drying in the Northern Pacific. The combined effect (Fig. 1d) is very similar to the total effect of all observations (Fig. 1a), highlighting the dominance of the SSM/I for the TCWV analysis.

One substantial difference between adding clear-sky and cloud/rain-affected observations to BL is the impact over land. Only BL + CLEAR minus BL (Fig. 1c) does not show substantial moisture analysis changes over land from the oceanic observations. Most likely, the reason is in the stronger impact of cloud/rain-affected observations on vertical motion and therefore on atmospheric circulation. In the Northern Hemisphere summer, there is strong moisture advection from the central Atlantic into the northern part of South America and from the western Indian Ocean into Central Africa producing a maritime precipitation climate in these areas (Kållberg et al. 2005, their Fig. C3, p. 36). This feature is amplified when cloud/rain-affected observations are added to BL. Over northern Africa, the moisture flux is directed from the eastern Mediterranean Sea into Egypt. The relative moisture increase in this area as well as the moisture decrease over the western Sahara from adding cloud/rain-affected observations, however, may not be considered significant because in this period atmospheric moisture contents reach their annual minimum so that relative increments may be substantial but absolute increments are small.

Figure 2 shows the zonal cross sections of analysis differences for relative humidity, R, in percent. The moistening with respect to BL in experiments BL + RAIN, BL + CLEAR, and BL + RAIN + CLEAR in the Northern Hemispheric tropics and subtropics extends throughout the entire troposphere. The BL + CLEAR moistens a shallow layer near 850 hPa (roughly 1.5-km altitude), while BL + RAIN moistens a deeper later through the intensification of the vertical moisture transport by convection. The vertical profile of increments is strongly driven by the shape of the humidity control variable background error standard deviation function. This peaks near 850 hPa and drops off rather sharply toward the surface and more smoothly toward the top of the atmosphere. The weak drying in the Southern Hemisphere is mainly produced by clear-sky observations and is weakened by the rain assimilation, mainly at levels above the boundary layer. The CTRL is much drier than BL in much of the tropics as seen in Fig. 2d. The combined effect of CLEAR and RAIN shown in Fig. 2a does not feature such extensive drying. This must therefore be the impact of other observations sensitive to moisture.

2) Data denial experiments

The effect on the average moisture analysis changes rather dramatically when SSM/I observations are withdrawn from CTRL. While the previous results were related to baseline experiments without moisture-sensitive observations, the denial experiments refer to a full observing system where only SSM/I data has been taken out. Figure 3 shows the effect on TCWV of the denial of rain-affected and clear-sky observations as differences with respect to CTRL. The differences have been normalized by CTRL and are presented in units of percent.

As already shown in Bauer et al. (2006b), large areas of systematic drying are produced when rain-affected SSM/I observations are assimilated (Fig. 3a). These patterns are statistically significant over large parts of the summer hemisphere with maximum values of 5%–10% in the Northern Pacific. The only large-scale moistening is found in the subtropical subsidence zones south of the ITCZ. These moisture differences are associated with mean sea level pressure (MSLP) and lower-level divergence changes in the analysis as also shown by Bauer et al. (2006b; only divergence shown). Drying coincides with MSLP increase and divergence while moistening produces a MSLP decrease and convergence. However, only little interaction with vertical motion could be identified. Neither short-range forecasts of convective available potential energy (CAPE) nor precipitation seemed to capture the initial moisture analysis signal. Only high cloud cover was found to systematically decrease (order 10%–20%) along the ITCZ near 10°–15°N along with the TCWV reduction.

When clear-sky SSM/I data is withdrawn (Fig. 3b), TCWV analysis patterns show a different behavior with patchy structures in the Northern Hemisphere and predominantly dryer analyses over the Southern Hemisphere oceans. These areas coincide with generally dry air masses and little cloud and precipitation occurrence at this time of the year. Evidently, no direct link between TCWV changes and divergence, clouds, instability, or precipitation could be found. It must be concluded that both observation types mainly work on systematic model biases that are related to areas where moist diabatic processes dominate (RAIN) or in areas well below saturation and with little cloud presence (CLEAR). The stronger impact of rain-affected observations on divergence indicates that, potentially, model biases are larger and that fewer observation types are available in these areas. The lack of direct impact of the moisture analysis on cloud and precipitation short-range forecasts suggests that the physics parameterizations dissipate the signal rather quickly.

b. Forecast

The first forecast evaluation is performed for TCWV. Figure 4 shows the summary of global root-mean-square (RMS) forecast errors for all experiments until day 6. The RMS errors are calculated by comparing the forecasts with the operational ECMWF analyses (Haseler 2004). When evaluating forecasts against analyses, the choice of analysis can affect the results. The operational, rather than the CTRL analyses, were chosen for their relative independence from the OSE experiment configuration so as to avoid spuriously amplifying the apparent impact of the observing system.

For both addition and denial experiments, the impact of SSM/I observations on TCWV forecast scores is always positive and slightly larger for clear-sky than for rain-affected observations. This is consistent with observing system experiments that have been performed over the continental United States. Lopez and Bauer (2007) replaced moisture-sensitive conventional and satellite observations with TCWV retrievals based on Next Generation Weather Radar (NEXRAD) rainfall observations and found significant improvement in both moisture analysis and forecast.

In Fig. 4, the gap between TCWV root-mean-square errors (RMSEs) from BL and BL + RAIN or BL + CLEAR is rather substantial and amounts to a forecast improvement of about 1 day over the first 48 h. This means that the skill of both BL + RAIN and BL + CLEAR at day 2 is as high as for BL at day 1. The combined effect of BL + RAIN + CLEAR adds another 3 h of skill. In the denial experiments, the loss of skill from withdrawing SSM/I data on TCWV forecast skill is visible but not significant in global terms beyond 24–48 h.

Figure 5 breaks up the global errors into zonal cross sections and displays relative RMSE with respect to the worst case (i.e., BL). This illustrates the improvement of all other experiments with respect to a poor system after 1 and 3 days, respectively. It becomes obvious that the strongest impact on forecast skill from adding clear-sky or rain-affected observations is in the tropics roughly between ±30° latitude but also in the Southern Hemisphere where the predominance of oceans allows the assimilation of much SSM/I data. Here, the error reduction with respect to BL amounts to 30% for day 1 and reduces to 10% and a less strong zonal gradient at day 3. In the tropics, both BL + RAIN and BL + CLEAR perform similarly while BL + CLEAR shows a stronger improvement at higher latitudes. In the denial experiments, withdrawing clear-sky observation always reduces TCWV day −1 forecast skill more (5% of error reduction) than withdrawing rain-affected observations; however, in the tropics their effect is rather similar. The effect is still noticeable at day 3 but rather insignificant.

If the forecast errors are stratified by level and latitude bands for relative humidity (Fig. 6) and vector winds VW (Fig. 7) the following picture emerges:

  1. The impact of data addition–denial experiments is much larger in the tropics (here between ±20° latitude). Outside the tropics, the impact is rather small but stronger in the Southern Hemisphere than in the Northern Hemisphere (i.e., the difference between hemispheres is due to the better constraint of the dynamics and humidity analysis by conventional observations over land surfaces in the Northern Hemisphere, as indicated by the smaller gap between BL and CTRL forecast scores).
  2. For R (Fig. 6), BL + RAIN performs better than BL + CLEAR at 200 and 500 hPa while they are very similar at 700 and 1000 hPa. The positive impact of BL + RAIN over a deeper atmospheric layer between boundary layer and tropopause is caused by the involvement of convection. The vertical distribution of the impact of BL + CLEAR is more affected by the shape of the background moisture errors (which peak between 700 and 950 hPa) than the moist physics because the observations are located in areas away from saturation.
  3. The combined effect in BL + RAIN + CLEAR is always equal to or better than either BL + CLEAR or BL + RAIN. This indicates that both observation types are mostly complementary despite small opposite mean contributions in areas like the southern subtropics (Fig. 1). This improvement increases toward the surface.
  4. For VW (Fig. 7), a similar behavior as for R can be observed. The improvement from BL to BL + CLEAR/BL + RAIN is still quite large demonstrating that moisture observations can have a direct and significant impact on wind forecasts (see also Mahfouf et al. 2005; Bauer et al. 2006b). As for TCWV and R, the gap between BL and BL + CLEAR/ BL + RAIN is about 1 day and more over the first 48 h. The gap is slightly reduced but significant for the BL-related experiments while it disappears for the CTRL-related experiments after day 3. This relates to the length of memory of the full system with respect to moisture observations at the initial time.
  5. If the verification statistics of Figs. 4 –7 are recalculated (not shown) having removed the June–August mean bias for TCWV (e.g., those shown in Fig. 1, and their equivalents in the forecasts), the RMSE does reduce a little in all experiments. However, there is no significant change to the shape of the results, except for tropical temperature and geopotential (not shown), where the results become very similar to those seen in wind vector and relative humidity. In other words, mean biases have no effect on wind vector and relative humidity, but do affect tropical temperature scores.

4. Conclusions

This paper presents the results from Observing System Experiments with the current ECMWF data assimilation and modeling system. The experiments were aimed at quantifying the impact of SSM/I observations sensitive to clear-sky moisture, to near-surface wind speed and, most importantly, to clouds and precipitation, on both analysis and forecast quality. The assimilation of rain-affected observations was implemented in operations in June 2005 and can be considered a major accomplishment for the first time adding satellite observations in otherwise data-sparse regions. The experiments were set up such that both clear-sky and rain-affected observations were either added to a poor baseline-observing system configuration or withdrawn from the full system. Experiment duration was 10 weeks of which the first 14 days were excluded from the evaluation to allow for model spinup.

The basic impact on the analysis is demonstrated by adding data to the baseline configuration. Together, clear-sky and rain-affected observations account for the bulk correction of moisture in the analysis, such that their combined effect almost entirely compensates for the humidity errors contained in the baseline when it is compared to the full system.

The impact of data denial and addition on forecast skill is significant. In terms of total column water vapor, SSM/I data adds 1 day of skill over the first 48 h. In the tropics, the rain-affected data contributes more skill to the moisture forecast than the clear-sky data for levels above 700 hPa while it is similar to clear-sky data below. In the Northern and Southern Hemispheres the effect is generally weaker and slightly in favor of clear-sky observations. A similar performance can be seen with respect to wind vector forecast skill that suggests the connection between moisture analysis and vertical motions in the tropics.

These results clearly demonstrate the important and unique role of microwave imager data in modern numerical weather prediction systems. Further developments for optimizing this observing system will be devoted to extending data coverage by adding other imagers such as the Special Sensor Microwave Imager Sounder (SSMI/S), Advanced Microwave Scanning Radiometer (AMSR-E), and the Tropical Rainfall Measuring Mission (TRMM) Microwave Imager (TMI) as well as by exploiting the use of combined microwave imager sounder data over land surfaces. This study also served as demonstration of the beneficial impact of rain-affected radiance data in modern forecasting systems and supports the requirement of adding data in areas where current systems have been only weakly constrained by observations.

The experiments have been performed with the latest version of the ECMWF modeling and data assimilation system and the denial experiments withdraw SSM/I data from a suite of about 30 different satellite instruments. Due to the limited interaction between moisture and other control variables in the analysis system, it was important to employ AMSU-A radiance observations from NOAA-16 in our baseline experiment to adjust the large-scale dynamic structures in the atmosphere. Otherwise, moisture observations alone may not show beneficial impact in data sparse areas like the Southern Hemisphere. Hence, the baseline-related experiments clearly show the strong contribution of SSM/I observations to forecast skill and the impact of humidity observations on forecast winds in the tropics well into the medium range. Therefore, the results have to be interpreted in this context when compared to those published by Bengtsson and Hodges (2005) and Andersson et al. (2007).

Andersson et al. (2007) show that the principal contribution to lower-tropospheric moisture over oceans originated from clear-sky SSM/I data, with the tendency for SSM/I observations to add moisture overall. They also showed that, in general, humidity observations were responsible for improved forecast scores, in marked contrast to the results of Bengtsson and Hodges (2005). While Andersson et al. did not consider the assimilation of rain-affected SSM/I observations, Bauer et al. (2006b) demonstrated that their assimilation improved forecasts of moisture in the tropics at 700 hPa, but all other forecast impacts were rather neutral.

The current study makes use of low baseline experiments, meaning that for the first time the impact of observation types can be properly examined in isolation, as well as within the complete observing system. Experiments were run for a longer period, making them more reliable, and the chosen approach toward bias correction eliminates one possible source of contamination from the results. The results for clear-sky SSM/I are generally consistent with those of Andersson et al. However, it was shown in this paper that the impact of rain-affected observations is very similar to the clear-sky observations, and like them, the impact is strong in the tropics, improving not just moisture but also the wind fields.

In the data denial experiments, the impact of SSM/I data withdrawal from the full observing system was also found to be mainly relevant in the tropics even though its magnitude was smaller and most effective over the first 1–2 days. This is partly explained by the number of other satellite observations that are sensitive to moisture, present in all denial and control experiments as well as the tendency for moisture changes to be rained out in the first few days of the forecast.

Acknowledgments

This study was mainly funded by the European Organisation for the Exploitation of Meteorological Satellites (EUMETSAT) through Contract EUM/MET/SOW/04/0290 and the EUMETSAT fellowship programme.

REFERENCES

  • Andersson, E., , R. Dumelow, , H. Huang, , J-N. Thépaut, , and A. Simmons, 2004: Space/terrestrial link. Outline study proposal for consideration by EUCOS management. (Available from EUCOS Secretariat, DWD Kaiserleistr, 29/35, 63067 Offenbach, Germany.).

  • Andersson, E., and Coauthors, 2005: Assimilation and modeling of the atmospheric hydrological cycle in the ECMWF forecasting system. Bull. Amer. Meteor. Soc., 86 , 387402.

    • Search Google Scholar
    • Export Citation
  • Andersson, E., , E. Hólm, , P. Bauer, , A. Beljaars, , G. A. Kelly, , A. P. McNally, , A. J. Simmons, , and J-N. Thépaut, 2007: Analysis and forecast impact of the main humidity observing systems. Quart. J. Roy. Meteor. Soc., 133 , 14731485. doi: 10.1002/qj.112.

    • Search Google Scholar
    • Export Citation
  • Auligné, T., , A. P. McNally, , and D. P. Dee, 2007: Adaptive bias correction for satellite data in a numerical weather prediction system. Quart. J. Roy. Meteor. Soc., 133 , 631642.

    • Search Google Scholar
    • Export Citation
  • Bauer, P., , G. Kelly, , and E. Andersson, 2002: SSM/I radiance assimilation at ECMWF. Proc. ECMWF/GEWEX Workshop on Humidity Analysis, Reading, United Kingdom, ECMWF, 167–175. [Available from the European Centre for Medium-Range Weather Forecasts, Shinfield Park, Reading RG2 9AX, United Kingdom.].

    • Search Google Scholar
    • Export Citation
  • Bauer, P., , P. Lopez, , A. Benedetti, , D. Salmond, , and E. Moreau, 2006a: Implementation of 1D+4D-Var assimilation of precipitation affected microwave radiances at ECMWF. Part I: 1D-Var. Quart. J. Roy. Meteor. Soc., 132 , 22772306.

    • Search Google Scholar
    • Export Citation
  • Bauer, P., , P. Lopez, , A. Benedetti, , D. Salmond, , S. Saarinen, , and M. Bonazzola, 2006b: Implementation of 1D+4D-Var assimilation of precipitation affected microwave radiances at ECMWF. Part II: 4D-Var. Quart. J. Roy. Meteor. Soc., 132 , 23072332.

    • Search Google Scholar
    • Export Citation
  • Bengtsson, L., , and K. I. Hodges, 2005: On the impact of humidity observations in numerical weather prediction. Tellus, 57A , 701708.

  • Dee, D. P., 2005: Bias and data assimilation. Quart. J. Roy. Meteor. Soc., 131 , 33233343.

  • Errico, R., , P. Bauer, , and J-F. Mahfouf, 2007: Issues regarding the assimilation of cloud and precipitation data. J. Atmos. Sci., 64 , 37853798.

    • Search Google Scholar
    • Export Citation
  • Gérard, E., , and R. Saunders, 1999: Four-dimensional variational assimilation of Special Sensor Microwave/Imager total column water vapour in the ECMWF model. Quart. J. Roy. Meteor. Soc., 125 , 30773102.

    • Search Google Scholar
    • Export Citation
  • Haseler, J., 2004: The early-delivery suite. ECMWF Tech. Memo. 454, 35 pp. [Available from European Centre for Medium-Range Weather Forecasts, Shinfield Park, Reading RG2 9AX, United Kingdom.].

  • Heckley, W. A., , G. Kelly, , and M. Tiedtke, 1990: On the use of satellite-derived heating rates for data assimilation within the tropics. Mon. Wea. Rev., 118 , 17431757.

    • Search Google Scholar
    • Export Citation
  • Kållberg, P., , P. Berrisford, , B. Hoskins, , A. Simmons, , S. Uppala, , S. Lamy-Thépaut, , and R. Hine, 2005: ERA-40 Atlas. ECMWF Re-Analysis Tech. Rep 19, ECMWF, 191 pp. [Available from European Centre for Medium-Range Weather Forecasts, Shinfield Park, Reading RG2 9AX, United Kingdom.].

    • Search Google Scholar
    • Export Citation
  • Klinker, E., , F. Rabier, , G. Kelly, , and J-F. Mahfouf, 2000: The ECMWF operational implementation of four-dimensional variational assimilation. Part III: Experimental results and diagnostics with operational configuration. Quart. J. Roy. Meteor. Soc., 126 , 11911215.

    • Search Google Scholar
    • Export Citation
  • Krishnamurti, T. N., , and H. S. Bedi, 1996: A brief review of physical initialization. Meteor. Atmos. Phys., 60 , 137142.

  • Krishnamurti, T. N., , K. Ingles, , S. Cooke, , T. Kitade, , and R. Pash, 1984: Details of low-latitude medium-range numerical weather prediction model using a global spectral model. Part 2: Effects of orography and physical initialization. J. Meteor. Soc. Japan, 62 , 613648.

    • Search Google Scholar
    • Export Citation
  • Lopez, P., , and E. Moreau, 2005: A convection scheme for data assimilation: Description and initial tests. Quart. J. Roy. Meteor. Soc., 131 , 409436.

    • Search Google Scholar
    • Export Citation
  • Lopez, P., , and P. Bauer, 2007: 1D+4D-Var assimilation of NCEP Stage IV radar and gauge hourly precipitation data. Mon. Wea. Rev., 135 , 25062524.

    • Search Google Scholar
    • Export Citation
  • Mahfouf, J-F., , and F. Rabier, 2000: The ECMWF operational implementation of four-dimensional variational assimilation. Part II: Experimental results with improved physics. Quart. J. Roy. Meteor. Soc., 126 , 11711190.

    • Search Google Scholar
    • Export Citation
  • Mahfouf, J-F., , V. Marécal, , M. Janisková, , F. Chevallier, , and J. Barkmeijer, 2001: Towards the use of satellite data for cloud and rain assimilation. Proc. ECMWF Seminar on Exploitation of the New Generation of Satellite Instruments for Numerical Weather Prediction, Reading, United Kingdom, ECMWF, 267–287. [Available from European Centre for Medium-Range Weather Forecasts, Shinfield Park, Reading RG2 9AX, United Kingdom.].

  • Mahfouf, J-F., , V. Marécal, , and P. Bauer, 2005: The assimilation of SSM/I and TMI rainfall rates in the ECMWF 4D-Var system. Quart. J. Roy. Meteor. Soc., 131 , 437458.

    • Search Google Scholar
    • Export Citation
  • Marécal, V., , and J-F. Mahfouf, 2000: Variational retrieval of temperature and humidity profiles from TRMM precipitation data. Mon. Wea. Rev., 128 , 38533866.

    • Search Google Scholar
    • Export Citation
  • Marécal, V., , and J-F. Mahfouf, 2002: Four-dimensional variational assimilation of total column water vapor in rainy areas. Mon. Wea. Rev., 130 , 4358.

    • Search Google Scholar
    • Export Citation
  • Marécal, V., , and J-F. Mahfouf, 2003: Experiments on 4D-Var assimilation of rainfall data using an incremental formulation. Quart. J. Roy. Meteor. Soc., 129 , 31373160.

    • Search Google Scholar
    • Export Citation
  • Moreau, E., , P. Lopez, , P. Bauer, , A. M. Tompkins, , M. Janisková, , and F. Chevallier, 2004: Variational retrieval of temperature and humidity profiles using rain rates versus microwave brightness temperatures. Quart. J. Roy. Meteor. Soc., 130 , 827852.

    • Search Google Scholar
    • Export Citation
  • Phalippou, L., 1996: Variational retrieval of humidity profile, wind speed and cloud liquid-water path with the SSM/I: Potential for numerical weather prediction. Quart. J. Roy. Meteor. Soc., 122 , 327355.

    • Search Google Scholar
    • Export Citation
  • Puri, K., , and M. J. Miller, 1990: The use of satellite data in the specification of convective heating for diabatic initialization and moisture adjustment in numerical weather prediction models. Mon. Wea. Rev., 118 , 6793.

    • Search Google Scholar
    • Export Citation
  • Rabier, F., , H. Järvinen, , E. Klinker, , J-F. Mahfouf, , and A. Simmons, 2000: The ECMWF operational implementation of four-dimensional variational assimilation. Part I: Experimental results with simplified physics. Quart. J. Roy. Meteor. Soc., 126 , 11431170.

    • Search Google Scholar
    • Export Citation
  • Tompkins, A. M., , and M. Janisková, 2004: A cloud scheme for data assimilation: Description and initial tests. Quart. J. Roy. Meteor. Soc., 130 , 24952518.

    • Search Google Scholar
    • Export Citation
  • Trémolet, Y., 2004: Diagnostics of linear and incremental approximations in 4D-Var. Quart. J. Roy. Meteor. Soc., 130 , 22332251.

  • Treadon, R. E., , H-L. Pan, , W-S. Wu, , Y. Lin, , W. S. Olson, , and R. J. Kuligowski, 2003: Global and regional moisture analysis at NCEP. Proc. ECMWF/GEWEX Workshop on Humidity Analysis, Reading, United Kingdom, ECMWF, 33–48. [Available from European Centre for Medium-Range Weather Forecasts, Shinfield Park, Reading RG2 9AX, United Kingdom.].

  • Tsuyuki, T., , K. Koisumi, , and Y. Ishikawa, 2003: The JMA mesoscale 4D-Var system and assimilation of precipitation and moisture data. Proc. ECMWF/GEWEX Workshop on Humidity Analysis, Reading, United Kingdom, ECMWF, 59–68. [Available from European Centre for Medium-Range Weather Forecasts, Shinfield Park, Reading RG2 9AX, United Kingdom.].

  • Uppala, S. M., and Coauthors, 2005: The ERA-40 re-analysis. Quart. J. Roy. Meteor. Soc., 131 , 29613012.

Fig. 1.
Fig. 1.

Mean analysis difference for TCWV as a percent of the TCWV content of the BL experiment: (a) CTRL − BL, (b) (BL + RAIN) − BL, (c) (BL + CLEAR) − BL, and (d) (BL + CLEAR + RAIN) − BL.

Citation: Monthly Weather Review 136, 7; 10.1175/2007MWR2292.1

Fig. 2.
Fig. 2.

Zonal cross sections of the mean relative humidity analysis difference: (a) (BL + CLEAR + RAIN) − BL, (b) (BL + RAIN) − BL, (c) (BL + CLEAR) − BL, and (d) CTRL − BL.

Citation: Monthly Weather Review 136, 7; 10.1175/2007MWR2292.1

Fig. 3.
Fig. 3.

Mean relative analysis difference for TCWV (percent) (a) CTRL − (CTRL − RAIN) and (b) CTRL − (CTRL − CLEAR). Hatched areas indicate regions where the calculated differences reach 95% statistical significance.

Citation: Monthly Weather Review 136, 7; 10.1175/2007MWR2292.1

Fig. 4.
Fig. 4.

RMSE (kg m−2) of TCWV forecasts from all experiments from verification against the operational model.

Citation: Monthly Weather Review 136, 7; 10.1175/2007MWR2292.1

Fig. 5.
Fig. 5.

Relative RMSE (% cf. BL = 100%) of TCWV forecasts from all experiments after (a) 24 and (b) 72 h.

Citation: Monthly Weather Review 136, 7; 10.1175/2007MWR2292.1

Fig. 6.
Fig. 6.

RMSE (%) of relative humidity (R) forecasts from all experiments stratified (top to bottom) by level at 200, 500, 700, and 1000 hPa and (left to right) by latitude between the Southern Hemisphere, tropics, and Northern Hemisphere from verification against the operational model.

Citation: Monthly Weather Review 136, 7; 10.1175/2007MWR2292.1

Fig. 7.
Fig. 7.

As in Fig. 6, but for vector wind (VW; m s−1).

Citation: Monthly Weather Review 136, 7; 10.1175/2007MWR2292.1

Save