Interpolation from GME vs LM nudging scheme: vertical structure

Marco Arpagaus, MeteoSwiss

Outline

  • Introduction
  • Analysis: temperature
  • Analysis: relative humidity
  • Analysis: wind direction
  • Analysis: wind speed
  • Forecast
  • Summary

    • Introduction

    To examine the impact of the nudging scheme, we use the upper air verification package operational at MeteoSwiss to check the vertical structure at forecast time +00 hours. Note that the sounding stations (c.f. figure 1) used for this study are also used by the nudging scheme itself (and, indirectly, for the assimilation derived from GME fields). Hence, the following is not an independent upper air verification, but rather a check of the vertical structure at locations where the impact of the nudging scheme should be largest.
     

    Figure 1: aLMo model domain with stations used for upper air verification.
     

    The statistics used for this investigation is based on aLMo runs for April to August 2001, so the results are certainly sound from a statistical point of view.
     

    Analysis: temperature

    The nudging scheme generally provides better temperature profiles (mean error [BIAS] & standard deviation [STD]) than an interpolation of the GME fields.
    However, there is a small bias towards negative values at midnight (c.f. figure 2).
     

    Figure 2: Verification (mean error and standard deviation) of temperature at midnight (forecast time +00 hours) for all stations. Nudging scheme in blue, interpolation from GME in red.
     

    Figure 3: Verification (mean error and standard deviation) of temperature at midday (forecast time +00 hours) for all stations. Nudging scheme in blue, interpolation from GME in red.
     

    The same can be seen when only considering stations close to the alps (Payerne, Lyon, Muenchen, Innsbruck, Udine, Milano; midnight, midday), where the bias at midnight is even more pronounced.
     

    Analysis: relative humidity

    Relative humidity is well assimilated by the nudging scheme (BIAS below 700 hPa & STD), except very close to the surface, where it suffers from the same positive bias as the profiles derived from the interpolated GME fields (c.f. figures 4 and 5).
     

    Figure 4: Verification (mean error and standard deviation) of relative humidity at midnight (forecast time +00 hours) for all stations. Nudging scheme in blue, interpolation from GME in red.
     

    Figure 5: Verification (mean error and standard deviation) of relative humidity at midday (forecast time +00 hours) for all stations. Nudging scheme in blue, interpolation from GME in red.
     

    Results are similar for the stations close to the alps (midnight, midday), where the effect near the ground is again stronger.
     

    Analysis: wind direction

    The two analysis schemes show small differences for wind direction.
    However, the nudging scheme shows some difficulty in the lowest part of the troposphere (BIAS & STD; c.f. figure 6) as well as at the tropopause height (STD; c.f. figures 6 and 7).
     

    Figure 6: Verification (mean error and standard deviation) of wind direction at midnight (forecast time +00 hours) for all stations. Nudging scheme in blue, interpolation from GME in red.
     

    Figure 7: Verification (mean error and standard deviation) of wind direction at midday (forecast time +00 hours) for all stations. Nudging scheme in blue, interpolation from GME in red.
     

    Again, the same conclusions can be drawn from the stations close to the alps alone (midnight, midday). Note the severe bias in the lowest part of the troposphere, especially at midnight.
     

    Analysis: wind speed

    Newtonian relaxation works fairly well for wind speed in the lower troposphere but shows severe problems at the tropopause height (BIAS, STD; c.f. figures 8 and 9).
     

    Figure 8: Verification (mean error and standard deviation) of wind speed at midnight (forecast time +00 hours) for all stations. Nudging scheme in blue, interpolation from GME in red.
     

    Figure 9: Verification (mean error and standard deviation) of wind speed at midday (forecast time +00 hours) for all stations. Nudging scheme in blue, interpolation from GME in red.
     

    The respective results for stations close to the alps (midnight, midday) exhibit the same results.
     

    Finally, note that all profiles (i.e. temperature, relative humidity, wind direction and wind speed) are virtually identical from some levels upwards, e.g. above 200 hPa for temperature or above 150 hPa for relative humidity, no matter whether they are results of the nudging scheme or are obtained from interpolated GME fields.
     

    Forecast:

    The improved analysis has an impact on the forecast error (BIAS) out to typically +12 hours. Fortunately, deficiencies (especially for wind direction and wind speed) seem to die out faster. An effect on the forecast spread (STD) can not be seen.

    As an example, we show the temperature profile at midday for +12 hours
     

    Figure 10: Verification (mean error and standard deviation) of temperature at midday (forecast time +12 hours) for all stations. Forecast from nudging scheme in blue, forecast from interpolated GME fields in red.
     

    as well as the relative humidity profile at midday for +12 hours.
     

    Summary