| Home > Newsevents > Training > Rcourse_notes > PARAMETRIZATION > SURFACE_ASSIMILATION > |
The role of the land surface in
the climate system
April 2002
European Centre for Medium-Range Weather
Forecasts
Table of contents >>
Next Section >>
Previous Section >>
6 . Conclusions
The examples presented above have shown that land surface can have a significant impact on the atmosphere at the synoptic/continental scale when it affects the partitioning of the surface energy into sensible/latent heat, via the soil water. This effect can be local or non-local. On the other hand, land surface has a significant impact on the atmosphere at the synoptic/continental scale when it affects the net energy at the surface, e.g., change of the albedo in snow-covered areas in spring. However, in winter, statically stable, conditions the land surface is decoupled from the atmosphere: Large variations in surface temperature affect only the lowest hundred metres of the atmosphere and do not have a significant impact on the circulation.
Recent experience at ECMWF, summarised above, shows that forecasts systems are sensitive to misrepresentations of longer timescales in the land-surface/atmosphere interaction. Progress can only be made with sustained efforts on validation of the model components. Data assimilation/forecast systems are ideal tools to validate parameterisations because of their constant confrontation with observations (in a "perfect synoptics" scenario) and a very large community of users requiring accurate diurnal cycle of weather parameters. In this context, GEWEX (Global Energy and Water Experiment) Continental Scale Experiments and the Coordinated Enhanced Observation Period planned for 2002 will play a prominent role. If the surface and upper-air observations arrive at the NWP centres they can be assimilated. Analysis and short term forecasts from those centres are the best hope of having a coherent picture of surface and atmosphere variables and of the surface fluxes.
An important feedback loop in the surface-atmosphere interaction is the link between the anomalies of soil moisture and precipitation. Its correct representation in the model relies on the interplay of several parameterisations: evaporation and soil moisture, boundary layer processes, clouds and convection. Most NWP models have their poorest scores on precipitation in the tropics and spring/summer extratropics and improvements on the understanding of the soil moisture-convection interaction will most likely greatly alleviate that deficiency. Validation of model results against results from the Large Scale Biosphere-Atmosphere Experiment in Amazonia (LBA) (see forthcoming Special Issue of J. Geophys.Res. 2001) observations of the diurnal cycle of precipitation and atmospheric thermal and humidity profiles over the Amazon River Basin will probably be a key for progress in that area.
Probably the best summary of the impact of land surface on weather can be shown on Fig. 8 , displaying the history of ECMWF operational short-range forecast errors of 2 m temperature over Europe as a time-series of monthly averages. These errors show a large annual cycle, are different for night and day (72 and 60 hour forecasts verifying at 12 and 00 UTC, respectively), and have a rich history of the many model changes that were made over the years. We will discuss only the model changes made from 1993 onwards.
In August 1993, a surface scheme with a climatological deep-soil boundary condition for temperature and moisture was replaced by the free-running four-layer scheme (Viterbo and Beljaars 1995), but the impact is not very obvious. The summer daytime bias of August 1993 was smaller than that of the previous year but, at that time, the soil scheme has been running freely for only two-months (including the July parallel test described earlier). The next summer showed a pronounced warm bias related to a gradual drying out of the soil which was reduced in July 1994 by resetting the soil moisture to field capacity over vegetated areas. A simple soil moisture analysis scheme was introduced in December 1994 (Viterbo 1996) with a clear beneficial impact on the daytime bias for summer 1995. The night-time temperatures have been biased cold for many years, related to an overly large amplitude of the diurnal cycle. The winter of 1995/1996 was particularly bad, mainly because the European area was blocked for most of the winter with easterly winds and very cold temperatures, although changes to the cloud scheme or the orographic drag might have had a negative impact on night-time temperatures. It is interesting that the reduction of the daytime bias actually increased the night-time bias by displacing the entire diurnal cycle to colder temperatures. Soil freezing and increased boundary layer diffusion in stable layers, introduced in September 1996, improved the monthly error statistics considerably. The wintertime bias was largely eliminated and the amplitude of the diurnal cycle was down to a reasonable level. The snow albedo reduction described in Section Boreal forests was introduced in December 1996, but its impact is not clear over Europe, due to the relatively small area covered by snow and the overall magnitude of the errors linked to the excessive soil cooling, corrected three months earlier. Finally, a much more selective way of initialising soil moisture (Douville et al. 2000), introduced in April 1999, might be responsible for a slight reduction of the standard deviation of temperature errors in that year. A new surface scheme was introduced in June 2000 (van den Hurk et al. 2000), but it is too early to assess its operational performance. It is fair to point out that the statistics presented in Figure 8 are averaged over a month and over a large area. The errors on a day-to-day basis can still be large, but are less systematic and are often related to errors in the forecasted clouds or the presence of snow.
Acknowldegments
This note summarises work done at ECMWF during the last ten years, in close collaboration with many colleagues, in particular Anton Beljaars and Alan Betts. Furthermore, the different contributions of Martin Miller, Jean-François Mahfouf, João Teixeira, and Eric Wood are gratefully acknowledged. Part of this work will also appear as a book chapter to be published by Springer-Verlag in 2002 (Viterbo and Beljaars 2002).
Training Course Notes Front Page >>
Table of contents >>
Next Section >>
Previous Section >>
Copyright © 2002, ECMWF. All rights reserved.
 |
12.06.2002 |
 |
© ECMWF |
