| Home > Newsevents > Training > Rcourse_notes > PARAMETRIZATION > RADIATION_TRANSFER > |
Radiation Transfer
March 2000
By Jean-Jacques
Morcrette
European Centre for Medium-range Weather
Forecasts, Shinfield Park, Reading Berkshire RG2 9AX, United Kingdom
Table of contents >>
Next Section >>
Previous Section >>
5 . Comparisons with observations
As for the representation of clouds, there is an imperative need to perform at least some of the validation on short-time and limited space scales, before the drift in the GCM climate makes comparison with observations very difficult to interpret. Relative to other physical processes, the existence of reference models (LbLs) that have been or can be validated against highly spectrally detailed measurements (such as those of spectrometer and interferometer, present on ARM sites, or embarked on Nimbus-3) offers a very useful tool for validation. However, although currently done for LW, this approach is not really used for SW, because of the relative lack of highly spectrally detailed measurements in the SW compared to LW, and therefore of carefully validated LbL models in the SW (Ramaswamy and Freidenreich, 1992).
The approach discussed above could also be used, albeit with more difficulties, for cloudy atmospheres. However, for years, most of the validation effort for radiation transfer in cloudy atmospheres has dealt with comparison of radiation fluxes at TOA with satellite-derived fluxes (Nimbus-7 Earth Radiation Budget, Earth Radiation Budget Experiment). More often than not, this was often done without considering at the same time a validation of the cloud characteristics so that tweaking the cloud properties (particularly, the relationships giving the cloud fraction and cloud water loading in diagnostic cloud schemes) allowed an easy agreement with TOA fluxes, specially over monthly mean time-scales. With the availability of carefully calibrated measurements of the radiation fluxes at the surface (Baseline Station Radiation Network (Ohmura et al., 1998), Global Energy Balance Archive, NOAA-ARL SURFace RADiation network), or as is the case with the ARM program, of radiation fluxes and relatively detailed information on the state of the overlying atmosphere and some information on the cloud structure, systematic verification of the radiation fields produced by the ECMWF model at various stages in the forecasts can now be carried out.
In the following, we present one example of such validation for a site where good quality surface radiation measurements are available with a high frequency (such sites are available as part of the BSRN, SURFRAD or ARM projects). For a weather forecast system, it is often preferable to use measurements not too remote in time, as the analysis and forecast system undergoes regular improvement over the years. In this respect, the NOAA-ARL SURFace RADiation network offers real-time availability of surface radiation measurements over 6 sites in the U.S. Figure 5.1 compares the downward and upward LW radiation over Goodwin Creek, Mississippi, over the period 17 November to 15 December 1997, when a new physical parametrisation package was being tested. Fig. 5.2 shows the corresponding downward SW radiation. As can be seen from the LW comparisons, the model is rather successful at producing the observed variability in temperature at the surface (as seen from the upward LW radiation in Fig. 5.1 , bottom panel) and in temperature, humidity, and cloud fraction in the lowest layers of the atmosphere (as seen from the downward LW radiation in Fig. 5.1 , top panel). On the other hand, it appears less successful at handling the amount of condensed water in clouds, with too small optical thickness translating in too large downward SW radiation at the surface (Fig. 5.2 ). Next stage in this type of comparisons is to make sure that the fluxes are computed for the proper vertical profiles of radiatively important quantities. Albeit obvious, this statement is very difficult to be put in practice, as a comprehensive knowledge of all the atmospheric profiles is rarely attained (and attainable). The ARM program is the best try in this direction, with, over the South Great Plains site in Oklahoma, long time-series of surface radiation flux measurements, radiosoundings, synoptic observations, vertically integrated water vapour and cloud water measurements from a microwave radiometer, and vertical profiles of temperature, humidity, and cloud water from the inversion of interferometer and cloud radar measurements. Figure 5.3 presents the total column water vapour and cloud liquid water in December 1997 as observed by the Microwave Radiometer and produced by the ECMWF model in a series of 31 operational forecasts all starting at 12 UTC. Between the 13 and 17 December, the sky is essentially clear. Fig. 5.4 compares the surface downward LW radiation observed during these 5 days and computations by different radiation schemes from the forecast fields of temperature and humidity. A difference of 5 to 8 W m-2 is systematically found between the pre-December'97 ECMWF radiation scheme (old_EC) and the RRTM scheme. However, even this state-of-the-art radiation scheme underestimates the observed flux by 10-15 W m-2, showing that, even this thought-to-be-simple comparison exercise is not straightforward. It is very likely that the model low-level temperature and humidity profiles are at the origin of these discrepancies.
Another validation exercise involving the model radiation scheme is the verification of the cloud signature on TOA radiances. Fig. 5.5 illustrates such comparison for the ERICA storm: the longwave window channel brightness temperature/radiance is simulated from the model cloud, temperature and humidity fields using the methodology of Morcrette (1991a).
and 
fields operationally produced from forecasts between
12 and 35 hours, all starting at 12 UTC.
Training Course Notes Front Page >>
Table of contents >>
Next Section >>
Previous Section >>
 |
12.06.2002 |
 |
© ECMWF |
