Adaptive radiative transfer parameterisation
This page presents two adaptive radiative transfer parameterisation schemes for numerical weather prediction (NWP) models, which illustrate the general concept of an adaptive parametrization scheme. More details on these schemes can be found in the articles listed at the end of this text. We developed these two adaptive parameterisation schemes for the radiative transfer (RT) modelling in the limited area model COSMO (used for short range numerical weather prediction and regional climate modelling).
Temporal perturbation method
The first method, a temporal perturbation scheme starts by computing the full radiation field with a delta-two-stream scheme. Then it regularly calculates the changes with a simple regression scheme. Large (local) biases typical for regression schemes are avoided because the regression algorithm is only employed to calculate changes. Furthermore, the regression method is employed to select the columns where the delta-two-stream scheme is called.
Spatial local-search method
Our second method is based on a local-search algorithm. Every radiation time step of, e.g. 5 minutes, it calls the delta-two-stream scheme in a small number of columns. The radiative tendencies of the other columns are determined by a local-search algorithm that searches for a column with similar radiative properties for which the tendencies are already known. The similarity is determined by a weighted index taking account of the differences in liquid water path, cloud cover, and albedo, and the time passed since the delta-two-stream scheme was called.
For the German weather service (DWD extramural research programm) we are currently working on an adaptive radiative transfer scheme that will combine the two above approaches. For this we will also develop smarter extrinsic parametrizations.
An illustration of the errors in the solar net radiative flux (all in W m-2) in the COSMO model at the surface for 19 September 2001 at 12.30 UTC. For orientation, the boundaries between water and land are indicated with a white line. In the north-west is the North Sea and The Netherlands, in the north-east the Baltic Sea and in the south large lakes around the Alps are visible. The left panel shows the error from keeping the radiation field constant for an hour (a typical way to reduce the computational costs), the middle panel the errors of the temporal adaptive scheme and the right panel the spatial adaptive scheme.
The UK MetOffice has also developed two adaptive RT schemes for their global numerical weather prediction and climate Unified Model (Manners et al., 2009). The first method computes the frequencies most influenced by clouds once per hour and the others once per three hours. The second parameterisation uses a simple scheme with a small number of frequency bands to compute the change in the radiative tendencies (similar to part of our temporal scheme) and is now operationally employed.
Venema, Victor, Annika Schomburg, Felix Ament, and Clemens Simmer. Two adaptive radiative transfer schemes for numerical weather prediction models. Atmospheric Chemistry and Physics, 7, 5659-5674, doi: 10.5194/acp-7-5659-2007, 2007.
A. Schomburg, V. Venema, F. Ament, and C. Simmer. Application of an adaptive radiative transfer scheme in a mesoscale numerical weather prediction model. Quarterly Journal of the Royal Meteorological Society, 138, pp. 91–102, doi: 10.1002/qj.890, 2012.
Manners, J., J.-C. Thelen, J. Petch, P. Hill, and J.M. Edwards. Two fast radiative transfer methods to improve the temporal sampling of clouds in numerical weather prediction and climate models. Q.J.R. Meteorol. Soc., 135, pp. 457-468, 2009
Schomburg, Annika. An adaptive radiative transfer parameterisation for the Lokal-Modell. Diploma (Master) thesis, 2006