Schomburg, Annika: Improving the simulation of small-scale variability in radiation and land-surface parameterizations in a mesoscale numerical weather prediction model. - Bonn, 2011. - Dissertation, Rheinische Friedrich-Wilhelms-Universität Bonn.
Online-Ausgabe in bonndoc: https://nbn-resolving.org/urn:nbn:de:hbz:5N-26327
@phdthesis{handle:20.500.11811/5033,
urn: https://nbn-resolving.org/urn:nbn:de:hbz:5N-26327,
author = {{Annika Schomburg}},
title = {Improving the simulation of small-scale variability in radiation and land-surface parameterizations in a mesoscale numerical weather prediction model},
school = {Rheinische Friedrich-Wilhelms-Universität Bonn},
year = 2011,
month = oct,

note = {For the simulation of subgrid-scale physical processes in mesoscale numerical weather prediction models various kinds of spatial and temporal sampling or averaging methods are employed to decrease their computational burden. These methods are applied both within the physical parameterizations, but also by restricting the number of calls to these parameterization schemes in time and space. This under-representation of small-scale variability can lead to systematic errors due to the nonlinearity of processes, and may cause inconsistencies between variables computed by the different parameterization schemes.
In this work two methods are presented, which provide an efficient spatial and/or temporal sampling of heterogeneities, in the atmosphere itself and at the earth’s surface as lower boundary of atmospheric models. The first method, called adaptive radiative transfer parameterization, provides an efficient technique to compute the radiative effects in the atmosphere and at the soil surface. The second method allows for a scale-consistent coupling of atmospheric and soil-surface models, by running a high-resolution soil-vegetationatmosphere transfer model coupled to the coarser atmospheric model, connected by a novel atmospheric disaggregation scheme. Both developments incorporate small-scale variability in radiative and soil/surface processes in an efficient and consistent way. Furthermore, both methods improve the representation of the energy budget at the earth’s surface; the first by giving more accurate radiation surface net fluxes, the second by improving the turbulent exchange fluxes of sensible and latent heat. Both approaches have been implemented into the COSMO numerical weather prediction model, and tested in the COSMO-DE model configuration on a 2.8 km grid.
The adaptive radiative transfer scheme takes advantage of the spatial and temporal correlations in the radiation characteristics of the atmosphere, and thus makes the parameterization computationally more efficient. The adaptive scheme generalizes the accurate radiation computations made in a fraction of the spatial and temporal space to the rest of the field. For validation three case studies with different synoptic conditions were carried out and the performance of the adaptive scheme is compared to the currently operational COSMO-DE radiation configuration, with quarter-hourly radiation computations on 2x2 averaged atmospheric columns. The reference for both schemes are frequent radiation computations on the full grid. The results show that the adaptive scheme is able to reduce the sampling errors in the surface radiation fluxes considerably and to conserve the spatial variability better, than to the operational scheme. Errors in the three-dimensional heating rates are reduced for larger averaging scales. Physical relations between the radiative quantities and cloud water or rain rates are captured better than with the operational scheme. It is shown, that these refinements also lead to improvements with respect to the dynamical development of the model simulation: the adaptive model runs show a smaller divergence from the reference model run than the currently operational scheme.
One approach to deal with subgrid-scale variability at the surface in atmospheric models is the so-called mosaic approach, in which the soil and the surface are modelled on an explicit higher horizontal grid resolution than the atmospheric part. In this work a statistical downscaling scheme for the atmospheric input variables needed to drive this higher resolved soil-vegetation-atmosphere-transfer model has been developed, ensuring a scale-consistent two-way coupling between the two sub-systems in the mosaic approach. The statistical downscaling combines deterministic with stochastic modeling in a stepwise approach. Downscaling rules between atmospheric variables as predictands and surface parameters as predictors, depending on the atmospheric state, have been developed. In order to model the small-scale variability correctly, the still missing variance is estimated, and added as autocorrelated noise. The disaggregation system has been built up and tested based on high-resolution model output (400m horizontal grid spacing). A novel automatic search-algorithm has been developed for deriving the deterministic downscaling rules. The approach has been extensively tested in an offline testbed by applying it to model output, but also “online” in the mesoscale COSMO model.
When applied to the atmospheric variables of the lowest layer of the atmospheric COSMO-model, the disaggregation is able to adequately reconstruct the reference fields. Applying the deterministic steps, root mean square errors are reduced. The stochastic step finally leads to a close match of the subgrid variability and temporal autocorrelation with the reference fields. These “offline” tests and also the “online“ application in fully coupled COSMO simulations in combination with the mosaic approach indicate that the mosaic approach is able to improve the performance of the turbulent surface exchange fluxes notably compared to simulations without any surface variability representation. Averaged over six case studies root mean square errors of sensible and latent heat fluxes were reduced by about 9 W/m2 and 13 W/m2, respectively, in the COSMO simulations using the 400m high-resolution COSMO model runs as reference. The application of the new downscaling scheme for the disaggregation of atmospheric forcing variables for the soil module, however, leads to only marginal improvements, despite the positive impact of the downscaling for the single terms in the flux equations. The explanation lies in a cancelling of errors for the computation of the fluxes in the standard mosaic approach, due to which the effect of the overall more realistic structure of the surface variables achieved by the distributed atmospheric forcing is mitigated.
In summary, the results indicate that for operational purposes the adaptive radiation parameterization can be recommended without restriction, because it has a large positive impact and does not lead to a significant increase in computation time. The effects of the novel atmospheric disaggregation scheme are small, both with respect to the improvement for the turbulent fluxes but also with respect to computational demands. Given the additional algorithmic complexity an operational application of this downscaling algorithm can currently not be advocated. An operational application of the mosaic approach itself, however, would be beneficial due to its considerable improvement for the representation of the turbulent heat fluxes and the dynamical model development. An increase in computation time would have to be accepted, however, depending on the chosen subgrid resolution.},

url = {https://hdl.handle.net/20.500.11811/5033}
}

The following license files are associated with this item:

InCopyright