Abstract
For driving soil–vegetation–transfer models or hydrological models, high-resolution atmospheric forcing data is needed. For most applications the resolution of atmospheric model output is too coarse. To avoid biases due to the non-linear processes, a downscaling system should predict the unresolved variability of the atmospheric forcing. For this purpose we derived a disaggregation system consisting of three steps: (1) a bi-quadratic spline-interpolation of the low-resolution data, (2) a so-called ‘deterministic’ part, based on statistical rules between high-resolution surface variables and the desired atmospheric near-surface variables and (3) an autoregressive noise-generation step. The disaggregation system has been developed and tested based on high-resolution model output (400mhorizontal grid spacing).Anovel automatic search-algorithm has been developed for deriving the deterministic downscaling rules of step 2. 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 downscaling step 1 and 2, root mean square errors are decreased. Step 3 finally leads to a close match of the subgrid variability and temporal autocorrelation with the reference fields. The scheme can be applied to the output of atmospheric models, both for stand-alone offline simulations, and a fully coupled model system.
Highlights
The earth’s surface is characterised by heterogeneity extending from microscopic to global scales
A standard bi-linear interpolation algorithm using the four surrounding values, which has been tested for comparison only, does not lead to such a strong error reduction; here the errors are reduced only slightly, i.e. the bi-quadratic spline interpolation leads to better results, see Table 5
Possible applications are the generation of high-resolution input for soil-vegetation-atmosphere transfer (SVAT) models and hydrological models, based on low-resolution atmospheric model output
Summary
The earth’s surface is characterised by heterogeneity extending from microscopic to global scales. Examples are threshold-dependent processes such as runoff production, snow melt and stomata control; or the turbulent exchange coefficients, which are nonlinear functions of the near-surface atmospheric stability For these reasons modelling of the exchange processes either needs to be performed at high resolutions, or has to account for this sub-grid heterogeneity in some other way. A more physically approach, based on a disaggregation of all forcing variables, is presented by a study by Boe et al (2007) They forced a hydrological model with differently downscaled atmospheric data and compared the resulting river discharges. Our approach addresses the scale gap between high-resolution surface models and atmospheric models running on a coarser grid It has been developed and tested on much smaller scales than the aforementioned studies.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
