550 Geowissenschaften
Refine
Year of publication
- 2019 (2) (remove)
Document Type
- Article (2)
Language
- English (2)
Has Fulltext
- yes (2)
Is part of the Bibliography
- no (2)
Keywords
- COSMO-CLM (2) (remove)
Institute
Convective shower characteristics simulated with the convection-permitting climate model COSMO-CLM
(2019)
This paper evaluates convective precipitation as simulated by the convection-permitting climate model (CPM) Consortium for Small-Scale Modeling in climate mode (COSMO-CLM) (with 2.8 km grid-spacing) over Germany in the period 2001–2015. Characteristics of simulated convective precipitation objects like lifetime, area, mean intensity, and total precipitation are compared to characteristics observed by weather radar. For this purpose, a tracking algorithm was applied to simulated and observed precipitation with 5-min temporal resolution. The total amount of convective precipitation is well simulated, with a small overestimation of 2%. However, the simulation underestimates convective activity, represented by the number of convective objects, by 33%. This underestimation is especially pronounced in the lowlands of Northern Germany, whereas the simulation matches observations well in the mountainous areas of Southern Germany. The underestimation of activity is compensated by an overestimation of the simulated lifetime of convective objects. The observed mean intensity, maximum intensity, and area of precipitation objects increase with their lifetime showing the spectrum of convective storms ranging from short-living single-cell storms to long-living organized convection like supercells or squall lines. The CPM is capable of reproducing the lifetime dependence of these characteristics but shows a weaker increase in mean intensity with lifetime resulting in an especially pronounced underestimation (up to 25%) of mean precipitation intensity of long-living, extreme events. This limitation of the CPM is not identifiable by classical evaluation techniques using rain gauges. The simulation can reproduce the general increase of the highest percentiles of cell area, total precipitation, and mean intensity with temperature but fails to reproduce the increase of lifetime. The scaling rates of mean intensity and total precipitation resemble observed rates only in parts of the temperature range. The results suggest that the evaluation of coarse-grained (e.g., hourly) precipitation fields is insufficient for revealing challenges in convection-permitting simulations.
The goal of limited area models (LAMs) is to downscale coarse-gridded general circulation model output to represent small-scale features of weather and climate. The LAM needs information from the driving coarse-gridded model passing through its lateral boundaries. The treatment of this information transfer causes inconsistencies between driving and nested models and, subsequently, issues in regional weather and climate simulations. This work examines errors arising from choices taken by the modeler (temporal update frequency of boundary data, spatial resolution jump, and numerical lateral boundary formulation) systematically in an idealized simulation environment. So-called Big-Brother Experiments were performed with the LAM COSMO-CLM (0.11° grid spacing). A baroclinic wave in a zonal channel was simulated over flat terrain with and without a Gaussian hill. The results reveal that the quality of the driving data, here represented by simulations only differing from the LAM simulations by reduced spatial resolution, dominates the performance of the nested model. Consequently, at the simulated mesoscale, the performance of the nested small-scale model simulations is weakly sensitive to the numerical lateral boundary formulation (Davies relaxation or the newly implemented, computationally less demanding Mesinger Eta-model formulation). The performance sensitivity to boundary update frequency and resolution jump is small when at least 6-hourly updates and a resolution jump factor of maximally six is used. Gaussian hill LAM simulations illustrated the strength of downscaling; they can represent small-scale features missing in the coarse-scale driving simulations. In the idealized simulation experiments, spectral nudging is not advisable as it imprints the driving models deficits on the nested simulation.