Original paper

Evaluation of daily maximum and minimum 2-m temperatures as simulated with the Regional Climate Model COSMO-CLM over Africa

Krähenmann, Stefan; Kothe, Steffen; Panitz, Hans-Jürgen; Ahrens, Bodo

Meteorologische Zeitschrift Vol. 22 No. 3 (2013), p. 297 - 316

published: Jul 1, 2013

DOI: 10.1127/0941-2948/2013/0468

BibTeX file


Open Access (paper can be downloaded for free)

Download paper for free


The representation of the diurnal 2-m temperature cycle is challenging because of the many processes involved, particularly land-atmosphere interactions. This study examines the ability of the regional climate model COSMO-CLM (version 4.8) to capture the statistics of daily maximum and minimum 2-m temperatures (Tmin/Tmax) over Africa. The simulations are carried out at two different horizontal grid-spacings (0.22° and 0.44°), and are driven by ECMWF ERA-Interim reanalyses as near-perfect lateral boundary conditions. As evaluation reference, a high-resolution gridded dataset of daily maximum and minimum temperatures (Tmin/Tmax) for Africa (covering the period 2008–2010) is created using the regression-kriging-regression-kriging (RKRK) algorithm. RKRK applies, among other predictors, the remotely sensed predictors land surface temperature and cloud cover to compensate for the missing information about the temperature pattern due to the low station density over Africa. This dataset allows the evaluation of temperature characteristics like the frequencies of Tmin/Tmax, the diurnal temperature range, and the 90th percentile of Tmax. Although the large-scale patterns of temperature are reproduced well, COSMO-CLM shows significant under- and overestimation of temperature at regional scales. The hemispheric summers are generally too warm and the day-to-day temperature variability is overestimated over northern and southern extra-tropical Africa. The average diurnal temperature range is underestimated by about 2°C across arid areas, yet overestimated by around 2°C over the African tropics. An evaluation based on frequency distributions shows good model performance for simulated Tmin (the simulated frequency distributions capture more than 80% of the observed ones), but less well performance for Tmax (capture below 70%). Further, over wide parts of Africa a too large fraction of daily Tmax values exceeds the observed 90th percentile of Tmax, particularly across the African tropics. Thus, the representation of processes controlling Tmax including cloud-solar interaction, radiation processes, and ground heat fluxes should be improved by further model developments. The higher-resolution simulation (0.22°) is on average about 0.5°C warmer with a more pronounced overestimation of the higher percentiles of Tmax, and yields no clear benefit over the lower-resolution simulation.


african climatedaily variabilityfrequency disributionmaximum temperatureminimum temperatureregional climate model