Volosciuk, Claudia D. (2012) The Spatial Scaling Behaviour of Extreme Precipitation in an Atmospheric General Circulation Model and in Observational Data. (Diploma thesis), Technische Universität Braunschweig, Braunschweig, Germany, 72, XXXIII pp.

Preview

Text Diplomarbeit_cvolosciuk.pdf - Published Version Available under License German copyright act UrhG. Download (8MB) | Preview

Abstract

Present-day simulations with the same boundary forcing but different resolutions (T213, T159, T106, T63, T42, T31) of the atmospheric general circulation model ECHAM5 were used to study the scaling behaviour of extreme precipitation in terms of return values, which are estimated quantiles of the generalised extreme value (GEV) distribution. Area-averaged return values of different regions in December, January, February (DJF) and June, July, August (JJA) of the highest resolution T213, averaged to coarser grids, were compared to area-averages of the coarser resolutions of ECHAM5. For the validation of return values, mean precipitation totals and the mean precipitation intensity of the different resolutions of ECHAM5 observational datasets of the UK and the USA were used. Additionally, the observational dataset E-OBS was validated in the UK. Different qualitative scaling behaviours were identified, depending on region and season. Extreme precipitation of different sources has to be compared with respect to spatial scale. However, mean precipitation totals do not show a scaling behaviour in the UK and the USA, indicating that for extreme precipitation in comparison to precipitation totals different processes and different spatial correlation lengths are responsible. T63 is the model resolution that is minimally necessary to represent extreme precipitation even though higher resolutions improve the result. The model ECHAM5 overestimates extreme precipitation and precipitation totals in the UK as well as in the USA. The bias varies in its amount and in the ratio of precipitation totals to return values. The qualitative scaling behaviours in the UK and the USA are captured by ECHAM5. The seasonal cycle of precipitation return values in the southeastern US is not well captured by the high resolutions of ECHAM5. The E-OBS dataset is not appropriate for the validation of climate models in the UK in JJA. Hence, in JJA the spatial correlation length of extreme precipitation is shorter than the distance between rain gauges included in the dataset. The responsible processes for return values in UK and US are on a considerably different spatial scale than those of precipitation totals.

Actions (login required)

View Item