02 Fakultät Bau- und Umweltingenieurwissenschaften
Permanent URI for this collectionhttps://elib.uni-stuttgart.de/handle/11682/3
Browse
Item Open Access Statistical downscaling of extremes of precipitation in mesoscale catchments from different RCMs and their effects on local hydrology(2011) Alam, Muhammad Mahboob; Bardossy, Andras (Prof. Dr. rer. nat. Dr. -Ing.)Global climate models are the only available comprehensive tools for studying the affects of climate change on our earth in terms of changes in different meteorological and hydrological variables in future. Precipitation and temperature are two of the most important meteorological variables with regards to their affects on other meteorological (e.g. humidity, evaporation etc.) and hydrological (e.g. river runoff) variables and on human life (e.g. food fibre production, economy etc.). Among other important local and large scale phenomenon that affects the occurrence and amount of precipitation (and severity of temperature), geographical and topographical conditions perhaps play most important role in the behaviour of these variables in certain area. This makes the two variables more or less local phenomenons that need to be specifically studied for each area of interest individually. Unfortunately the scale at which global climate models (GCMs) operate is too large for any meaningful study to be performed related to future patterns of these two variables on local scale. Different methodologies have thus been developed to downscale (i.e. to increase the resolution of) GCM data to the local scale. The two broad categories of downscaling methodologies are statistical and dynamical downscaling. In statistical downscaling methodology, an attempt is made to develop a relationship between large scale GCM modelled variable (called predictor) and local scale observed/measured variable (called predictant). Assuming that in future this relationship will hold, the relationship is used to predict local scale predictand for future simulated scenarios of predictor. In dynamical downscaling (the so called regional climate models (RCMs)) on the other hand, an attempt is made to embed a complete physical model of more or less the same complexity as GCM, in a GCM and upon receiving values from GCM at its boundaries, recalculate all possible physical formulations at a much finer scale. The local conditions are thus taken in to account and the results are believed to be more suitable for local scale studies. Both downscaling methodologies have been extensively applied in climate change and impact studies around the world with varying degree of success and new techniques are consistently being developed to improve upon them. Both methodologies have associated advantages and disadvantages. While statistical downscaling is computationally much cheaper than RCMs, statistical downscaling is based on basic assumption of stationarity which is sometimes hard to justify. RCMs on the other hand although attempt to solve physical equations at local scale, does also inherit bias from the parent GCM. This thesis presents statistical downscaling methodology which attempts to correct for the biases that are inherited by different RCMs. Three different RCMs are considered for German part of Rhine basin and using bias correction methodology based on correction of quantiles of precipitation (and temperature for some studies), new scenarios of precipitation are developed. Further, a distributed version of conceptual hydrological model HBV is calibrated and validated for German part of Rhine basin and raw and downscaled RCM scenarios of precipitation are fed into the model to ascertain the future hydrological regime in face of climate change for this important river. The downscaling procedure briefly discussed above was applied in two ways. In the first case the statistical downscaling methodology was performed on RCM data without considering any constraint during quantile-quantile exchange between RCM control and scenario runs. In the second case, the quantile-quantile exchange was conditioned on occurrence of certain circulation pattern. It was briefly discussed above how precipitation (occurrence and amount) is conditioned by certain phenomenon. In addition to geographical and topographical location, precipitation also depends upon large scale circulation patterns. Thus it was assumed that conditioning the downscaling methodology also on circulation patterns would bring about better results. To realize above concept, classification of circulation patterns is performed. Fuzzy rule based classification methodology is used to classify circulation patterns. Two new methodologies of classification of circulation patterns are presented in this thesis. One is based on low flow conditions in rivers in the study area and the other is based on clustering of precipitation stations. The new classification methodology is believed to provide better classification of circulation patterns in that the difference between the individual classes is enhanced and similarity among the same class intensified. A classification analysis measure called wetness index was developed and used to identify critical circulation patterns among the classified circulation patterns. Critical circulation patterns were identified for extreme wet and dry conditions and it was shown that all extreme cases of floods and droughts are caused by identified critical CPs. This thesis also presents and applies another statistical downscaling methodology based on multivariate autoregressive model of order 1 (one). The methodology makes use of the classification of circulation patterns described above. The parameters of the autoregressive model depend upon the circulation patterns. The methodology is used for number of head catchments in southern and eastern Germany. Head catchments by definition have very quick response time to any significant precipitation event. They contribute quickly to the surface runoff and if they are head catchments of larger rivers, may also result in bigger flood events. Downscaling of precipitation was performed for these catchments by using mean sea level pressure (MSLP) as predictor and local station precipitation as predictant. The model was developed such that ensemble of daily precipitation could be produced. Thereby enabling one to estimate associated uncertainty. Finally drought analysis are performed for German part of Rhine basin using Palmer drought severity index. A FORTRAN routine is developed which can calculate different kind of drought indices such as Palmer drought severity index, Palmer hydrological drought index, and monthly moisture anomaly index for certain catchment. The program developed is also capable of simultaneously mapping the results. The mapping of results makes it possible to ascertain the severity of drought over the larger area. The analysis of drought is performed for observational gridded data set and for control and A1B scenarios of three different RCMs.