Statistical modelling of observed precipitation and its application to extreme value statistics in different spatiotemporal scales (Modex)
Final Report Abstract
Statistical modelling of precipitation time series. Changes in the probability of occurrence and intensity of extreme events attract attention due to their high and immediate social and economic impact. In this project the variability and long term changes in both, the precipitation amount and the occurrence of a precipitation event based on daily and monthly station data, are investigated. The precipitation amount NN is interpreted as a realization of a probability density function (pdf). Depending on the temporal resolution and rain regime, the Weibull, Gumbel or Gamma distribution are used for its statistical description. Structured components like trends and seasonal cycles are detected in all parameters of the respective pdf; thus they vary in time (t). The probability of occurrence p(t) of a precipitation event is described as a point process and tested for trends and seasonal changes. We provide a two-part statistical description in the form f(NN, t)=p(t) ∙ pdf(NN, t) which provides the probability for the precipitation sum NN (≥ 0 mm) for each point in time t. This more complete description allows for the assessment of over- and undershooting probabilities for arbitrary thresholds at any time. We provide results for daily precipitation time series in Germany and Europe and for the monthly global GPCC data set via maps of trends and probability assessments for the precipitation amount and the occurrence of precipitation events. Especially in Germany we observe a change in the seasonal cycle to more precipitation and extremes in late winter and less in summer. Regionalisation of climate variability in Germany from 1951-2000 using multivariate indices. We introduce a multi-index approach for the assessment of climate change over Germany over the past 50 - 60 years based on in total 47 climate indices. In a first step we delineate climate regions characterized by similar temporal behaviors of seasonal and annual climate indices using a principal component analysis (PCA) in S mode with Varimax-rotation based on the correlation matrix of the detrended indices time series. The PCA is applied to the four seasons separately but also to the indices describing the entire year. Three to five climate regions are detected in the differently filtered data set, representing distinct geographical regions and roughly dividing the country in a northern, middle and southern part. The number of regions varies with seasons and considered indices. As expected, the inclusion of indices representing extreme events increases the number of delineated climate regions. The mean values and trends of the spatially-weighted average time series of the single indices for the delineated regions provide a first characterization of regional climate and its change. In a second step regional climate variability and change are analyzed via regional multi-indices. Multiindices constitute synthetic time-series of a group of well-correlated single indices, which are clustered by applying the PCA in P-mode to the weighted regional time series of the indices. Depending on season and region, 8-10 multi-indices are found, which can be related to typical weather situations. Several significant trends on a 95 percent significance level are detected in the weighted time series of the multiindices and indicate a change of climate particularly in the summer season. Generally, the trends suggest a change to more 'nice' sunny, warm and dry summer weather with less cloudiness and lower relative humidity, a larger anticyclonic influence, longer dry periods, less snow, and higher minimum temperatures. Weak changes in precipitation characteristics are observed in summer. While several authors suggest an increase in extreme precipitation in winter, our study reveals a not significant trend to more extreme precipitation in winter.
Publications
- 2015: Statistical modeling of observed precipitation and its application to extreme value statistics in different spatiotemporal scales. 10. Deutsche Klimatagung, 21.-24. September 2015 in Hamburg
Stockhausen, B., S. Trömel, and A. Hense
- 2017: More extreme precipitation over Africa - a statistical analysis of observational and reanalysis data for probability assessments. Climate Change in Africa: Evidence, mechanisms and Impacts, Past and Present (CCA), 6-11 November 2017 in Marrakesh, Morocco
Trömel, S., B. Stockhausen, C. Simmer
