Statistical modeling of multi-site daily precipitation processes in the context of climate change

Thursday 2 july 2015

17:39 - 17:42h at North America (level 0)

Themes: (T) Extreme events, natural variability and climate change, (ST) Learning from disasters, Poster pitches

Parallel session: Poster pitches: 13I. Extreme - Lessons Disaster

Climate change has been recognized as having a profound impact on the hydrologic cycle; and Global Climate Models (GCMs) have been extensively used in many studies for assessing this impact. However, outputs from these models are usually at resolutions that are too coarse and not suitable for the hydrological impact assessment at a regional or local scale. Downscaling methods have been hence proposed for developing the linkage between GCM predictions of climate change to hydrologic processes at the relevant space and time scales for these impact studies. In particular, statistical downscaling (SD) methods have been commonly used because of their simple computational requirement. However, their main limitation is related to the stationary assumption of the SD model parameters. Another limitation of many SD procedures is that these procedures have been mainly dealing with downscaling of climatic processes for a single site, but very few studies are concerned with the downscaling of these processes for multi-sites because of the complexity in describing accurately both observed at-site temporal persistence and spatial dependence between different locations; this could have an important effect on the accuracy of the impact assessment results. The main objective of the present study is therefore to develop an efficient statistical downscaling (SD) approach for simulating simultaneously and concurrently daily precipitation series at many sites. The proposed approach consists of a combination of two distinct multiple regression models to represent the linkage between global climate predictors and and the probability of local daily rainfall occurrences and the daily rainfall amounts, and the singular value decomposition (SVD) technique to represent the observed statistical properties of the stochastic component of the proposed combined model. Results of an illustrative application using observed daily precipitation data available at ten weather stations in Canada and the climate predictors estimated from the NCEP re-analysis data set for the period from 1961 to 2000 have indicated the feasibility and accuracy of the proposed SD approach.

More information