Defining Homogeneous Precipitation Regions by Means of Principal Components Analysis

Dirk Mallants Laboratory of Land Management, Faculty of Agricultural Sciences, Catholic University of Leuven, Leuven, Belgium

Search for other papers by Dirk Mallants in
Current site
Google Scholar
PubMed
Close
and
Jan Feyen Laboratory of Land Management, Faculty of Agricultural Sciences, Catholic University of Leuven, Leuven, Belgium

Search for other papers by Jan Feyen in
Current site
Google Scholar
PubMed
Close
Restricted access

Abstract

The spatial patterns of precipitation over the IJzer watershed (in western Belgium and northern France) are investigated for three years: 1973 (dry), 1977 (wet), and 1978 (average). Analyses were performed using daily precipitation data for the three years together and for the three years separately. The first principal component explains most of the variance (about 75%) and is uniformly distributed over the whole region. Higher-order components delineate subregions of consistent rainfall. No essential difference in pattern in the three years occurs between the first four components. The components patterns seem to express maritime and topographic effects, and form a basis for dividing the Ijzer catchment into four coherent subregions. For representative stations inside each subregion, cross correlations with the remaining stations over the entire watershed indicate that at a level of r = 0.85 three stations define three homogeneous precipitation regions. The result is three representative time series of daily rainfall, one for each region.

Abstract

The spatial patterns of precipitation over the IJzer watershed (in western Belgium and northern France) are investigated for three years: 1973 (dry), 1977 (wet), and 1978 (average). Analyses were performed using daily precipitation data for the three years together and for the three years separately. The first principal component explains most of the variance (about 75%) and is uniformly distributed over the whole region. Higher-order components delineate subregions of consistent rainfall. No essential difference in pattern in the three years occurs between the first four components. The components patterns seem to express maritime and topographic effects, and form a basis for dividing the Ijzer catchment into four coherent subregions. For representative stations inside each subregion, cross correlations with the remaining stations over the entire watershed indicate that at a level of r = 0.85 three stations define three homogeneous precipitation regions. The result is three representative time series of daily rainfall, one for each region.

Save