Estimation models for precipitation in mountainous regions: the use of GIS and multivariate analysis

Abstract

Using multiple linear regression and Geographic Information System techniques, we modelled the spatial distribution of mean monthly precipitation for the seasonal and annual periods in a mountainous region of 10,590 km², located in the central area of the Cantabrian Coast, Spain. We used precipitation data measured at 117 stations for the period 1966–1990, using 84 stations for function development and reserving 33 for validation tests. The best model developed used five topographic descriptors as independent variables: elevation, distance from the coastline, distance from the west, and a measurement of elevation and slope means into homogeneous areas. These topographic variables were calculated as raster models with 200 m resolution.

The model accounted for most of the spatial variability in mean precipitation, with an adjusted R² between 0.58 and 0.67. The standard error was approximately 10% and the mean absolute error ranged from 8.1 to 26.1 mm, which represented 13–19% of observed precipitation. Regression enabled us to estimate precipitation in areas where there are no nearby stations and where topography has a major influence on the precipitation.

Introduction

Spatial modelling of a climate variable is of interest because many other environmental variables depend on climate. Accurate climate data only exist for point locations, the meteorological stations, as a result of which values at any other point in the terrain must be inferred from neighbouring stations or from relationships with other variables.

Many studies (Kurtzman and Kadmon, 1999, Oliver and Webster, 1990, Philip and Watson, 1982, Mitas and Mitasova, 1988) model the spatial distribution of a climate variable using interpolation methods. These techniques can obtain satisfactory results from limited data, based mainly on the geographic situation of the sampling points, on the topological relationships between these points, and on the value of the variable to be measured. However, interpolation methods only consider spatial relationships among sampling points, and do not take into account other properties of the landscape.

Precipitation generally increases with elevation (Spreen, 1947, Smith, 1979), and so many authors have incorporated elevation into geostatistical approaches (Martı́nez-Cob, 1996, Prudhomme and Duncan, 1999, Goovaerts, 2000). Others have developed relationships between precipitation and various topographic variables such as altitude, latitude, continentality, slope, orientation or exposure, using regression (Basist et al., 1994, Goodale et al., 1998, Ninyerola et al., 2000, Wotling et al., 2000, Weisse and Bois, 2001). Nevertheless, accuracy of the results obtained by these methods in mountainous regions is still very limited (Basist et al., 1994, Daly et al., 1994).

This study sought to develop the relationships between precipitation and a range of topographic variables in order to map precipitation across a mountainous region of Northern Spain. We chose topographic variables that had a great influence on precipitation within the area under study: previous climatic studies in the Cantabrian zone (Mounier, 1979, Sitges Menéndez et al., 1982, Felicı́simo Pérez, 1992, Fernández Alvarez et al., 1996) point to the relevance of the position from the west and from the coastline as a predictor of precipitation, due to the dominance of fronts from the NW. We included two different measurements relative to these distances: a Euclidean length from each point to the coastline, and a Euclidean length from each point to a relative west. We also included commonly used topographic variables cited in the literature, such as elevation (Goodale et al., 1998, Katzfey, 1994, Basist et al., 1994, Bradley et al., 1998, Kurtzman and Kadmon, 1999, Wotling et al., 2000) and slope (Basist et al., 1994, Weisse and Bois, 2001), which we measured in a sub-basin area around each reference point, in order to integrate the effect of the orography on a local scale.