Interpolation processes using multivariate geostatistics for mapping of climatological precipitation mean in the Sannio Mountains (southern Italy)

Abstract

The spatial variability of precipitation has often been a topic of research, since accurate modelling of precipitation is a crucial condition for obtaining reliable results in hydrology and geomorphology. In mountainous areas, the sparsity of the measurement networks makes an accurate and reliable spatialization of rainfall amounts at the local scale difficult. The purpose of this paper is to show how the use of a digital elevation model can improve interpolation processes at the subregional scale for mapping the mean annual and monthly precipitation from rainfall observations (40 years) recorded in a region of 1400 km 2 in southern Italy. Besides linear regression of precipitation against elevation, two methods of interpolation are applied: inverse squared distance and ordinary cokriging. Cross-validation indicates that the inverse distance interpolation, which ignores the information on elevation, yields the largest prediction errors. Smaller prediction errors are produced by linear regression and ordinary cokriging. However, the results seem to favour the multivariate geostatistical method including auxiliary information (related to elevation). We conclude that ordinary cokriging is a very flexible and robust interpolation method because it can take into account several properties of the landscape; it should therefore be applicable in other mountainous regions, especially where precipitation is an important geomorphological factor.

Seasonal precipitation interpolation at the Valencia region with multivariate methods using geographic and topographic information

“…Of the exposed issues, it can be deduced that, despite the indicated limitations, the obtained results prove that the use of geographical and topographical information in most of the cases improve the capacity to deduce the rainfall in those places where there is a lack of data. This conclusion coincides with that expressed by many authors (Pardo-Igúzquiza, 1998;Goovaerts, 2000;Ninyerola et al, 2000;Wilk and Anderson, 2000;Johansson and Chen, 2003;Marquínez et al, 2003;Vicente-Serrano et al, 2003;Diodato and Ceccarelli, 2005). However, it can be observed that this improvement is evident in the case of summer but limited for the rest of seasons.…”

“…For example, although due to complexities of the precipitation process there are many exceptions to simple rules (Andersson, 1980), precipitation usually increases with elevation and several authors have introduced this into geostatistical interpolation methods (Goovaerts, 2000;Ninyerola et al, 2000;Brunsdon et al, 2001;Johansson and Chen, 2003;Marquínez et al, 2003;Vicente-Serrano et al, 2003;Diodato and Ceccarelli, 2005). Other variables having an effect on precipitation are the longitude and latitude (having positional effects in our case), relief, slope, aspect, and distance to the sea.…”

“…of Catalunya (Spain); Wilk and Andersson (2000) compared a Thiessen based method with a regression based method for a mountainous river basin with monsoon climate (southern India), obtaining significant differences in seasonal rainfall analyses; Wotling et al (2000) used stepwise regression -that included some topographical relief features -applied to a regular grid that was later interpolated with spline functions; Lin and Chen (2003) proposed a radial basis function network incorporating a semivariogram model to estimate rainfall distribution at the Tanshui river basin (Taiwan); Johansson and Chen (2003) found a regression model that included the wind variable for Sweden; Marquínez et al (2003), used a regression model performed with topographic variables for Cantabria (Spain); Vicente-Serrano et al (2003) concluded that the best results were obtained by geostatistical methods and a multiple regression (MR) model using four geographic and topographic variables for the central part of the Ebro Valley (Spain); Diodato and Ceccarelli (2005) compared the inverse squared distance method with linear regression and ordinary cokriging (OCK) for the Sannio Mountains (southern Italy), obtaining the best results for the cokriging that included the elevation as secondary information; finally, Diodato (2005) centers his study on Benevento's province (Italy), where the elevation factor is low correlated with the rainfall. He compares the methods of the ordinary kriging (OK) and cokriging, using as secondary information a function that relates the level of greenery of the existing vegetation with the elevation, and concludes that the better results are for the cokriging approach.…”

Espacialização da precipitação pluvial por meio de krigagem e cokrigagem

“…Estas variáveis têm baixo custo como fonte de informação secundária, se obtidas de modelos digitais de elevação (Goovaerts, 2000). Dessas variáveis, a altitude tem sido mais utilizada como covariável, e estudos têm mostrado melhorias na acurácia da predição de chuvas com sua utilização na cokrigagem (Diodato & Ceccarelli, 2005;Moral, 2010;Viola et al, 2010).…”

“…Alguns autores têm levado em conta a influência marítima sobre a precipitação, em áreas costeiras, e sugerido a distância da costa como uma importante variável no ajustamento de modelos de regressão (Johansson & Chen, 2003). No entanto, diferentemente da altitude, cujos bons resultados como variável auxiliar na estimativa da precipitação pluvial são conhecidos (Diodato & Ceccarelli, 2005;Viola et al, 2010), para a variável distância da costa há ainda necessidade de avaliar a utilização como auxiliar na cokrigagem.…”