Premium
Application of several spatial interpolation techniques to monthly rainfall data in the Calabria region (southern Italy)
Author(s) -
Pellicone G.,
Caloiero T.,
Modica G.,
Guagliardi I.
Publication year - 2018
Publication title -
international journal of climatology
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.58
H-Index - 166
eISSN - 1097-0088
pISSN - 0899-8418
DOI - 10.1002/joc.5525
Subject(s) - kriging , inverse distance weighting , multivariate interpolation , interpolation (computer graphics) , environmental science , watershed , precipitation , geostatistics , spatial distribution , spatial dependence , bayesian probability , spatial variability , climatology , meteorology , statistics , geology , mathematics , geography , computer science , animation , computer graphics (images) , machine learning , bilinear interpolation
The spatial distribution of rainfall is paramount for water‐related research such as hydrological modelling and watershed management. The use of different interpolation methods in the same area may cause large differences and deviations from the real spatial distribution of rainfall; these differences depend on the type of chosen model, its mode of geographical management and the resolution used. In this study, different algorithms of spatial interpolation of rainfall in a region of southern Italy (Calabria) were applied and the results of geostatistical and deterministic approaches were compared in order to choose the best method for reproducing the actual precipitation field surface. In particular, inverse distance weighting (IDW), ordinary kriging (OK), kriging with external drift (KED), ordinary cokriging (COK) and empirical Bayesian kriging (EBK) were applied to produce the monthly rainfall maps of Calabria. The maps were obtained from a rainfall data set of 129 monthly rainfall series (about one station per 117 km 2 ) collected in the period 1951–2006. Cross‐validation and visual analysis of the precipitation maps were performed to examine the results of these different models. Results clearly indicate that geostatistical methods outperform inverse distance. Moreover, among these methods, the kriging with an external drift showed the smallest error of prediction.