Spatial interpolation methods for monthly rainfalls and temperatures in Basilicata

Spatial interpolated climatic data on grids are important as input in forest modeling because climate spatial variability has a direct effect on productivity and forest growth. Maps of climatic variables can be obtained by different interpolation methods depending on data quality (number of station, spatial distribution, missed data etc.) and topographic and climatic features of study area. In this paper four methods are compared to interpolate monthly rainfall at regional scale: 1) inverse distance weighting (IDW); 2) regularized spline with tension (RST); 3) ordinary kriging (OK); 4) universal kriging (UK). Besides, an approach to generate monthly surfaces of temperatures over regions of complex terrain and with limited number of stations is presented. Daily data were gathered from 1976 to 2006 period and then gaps in the time series were filled in order to obtain monthly mean temperatures and cumulative precipitation. Basic statistics of monthly dataset and analysis of relationship of temperature and precipitation to elevation were performed. A linear relationship was found between temperature and altitude, while no relationship was found between rainfall and elevation. Precipitations were then interpolated without taking into account elevation. Based on root mean squared error for each month the best method was ranked. Results showed that universal kriging (UK) is the best method in spatial interpolation of rainfall in study area. Then cross validation was used to compare prediction performance of tree different variogram model (circular, spherical, exponential) using UK algorithm in order to produce final maps of monthly precipitations. Before interpolating temperatures were referred to see level using the calculated lapse rate and a digital elevation model (DEM). The result of interpolation with RST was then set to originally elevation with an inverse procedure. To evaluate the quality of interpolated surfaces a comparison between interpolated and measured temperatures at eight sites from an independent dataset was done. There was a good agreement with mean R2 = 0.99 (mean RMSE = 0.6 °C). Based on this results universal kriging estimates and RST were used to produce monthly rainfall and temperature maps for Basilicata region aimed at using as quality input in forest modeling.