Comparación de métodos de interpolación geoestadísticos con y sin varibles auxiliares para la precipitación anual (2003-2008) en Ecuador

Abstract

The study compares two geostatistical methods for interpolation, ordinary kriging (without secondary variables) and universal kriging (with secondary variables), in the analysis of the distribution of the annual precipitation of Ecuador in the period 2003-2008. The secondary variables explored are satellite images, being selected for the study images of precipitation TRMM, cloud fraction, cloud water content, cloud top temperature, cloud top pressure, soil temperature and digital elevation model. The method of study consist in three stages. First, the interpolation without secondary variables of annual precipitation is made (ordinary kriging). Second, the interpolation with secondary variables (universal kriging). To select the right secondary variables, multiple linear regression is applied. Third, cross validation is used in order to evaluate the goodness of the interpolations. The cross validation results showed that the best method for interpolation was universal kriging with TRMM, followed nearly by the combination of TRMM and cloud water content. With these secondary variables, also, the standard deviation of the interpolation is minimized, generating a homogenous deviation all across the country. This deviation is inferior at the one generated by the interpolation without secondary variables. Finally, the secondary variables allowed to represent the variability of the precipitation in the Amazon, region with a lack of stations.