Evaluation of different regressor combinations within the GMET algorithm for the generation of precipitation grids in the transition region of the Andes - Amazonas, Bolivia

The GMET (Gridded Meteorological Tool) has been used for the elaboration of the Surface Water Balance of Bolivia 2017, with satisfactory results on a national scale in terms of estimation of spatial distribution of precipitation. GMET uses two major calculation processes: logistic regression and determination of correlated random fields. This methodology, currently only considers geographic regressors (Elevation, Latitude, Longitude and Slope). However, it was identified that the estimation method must be modified to obtain a better spatial distribution of precipitation on a local scale for specific areas in the country.

The orographic configuration and location of the transition between the Andes and the Amazon generates a spatial distribution of precipitation of high complexity. The inter-Andean valleys resulting from the complicated geography of the region, are sectors of very high precipitation that is only possible to characterize with rain gauges so far. The modified logistic regression structure proposed in this study, tackles with the lack of information from rain gauges and take advantage of different factors which help to improve the estimated spatial distribution of precipitation in the regions.

From our assessment in this study, it has been established that the following regressors: altitude, vegetation index and indirect estimation of precipitation (remote sensing) are the optimal combination of regressors at local level in the transition region between the Andes and Amazonia.