Comparação de dois interpoladores na análise espacial da precipitação anual do estado do Rio Grande do Sul

Abstract

Precipitation is one of the variables with the most variation within the water cycle and one of the main reasons for hydrologic modelling error. An accurate analysis of rainfall spatial distribution is of utmost importance for managing watershed’s water resources, in addition to giving support to climatological and meteorological studies. The goal of this work is to compare the performance of two interpolation methods: Inverse distance weighted (IDW) and Kriging in the analysis of rain spatial distribution. To do so, annual precipitation data from the state of Rio Grande do Sul (Brazil) spanning from 1961 through 2017 were used. To determine which sample proportion results in more precise rainfall distribution maps, a certain amount of near points to the estimated point were used. For that, Mean Squared Error (MSE), Coefficient of Determination (R²) and Root-Mean-Squared Error (RMSE) were used. To support the results, it was conducted a random fields simulations study and the performance of the geoestatistic and classic methods for exposed case was evaluated in terms of precision and accuracy obtained by Monte Carlo simulation. The results indicate that the Ordinary kriging interpolator shows the lowest MSE and higher R². It can be concluded that the geostatistical method of kriging utilizing nine closer points was the one that better represented rainfall spatial distribution in the state of Rio Grande do Sul.