TESTING THE PERFORMANCE OF DIFFERENT SPATIAL INTERPOLATION TECHNIQUES ON MAPPING SHORT DATASERIES OF PRECIPITATION PROPRETIES

patial interpolation, in the context of spatial analysis, can be defined as the derivation of new data from already known information, a technique frequently used to predict and quantify spatial variation of a certain property or parameter. In this study we compared the performance of Inverse Distance Weighted (IDW), Ordinary Kriging and Natural Neighbor techniques, applied in spatial interpolation of precipitation parameters (pH, electrical conductivity and total dissolved solids). These techniques are often used when the area of interest is relatively small and the sampled locations are regularly spaced. The methods were tested on data collected in Iasi city (Romania) between March – May 2013. Spatial modeling was performed on a small dataset, consisting of 7 sample locations and 13 different known values of each analyzed parameter. The precision of the techniques used is directly dependent on sample density as well as data variation, greater fluctuations in values between locations causing a decrease in the accuracy of the methods used. To validate the results and reveal the best method of interpolating rainfall characteristics, leave-one – out cross-validation approach was used. Comparing residues between the known values and the estimated values of pH, electrical conductivity and total dissolved solids, it was revealed that Natural Neighbor stands out as generating the smallest residues for pH and electrical conductivity, whereas IDW presents the smallest error in interpolating total dissolved solids (the parameter with the highest fluctuations in value)