Computer models, e.g. hydrological models, are one of the key components in determining effect policy and estimating impact of water management. Many of these models require gridded information on meteorological variables such as evaporation and precipitation. Creating such gridded maps from observations is done using interpolation methods such as splines or linear regression. The aim of this study was to find the optimal interpolation method for Makkink reference evaporation.
The evaporation dataset consisted of three temporal support levels: yearly normalen, monthly normalen and daily observations. For each level we interpolated using a number of interpolation methods: inverse distance weighted interpolation, kriging, linear regression, interpolation with the mean, nearest neighbor and thin plate splines. To choose which method was optimal we looked at two aspects: summary statistics of cross-validation and the visual appearance of the map.
From our results we conclude that for this dataset thin plate splines is the optimal interpolation method.
Paul Hiemstra and Raymond Sluiter. Interpolation of Makkink evaporation in the Netherlands
KNMI number: TR-327, Year: 2011, Pages: 86