[en] Spatial interpolation of precipitation data is of great importance for hydrological modelling. The methods of
geostatistics (krigings) become more popular to make spatial interpolation from point measurement to distributed
hydrological models. However, most of existing geostatistic algorithms are available only for single-moment
data. The first step of Kriging computation is the semi-variogramme modelling which usually uses only one
variogramme model for all-day data. The objective of this paper is to review the implementation of an algorithm
of spatial interpolation methods for daily rainfall and to compare the results of geostatistic and deterministic
approaches. In this study, we will use daily rainfall data from 70 rain gauges in the hilly landscape of Ourthe and
Ambleve Basins in Belgium (2751 km2). This area lies between 35 and 690 m in elevation and consists of river
networks which are the tributaries of the Meuse River. The proposed algorithm will use the method of Cressie’s
Approximate Weighted Least Squares to fit among sevens semi-variogramme models (logarithmic, power, exponential,
Gaussian, rational quadratic, spherical and penta-spherical) to daily sample semi-variogrammes. These
seven models are computed on a daily basis. Firstly, one model is chosen by considering the minimum of least
squares coefficient. Secondly, if the chosen model gives negative interpolated values, other models will be chosen
again until the result become positive. Cross validation will be used to compare the interpolation performance
of geostatistic to deterministic methods usually known as Thiessen polygon and Inverse DistanceWeighting (IDW).
