Estimation of rainfall spatial distribution is an important step in water resources studies. Preliminary
assessment of existing rain gage networks show that the spatial distribution of the gages is not suitable as
many are just located in populated areas. Optimization of rain gage locations can improve the accuracy of
water balance studies. Several methods have been proposed for such tasks. Kriging is a geostatistical
method which can estimate the variance error based on a pre-determined semivariogram with no reliance
on historical record. One can therefore apply Kriging method to estimate the reduction in error variance
due to any change in the network such as addition of a new gage or moving an existing gage. In this paper,
the objective is to optimize gage locations based on error variance and topography of the region. This is
done such that no extra cost is incurred i.e. only omission or moving the stations is allowed. The results
indicated that the number and the network structure is sufficient for estimation of rainfall distribution in
dry months. However, in case of wet months and the annual rainfall, moving some of the stations is necessary.
In all, 17 stations may be moved to reduce the average error by 10 percent. It must be noted that
the error is quite high in extrapolation regions, as expected.