Nonparametric Spatial Rainfall Characterization Using Adaptive Kernel Estimator
1 May 1998
A nonparametric statistical tool based on kernel function estimation is developed for spatial rainfall characterization. In this method, observations closer to the point of estimate are weighted higher using kernel function with a prescribed bandwidth. The kernel bandwidth is local and it extends only to the Kth Nearest Neighbor, KNN, observation. An optimal value for KNN is selected by cross validation. Unlike Kriging, the underlying stochastic process is not assumed to be stationary. An application of this model using rainfall data is presented.
Journal of Geographic Information and Decision Analysis, Vol. 2, No. 2, pp. 34-43, 1998.
KEYWORDS: nonparametric statistics, Kth nearest neighbor, kernel estimator, nonstationarity, cross validation