Interpolating spatial distribution using Kriging combined with environmental condition

Ming Luo,Tao Pei,Qun Dai

2011-01-01

Abstract:Environmental condition usually impacts the spatial distribution of geographical features, and this environmental condition information can be utilized to help improve interpolating the spatial distribution of geographical feature. This paper proposes a new spatial interpolation approach combining Kriging with environmental condition. The proposed approach estimates the values at unknown locations by a weighted summation of observations at sampled neighborhood locations. The weights of estimation consist of two parts: Kriging weights and environmental similarity. Ordinary Kriging interpolation was used to solve the Kriging weights related to spatial autocorrelation, and similarity-based fuzzy set approach was used to solve the environmental similarity related to environmental condition. Then two parts of weights were integrated by introducing an adjustment coefficient under the condition of minimizing error of sample learning. An experiment, which was performed in a study area located in Northeast China to map soil organic matter content in the topsoil, showed that, compared with several other conventional approaches (e.g., Inverse Distance Weighting, Ordinary Kriging, Ordinary CoKriging, etc), the proposed approach improved mapping accuracy distinctly in terms of the smallest Root Mean Squared Error and Mean Absolute Error.