GPU-based parallel Kriging and its application in reserve estimation

Kriging is one of the most important and basic local estimation methods in spatial information statistics, which uses the spatial distribution of the regionalized variables to estimate unknown regions. To solve the problem of high time complexity of Kriging algorithm, we propose a GPU-based parallel Kriging algorithm which improves Kriging interpolation algorithm. This method is a good solution to hightime complexity of Kriging with no reduction in accuracy. The estimation result of copper resource reserve in Tibet Jiama in dicates that parallel Kriging method has good speed-up and efficiency of parallel computing, which verifies the practicability of the method. Comparative experiments with pure CPU computing show that GPU based parallel computing results are correct and reliable。