The Research On Hybrid Parallel In Climate Resources Interpolation Algorithm And Zoning Technology

In recent years, the development of agriculture is facing the ever-changing climate conditions, so it puts forward a higher demand for the research and application of agro-climatic resources. Through meticulous research of agricultural climate resources, it is necessary to support the development of China’s agricultural production. In order to get a fine climate division data, two commonly used techniques are the small grid interpolation technology and agricultural zoning of climatic resources. However, due to the the climate divisions need to handle huge historical meteorological data, or time series data, the small grid interpolation using the algorithm based on a single-processor serial execution efficiency is not high. The calculation of the time consumes very large. In practical applications the system waits for a long time. The clustering technique is also true. Especially for complex multi-attribute data, clustering techniques of computation time consumption is even greater.In order to improve the response speed and efficiency of information processing of the application of the system as a whole, we use a hybrid parallel strategy to transform the grid interpolation technique is the key technology of Kriging algorithm. According to the characteristics of the Kriging algorithm, first segmentation algorithm can be coarse-grained and fine-grained part of its package, and then modify the various parts of the data storage structure to accommodate parallel data storage and interaction, followed by the establishment of more than one parallel processing cluster compute nodes, based on data from MPI parallel strategies and master-slave mode coarse-grained parallel processing part of the multi-machine, each compute node within the fine-grained part based on OpenMP parallel processing. The experimental data show that compared to the serial algorithm Kriging algorithm based on hybrid parallel strategy improves the computational efficiency.For climatic zoning of cluster analysis, in order to simplify the calculation process, and in particular to improve the efficiency of massive data clustering extended to multi-attribute, this article will add pruning strategy to the K-means algorithm, and propose a pruning standard, known as PK-means algorithm based on the strategy. First, we transform the input data storage structure of the K-means algorithm, and randomly select k points as the initial centroid, using the standard kd tree structure to store other cluster data, and then prune the candidate set of the centroid for each node of the tree structure. The minimum distance and the maximum distance were calculated for each centroid of the node. The minimum distance is obtained by the kd tree nearest neighbor method, and the maximum distance segmentation hyperplane represented by the node represented by the most distant. Delete those centers that whose minimum distance is greater than the maximum distance of a minimum. Thus it can form a qualitative center candidate cluster, with which the node only calculation Euclidean distance, and it is assigned to the nearest centroid. When pruning the nodes on the sub-node, its candidate qualitative center cluster inherits from parent node. Then repeatedly assign the node and calculate the center, until the end of the cluster.So the improved algorithm based on the pruning strategy, for multi-dimensional data, compared with the classic K-means algorithm, operating efficiency is more superior.In this paper, the parallel transformation of small grid interpolation techniques Kriging algorithm, using the hybrid MPI and OpenMP parallel strategy has good parallel efficiency PK-means algorithm for clustering algorithm based on pruning strategy, in dealing with large-scale and multi-attribute data, the computational efficiency is very obvious, and greatly improve the computing speed.