Regional Rartitioning of Agricultural Non-point Source Pollution in China Using a Projection Pursuit Cluster Model

At present, agricultural non-point source pollution has become the dominan source of water pollution in China, and an important factor in air pollution. Facing these serious environmental problems, researchers have made many suggestions to control the non-point source pollution in China. However, China covers a vast area, and the degree of overuse of fertilizer and pesticide varies regionally. Remediation should differ in each area, so the regional partitioning of agricultural non-point source pollution in China is very necessary.

A projection pursuit cluster (PPC) model was used by Dr. LI Xinhu to analyze the regional partitioning of agricultural non-point source pollution in China. The environmental factors impacting the agricultural non-point source pollution were compiled into a projection index to set up the projection index function. A novel optimization algorithm called Free search (FS) was introduced to optimize the projection direction of the PPC model. By making the appropriate improvements as they explored the use of the algorithm, it became simpler, and developed better exploration abilities. Thus, the multi-factor problem was converted into a single-factor cluster, according to the projection, which successfully avoided subjective disturbance and produced objective results. The results show that five regions have been partitioned, and Shandong province was the worst area for agricultural non-point source pollution. In conclusion, the cluster results of the PPC model mirror the actual regional partitioning of the agricultural non-point source pollution in China, indicating that the PPC model is a powerful tool in multi-factor cluster analysis, and could be a new method for the regional partitioning of agricultural non-point source pollution.

This work has been published on Journal of Arid Land, 2011, 3(4): 278-284. This paper is also archived at http://jal.xjegi.com/EN/abstract/abstract102.shtml.