Evaluation Of Soil And Water Conservation Function In Seabuckthorn Forest Based On Genetic Algorithms And Neural Networks In Soft Sandstone Areas

The sandstone areas was the erosion center of the loess plateau in China,with an average annual sediment yield into the Yellow River up to350million tons. The bad ecological environment has caused serious threat and obstacles to the healthy development of the surrounding cities and Yellow River. Our country attached great importance to the management. The project of Hippophae rhamnoides Linn afforestation in Shanxi Shaanxi and Inner Mongolia soft sandstone area chose to seabuckthorn as the pioneer species controlling soil and water loss since1999,which has played a great role in inproving the local ecological environment. In order to assess the achievement of the project, it’s important and significant to evaluate the soil and water conservation function in seabuckthorn forest.Traditional comprehensive evaluation methods usually have two defects:the weight is subjective and the nonlinear relation is difficult to express using mathematical model. Based on BP artificial neural network technology and genetic algorithm, this paper built a GA-BP grade evaluation model which topology structure was6-6-1(6input layer nodes,6cryptic layer nodes and1output layer node). The model used index values of soil and water conservation function of sea-buckthorn as input data and the comprehensive evaluation rank values as output data. The research focused on some difficulties such as how to expand samples with interpolation methods, ascertain network topological structure and the training function, optimize weights and threshold. At the same time, the research built a projection pursuit grade evaluation model to evaluate the same case in order to test and verify the evaluation result of the other model.The results showed that the status of water and soil erosion of five regions in Ordos was well controlled by sea-buckthorn after ten years’ construction of ecological projects.The evaluation value of GA-BP grade evaluation model for five regions had raised two grand. The order of soil and water conservation function of sea-buckthorn in different areas was Dongsheng District (0.84,Grand Ⅱ)> Zhungeer Banner (0.63, Grand Ⅲ)>Ejin Horo Banner (0.59, Grand Ⅲ)>Dalad Banner (0.55, Grand Ⅲ)>Hanggin Banner (0.29, Grand Ⅲ). This model which had selected appropriate parameters and training function accorded with the requirement of accuracy on the basis of a lot of test repeatedly. Because the evaluation result was objective stable, the model will be used in other similar areas. Compared to the result of GA-BP grade evaluation model, the one of projection pursuit grade evaluation model was more than eighty percent of the same degree. The order of soil and water conservation function of sea-buckthorn in different areas was Dongsheng District (1.63,Grand Ⅱ)>Zhungeer Banner (2.81, Grand Ⅲ)>Ejin Horo Banner (2.90, Grand Ⅲ)>Dalad Banner (2.95, Grand Ⅲ)>Hanggin Banner (4.58, Grand Ⅴ).Two models met the accuracy requirements and the results were proved to be objective and stable, which can offer effective recommendations to development and management of sea-buckthorn ecological project in sandstone areas.