自由搜索算法的投影寻踪模型在湿地芦苇调查中的应用

Application of a free search-based projection pursuit model in investigating reed in wetlands

针对传统湿地芦苇调查统计方法中的问题,建立了自由搜索的投影寻踪模型,将芦苇的生理特征指标和环境因子指标作为投影指标构建了投影指标函数,通过自由搜索算法优化得到最佳投影方向,由最佳投影指标函数来反映各类芦苇的特征,避免了人为赋予权重的干扰,客观性强,数学概念清晰,并在博斯腾湖芦苇调查统计中进行了应用,结果表明:博斯腾湖芦苇的种类及群落特征没有发生显著的变化,在8个环境因子中水质及水量是影响芦苇产量的主要环境因子,土壤有机质对芦苇产量的影响最小;通过实际应用表明模型应用效果较好,为湿地芦苇资源调查及湿地芦苇的生态保护提供了新的思路。

A projection pursuit model （PP） was used to solve the multi-factor problem in the synecological investigation of reed in Bosten Lake. A novel optimization algorithm called free search was introduced to optimize the projection direction of the PP model. The classification of reed and analysis of environmental factors that impact the growth of reed were compiled into a projection index to set up the projection index function. The multi-factor problem was converted into a single-factor problem according to the projection, which successfully avoids subjective disturbance and ensures objective results. In 2006, soil samples at 78 sampling sites in 11 areas were collected according to the reed distribution in Bosten Lake. Of these sites, 41 were in the region of Small Lake, 22 in the region of Huangshui Ditch, and 15 in the West Bank region of Big Lake. Plant height, stalk diameter, biomass density, number of node, wall thickness, and cellulose were observed. Organic matter, total and available nitrogen and phosphorus, and total salt in soil samples as well as the chemical oxygen demand and pH of water were selected for the analysis. Then, the relationship between reed biomass and these environmental factors were analyzed. The results showed that the reed in Bosten Lake could be divided into four classes （grades） of growth. Reeds grow best in the West Bank region of Big Lake, followed by the Southwest Small Lake Region, and growth is worst in the Huangshui Ditch Region. No significant change in community characteristics was observed. Soil and water quality is a key factor that influences the growth of reed. The effect of phosphorus on biomass is greater than nitrogen＇s, and soil organic matter has the least impact on reed biomass in Bosten Lake. To improve reed biomass, thedischarge of upstream effluents, both agricultural and industrial, needs to be controlled ; this method can also serve as the first step in protecting reed wetlands. The projection pursuit model can be used to analyze the relationship between reed biomass and its environmental impact factors, and the findings can reflect the main ecological and environmental problems in the Bosten Lake wetland as well as provide useful information for the environmental management of reed wetlands. The results of the projection pursuit model agree well with the actual situation, which indicates that this model is a powerful tool in research that involves multi-factor problems and could be a new method to conduct a synecological investigation of reed.