基于随机森林模型长汀县植被覆盖度的影响因子

应用随机森林回归模型对水土流失典型区域——长汀县2000—2010年植被覆盖度变化及主要影响因子进行分析.研究表明:长汀县近11 a植被覆盖度以每年4.11%的速率增加,其中2005年植被覆盖度最低(0.12),2010年植被覆盖度最高(0.70),且2006年之后植被覆盖度均发生正向突变;随机森林模型对植被覆盖度主要驱动因子的拟合效果较好,其观测值和预测值的方差解释率均在78.30%以上,相关性为0.885~0.939,极显著相关(P0.01),且均方残差都小于0.001;当地财政支出和农村人均纯收入、年平均地表比辐射率、年平均风速等人为因子和自然因子是影响长汀县植被覆盖度变化的主要因子.

Factors affecting vegetation coverage in Changting County based on random forest algorithm

Random forest regression model was used to analyze changes on vegetation coverage and its dominant factors influencing soil and water loss area in Changting County from 2000 to 2010. The results showed that vegetation coverage increased at an annual rate of 4.11% in recent 11 years, with the lowest vegetation coverage being 0.12 in 2005 and the highest value being 0.70 in 2010. Vegetation coverage mutated positively after 2006. Random forest model had a satisfactory fitting effect on main driving factors of veg-etation coverage. The interpretation rate of the observed and predicted values was above 78.30%, with correlation ranging between 0.885-0.939 (P<0.01). The mean square residuals were less than 0.001. To summarize, human factor, including local fiscal ex-penditure, rural per capita, net income, and natural factors including annual average surface area emissivity and annual average wind speed, were the dominant factors determining the trend of vegetation coverage in Changting County.