基于meta分析的中国森林生态系统服务价值评估

meta分析作为价值转移方法中最有效的手段之一,已在生态系统服务价值评估中得到了较为广泛的应用。然而国内相关研究的meta回归模型大多依据传统的最小二乘法建立,忽略了原始数据的层次结构特征。通过收集关于中国森林生态系统服务已有实证研究的价值评估结果,建立meta分析数据库;通过面板数据回归方法构建meta回归模型,对模型的价值转移有效性进行评估;在构建的meta回归模型基础上,根据IPCC SRES中的四类情景——A1B、A2、B1、B2,计算中国森林生态系统2010-2100年的生态系统服务价值。研究结果表明：（1）相较基于普通最小二乘法和加权最小二乘法建立的回归模型,基于面板数据回归方法建立的模型平均转移误差最小（11.57%）,模型有效性较高,因此为适合本研究的meta回归方法;对于不同价值观察值,转移误差存在较大差异,随着观察值的增大,模型预测值由高估逐渐转变为低估,转移误差趋向减小。（2）基于面板数据回归方法建立的meta回归模型能够有效揭示中国森林生态系统服务的价值转移规律,生态系统服务类型、植被区划、森林面积、森林丰度、人均GDP和铁路长度是中国森林生态系统服务价值的重要影响因素。（3）中国森林生态系统2010-2100年价值变化的情景分析表明,情景A1B和B1下森林面积和生态系统服务总价值持续增加,情景A2下森林面积和生态系统服务总价值持续下降,情景B2下森林面积和生态系统服务总价值先上升后下降;其中情景B1下中国森林生态系统服务总价值增长最大,至2100年达到41.58万亿元,情景B2下价值损失最为显著,至2100年降至22.97万亿元。

Evaluation of forest ecosystem services value in China based on meta-analysis

As one of the most effective methods of value transfer, meta-analysis has been widely used in ecosystem services evaluation in recent years. However, most of the meta-regression models used in relevant studies in China are based on the traditional least squares method, which ignores the hierarchical structure of the original data. In this research, considering the data correlation and hierarchy in relevant studies, we established a meta-analysis database by collecting the results of forest ecosystem services evaluation of existing empirical studies in China from 1990 to 2019. Then we used the panel data regression method to build a meta-regression model and assessed the model''s effectiveness for value transfer. Finally, on the basis of the established meta-regression model, we calculated the forest ecosystem services value in China from 2010 to 2100 under the four scenarios in the IPCC SRES:A1B, A2, B1, and B2. The results showed the following:(1) The model based on the panel data regression method was the most effective with the average transfer error of 11.57%, compared with the model based on the ordinary least squares method and the weighted least squares method. Therefore, the panel data regression was considered the most suitable method for this research. What''s more, this model performed better for higher observed values. When the observed value was small, the predicted value of the model was mostly higher than the observation and the transfer error was relatively high; when the observed value was large, the predicted value of the model was mostly lower than the observation and the transfer error was relatively low. (2) The meta-regression model based on the panel data regression method could effectively reveal the law of value transfer of forest ecosystem services in China. It affirmed that ecosystem service types, vegetation zoning, forest area, forest abundance, per capita GDP, and railway length were important factors affecting forest ecosystem services value in China. (3) The scenario analysis of the forest ecosystem services value in China from 2010 to 2100 revealed that the forest area and the total ecosystem services value continued to increase (decrease) under scenario A1B and scenario B1 (scenario A2), which both increased first and then decreased under scenario B2. Among these four scenarios, the total ecosystem services value in China showed the most significant increase under scenario B1, reaching 41.58 trillion yuan by 2100. On the contrary, the total ecosystem services value lost most under scenario B2, which decreased to 22.97 trillion yuan by 2100.