三峡库区森林郁闭度的遥感定量估测

Remote sensing estimation of forest canopy closure in forests of Three Gorges Reservoir Region

以三峡库区为例,利用森林资源一类清查数据和TM遥感影像为基础数据,分析了遥感因子和地形因子与森林郁闭度的定量关系,在此基础上,利用逐步回归、最小二乘回归、主分量回归和偏最小二乘回归构建了郁闭度定量估测模型。结果表明,TM1、TM2、TM3、TM5、TM7、TM7/3、TM4*TM3/TM7、亮度和湿度9个因子与森林郁闭度具有极显著相关,可以作为建模的自变量;所建4种森林郁闭度估测模型的各项统计指标差异不显著,但检验结果反映最小二乘回归模型效果略差;从参数可解释性和模型实用性角度考虑,主分量回归和偏最小二乘回归方法具有普遍的推广意义。本研究为今后森林郁闭度定量估测提供了一定的理论依据,也为其它基于遥感信息估测模型的建立提供了可行方法。

Based on the national forest inventory data and TM images in Three Gorges Reservoir Region, firstly, the quantitative relations among forest canopy closure and remote sensing factors and topographical factors were analyzed. Then the quantitative estimation models for forest canopy closure in Three Gorges Reservoir Region were setup by using stepwise regression, least squares regression, principal component regression and partial least squares regression. The results show that the 9 factors, such as TM1, TM2, TM3, TM5, TM7, TM7/3, TM4＊TM3/TM7, brightness and wetness, had extremely significant relationships at 0.01 level to the forest canopy density, which were selected as independent variables for the models; the differences among the statistical indicators of the four models were not significant, but the test results showed that the least squares regression model＇ s effect was not as well as the other three models in the test of validation data; considering the explanatory of parameters and applicability of models, the principal component regression model and partial least squares regression model were worthy to be recommended for popularization and application. The findings provide a theoretical basis for quantitative estimation of forest canopy closure and a feasible approach for constructing other regression models based on remote sensing information.