以江西省泰和县为研究区,探讨利用Landsat TM影像进行亚热带森林郁闭度森林遥感估算的可行性.首先,利用赤池信息准则(AIC)筛选了用于建模的遥感变量因子,通过模拟试验确定了估算森林郁闭度的前馈神经网络(BP)模型参数,利用检验样本数据验证了模型结果的可靠性.结果表明,遥感变量因子之间存在着强的相关关系,为了保证所建模型的效果需要剔除部分因子,利用BP神经网络建立的森林郁闭度模型的模拟和拟合精度都优于逐步回归模型,研究区内森林郁闭度较高,有47.8%的森林郁闭度大于0.7,但空间差异明显,西部和东部山区森林郁闭度高于中部丘陵森林的郁闭度.%This paper is to explore the applicability of Landsat TM data in the retrieval of forest canopy closure (FCC) in the subtropical region. Taihe county, Jiangxi province was taken as the studied area. Factors from remote sensing data were selected for modeling FCC by using the Akaike information criterion (AIC). The parameters in the BP - artificial neural network were determined through simulation experiments. The performance of the model was validated by the FCC measured with the TRAC instrument. The results showed that strong correlations existed among remote sensing factors. Some factors should be excluded to make the modeling results reliable. The simulative accuracy and predictive accuracy of the BP - artificial neural network model for estimating the FCC from remote sensing were higher than those of the stepwise regression model. FCC in the studied area was high and above 0. 7 for 47.8％ of forests. The spatial variability of FCC was noticeable, and FCCs were higher in the western and eastern mountainous areas than those in middle hilly areas.
展开▼
