摘要:雪岭云杉林是新疆天山山脉重要的水源涵养林，精确估算雪岭云杉林生物量及准确表征空间格局特征对其生态系统的生物生产力和生态服务功能的评估具有重要作用。结合Landsat 8 OLI遥感数据和66块天山雪岭云杉林样地调查数据，选择包括各波段灰度值、不同波段灰度值之间的线性和非线性组合（包括5种植被指数）以及环境因子在内的42个自变量，分别采用多元逐步回归分析法、偏最小二乘法和主成分分析法建立天山雪岭云杉林生物量估测模型。结果表明：多元逐步回归法采用3个自变量所建模型平均拟合精度为69.07%，绝对误差为64.50 t/hm²，平均相对误差为10.89%，样地生物量实测值与预测值相关系数为0.465；偏最小二乘回归法采用11个自变量所建模型平均拟合精度为74.36%，绝对误差为144.94 t/hm²，平均相对误差为28.78%，相关系数为0.717；主成分分析方法提取3个主成分，所建模型平均拟合精度为71.22%，相关系数为0.730；因此偏最小二乘法要优于主成分分析法和多元逐步回归法。天山雪岭云杉林生物量随经纬度的增加而降低，整体呈现西部高，中东部低的趋势；随海拔的升高呈"单峰"型变化；样地生物量主要分布在山脊位置，随坡度的增加呈先降低后升高，然后再下降的趋势；随着阴坡-阳坡的坡向变化，样地生物量逐渐降低。



Abstract:Picea schrenkiana (Schrenk's spruce) forest is a major water conservation forest located in the Tianshan Mountains of Xinjiang Province.Therefore, estimating the biomass of Picea schrenkiana forest accurately and characterizing its spatial pattern exactly are significant for assessing the biological productivity and ecological service function of its ecosystem. Based on the data acquired by Landsat 8 OLI in remote sensing as well as a survey of 66 sample plots of Picea schrenkiana forest in the Tianshan Mountains, this study selects 42 independent variables including the gray value of each band,the linear and non-linear combinations of gray values in different bands (including 5 vegetation indexes), and adopts the multiple stepwise regression analysis method, the partial least squares method, and principal component analysis method to establish a biomass estimation model for the Picea schrenkiana forest in the Tianshan Mountains. The results reveal as follows: the multiple stepwise regression method adopts three independent variables to establish a model. In this model, the average fitting accuracy is 69.07%, the absolute error is 64.50 t/hm², the average relative error is 10.89%, and the correlation coefficient between the measured value and predicted values of the biomass in the sample plots is 0.465; the partial least squares regression method applies 11 independent variables to establish a model. In this model, the average fitting accuracy is 74.36%, the absolute error is 144.94 t/hm², the average relative error is 28.78%, and the correlation coefficient is 0.717; the principal component analysis method extracts 3 principal components to establish a model. In this model, the average fitting accuracy is 71.22%, and the correlation coefficient is 0.730. Hence, the partial least square method is better than the principal component analysis method and the multiple stepwise regression method. The biomass of the Picea schrenkiana forest in the Tianshan Mountains decreases with the increase of latitude and longitude. The overall trend is high in the west and low in the middle and the east; it changes in a ″single peak″ shape with increasing altitude; the biomass of the sample plots is mainly distributed in the ridge position, which tends to decrease first, then increases with the increase of slope, and decreases ultimately; with the change of slope direction (from the shaded slope to the sunny slope), the biomass of the sample plot decreases gradually.





PDF全文下载地址:

https://www.ecologica.cn/stxb/article/pdf/stxb201904240841