Estimating the age of deciduous forests in northeast China with Enhanced Thematic Mapper Plus data acquired in different phenological seasons
Estimating the age of deciduous forests in northeast China with Enhanced Thematic Mapper Plus data acquired in different phenological seasons
复制标题
DOI：
10.1117/1.jrs.8.083670
复制
发表时间：
2014
影响因子：
1.7
通讯作者：
Dengqiu Li;W. Ju;W. Fan;Z. Gu
中科院分区：
文献类型：
--
作者：
Dengqiu Li;W. Ju;W. Fan;Z. Gu
摘要本文研究了中国东北地区落叶松林和白桦林分年龄的估算能力，并利用Landsat Enhanced Thematic Mapper Plus遥感数据对落叶松林和白桦林分年龄进行了估算。研究了六个波段反射率、九个植被指数和六个质地指标与林龄的关系。利用线性回归和多元回归模型以及多层感知器神经网络（MLP NN）对林龄进行了估计。结果表明，在开叶图像中，短波红外波段反射率和湿度与林龄的相关性较强，而在关叶图像中，蓝、绿色波段反射率和绿度与林龄的相关性较强。MLP神经网络模型可以有效地用于林分年龄的反演，落叶松和白桦的最大决定系数和最小均方根误差值分别为0.47和21.3年，白桦分别为0.60和10.1年。落叶松和白桦的林龄分别在> 100年和> 50年时，预测误差显著增大。在研究区，在开叶季节获得的遥感数据比在休叶季节获得的数据更适合于估算林龄。
Abstract This study investigated the ability of Landsat Enhanced Thematic Mapper Plus data acquired in leaf-on and leaf-off seasons to estimate stand age of Larix gmelinii and Betula platyphylla in northeast China. The relationships of six band reflectances, nine vegetation indices, and six texture measures with stand age were examined. Linear and multivariable regression models and multilayer perceptron neural network (MLP NN) were employed to estimate forest age based on these variables. The results indicate that reflectance in short-wave infrared bands and wetness are more significantly correlated with stand age in the leaf-on image, while reflectance in blue and green bands and greenness are more sensitive to stand age in leaf-off image. The MLP NN model can be effectively used to retrieve the stand age; the highest coefficient of determination and minimum root mean square error values of retrieved age are 0.47 and 21.3 years for Larix gmelinii, and 0.60 and 10.1 years for Betula platyphylla, respectively. The predicted age errors increased significantly when stand ages were > 100 and > 50 years for Larix gmelinii and Betula platyphylla, respectively. Remote sensing data acquired in the leaf-on season is more suitable for estimating forest age than that acquired in the leaf-off season over the study area.