基于SPOT-7影像的马尾松生物量估测模型研究

Study on biomass estimation model of masson pine based on SPOT-7 image

森林生物量的定量估算为全球碳储量循环及气候变化研究提供了重要的参考依据。以肇庆市马尾松生物量为研究对象,基于SPOT-7影像数据,提取出单波段、波段比值、纹理特征、地形等因子并进行优选,对优选后的变量因子进行分组,生成光谱特征、纹理特征、光谱和纹理特征相结合的3种自变量集,采用偏最小二乘回归法和3种自变量集构建马尾松生物量估测模型并进行优选。结果显示:基于光谱特征构建的马尾松生物量估测模型的决定系数R2为0.81,均方根误差为15.30 t/hm2,误差平均值为-5.65 t/hm2,总预报偏差的相对误差为9.30%;基于纹理特征构建的马尾松生物量估测模型的决定系数R2为0.84,均方根误差为15.69 t/hm2,误差平均值为-4.73 t/hm2,总预报偏差的相对误差为7.78%;基于光谱和纹理特征相结合构建的马尾松生物量估测模型的决定系数R2为0.83,均方根误差为15.24 t/hm2,误差平均值为-5.27 t/hm2,总预报偏差的相对误差为8.67%。由此可知,基于纹理特征构建的马尾松生物量模型要好于其他两种方法,估算精度更高。

The quantitative estimation of forest biomass provides an important reference for the study of global carbon storage cycle and climate change. Taking the biomass of Masson pine in Zhaoqing city as the research object, single band,band ratio, texture feature and terrain factor were extracted from the SPOT-7 image and DEM data, meanwhile, the optical selection of these factors are carried out and the preferred factors were divided into three categories： spectral feature, texture feature, and integration of spectral and texture feature. The estimation models of Masson pine biomass were built and analyzed based on Partial Least Squares Regression（PLSR） method and three different sets of factors. The results showed that the determination coefficient（R2）, the root mean square error（RMSE）, average error（AE） and relative error（RE） between estimated and measured values based on spectral feature and PLSR were 0.81, 15.30 t/hm2,-5.65 t/hm2, 9.30%, the R2, RMSE, AE and RE values of model based on texture feature and PLSR were0.84, 15.69 t/hm^2,-4.73 t/hm^2, 7.78%, the R2, RMSE, AE and RE values of model based on integration of spectral feature and texture feature and PLSR were 0.83, 15.24 t/hm^2,-5.27 t/hm^2, 8.67%. Therefore. The Masson Pine biomass model constructed based on texture feature was better than the other two methods, and the higher estimation accuracy was also acquired.