基于GF-1影像的普达措国家公园森林地上生物量遥感估算

精确估算森林地上生物量有利于掌握森林资源碳储量的分布特征,该研究以普达措国家公园为研究区,基于国产高分一号(GF-1)全色多光谱(Panchromatic Multispectral Sensor,PMS)卫星影像和数字高程数据,提取波段信息、植被指数、纹理信息和地形因子,利用多元线性逐步回归、支持向量机、神经网络和随机森林模型,估算森林地上生物量。研究结果表明,基于GF-1影像构建的随机森林模型的精度效果最佳,决定系数为0.77,均方根误差为27.53 t/hm2;普达措国家公园森林地上生物量为7085614t,平均生物量达136.01t/hm2,表明公园内寒温性针叶林发育完好;海拔3 500～4000m区域森林生物量平均值最高,为126.56t/hm2,与生态保护目标分布范围相符;不同坡向生物量存在差异,阴坡和半阴坡平均生物量高出其他坡向20.48%,立地条件较优。研究结果证实基于GF-1优化的生物量经验模型具有对亚高山天然林地上生物量的估算潜力,对区域森林资源的有效科学管理和维护森林生态环境具有重要意义。

Remote sensing estimation of forest aboveground biomass in Potatso National Park using GF-1 images

Abstract: Potatso National Park is an important ecological functional area in the northwest Yunnan Plateau. The study of forest aboveground biomass in the Potatso National Park is conducive to the understanding of forest resources and biomass distribution characteristics in the subalpine regions, which is of great significance to the monitoring of regional forest resources. In this study, four empirical models including the multiple Linear Step Regression (MLSR), Support Vector Machine (SVM), Back Propagation Neural Network (BPNN), and Random Forests (RF) were established to estimate the aboveground biomass of forest land in the Potatso National Park. Biomass samples were obtained by empirical conversion formulas from a forest resources survey. 105 factors of forest biomass were obtained from domestic GF-1 satellite images and classified into four categories (band information, vegetation indices, texture information, and topography factors). Then, the significant importance variables were introduced into four empirical models as independent variables, and the estimation models of aboveground biomass of forest in the region were established. In addition, models were compared and the optimal model was selected to estimate the Aboveground Biomass (AGB), and the aboveground biomass distribution of the region forests was analyzed and compared. The results showed that 1) GF-1 images achieved a high precision in the estimation of aboveground biomass of forests in the Potatso National Park, and the non-parametric models were superior to the linear model, and the random forest model (the coefficient of determination was 0.77, and the root mean square error was 27.53 t/hm2) with the best comprehensive performance and reliable estimation results. 2) The total biomass of the main forest in the Potatso National Park was estimated by the random forest model to be 7 085 614 t, with an average of 136.01 t/hm2. And the sum areas of slightly high and medium biomass accounted for 67.1% of the forest area in the study area, indicating that the alpine and subalpine cord-temperate needle-leaved forest in the park was well developed, and there was the most primitive spruce forest. 3) The elevation range of the park was 2 308-4 550 m, and the forest biomass at an altitude of >3 500-4 000 m was the highest, with an average of 126.56 t/hm2, accounting for 62% of the total area, which was consistent with the elevation distribution range of the protection target "natural cord-temperate needle-leaved forest and evergreen broad-leaved forest". 4) the Potatso National Park was dominated by spruce forest and fir forest, and the forest community remained in the original state. There were differences in forest biomass on different slopes. The aboveground biomass on shady and half shady slopes was 20.48% higher than that on other slopes, and the site conditions were relatively better. These results confirmed that the biomass empirical model based on GF-1 optimization could quickly and accurately estimate the aboveground biomass of natural forests, and could be used as a reference for estimation of forest biomass by high-resolution satellite remote sensing in subalpine areas.