地形对植被生物量遥感反演的影响——以广州市为例

目前植被生物量遥感反演研究中的地形校正主要是校正地形变化对地表反射率的影响,较少考虑地形起伏引起的像元面积与实际地表面积的差异,而这种差异将导致植被生物量估算结果的偏差.在生物量遥感反演的基础上,结合地表面积计算模型和物质守恒定律,建立了生物量地形校正模型,定量分析和讨论了地形起伏对广州市植被类型面积提取和生物量准确估算的影响.结果表明:地形校正前后全市针叶林、阔叶林、草地、灌木林和园地面积分别增加6.18％、3.70％、2.86％、1.92％和1.29％；在综合分析区域生物量遥感反演中的各种不确定性的基础上,建立的各植被类型的生物量模型均具有较高精度,相关系数均接近或者超过0.9,可以满足生物量反演的要求；全市植被生物量呈现出东、北高,西、南低的分布格局,像元实际代表的林地(阔叶林和针叶林)平均生物量为61.86t/hm2,高于珠三角区域生物量平均值,但与亚热带林的顶级群落生物量水平有较大差距,林地生物量还有较大的增长空间；经过校正地形变化引起的像元面积和实际地表面积差异对生物量提取结果的影响后,植被总生物量比校正前增加了5.82％,5种植被类型的总生物量有不同程度的增加,阔叶林、针叶林、草地、灌木林和园地分别增加了7.74％,4.76％、3.34％、2.50％和1.58％.与其它的表面积计算模型相比,利用的像元地表面积模型具有较高的精度,可以满足生物量遥感估算中地形校正的需要.

Influence of terrain on plant biomass estimates by remote sensing: a case study of Guangzhou City, China

Terrain correction is typically used to adjust surface reflectivity for the influence of slope and aspect when plant biomass is estimated by remote sensing. There are very few studies on the differences between the pixel area of a remote sensing image and its real surface area. However, this difference inevitably influences the precision of quantitative remote sensing when the terrain is undulating. This paper addresses the effects of the difference between the pixel area of a remote sensing image and the real surface area of plant biomass obtained with remote sensing, when biomass-terrain correction model was calibrated according to surface-area calculation model and the law of the conservation of mass. The results show that the pixel area, influenced by undulating topography, was different from the actual surface area for Guangzhou city, China, and this accounted for 33.4% of the variability. Statistical values for the areas in different vegetation types were much higher after the terrain correction than before. The biggest proportions were for the coniferous forest and broad-leaved forest areas, at 6.18% and 3.70%, respectively. The grassland, shrubland, and orchard areas also increased by 2.86%, 1.92%, and 1.29%, respectively. Based on the uncertainty analysis for forest biomass estimation with remote sensing, remote sensing models with high accuracy were built for different vegetation types, and their correlation coefficients were close to or more than 0.9. Therefore, these models can be applied to estimating of plant biomass. Biomass of vegetation in the eastern and northern regions was much greater than that in the western and southern regions of Guangzhou. The average biomass of woodland (broad-leaved forest and coniferous forest) was 61.86 t/hm². This value was more than the average biomass in Pear River Delta but much lower than the biomass of the climax community of southern subtropical forest in Dinghushan. This suggests that the studied woodland biomass has a large potential for growth. The difference in biomass distribution due to terrain undulation was given by the biomass model obtained from remote sensing and the biomass-terrain correction model. The biomass of different vegetation types increased with terrain correction. The broad-leaved forest biomass increased by 7.74% which was the highest rate of change. Biomass of coniferous forest increased by 4.76%, that of grassland by 3.34%, that of shrubland by 2.50%, and that of orchard by 1.58%. Total vegetation biomass of Guangzhou increased by 5.82%. We conclude that the impact of terrain cannot be ignored because it was an important factor affecting the precision of biomass conversions. Compared with other models of surface area calculation, the pixel surface area model could be used with high accuracy to correct for the terrain impact of plant biomass estimation with remote sensing.