云南草地资源地上生物量遥感分类建模方法研究

Research on remote sensing classification and modeling methods for aboveground biomass of grasslandresources in Yunnan

摘要:

【目的】 草地地上生物量（AGB）是衡量草地生产力和生态系统健康状况的重要指标。精准评估草地AGB 对于科学指导草地资源的开发利用、生态功能维持与修复至关重要。云南省地形复杂、气候多样，草地资源丰富且类型多样，探索利用卫星遥感数据构建云南省的AGB 分类建模方法。【方法】 基于草地资源调查收集的4 种草地类型实测样地数据，构建了针对云南全域的NDVI-VFCAGB分类建模方法体系：首先，利用坡向、海拔、纬度因子对全域草地资源进行4 种类型划分；其次，对4 种草地类型影像像元的NDVI 与样方VFC 建立NDVI-VFC 反演模型；接着，利用样方实测数据建立4 种类型的VFC-AGB 拟合模型；最后，对反演出的AGB 叠加全域草地资源图斑进行空间统计，得到各统计单元的AGB 数据。【结果】 基于样本统计进行简单4 类划分取得了约82% 的分类精度，基于此进行的VFC 与AGB 遥感建模反演，经样本抽样检查偏差分别为17. 21% 和18. 87%，取得了全省范围内草地资源AGB 的数量与分布，其统计结果与纯样地调查平均值基本一致。【结论】 卫星遥感NDVI反映了植物的覆盖度与长势，能够有效用于NDVI-VFC 建模并取得较高的精度。VFC-AGB 分类建模较之于单一建模方法，能够显著提升卫星遥感反演AGB 的精度。

Abstract:

【Objective】 AboveGround Biomass（ AGB） is an important indicator for measuring grassland productivityand ecosystem health. Accurately assessing grassland AGB is crucial for scientifically guiding the developmentand utilization of grassland resources，as well as for maintaining and restoring ecological functions. Yunnan has a complexterrain and diverse climate，with abundant and varied grassland resources. Exploring the use of satellite remotesensing data to develop an AGB classification modeling method for Yunnan Province.【Method】 Based on the fielddata of four grassland types collected from grassland resource surveys，a "NDVI-VFC-AGB" classification andmodeling system was developed for the entire Yunnan region. First，the grassland resources were classified into fourtypes using aspect，elevation，and latitude factors. Second，an "NDVI-VFC" inversion model was established by correlating the NDVI of image pixels with the VFC from the sample plots for the four grassland types. Next，a "VFCAGB"fitting model was constructed using field‐measured data from the sample plots for the four grassland types. Finally，spatial statistics were performed by overlaying the inverted AGB onto the grassland resource pattern across theentire region，yielding AGB data for each statistical unit.【Results】 A simple four‐class classification based on samplestatistics achieved a classification accuracy of approximately 82%. Utilizing this classification，remote sensing modelsfor VFC and AGB were developed and inverted. Sampling inspection of the model results showed biases of 17. 21%and 18. 87% for VFC and AGB，respectively. The models provided estimates of the quantity and distribution of AGBfor grassland resources across the entire province，with the statistical results being largely consistent with the averagevalues obtained from pure field plot surveys.【Conclusion】 Satellite remote sensing NDVI reflects vegetation coverageand growth and can be effectively used for "NDVI‐VFC" modeling with high accuracy. Compared to single modelingmethods，the "VFC-AGB" classification modeling can significantly improve the accuracy of satellite remote sensinginversion for AGB.