基于多植被指数模型的草地地上生物量协同估算

以青海省金银滩草原为研究区，采用Sentinel-2卫星影像结合地面实测数据进行草地地上生物量估算研究。分析了18种典型植被指数与生物量的拟合关系，通过精度评价和敏感性分析确定了不同植被指数模型的适用范围，并提出基于多植被指数模型的协同估算方案来提高草地生物量的制图精度，尝试克服传统单变量植被指数模型适用范围受限的问题。结果表明：18种植被指数与生物量的最优拟合模型呈现幂函数和指数函数两种类型，其中幂函数模型中CI_green (Green chlorophyll index)所对应的估算精度最高，且当生物量高于0.65 kg·m^-2时适用性最强；指数函数模型中NDII（Normalized difference infrared index）所对应的估算精度最高，且当生物量低于0.65 kg·m^-2时适用性最强，且NDII与CI_green模型的适用范围具有互补性。提出的多植被指数协同估算模型对应的R²_cv达到了0.61，RMSE_cv为0.226 kg·m^-2，相对于单植被指数模型精度明显提高，R²_cv增加7.0%以上，RMSE_cv减小超过3.8%。综上，提出的多指数模型协同估算方案充分考虑了不同指数模型的适用范围，提高了牧草生物量的估算精度。

Collaborative estimation of aboveground biomass in grassland based on multiple vegetation index models

This study selected Jinyintan grassland in Qinghai Province as the study area. The study used Sentinel-2 satellite images combining with ground measurement data to estimate the aboveground biomass of grassland. Specifically, the study analyzed the fitting relationship between 18 typical vegetation indices and biomass, determined the applicability of different vegetation index models through accuracy evaluation and sensitivity analysis, and then proposed a collaborative estimation scheme based on multiple vegetation index models to improve the accuracy of grassland biomass mapping. The results showed that the optimal fitting models of 18 indices presented two trends of power function relation and exponential function relation. The CI_green (Green chlorophyll index) had the best estimation accuracy among power function models, which was the most applicable when the biomass was higher than 0.65 kg·m^-2. The NDII (Normalized differential infrared index) showed the best verification accuracy in exponential function models and had the strongest applicability when the biomass was lower than 0.65 kg·m^-2. Moreover, the NDII and CI_green models had complementary applicability ranges. The R²_cv and RMSE_cv of the proposed method reached 0.61 and 0.226 kg·m^-2, respectively. Compared with the single index models, the R²_cv increased about 7.0% and the RMSE_cv decreased around 3.8%, which significantly improved in precision. In summary, the proposed collaborative estimation scheme based on multiple vegetation index models fully considered the applicability scope of different index models and improved the accuracy of forage biomass estimation.