基于植被指数估算天山牧区不同利用类型草地总产草量

针对天山牧区草地面积广、生长环境差异大等特点,选用MODIS/MOD13Q1数据,生成归一化植被指数、增强植被指数、土壤调节植被指数和差值植被指数4种从不同角度反映牧草长势的植被指数,对4个不同牧草利用类型分区分别建立了4种指数及其组合与总产草量的4类回归方程。利用留一交叉检验法评价各模型精度,最终获得适合不同草地利用类型的产草量遥感估算模型。结果表明:4种植被指数都可用于产草量估算,不同植被指数估算模型的拟合精度有区域差异性。当采用SAVI和二次多项式拟合时,RMSE最大值出现在I区,为5 857 kg/hm2,当采用NDVI和二次多项式拟合时,最小值出现在III区,仅为616.487 kg/hm2。值得注意的是采用差值植被指数、土壤调节植被指数估算产草量需考虑植被覆盖状况。其次,多个植被指数组合有信息互补的优势,采用线性回归模型估产时,多个植被指数组合精度高于单一植被指数。该研究可为天山地区草地总产草量遥感估算提供有意义的借鉴。

Estimation of total yield of different grassland types in Tianshan pastoral area based on vegetation index

Abstract: The rapid and precise estimation model of herbage yield using remote sensing technology is of great significance to maintain grassland ecological balance, reasonably manage the yield of animal husbandry and determine carrying capacity. Taking the Tianshan pastoral area of Xinjiang into account, the vegetation index NDVI (normalized difference vegetation index), the vegetation index EVI (enhanced vegetation index), the soil conditioning index SAVI (soil adjusted vegetation index) and the difference index DVI (soil adjusted vegetation index) were calculated using the MODIS/MOD13Q1 data with 250 m spatial resolution. And the 4 indices and their different combinations were used to establish a regression model related to total herbage yield (fresh weight) of grassland for each subregion divided by herbage use type. Then the leave-one-out cross validation method was used to evaluate the precision of the model and the difference in simulation of different vegetation indices in the same pastoral area so as to obtain the pasture herbage yield - optimal vegetation index estimation model in the cases of different pasture use types. So, spatial distribution and difference of the herbage yield (fresh weight) in Tianshan mountainous area could be estimated. The results showed that estimation precision using nonlinear regression equation with respect to single vegetation index is higher than that of the linear regression equation. There is a nonlinear relationship between the herbage yield and the vegetation index derived from remote sensing image. Secondly, the precision of grass yield estimation model based on the combination of multiple vegetation indices is higher than that of the model based on one vegetation index in that the use of the combination of complementary information from different vegetation indices is preferred to improve the estimation precision of grass yield with the remote sensing technology. Additionally, the 4 indices including NDVI, EVI, DVI and SAVI were correlated to the herbage yield of pasture and could be used to estimate the herbage yield. However, there is a difference in estimation precision for different vegetation indices, the maximum of RMSE (root mean square error) when fitting with quadratic polynomial and SAVI occurs in the Subregion I, which is 5 857 kg/hm2, and the minimum of RMSE when fitting with quadratic polynomial and NDVI occurs in the Subregion III, which is 616.487 kg/hm2. It is noteworthy that both DVI and SAVI considering environment factors especially like vegetation cover circumstance in the study area are most suitable for herbage yield monitoring. The work can provide significant suggestion for the estimation of total production of herbage in Tianshan Mountains.