NDVI相似性分区下天山地区草地总产草量遥感估算

Estimation of the total production of the herbage in the Tianshan Mountain Area using remote sensing technology with NDVI similarity zoning

山区草地总产草量遥感估算是定量评价区域牧业生产力的有效手段。常规总产草量地面观测数据准确性较高,但无法覆盖整个天山山区,且耗时耗力。针对此问题,以新疆天山山区为研究区,选取MODIS/MOD13Q1 250m植被指数(normalized difference vegetation index,NDVI)产品数据,以县(市)为单元,基于巴氏距离定量评价研究区植被指数分布区域相似性以得到有效遥感建模分区,在此分区基础上,结合草地总产草量实测数据,建立研究区植被指数-草地总产草量遥感估算模型。结果显示,1)基于各县(市)2009-2015年7月底至8月初植被生长期多年NDVI均值直方图计算巴氏距离,以巴氏距离d>0.5为阈值,研究区被划分为7个遥感建模区;2)各分区内NDVI-草地总产草量数据拟合方程形式不同,有线性、指数、幂指数和多项式回归方程几种形式。总体来看,各分区NDVI-草地总产草量拟合相关系数在0.784~0.836。交叉检验除天山北坡西段-伊犁河谷草原畜牧业区RMSE值在2 951kg·hm-2外,其他分区RMSE值均在266~928kg·hm-2,原因在于伊犁河谷草原畜牧业区实测草地总产量在10 000~30 000kg·hm-2的样点居多,区域草地总产量较其他区域多。

An estimation of the total production of herbage in Xinjiang using remote sensing technology is an effective method for quantitative evaluation of regional animal husbandry productivity.It is time-consuming and cannot cover the whole area of the Tianshan Mountains,although the measured accuracy of the total herbage yield using conventional means is very high.To address this problem,we used MODIS/MOD13 Q1 vegetation index products,with 250 mGSD as experimental data and city or county as a basic unit for analysis.The Bhattacharyya distance was used to quantitatively evaluate the distribution of the similarity of the vegetation index in the Tianshan Mountain Area as a study case.The purpose was to obtain an effective remote sensing modeling zoning,and then to construct an estimation model of the total production of herbage with respect to the vegetation index using remote sensing technology.Finally,the spatial distribution and feature analysis of the total herbage yield（fresh weight）in the Tianshan Mountains from the year 2009 to 2015 were determined based on the analysis of the spatial distribution and characteristics obtained under the GIS platform.The results showed the following：the seven modeling zones were derived from an analysis of the mean histogram of NDVI data collected during the optimal period（i.e.,July and August each year）of vegetation growth for each city or county in the study area via the Bhattacharyya distance with a threshold greater than 0.5.Secondly,the constructed estimation model of the total production of the herbage showed different fitting relationships to the vegetation index,and there were three forms,including the exponential,power index,and a unary regression equation with the second order.On the whole,the fitting correlation coefficient of the constructed estimation model could reach between 0.754 and 0.836 for each zones.The RMSE value of cross-validation in the northern slope of the Tianshan Mountains-Yili Valley was 2 951 kg·ha^-1,and the RMSE value were between 266 and 928 kg·ha^-1 in the other zones.This was because more measured samples which total production of the herbage was between 10 000 and 30 000 kg·ha^-1 were collected in the zone.Also there was higher total production of the regional herbage in the zone than others.