Received date: 2019-07-25
Revised date: 2019-12-27
Online published: 2020-05-25
Qinghai Province is one of the four major pastoral areas in China.It is of great significance to monitor the growth of grassland vegetation and accurately estimate grass yields for the sustainable development and ecological protection of the Qinghai pastoral area.Remote sensing estimation of grass yields is mainly based on statistical relationship between vegetation index and ground measured data.However,the selection of different vegetation indexes,various statistical models,and optional modeling metrics affect the accuracy of the estimation results.Based on the ground measured grass yield data of the Qinghai Province and the Moderate Resolution Imaging Spectroradiometer (MODIS) data,this paper selected six vegetation indexes,five statistical models,and three modeling metrics to test the accuracy of different estimation models.The vegetation indexes included the Normalized Difference Vegetation Index (NDVI),Enhanced Vegetation Index (EVI),Ratio Vegetation Index (RVI),Difference Vegetation Index (DVI),Renormalized Difference Vegetation Index (RDVI),and Modified Soil Adjusted Vegetation Index (MSAVI).The statistical models consisted of a simple linear model,quadratic polynomial model,power function model,exponential function model,and logarithmic function model.The modeling metrics included the annual maximum value of vegetation index,cumulative value of vegetation index in growing season,and annual cumulative value of vegetation index.First,we explored which vegetation index and estimation model were the most suitable for estimating the grass yields of different grassland types in the Qinghai Province.Then,we compared the accuracy of estimated grass yields from the three modeling metrics.Finally,we selected the optimal grass yield estimation models to estimate grass yields of the Qinghai Province in 2015 and 2016 and analyzed grass yield change during 2015-2016.The results showed as follows:(1) among the six vegetation indexes,NDVI was the most suitable vegetation index to estimate grass yields in the Qinghai Province.The accuracy of nonlinear models was generally higher than that of linear models,especially the exponential models,which are suitable for estimating grass yields of most grassland types.The estimation models based on the annual maximum of NDVI showed the highest coefficient of determination (R2) for almost all grassland types.(2) from the perspective of hay yields,the high-yield area (>1 200 kg·hm-2) was mainly located in the alpine steppes in eastern Qinghai,and the middleclass area (600-1 200 kg·hm-2) was located in the grass steppes and the alpine steppes in the north and east of Qinghai.Grasslands with lowyield (<600 kg·hm-2) were located in the alpine meadows,alpine steppes,alpine deserts,and salt meadows in the western and northern part of Qinghai.(3) the total hay production in the Qinghai Province was 31.60 × 104 t in 2016,showing 1.36% decrease compared to 2015.Among all,the grass steppes and alpine meadows showed the highest yield loss,while in the desert steppes and salt meadows,the yield increased.This study not only provided a reference for ecological protection and sustainable development in the pastoral area of Qinghai,but also provided research and practices for estimating grassland yield using remote sensing technology.
Key words：
grass yield
HUANG Xin, LIU Jian-hong, SHEN Ke-jian, LIU Yong-mei, WANG Lei
.
Grassland yield change in Qinghai Province based on
MODIS data[J]. Arid Land Geography, 2020
/
| 〈 |
|
〉 |
