引入地形特征的田块尺度玉米遥感估产与空间格局分析  

Maize Yield Estimation and Spatial Pattern Analysis Based on Topographic Features by Remote Sensing at the Field Scale

作　　者：叶强[1] 杨凤海[1] 刘焕军[1,2] 孟祥添 官海翔 崔杨 YE Qiang;YANG Feng-hai;LIU Huan-jun;MENG Xiang-tian;GUAN Hai-xiang;CUI Yang(School of Public Administration and Law, Northeast Agricultural University, Harbin 150030, China;Northeast Institute of Geography and Agroecology, Chinese Academy of Sciences, Changchun 130012, China)

机构地区：[1]东北农业大学公共管理与法学院,哈尔滨150030 [2]中国科学院东北地理与农业生态研究所,长春130012

出　　处：《科学技术与工程》2021年第24期10215-10221,共7页Science Technology and Engineering

基　　金：国家自然科学基金(41671438);吉林省发改委创新能力建设项目(2021C044-10)。

摘　　要：田块尺度作物估产引入地形特征提升精度与产量空间格局分析对规模经营具有重要意义。以黑龙江省规模化种植的春玉米为研究对象,测定无人机(unmanned aerial vehicle,UAV)高精度地形与变量测产数据,基于多时期SPOT-6影像提取7种植被指数;采用最小二乘法,构建不同时期植被指数与春玉米实测产量的经验统计模型,确定遥感估产最佳时期和最优植被指数;提取6种地形因子,使用多元逐步回归评价引入地形因子的遥感估产模型,应用空间统计分析探索产量空间分布格局。结果表明:春玉米灌浆期是遥感估产的最佳时期,决定系数R 2达到0.6以上的植被指数共6种,比值植被指数(ratio vegetation index,RVI)为最优植被指数,其余依次为修正比值植被指数(modified simple ratio,MSR)、无蓝色波段增强型植被指数(enhanced vegetation index without a blue band,EVI2)、归一化植被指数(normalized difference vegetation index,NDVI)、次生修正土壤调节植被指数(modified secondary soil adjusted vegetation index,MSAVI2)、绿度归一化植被指数(green normalized difference vegetation index,GNDVI);最佳估产模型引入地形辅助信息后R^(2)提升5.6%,达到0.79,均方根误差(root mean squared error,RMSE)为347.03 kg/hm^(2);高海拔与高坡度区域产量均值最低为7502.64 kg/hm^(2),中海拔与低坡度区域产量均值最高为9157.63 kg/hm^(2)。优化后的遥感估产模型可以快速评估作物产量,确定春玉米最佳生长区域,为规模化农业精细管理、土地整治与作物种植结构调整提供科学依据。Improving the accuracy of remote sensing yield estimation and spatial pattern analysis of yield by introducing topographic factors is important for scale management.The research object is spring maize grown on large-scale in Heilongjiang Province.The high-precision terrain was obtained by the unmanned aerial vehicle(UAV)platform,and the variable yield data were acquired by the inte-lligent yield measurement system installed on large-size harvester.Seven vegetation indices were extracted based on multi-period SPOT-6 satellite images,and six terrain factors were calculated from high-precision terrain data.In order to identify the optimal period and the optimal vegetation index for remote sensing yields estimation.The least squares method was utilized to construct empirical statistical models based on vegetation indices at different time periods and the measured yield of spring maize.And multiple stepwise regression models were used to assess remote sensing yield estimation models with the introduction of topographic factors.Spatial statistical methods were used to quantitatively describe the spatial distribution of yields.The results show that the filling stage is the best period for corn yield estimation by remote sensing.The coefficient of determination R2 of six vegetation index models reached above 0.6,and ratio vegetation index(RVI)is the optimal vegetation index,followed by modified simple ratio(MSR),enhanced vegetation index without a blue band(EVI2),normalized difference vegetation index(NDVI),modified secondary soil adjusted vegetation index(MSAVI2)and green normalized difference vegetation index(GNDVI).The R^(2) of the optimal yield estimation model integrates with topographic factor and vegetation index increase by 5.6%to 0.79,and root mean squared error(RMSE)is 347.03 kg/hm^(2).The results of spatial pattern analysis show that the lowest average yield is 7502.64 kg/hm^(2) in high altitude and high slope areas,and the highest average yield is 9157.63 kg/hm^(2) in middle altitude and low slope areas.The o

关 键 词：农作物 产量估算 地形 植被指数 空间格局 

分 类 号：S513[农业科学—作物学] S127