论文部分内容阅读
为提高棉花遥感估产精度,该文选取加州San Joaquin Valley地区2个棉花地块作为研究区,利用时间序列Landsat_5_TM、Landsat_7_ETM遥感影像数据,结合野外实测产量数据,进行棉花产量遥感预测模型研究。结果表明:基于Landsat影像纯像元的植被指数时间序列准确地揭示了棉花整个生长期的长势情况,不同长势的棉花植被指数随时间变化在花铃期差异比较显著;整个花铃期植被指数与产量之间的相关系数均大于0.80,最大相关系数达0.90,花铃期NDVI平均值建模决定系数为0.82,均方根误差为463.69,证明花铃期比其他生长期更适用于棉花产量预测;单一时期最优模型为第206天(7月25日),多时期最优模型以NDVI最大值前三期NDVI平均值为自变量;整个花铃期NDVI最大值建模决定系数为0.81,均方根误差为477.82,该模型具有普适性。该文的研究成果为基于MODIS_NDVI最大值合成法的相关研究提供了理论依据,并且为其他农作物的估产模型建立提供借鉴。
In order to improve the yield accuracy of cotton remote sensing, two cotton plots in San Joaquin Valley, California were selected as the study area. Using the time series Landsat_5_TM and Landsat_7_ETM remote sensing image data, combined with the field measured yield data, the cotton yield prediction model was studied. The results showed that the time series of vegetation index based on Landsat pure pixel accurately revealed the growing status of cotton during the whole growth period. The cotton growing index under different growing time was significantly different at the flowering and boll stage; The correlation coefficient between the two cultivars was more than 0.80, the maximum correlation coefficient was 0.90, the average coefficient of NDVI was 0.82 and the root mean square error was 463.69 in flowering and boll stage, which proved that flowering period was more suitable for cotton yield forecast than other growing periods ; The optimal model for the single period was Day 206 (July 25). The best model for multiple periods was the average NDVI of the first three periods of NDVI as the independent variable; the maximum coefficient of determination of NDVI for the entire flower-boll stage was 0.81, The root mean square error is 477.82, which is universal. The research results of this paper provide a theoretical basis for the related research based on MODIS_NDVI maximum value synthesis method, and provide reference for the establishment of other crop yield estimation models.