耕地质量数据在冬小麦遥感估产中的应用研究

Wheat Yield Estimation Based on Remote Sensing and the Cultivated Land Quality Data

耕地质量作为非遥感信息的一种,对农作物产量的影响至关重要.随着耕地质量等级调查与评定工作的完成,全国可比的耕地质量数据库的构建,为耕地质量在农作物遥感估产的应用提供基础.本研究利用耕地质量等级调查成果,建立冬小麦估产的耕地质量背景数据库,并基于MODIS/EVI和野外实地调查数据,将耕地质量因素作为一种背景因素引入到河南省冬小麦遥感估产工作中.结果表明:反映耕地质量的自然等指数与冬小麦单产呈显著相关,但单一利用自然等指数来进行冬小麦估产,精度明显不够;通过将自然等指数引入到冬小麦的遥感估产模型中,模型的相关系数提升到0.715,方程的标准误差也降低到621.80 kg·hm^(-2),估产模型精度有所提高;另外通过不同类型区样点调查产量和各地市统计单产数据的验证,引入耕地质量的遥感估产模型不仅精度高于原模型,在全省范围的适宜性也更优.

The cultivated land quality,one kind of non-remote sensing information,has important impact on crop yield. The completion of the cultivated land quality grading investigation and evaluation,and the building of comparable country’s cultivated land quality database,provided the basis for the application of cultivated land quality in crop yield estimation remote sensing. Survey results of cultivated land quality grading,were used for establishing cultivated land quality background database of the winter wheat yield estimation. And based on MODIS/EVI and field survey data,cultivated land physical quality considered as a background factor,was introduced into the remote sensing estimation of winter wheat yield in Henan province. Results showed that the land physical quality grade index reflecting cultivated land quality was significantly related to winter wheat yield,however,the precision of the results is not sufficiently high if only used land physical quality index for winter wheat yield estimation. After introducing land physical quality index to remote sensing estimation model of winter wheat yield,the correlation coefficient of the model rose up to 0.715,and the equation standard error reduced to 621.80 kg·hm-2,also the yield estimation model accuracy improved. In addition,according to the yield survey on sample areas of different types and statistical yield data validation of cities,the accuracy of remote sensing yield estimation model with cultivated land quality data was higher than the original model,and the suitability of the whole province yield estimation was more optimal.