| 摘要: |
| 【目的】 及时、准确、无损地估算冬小麦产量有助于粮食生产管理和粮食安全。【方法】 文章使用Sentinel-2 的红光波段和短波红外数据及MOD09Q1数据，使用ESTARFM融合方法，生成冬小麦生长期（3~6月）内8 d的NDVI高空间分辨率时间序列数据。结合MERRA-2气象同化数据，使用EC-LUE模型进行农作物总初级生产力（GPP）的模拟估算，并使用收割指数方法将之转化为冬小麦产量，将估算结果与美国农业部门公布的县级产量数据进行比较验证。【结果】 实验表明，Sentinel-2与MOD09Q1融合 NDVI具有良好的融合精度，相关系数在0.60~0.87之间。基于融合NDVI 估算的GPP相比MOD17A2H具有更好的空间细节和纹理。2017—2020年估算产量平均绝对误差MAE为8.41 bu/acre，平均相对误差为18.4%，均方根RMSE为9.7 bu/acre。【结论】 基准影像数量及其与预测日期的时间差会影响融合的精度，总体上能用于后续GPP模拟；EC-LUE模型较好地模拟了农作物的GPP水平和产量，在土地覆盖类型复杂的区域，可以提供更好的GPP空间变异信息，具有可移植性；基于收割指数方法将生长期内累计的GPP能转换为产量信息，能满足在作物收割之前的产量估算需求。 |
| 关键词: ESTARFM GPP 光能利用率模型 冬小麦 产量估算 |
| DOI：10.12105/j.issn.1672-0423.20230103 |
| 分类号: |
| 基金项目:国家自然科学基金面上项目“基于高斯过程回归和综合遥感干旱指数的农业干旱预测模型研究”（42271321） |
|
| Winter wheat yield estimation based on multi-source remote sensing data and EC-LUE model |
|
Ouyang Hongda1,2, Zhao Shuhe1,2, Zhang Xinming3
|
|
1.School of Geography and Ocean Science， Key Laboratory for Land Satellite Remote Sensing Applications of Ministry of Natural Resources， Jiangsu Provincial Key Laboratory of Geographic Information Science and Technology，Nanjing 210023， China;2.Jiangsu Center for Collaborative Innovation in Geographical Information Resource Development and Application， Nanjing 210023， China;3.Weifang Planning Development Research Centre， Shandong Province， Weifang 261041， China
|
| Abstract: |
| 【Purpose】 The timely，accurate and non-destructive estimation of winter wheat yield is important for grain production management and food security.【Method】 This study used Sentinel-2 red and shortwave infrared data and MOD09Q1 data，and the ESTARFM fusion method to generate an 8-day high spatial resolution time series of NDVI data during the winter wheat growth period（March-June）. In combination with MERRA-2 meteorological assimilation data，the EC-LUE model was used to simulate GPP and convert it to winter wheat yield using the harvest index method. The estimation results were compared and verified with county-level yield data published by the United States Department of Agriculture.【Result】 The experiment results showed that the fusion of Sentinel-2 and MOD09Q1 NDVI had good fusion accuracy，with a correlation coefficient between 0.6 and 0.87. The GPP estimated based on fusion NDVI had better spatial details and texture compared with MOD17A2H. The mean absolute error（MAE）of the estimated yield from 2017 to 2020 is 8.41 bu/acre，the mean relative error（MER）is 18.4%，and the root mean square error（RMSE）is 9.7 bu/acre.【Conclusion】 The results of this study show that the number of baseline images and the time difference between them and the predicted date would affect the accuracy of fusion. However，overall，this method could be used for subsequent GPP simulation. The EC-LUE model demonstrates strong performance in simulating crop GPP levels and yields. In regions with complex land cover types，it provides better spatial variability information of GPP with portability. The harvest index method could convert the accumulated GPP during the growth period into yield information，which could meet the demand for yield estimation before crop harvest. |
| Key words: ESTARFM GPP light use efficiency model winter wheat yield estimation |
