Journal of Integrative Agriculture ›› 2014, Vol. 13 ›› Issue (7): 1538-1545.DOI: 10.1016/S2095-3119(14)60817-0

• 论文 • 上一篇    下一篇

Corn Yield Forecasting in Northeast China Using Remotely Sensed Spectral Indices and Crop Phenology Metrics

 WANG Meng, TAO Fu-lu , SHI Wen-jiao   

  1. 1、Institute of Geographical Sciences and Natural Resources Research, Chinese Academy of Sciences, Beijing 100101, P.R.China
    2、University of Chinese Academy of Sciences, Beijing 100049, P.R.China
  • 收稿日期:2013-10-29 出版日期:2014-07-15 发布日期:2014-07-16
  • 通讯作者: TAO Fu-lu, Tel: +86-10-64888269, E-mail: taofl@igsnrr.ac.cn
  • 作者简介:TAO Fu-lu, Tel: +86-10-64888269, E-mail: taofl@igsnrr.ac.cn
  • 基金资助:

    This study was supported by the National Basic Research Program of China (2010CB950902), the National Natural Science Foundation of China (41371002), and the Strategic Priority Research Program of the Chinese Academy of Sciences (XDA05090310).

Corn Yield Forecasting in Northeast China Using Remotely Sensed Spectral Indices and Crop Phenology Metrics

 WANG Meng, TAO Fu-lu , SHI Wen-jiao   

  1. 1、Institute of Geographical Sciences and Natural Resources Research, Chinese Academy of Sciences, Beijing 100101, P.R.China
    2、University of Chinese Academy of Sciences, Beijing 100049, P.R.China
  • Received:2013-10-29 Online:2014-07-15 Published:2014-07-16
  • Contact: TAO Fu-lu, Tel: +86-10-64888269, E-mail: taofl@igsnrr.ac.cn
  • About author:TAO Fu-lu, Tel: +86-10-64888269, E-mail: taofl@igsnrr.ac.cn
  • Supported by:

    This study was supported by the National Basic Research Program of China (2010CB950902), the National Natural Science Foundation of China (41371002), and the Strategic Priority Research Program of the Chinese Academy of Sciences (XDA05090310).

摘要: Early crop yield forecasting is important for food safety as well as large-scale food related planning. The phenology-adjusted spectral indices derived from Moderate Resolution Imaging Spectroradiometer (MODIS) data were used to develop liner regression models with the county-level corn yield data in Northeast China. We also compared the different spectral indices in predicting yield. The results showed that, using Enhanced Vegetation Index (EVI), Normalized Difference Water Index (NDWI) and Land Surface Water Index (LSWI), the best time to predict corn yields was 55-60 days after green-up date. LSWI showed the strongest correlation (R2=0.568), followed by EVI (R2=0.497) and NDWI (R2=0.495). The peak correlation between Wide Dynamic Range Vegetation Index (WDRVI) and yield was detected 85 days after green-up date (R2=0.506). The correlation was generally low for Normalized Difference Vegetation Index (NDVI) (R2=0.385) and no obvious peak correlation existed for NDVI. The coefficients of determination of the different spectral indices varied from year to year, which were greater in 2001 and 2004 than in other years. Leave-one-year-out approach was used to test the performance of the model. Normalized root mean square error (NRMSE) ranged from 7.3 to 16.9% for different spectral indices. Overall, our results showed that crop phenology-tuned spectral indices were feasible and helpful for regional corn yield forecasting.

关键词: remote sensing , yield , corn , MODIS , phenology

Abstract: Early crop yield forecasting is important for food safety as well as large-scale food related planning. The phenology-adjusted spectral indices derived from Moderate Resolution Imaging Spectroradiometer (MODIS) data were used to develop liner regression models with the county-level corn yield data in Northeast China. We also compared the different spectral indices in predicting yield. The results showed that, using Enhanced Vegetation Index (EVI), Normalized Difference Water Index (NDWI) and Land Surface Water Index (LSWI), the best time to predict corn yields was 55-60 days after green-up date. LSWI showed the strongest correlation (R2=0.568), followed by EVI (R2=0.497) and NDWI (R2=0.495). The peak correlation between Wide Dynamic Range Vegetation Index (WDRVI) and yield was detected 85 days after green-up date (R2=0.506). The correlation was generally low for Normalized Difference Vegetation Index (NDVI) (R2=0.385) and no obvious peak correlation existed for NDVI. The coefficients of determination of the different spectral indices varied from year to year, which were greater in 2001 and 2004 than in other years. Leave-one-year-out approach was used to test the performance of the model. Normalized root mean square error (NRMSE) ranged from 7.3 to 16.9% for different spectral indices. Overall, our results showed that crop phenology-tuned spectral indices were feasible and helpful for regional corn yield forecasting.

Key words: remote sensing , yield , corn , MODIS , phenology