Use time series NDVI and EVI to develop dynamic crop growth metrics for yield modeling
The normalized difference vegetation index (NDVI) and enhanced vegetation index (EVI) derived from Moderate Resolution Imaging Spectroradiometer (MODIS) satellite imagery are widely used for crop yield analysis. However, the growth metrics derived from the MODIS NDVI or EVI have so far not been expl...
Guardado en:
| Autores principales: | , |
|---|---|
| Formato: | article |
| Lenguaje: | EN |
| Publicado: |
Elsevier
2021
|
| Materias: | |
| Acceso en línea: | https://doaj.org/article/98b1358d188e428fab46feb1a9280479 |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| Sumario: | The normalized difference vegetation index (NDVI) and enhanced vegetation index (EVI) derived from Moderate Resolution Imaging Spectroradiometer (MODIS) satellite imagery are widely used for crop yield analysis. However, the growth metrics derived from the MODIS NDVI or EVI have so far not been explored and applied to crop yield yet. To the best of our knowledge, this study is the first to design NDVI- and EVI-based crop growth metrics, which biometrically capture the status and trend of crop growth and thus could be more powerful for growth yield management. We developed 19 NDVI- and EVI-based growth metrics, respectively, to monitor crop growth and yield, which is based on a time series of MODIS Terra 16-day 250 m data product from 2000 to 2018. Among the NDVI- and EVI-based vegetation growth metrics (VGM), the maximum (VGMmax), the integrated (VGMinteg), the sum of green-up (VGMsumgrn), the 70 days growth stage (VGM70), 85 days growth stage (VGM85), and 98 days growth stage (VGM98), the sum of 85 days growth stage (VGM85total), and the sum of 98 days growth stage (VGM98total) are mentionable. In this study, we implemented these crop growth metrics for soybean crop yield modeling at Mississippi Delta, Mississippi, USA. Soybean is a major crop cultivated in this region that is consisted of a total of 18 counties with similar agricultural cropping patterns. We observed that NDVI- and EVI-based VGMmax, VGM70, VGM85, VGM98total fitted models best with R-Square about 0.95. Using cross-validation of 80% train and 20% test size, we found NDVI-based VGM85 (e.g., normalized mean prediction error (NMPE) = 0.034) and EVI-based VGMmax (NMPE = 0.033) were the best fit linear yield models for this region. Designing novel crop growth indices based on crop phenological and ecological characteristics, this study further showed NDVI- and EVI-based growth metrics for crop growth monitoring and yield modeling. These growth metrics can be applied to other types of crop monitoring in different climate zones. |
|---|