Reconstructing daily 30 m vegetation index over complex agricultural landscapes using crop reference curves approach

Authors: SUN, L. - Chinese Academy Of Agricultural Sciences; Gao, Feng; XIE, D. - Beijing Normal University; Anderson, Martha; CHEN, R. - Chinese Academy Of Agricultural Sciences; YANG, YUN - US Department Of Agriculture (USDA); YANG, YANG - US Department Of Agriculture (USDA); CHEN, Z. - Chinese Academy Of Agricultural Sciences

Submitted to: Remote Sensing of Environment
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 10/20/2020
Publication Date: 11/5/2020
Citation: Sun, L., Gao, F.N., Xie, D., Anderson, M.C., Chen, R., Yang, Y., Yang, Y., Chen, Z. 2020. Reconstructing daily 30 m vegetation index over complex agricultural landscapes using crop reference curves approach. Remote Sensing of Environment. 253:112156. https://doi.org/10.1016/j.rse.2020.112156.
DOI: https://doi.org/10.1016/j.rse.2020.112156

Interpretive Summary: High spatial and temporal resolution remote sensing data are required for monitoring crop progress and conditions at the sub-field scale. Data fusion approaches have been developed to fuse remote sensing imagery from different sensors to generate frequent observations at a high spatial resolution. However, these approaches have been challenging to apply in highly heterogeneous areas, especially in complex agricultural landscapes. This paper presents a novel method to reconstruct daily 30 m Normalized Difference Vegetation Index (NDVI) using a crop reference curve (CRC) extracted from 500 m pure MODIS (Moderate Resolution Imaging Spectroradiometer) pixels. The CRC-based method was applied over a complex agricultural landscape in the Choptank River watershed on the eastern shore of Maryland. Results show that the CRC method outperforms the image pair-based data fusion algorithm when clear Landsat images are scarce. The resulting 30 m NDVI time-series data produced by this approach would enable accurate crop monitoring at the sub-field scale.

Technical Abstract: Multi-sensor remote sensing data fusion technologies have been widely developed and applied in recent years, providing a feasible and economic solution to increase availability of high spatial and temporal resolution data. These methods, however, have been challenging to apply in highly heterogeneous areas, especially in complex agricultural landscapes where there are rapid changes at small scales, while features at larger scales change more slowly. In this study, we developed a novel method to reconstruct daily 30 m Normalized Difference Vegetation Index (NDVI). This method utilized crop reference curve (CRC) which is timeseries NDVI extracted from pure MODIS pixels, and then used to fit Landsat observations. The CRC based method was applied over a complex agricultural landscape in the Choptank River watershed on the eastern shore of Maryland, and reconstructed 30 m daily NDVI maps were validated using a leave-one-out approach in comparison to the original Landsat images. Results show that the relative error (RE) in NDVI is around 6-8% during periods of rapid crop growth, and 3-5% during peak periods when growth is slow. The accuracy outperforms a standard image pair based data fusion algorithm (Spatial and Temporal Adaptive Reflectance Fusion Model- STARFM) with RE of 4-9% in slow growth periods and 10-16% in fast growth periods when clear Landsat images are scarce. The reconstructed NDVI time series for corn, soybean, winter wheat/soybean and forest at 30 m resolution show clear phenological patterns at sub-field scale. The resulting 30 m NDVI time-series data provide useful information for mapping crop phenology and monitoring crop condition in complex agricultural landscapes.