Challenges of Cross-Sensor Application of MODIS EVI: A Case Study of Landsat-8

Objectives: This study aims to investigate the applicability of the MODIS Enhanced Vegetation Index (EVI) when directly applied to data from other satellite sensors, an important challenge that has remained unexplored for nearly 30 years since the introduction of EVI. It focuses on whether the four empirical parameters of EVI developed based on MODIS data are suitable for other satellite datasets, using Landsat-8 data as a case study. Methods: We systematically examined the relationship between MODIS EVI and Landsat-8 NDVI across six test areas. Given that the NDVI formula does not include parameters, the NDVI relationship between MODIS and Landsat-8 served as a reference for assessing the suitability of MODIS EVI for Landsat-8 data. Additionally, a modified EVI, referred to as EVI_b, was constructed to evaluate the influence of the blue band on the MODIS EVI application in Landsat-8 data. Results: The results revealed that the discrepancies between EVI data from MODIS and Landsat-8 were significant, which is, on average, nearly three times larger than those observed for NDVI data from the two satellites, with an 11.8% difference in (R²). The analysis indicated that the source of these discrepancies was not the blue band, but rather the four empirical parameters used in MODIS EVI. Sensitivity testing further confirmed that the interactions among the four parameters in the EVI formula significantly affected EVI values, particularly highlighting the strong sensitivity of the soil adjustment factor (L). Conclusions: Therefore, we recommend calibrating these parameters when applying EVI to Landsat-8. The cross-calibration conducted in this study has reduced the average bias rate between the MODIS and Landsat-8 EVI from 12.8% to nearly 0% across the six test areas. This research provides significant guidance for the cross-sensor application of the EVI index and holds important practical value for the precise monitoring of vegetation changes at both regional and global scales.