Extending NDVI time series in Mongolia using spatial correlation analysis between AVHRR-GIMMS and MODIS TERRA data

Abstract

Accurate and long-term monitoring of vegetation dynamics is critical for understanding how ecosystems respond to climate change and human-induced pressures, especially in environmentally sensitive regions like Mongolia. This study examined the relationship between the Normalized Difference Vegetation Index (NDVI) from the Moderate Resolution Imaging Spectroradiometer (MODIS) and the third-generation Global Inventory Modeling and Mapping Studies (GIMMS) dataset to generate an extended NDVI time series for Mongolia. Spatial correlation analysis was performed for the growing season (April to September) from 2002 to 2015, with MODIS NDVI resampled to the 8 km resolution of the GIMMS dataset to ensure comparability. The results revealed a strong correlation between the two datasets, with the coefficient of determination (R²) ranging from 0.776 to 0.905 and root mean square error (RMSE) values between 0.034 and 0.070. Higher agreement was observed in steppe and forest-steppe regions, while reduced consistency in mountainous and arid areas highlighted challenges in monitoring vegetation in complex or sparsely vegetated environments. The application of the Partial Least Squares Regression (PLSR) model demonstrated that MODIS NDVI could be reliably reconstructed from GIMMS NDVI values. This integration enables the extension of MODIS-like NDVI back to the 1980s, capitalizing on the longer temporal coverage of the GIMMS dataset for comprehensive vegetation monitoring and trend analysis. Overall, the study supports the combined use of MODIS and GIMMS NDVI for robust, long-term ecological assessments and provides a solid foundation for future research focused on cross-sensor harmonization using advanced statistical and machine learning approaches.