A vegetation phenology dataset developed by integrating multiple sources using the reliability ensemble averaging method

Abstract

Global change has substantially shifted vegetation phenology, with important implications in the carbon and water cycles of terrestrial ecosystems. Various vegetation phenology datasets have been developed using remote sensing data. However, the significant uncertainties in these datasets limit our understanding of ecosystem dynamics in terms of phenology. It is therefore crucial to generate a reliable large-scale vegetation phenology dataset, by fusing various existing vegetation phenology datasets, to provide a comprehensive and accurate estimation of vegetation phenology with a fine spatiotemporal resolution. In this study, we merged four widely used vegetation phenology datasets to generate a new dataset using the reliability ensemble averaging (REA) fusion method. The new dataset has a spatial resolution of 0.05° and covers the period from 1982 to 2020, with geographic coverage extending above 30° N in the Northern Hemisphere. The evaluation using ground-based phenocam data from 280 sites indicated that the accuracy of the newly merged dataset was substantially improved compared to the four original datasets. The start and end of the growing season (SOS and EOS) in the newly merged dataset showed the highest correlation with ground-based phenocam observations, compared to the original datasets (0.84 and 0.71, respectively) and accuracy in terms of the root mean square error (RMSE) between phenocam data and merged datasets (12 and 17 d, respectively). Using the new dataset, we found that the SOS is occurring approximately 0.19 d earlier per year (p < 0.01), while the end of the growing season is occurring 0.18 d later per year (p < 0.01) over the period 1982–2020 across regions north of 30° N. This dataset offers a unique and novel source of vegetation phenology data for global ecology studies. This study uses multiple phenology datasets, including the MCD12Q2 dataset (Friedl et al., 2022, https://doi.org/10.5067/MODIS/MCD12Q2.061), the VIP dataset (Didan and Barreto, 2016, https://doi.org/10.5067/MEaSUREs/VIP/VIPPHEN_NDVI.004), the GIMMS NDVI3g dataset (Wang et al., 2019, http://data.globalecology.unh.edu/data/GIMMS_NDVI3g_Phenology/), the GIMMS NDVI4g dataset (Chen and Fu, 2024, https://doi.org/10.5281/zenodo.11136967), the PhenoCam dataset (Richardson et al., 2018, https://doi.org/10.3334/ORNLDAAC/1674), the Internet Nature Information System of Japan (Ministry of the Environment of Japan, 2024, http://www.sizenken.biodic.go.jp), the Phenological Eyes Network (PEN, 2024, http://www.pheno-eye.org), and the MCD12Q1 land use dataset (Friedl and Sulla-Menashe, 2022, https://doi.org/10.5067/MODIS/MCD12Q1.061). The REA phenology dataset developed in this study is available at https://doi.org/10.5281/zenodo.15165681 (Cui and Fu, 2024).