ABSTRACT Vegetation phenology is a critical indicator of ecosystem responses to climate change. While satellite‐derived vegetation phenology products provide large‐scale, long‐term observations, their integration is hampered by substantial discrepancies arising from variations in sensor platforms, temporal resolution, and retrieval algorithms. This challenge is acutely evident in China, where ecological conditions are highly diverse and a systematic assessment of product consistency and uncertainty remains lacking. To address this gap, we conducted a comprehensive evaluation across China (2001–2016) comparing three vegetation phenology datasets: MCD12Q2 (utilising 15%, 50%, and 90% thresholds), VIPPHEN, and GIMMS. Ground observations and the extended triple collocation (ETC) method were employed to assess the spatial and temporal consistency and random errors in start (SOS) and end (EOS) of season estimates. Results indicate that MCD12Q2 demonstrated superior overall performance. Specifically, the 15% threshold achieved the highest SOS accuracy, while the 50% threshold yielded optimal EOS estimates. In general, SOS estimates were more reliable than EOS. Spatially, products exhibited higher consistency and lower errors in deciduous forests and low‐to‐mid elevation regions, compared to evergreen forests and high‐altitude areas. ETC analysis further confirmed the advantages of MCD12Q2 in phenological extraction. This study provides critical insights for selecting and integrating satellite‐derived vegetation phenology datasets, thereby enhancing our understanding of vegetation–climate interactions and supporting more accurate terrestrial carbon cycle simulations, Earth system modeling, and climate policy formulation.
Meng et al. (Thu,) studied this question.