Spatiotemporal variation of spring phenology and the corresponding scale effects and uncertainties: A case study in southwestern China

• All sensors suggest an advancing trend of LSP at a rate of 6–8 days per decade. • Significant variations of the advancing trend exist in different satellite sensors. • Data quality adds more uncertainty in phenological dynamics than retrieval methods. • The MODIS-Landsat LSP bias may not apply to infer a coarse–fine LSP relationship. Understanding terrestrial vegetation phenology—the timing of life-cycle events—is crucial for insights into ecosystem energy and material cycles. Land surface phenology (LSP) derived from satellite observations has become a critical tool for tracking vegetation phenology across large spatial scales. However, LSP data from coarse spatial resolutions often mix phenological signals from multiple land cover types, a limitation that fine-resolution satellite data can help overcome. Recent studies indicate that spring phenology derived from the 500-m Moderate Resolution Imaging Spectroradiometer (MODIS) data tends to be biased earlier than that from the 30-m Landsat data due to scale effects. The extent of this bias across other satellite sensors and its impact on long-term phenological trends remains unclear. Additionally, few studies have used medium- to high-resolution LSP data to investigate southwestern China, partly due to limited data availability, which may exacerbate uncertainties related to scale effects in LSP observations. To address these gaps, we selected Jinfo Mountain in southwestern China—a region with high spatial heterogeneity—to analyze the spatiotemporal patterns of spring phenology and examine associated scale effects and uncertainties. We applied two phenology retrieval methods to multi-resolution LSP data from various sensors: 30-m Landsat (1984–2023), 250-m MODIS (2002–2021), 500-m MODIS (2000–2023), 1-km SPOT (1999–2019), and 8-km AVHRR (1982–2022). Our findings revealed that all sensors consistently captured the spatial patterns of spring phenology, indicating an advancing trend of 6–8 days per decade, though the trend’s magnitude varied notably across sensors. Data quality, rather than retrieval methods, emerged as the primary source of uncertainty in characterizing phenological dynamics, with elevation contributing significantly to bias due to its negative correlation with the number of available clear observations. Moreover, we found that the MODIS-Landsat bias in spring phenology may not generalize across other coarse-to-fine LSP comparisons. This study provides valuable insights into phenology in the understudied region of southwestern China, highlighting the importance of spatial resolution and sensor characteristics for accurate plant phenology mapping and monitoring.