Varying effects of tree cover on relationships between satellite-observed vegetation greenup date and spring temperature across Eurasian boreal forests

Earth observation satellites have facilitated the quantification of how vegetation phenology responds to climate warming on large scales. However, satellite image pixels may contain a mixture of multiple vegetation types or species with diverse phenological responses to climate variability. It is unclear how these mixed pixels affect the statistical relationships between satellite-derived vegetation phenology and climate factors. Here, we aim to investigate the impacts of percent tree cover (PTC), a measure of mixed pixel, on the statistical relationships between satellite-derived vegetation greenup date (GUD) and spring air temperature across Eurasian boreal forests at a 0.05° spatial resolution. We estimated GUD using Moderate Resolution Imaging Spectroradiometer (MODIS) time series data. The responses of GUD to interannual variation in spring temperature (April to May) during 2001-2020 were characterized by correlation coefficient (RT_AM) and sensitivity (ST_AM). We then evaluated the local impacts of PTC on spatial variations in RT_AM and ST_AM using partial correlation analysis through spatial moving windows. Our results indicate that, for most areas, forests with higher PTC were associated with stronger RT_AM and ST_AM. Moreover, PTC had stronger local impacts on RT_AM and ST_AM than mean annual temperature and temperature seasonality for 37.3% and 27.4% of the moving windows, respectively. These impacts were spatially varying and different among forest types. Specifically, deciduous broadleaf forests and deciduous needleleaf forests tend to have a higher proportion of these impacts compared to other forest types. Our findings demonstrate the nonnegligible effects of PTC on the statistical responses of GUD to temperature variability at coarse spatial resolution (0.05°) across Eurasian boreal forests.