Zhao, Yaohua, Peng, Xiaoqing, Frauenfeld, Oliver W., Cui, Xia, Bi, Jian, Ma, Xuanlong, Wei, Gang, Mu, Cuicui, Sun, Hao, and Sui, Jia
Vegetation phenology interacts strongly with climate through the exchange of carbon, water, momentum, and energy between terrestrial ecosystem and atmosphere. These vegetation dynamics in Northern Hemisphere permafrost regions have substantial uncertainties in previous studies, partly due to differences in datasets. Thus, reliable land surface phenology (LSP) retrievals are crucial for understanding the effects of climate change on ecosystems and biosphere‐atmosphere‐hydrosphere interactions. We assessed various LSP datasets at different spatial resolutions for 2001–2014, generated by different methods based on different satellite observations. We also assessed the accuracy of LSP by comparing with CO2 flux phenology. For start of growing season (SOS), the comprehensive evaluation indicated MODIS phenology showed better consistency and higher accuracy with flux‐derived phenology observations (R = 0.54, RMSE = 36.2 days, bias<5 days). For end of growing season (EOS), we cannot conclusively determine which LSP performs best. In the Northern Hemisphere permafrost regions, SOS occurred at 100–150 days (April–May), and EOS occurred at 260–320 days (September–November). During 2001–2014, SOS occurred earlier by 0.33 ± 0.30 days/yr. Significant trends were observed for 6.4%–27.6% of pixels, averaging −1.30 ± 1.16 days/yr. EOS occurred earlier by 0.25 ± 0.43 days/yr, with significant trends averaging −0.52 ± 0.94 days/yr. Among LSPs, variability in EOS trends was significant, with even the direction of trends differing. This study may provide insights into LSP data selection and further understanding of vegetation dynamics and its mechanisms in permafrost regions. Plain Language Summary: To better understand permafrost ecosystem dynamics associated with global change, the changes in vegetation and their spatial patterns in the Northern Hemisphere permafrost regions must be carefully evaluated. Thus, reliable land surface phenology (LSP) retrievals are very important. We therefore collected a total of 13 LSPs to understand how these products perform relative to each other and to ground observations and to quantify their uncertainties. MODIS and different GIMMS products are slightly more consistent overall. Considering different metrics, regions, and parameters, the temporal trends and spatial patterns of phenology indicate an earlier start to the growing season, but the end of the growing season is inconsistent. Bigger vegetation changes are found in the continuous permafrost region. These findings allow us to clarify the applicability of the LSP products, to provide a reference for data selection, and insights into the understanding of vegetation dynamics and their mechanisms in permafrost regions. Key Points: The average start and end of the growing season occurs on 131 ± 18 days and 291 ± 4 days in Northern Hemisphere permafrost regionsComparison of 13 land surface phenology products to CO2 flux phenology indicates that the RMSEs ranges from 27.25ays to 77.84 daysBoth the start and end of the growing season are earlier by 1.30 ± 1.16 and 0.52 ± 0.94 days/yr in Northern Hemisphere permafrost regions [ABSTRACT FROM AUTHOR]