Comparative Study On Land Surface Phenological Parameters And Its Climate Responses Retrieved From Different Data Sources

In the context of global climate change,when retrieving phenological information based on multi-source remote sensing data,the selection of data sources leads to uncertainty in the phenological results.Based on the AVHRR GIMMS3g(1982–2015)and Terra MODIS(2000–2015)datasets,this paper used the Savitzky-Golay(S-G)filtering method to reconstruct the time series curve of the Normalized Difference Vegetation Index(NDVI),used the derivative method to extract the Land Surface Phenology(LSP)parameters of the Northern Hemisphere,namely the start of the growing season(SOS)and the end of the growing season(EOS).In this study,the same method was selected to extract the LSP parameters of the weekly average GPP data in the FLUXNET2015 dataset to avoid the uncertainty of the extraction of phenological parameters caused by multiple methods and to compare the similarities and differences between remote sensing phenology and ground phenology.The partial correlation analysis method is used to calculate the maximum partial correlation coefficients between the LSP parameters and the preseason temperature and precipitation,and the period corresponding to the maximum absolute value is taken as the preseason(June to November of the previous year,April to November)length,using the unary linear regression method to calculate the sensitivity coefficients of LSP and preseason temperature and precipitation,and comparing the sensitivity characteristics of the SOS and EOS from the two datasets to climate change.It reveals the inconsistency of the LSP of different vegetation types in the Northern Hemisphere based on the AVHRR GIMMS3 g and MODIS data sets,as well as the difference in the sensitivity characteristics of LSP to pre-season temperature and pre-season precipitation.The results showed that(1)Based on the GIMMS3g(1982–2015 and 2000–2015)and MODIS(2000–2015)datasets to retrieve the single-season vegetation LSP in the northern hemisphere,SOS retrieved by GIMMS3g(G?SOS)and EOS retrieved by GIMMS3g(G?EOS)showed a significant delay(0.1716 days/year)and advanced(0.9172 days/year)trend respectively,and the SOS retrieved by MODIS(M?SOS)and EOS retrieved by MODIS(M?EOS)showed a significant advanced(0.5861 days/year)and delayed(0.6305 days/year)trend respectively in most parts of the Northern Hemisphere from 2000 to 2015;From 2000 to 2015,pixels with a common significant trend of SOS and EOS of the two datasets accounted for 1.33% and 1.17% of the total number of pixels in vegetation regions of the Northern Hemisphere,respectively.Among the different vegetation types in high latitude regions,M?SOS has a more obvious trend in advance than G?SOS.G?EOS has a significant trend in advance,while M?EOS has a significantly delayed trend in different vegetation types.The phenological parameters retrieved by GIMMS3 g are closer to the ground phenology than those of MODIS.(2)The most relevant preseason lengths of SOS to temperature and precipitation are the current season and January,February,and March,while EOS is the current season and January,February,and May.2000-2015 are in the Northern Hemisphere.In most of the regions,the SOS retrieved has less correlation with preseason temperature and precipitation than EOS from the GIMMS3 g and MODIS datasets.Besides,the partial correlation between preseason precipitation and SOS and EOS is better than that of preseason temperature.The sensitivity of SOS from the two datasets to preseason temperature decreases with increasing temperature,and the sensitivity of EOS to preseason temperature increases with temperature from 2000 to2015;while SOS and EOS are sensitive to seasonal temperature,and the sensitivity of preseason precipitation decreases with increasing precipitation.The sensitivity of LSP to preseason temperature and preseason precipitation has latitude regularity,and LSP is more sensitive to climate change in the mid-high latitudes of the Northern Hemisphere.Based on different remote sensing data sources,this study demonstrates the differences in the temporal and spatial characteristics of LSP parameters and their sensitivity to preseason temperature and precipitation,thereby providing a reference basis at a hemispherical scale for the selection of multi-source data in the process of vegetation monitoring.