The Spatio-temporal Characteristics Of Grasslands Greenup Onset Date And Its Sensitivity To Temperature And Precipitation In The Northern Hemisphere

The response of terrestrial ecosystems to global climate change is a research hotspot in ecology and environmental science.Vegetation phenology,as an important biological indicator of vegetation growth in response to climate change is sensitive to climate change.Climate warming has significantly advanced the onset of spring phenology for vegetation in the Northern Hemisphere.This vegetation phenological change,driven by climate change,affects the structure and function of terrestrial ecosystems with the exchange of water,carbon and energy.However,in the study of vegetation phenology,many large-scale studies are limited to phenological comparisons among different vegetation types（such as evergreen forest,deciduous forest,tundra,etc.）,ignoring the variability of phenological events within single vegetation type on a large scale.Based on the GIMMS NDVI3g（1982-2015）dataset,the study uses harmonic analysis of time series method,dynamic-threshold method and double logistic function method to extract the greenup onset date in the middle and high latitude grasslands of the Northern Hemisphere,and compares the temporal and spatial differences of the rejuvenation period obtained by the three methods.Then,through partial least square regression（PLSR）method and random forest method,the three dominant temperature indexes of greenup onset date in grasslands were determined.Finally,based on the correlation coefficient results of the dominant temperature index and precipitation index on the greenup onset date of grasslands,the sensitivity differences of preseason T_max and preseason cumulative precipitation in the greenup onset date of different grassland types and different climate types were quantified by using multiple linear regression.The main conclusions are as follows:（1）There are some differences in the average annual rejuvenation period and the inter annual variation trend obtained by the three remote sensing phenological extraction methods,which are specifically shown as follows:the greenup onset date extracted by the dynamic-threshold method in the grasslands of the Northern Hemisphere occurs the earliest,with the most obvious advance trend.And the greenup onset date extracted by harmonic analysis of time series method and double logistic function method occurs later,with a similar interannual variation characteristics largest advance rate,and the advance trend is not significant.In order to reduce the difference between different phenological extraction methods,the average values of the three methods are taken as the final results of the greenup onset date in grasslands of the Northern Hemisphere.（2）In this study,the spatial and temporal distribution characteristics and interannual variation trend of grassland rejuvenation in the Northern Hemisphere were quantified.In general,it was found that most of the pixel（81.36%）in grasslands in the Northern Hemisphere had a greenup onset date of 110 to 170 days,and the regional interannual fluctuations were relatively large where the rejuvenation period occurred earlier.The rejuvenation period was delayed with the increase of altitude.During the period from 1982 to 2015,62.6%of the pixel rejuvenation period showed an advance trend,of which 18.8%was significantly ahead of schedule,with an average advance rate of 0.38±0.22 d/10a;The remaining 37.4%of the pixels showed a delayed trend,of which 6.8%were significantly delayed,with an average delay rate of 0.5±0.32 d/10a.From the perspective of interannual change trend,the greenup onset date in the study area showed an early trend（0.28d/10a）,but this early trend reversed and was delayed at a rate of 0.31d/10a in 2001-2015.（3）According to the analysis results of PLSR method and random forest method,the greenup onset date is mainly affected by the minimum temperature,maximum temperature and average temperature,and similar conclusions are obtained in most grassland types.According to the correlation analysis results between the greenup onset date and various meteorological factors,it is found that the influence of preseason temperature in the study area on the greenup onset date is mainly promoting,while precipitation is mainly delayed.The relative contribution of temperature and precipitation changes to the greenup onset date is quantified according to the absolute value of pixel correlation coefficient.About 33%of the greenup onset date of pixels is dominated by the preseason maximum temperature,25%by the preseason accumulated precipitation.And 22%is dominated by the preseason minimum temperature,mainly located in the Qinghai-Tibet Plateau region.Finally,based on the multiple regression analysis,the response of the greenup onset date to the preseason T_max and preseason accumulated precipitation under different grassland types and climate types was quantified.The overall greenup onset date showed a significant advance trend with the increase of preseason temperature（2.02 day/℃）.Among different grassland types,the warm semi-desert grassland（WG）is the least sensitive to temperature changes,with only4.09%of the pixels passing the significance test,while the boreal grassland has the highest temperature sensitivity（2.62 day/℃）.The greenup onset date in different climate types also exhibit the same trend,with the arid zone（B）being the least sensitive to preseason T_max,while the temperate zone（C）being the most sensitive to preseason T_max.The sensitivity of the greenup onset date to the preseason accumulated precipitation decreases with the increase of precipitation（0.4day/mm）.Among different grassland types,the precipitation sensitivity of temperate grasslands（TG）and warm semi-desert grassland（WG）are more sensitive,and the greenup onset date was 1.28 and 4.04 days earlier for every 1mm increase in preseason accumulated precipitation.Among different climate types,temperate zone（C）,dry and cold winter climate（Dw）and arid,steppe climate（BS）are more sensitive to precipitation changes.For every 1mm increase in preseason accumulated precipitation,C,Dw and BS are 0.48,1.31and 0.51 days ahead of schedule respectively.These results are of great significance for parameterization of global spring phenological models and prediction of vegetation phenological dynamics.