Validation For Phenology Metrics From Satellite Datasets And Monitoring Phenology Dynamics In The Northern Hemisphere And Typical Regions

Phenology is a sensitive and critical feature of vegetation change that has regarded as a good indicator in climate change studies.Studying the response of vegetation phenology to climate change at different temporal and spatial scales is important for understanding and predicting future terrestrial ecosystem dynamics and the adaptation of ecosystems to global change.In recent years,satellite sensors have become an important tool for understanding patterns of vegetation phenology,and for providing temporally continuous observations from regional to global scales.So far,varities of remote sensing data sources and phenology extraction methods from satellite datasets have been developed to study the spatial-temporal dynamics of vegetation phenology.However,the differences between vegetation phenology results caused by the varies satellite datasets and phenology extraction methods are not clear,and the reliability for different phenology results extracted from remote sensing datasets is not verified and compared using the ground observation data.Quantitative validation and difference comparision for varies vegetation phenology results is an urgent problem to be solved.It is important to analyze the differences among all these phenology results version to link existing studies with future applications of the remote sensing technology in monitoring vegetation phenology.And this is important for the future's monitoring about vegetation's response to global changs in large spatial scale and with high accuracy.In this study,we selected two kinds of long time series satellite datasets: GIMMS NDVIg and GIMMS NDVI3 g.Based on three most popular remote sensing phenology extraction methods,we calculated six versions of phenology metrics for each pixels in the Northern Hemisphere.The three methods used in this research are: maximum increase method,dynamic threshold method and midpoint method.Then based on these six methods,we quantitatively explored the existed differences and reasons among different phenology results both caused by the difference of satellite source and diversity of extraction methods.Additionally,this research evaluated the reliability of phenology results extracted from two GIMMS NDVI versions using MODIS MOD13Q1 NDVI dataset which with the higher spatial resolution.Most importantly,this study uses many kinds of ground observation datasets to validate the reliability of six phenology results calculated from remote sensing datasets.These ground observation datasets include the ground phenology data observed by the US phenology observation network,the vegetation phenology data provided by agricultural-metrological stations in the Tibetan Plateau,and the carbon flux datasets monitored by global flux website.At last,we selected the most reliable remote sensing long time series datasets and phenology extraction method to study the spatial-temporal dynamics of vegetation's response to global changes in the Northern Hemisphere and some typical areas like the Tibetan Plateau,and also,explored vegetation's accumulation characteristic for climate consistently change.The main conclusions are listed as below.(1)In most areas of the Northern hemisphere,the spring phenology metric – the start of growing season extracted from GIMMS NDVIg(SOSg)using three different methods are all later than that from GIMMS NDVI3g(SOS3g).The advance change rate based on SOSg is faster than SOS3 g during 1982-2012.Compared with phenology results from MODIS NDVI,we included that GIMMS NDVI3 g is more sensitive in monitoring vegetation dynamics than previous GIMMS NDVIg,and is more reliable for phenology research.(2)Only the remote sensing phenology result extracted from GIMMS NDVI3 g using midpoint method are significantly related with all the three kinds of ground observation datasets: ground phenology data observed by the US phenology observation network,vegetation phenology data provided by agricultural-metrological stations in the Tibetan Plateau,and the carbon flux datasets monitored by global flux website.That's to say,studying phenology dynamics in the Northern hemisphere by using GIMMS NDVI3 g and midpoint method is the most reliable way.(3)The start of growing season(SOS)for natural vegetation in the Northern hemisphere advanced by 0.066d/y during 1982-2012,and the end of growing season delayed by 0.074d/y.SOS advanced is main caused by temperature rises in spring,but EOS delayed is impacted by both temperature and precipitation changes in autumn.(4)Response of vegetation phenology to meteorological factors varies among ecogeographical zones in the Tibetan Plateau.Because of the gradual adaptability of vegetation to the global change in the Tibetan Plateau,the effect of spring warming on the growing season(SOS)of the vegetation is becoming weaker and weaker.