Vegetation phenology

the study of recurring vegetation cycles and their connection to climate

is an important variable in a wide variety of earth and atmospheric science applications.Vegetation phenology is also an integraph of global changes and a comprehensive indicator of landscape and environment changes

and the studies on its response to global environment changes have become a focus of global changes field.Vegetation phenology detection methods based on remote sensing overcome conventional ground observation’s shortcomings

such as limited observation sites and missing data

and realize the spatial scale transition of observation methods from points to coverage.Remote sensing technology greatly promotes the study on vegetation ecosystem response to climate changes at regional

continental

even global scales.In order to keep consistent with the character of remote sensing-based vegetation phenology detection

the paper uses "vegetation greenness period" to replace "vegetation growing season"

and chose leaf unfolding and leaf coloration of local plant communities as indicator events to show the start and end of vegetation greenness period.Then

based on NOAA/AVHRR dataset

meteorological data

ground phenology observation data

and so on

the paper builds a remote sensingbased vegetation greenness period detection model

namely

Logistic fitting model on cumulative frequency of NDVI to determine the beginning date of greenness period（BGP） in spring and the end date of greenness period（EGP） in autumn of China since 1982.BGP and EGP are utilized to reflect the leaf-unfolding stage and leaf-coloring stage of the terrestrial vegetation

respectively.The computed results indicate that BGP appeared to delay and EGP have an advance trend from south to north.Finally

through comparing the results of the model with the 9 ground observation sites and other remote sensing-based detection models

it is found that BGP and EGP computed by the model have differences of 9—21 days and 0—13 days

respectively

with the ground observation in Mudanjiang

Huhehot

Beijing

Luoyang and Xi’an observation sites.The model is more precise than the other remote sensing-based detection models and the annual BGP and EGP fluctuations are comparatively small.It is obvious that BGP and EGP estimated by the model are reliable in the north temperate regions.In Tunxi

Guiyang

Renshou and Guangzhou observation sites

the differences of BGP and EGP with the ground observation are more obvious than those in Mudanjiang

Huhehot

Beijing

Luoyang and Xi’an observation sites in spite of any remote sensing-based detection model.Tunxi

Guiyang

Renshou and Guangzhou observation sites locate in the southern subtropical evergreen region.The vegetation has no obvious and consistent leaf-unfolding stage and leaf-coloring stage.However

obvious BGP and EGP can be computed by remote sensing-based detection models

which are mainly related to continuously overcast

rainy and foggy days during the rainy season in the sites.Therefore

BGP and EGP estimated by the model are not the real start and end of the vegetation growing season

but reflect vegetation’s response to regional climate changes.In a word

compared with other remote sensing-based detection model

the logistic fitting model on cumulative frequency of NDVI in China can be characterized in three ways.（1） NDVI data needn’t be exceedingly smoothed

which can remain more temporal details;（2） Logistic model only includes three fitting parameter.So

computing process is relatively simple;（3） Multi-model of NDVI arisen from multiple growth cycles（e.g.

double or triple-crop agriculture

semiarid systems with multiple rainy seasons

etc.） is considered.BGP and EGP can be straightly determined by fitting the cumulative frequency of NDVI.The logistic fitting model on cumulative frequency of NDVI is more suitable for China than the other remote sensing-based detection models and can be applied to different spatial scales.