Extraction And Verification Of Urban Vegetation Phenology In China

Urban vegetation is an important component of the environment,and the dynamics of urban vegetation phenology play an important role in influencing human activities.Previous studies have shown high-resolution remote sensing as a tool for urban vegetation mapping,but the low temporal resolution of these data limits their use for phenological modeling.Therefore,it is of great significance to evaluate MODIS imagery for urban vegetation phenology monitoring.Besides,different types of remote sensing data and fitting and extraction methods lead to uncertainties in phenological results.Here,to account for this potential uncertainty,we applied three identification methods to extract the start and end of growing season(SOS and EOS)in urban ecosystems based on both the Normalized Difference Vegetation Index(NDVI)and the Enhanced Vegetation Index(EVI)from the 250 m MODIS vegetation indices product(MOD13Q1).Then the accuracies of the satellite-derived SOS and EOS were evaluated through comparing with phenological observations at 18 ground sites.The preliminary conclusions are as follows:(1)SOS was most consistent with the prime of leaf unfolding(PLU),and EOS was most consistent with the beginning of leaf coloring(BLC)date.(2)EVI was found to have greater advantages than NDVI in detecting urban vegetation phenology in terms of both higher correlation coefficients and lower root mean square errors.The NDVI and EVI datasets used the first method(threshold method)to extract the urban vegetation greening date(SOS)were more concentrated than the latter two methods(piecewise logistic function,double logistic function).(3)Spring phenology extracted by the first method(threshold method)is more consistent with ground phenology observation(R=0.66,Bias=7.5,RMSE=18.3).The autumn phenology extracted by the third method(piecewise logistic function)is close to the ground phenology observation(R=0.34,Bias=-0.5,RMSE=22).(4)The changes of LSP showed certain differences in each city.The SOS extracted from the same data(NDVI?EVI)by various methods was also very different.Under the background of general delay of vegetation EOS,vegetation EOS extracted by diverse methods using EVI dataset in Mudanjiang,Changchun and Beijing showed an advance trend,while vegetation EOS extracted by diverse methods using NDVI dataset showed a delay trend.We should pay more attention to the selection of vegetation index and algorithms when studying urban vegetation phenology by remote sensing data.