Abstract: NDVI (Normalized Difference Vegetation Index) time series analysis can reveal the phenological changes of vegetation cover. With the characteristics of high spatial resolution and high temporal resolution, HJ-1 A/B satellite can provide abundant time series optical data for the application and analysis of phenological characteristics of tree species at the medium and small scales. In this study, we used Savizky-Glolay (S-G) filtering method and Harmonic Analysis of NDVI Time-Series (Hants) to reconstruct 58 HJ-1 NDVI time series data curves of eight typical tree species in Xiamen City. Mean value, mean absolute error and correlation coefficient were selected to quantitatively evaluate the filtered results. Phenological characteristics of the eight tree species were analyzed using NDVI ratio. We further examined the effects of meteorological factors including temperature and precipitation on NDVI time series fluctuation of the eight tree species. The results showed that both S-G and Hants filtering methods can restore the NDVI fluctuation shape of the vegetation characteristic of obvious phenological changes, with Hants filter being the best choice due to its ideal smoothness effect. The filtered NDVI curves showed that NDVI values of all the tree species except Eucalyptus urophylla were at a high level. The intra-annual fluctuation of NDVI time series was not intense, with obvious double peaks. Moreover, the NDVI values peaked firstly in May and June, and then declined in July and August. The second peak occurred in October, which was lower than the first peak. The growth period of all the species lasted for 7-8 months. Except that E. urophylla entered growth period at the end of February, the other tree species entered growth period at the end of April, ended in November, and reached the growth peak in July. In addition, tree growth was sensitive to the abrupt drop of precipitation in July. The lag period of such responses differed among tree species, with differences in the speed of growth reduction, but most of the species had low values of NDVI ratio in August. Our results provide reference for the selection of filtering methods for HJ-1 A/B NDVI time series data, characterizing the phenology of typical tree species, and fine classification of tree species in subtropical regions.

Key words: landscape pattern, ethnic township, urbanization, driving force., Aerla Town