Phenological, meteorological, and time-integrated normalized difference vegetation index (TI NDVI) data from 1982 to 1993, at three sample stations, were used to investigate the response of the growing season of local plant communities to climate change and the linkage of satellite sensor-derived greenness to the surface growing season. Results suggest that mean air temperature and growing degree days (GDDs) above 5 °C during late winter and spring, and precipitation in autumn, are the most important controls on the beginning and end dates of the growing season (BGS and EGS). In contrast, annual mean air temperature, annual GDD totals, mean air temperature during late winter and spring, and growing season TI NDVI are the most important controls on length of the growing season (LGS).
Using correlation and regression analysis, simple and multiple linear regression models were developed for individual and all stations. Since the standard error of the estimates (SE) of the BGS models for all stations are smaller than those of the EGS models, estimates of the beginning date of the growing season are probably more reliable than estimates of the end date. On average, if the mean air temperature in late winter and spring increases by 1° C, then the beginning date of the growing season will advance 5–6 days, and the end date will be 5 days later. Moreover, if autumn precipitation increases 100 mm, then the end date will advance 6–8 days. In terms of the LGS models, mean air temperature in late winter and spring, annual mean air temperature, and annual GDD totals have significant positive correlations with growing season duration, whereas growing season TI NDVI has a negative relationship. Comparing the SE of different LGS models, those developed with each of the three temperature variables fit the observed growing season duration data much better than those using growing season TI NDVI. Copyright © 2002 Royal Meteorological Society.