A new algorithm predicting the end of growth at five evergreen conifer forests based on nighttime temperature and the enhanced vegetation index

Accurate estimation of vegetation phenology (the start/end of growing season, SOS/EOS) is important to understand the feedbacks of vegetation to meteorological circumstances. Because the evergreen forests have limited change in greenness, there are relatively less study to predict evergreen conifer forests phenology, especially for EOS in autumn. Using 11-year (2000-2010) records of MODIS normalized difference vegetation index (NDVI) and enhanced vegetation index (EVI), together with gross primary production (GPP) and temperature data at five evergreen conifer forests flux sites in Canada, we comprehensively evaluated the performances of several variables in modeling flux-derived EOS. Results showed that neither NDVI nor EVI can be used to predict EOS as they had no significant correlation with ground observations. In comparison, temperature had a better predictive strength for EOS, and R-2 between EOS and mean temperature (T-mean), the maximum temperature (T-max, daytime temperature) and the minimum temperature (T-min, nighttime temperature) were 0.45 (RMSE = 5.1 days), 0.32 (RMSE = 5.7 days) and 0.58 (RMSE = 4.6 days), respectively. These results suggest an unreported role of nighttime temperature in regulating EOS of evergreen forests, in comparison with previous study showing leaf-out in spring by daytime temperature. Furthermore, we demonstrated that it may be because nighttime temperature has a higher relationship with soil temperature (T-s) (R-2 = 0.67, p < 0.05). We then developed a new model combining T-min and EVI, which improved EOS modeling greatly both for these five flux sites and also for data collected at nine PhenoCam sites. Our results imply that the accuracy of current remote sensing VI estimated EOS should be used cautiously. In particular, we revealed the usefulness of nighttime temperature in modeling EOS of evergreen forests, which may be of potential importance for future ecosystem models.