Dramatic ecosystem responses to global climate change has been reported around the world. Land Surface Phenology (LSP) derived from satellite imagery provides a powerful tool to detect phenological responses of vegetation to climate change at the landscape scale. Most previous studies of remotely sensed LSP used double logistic method or thresholds to determine phenological transitions and growing season lengths based on 8-day or 16-day satellite imagery. However, there is a large (and unquantified) uncertainty in estimated phenological dates due to the relatively coarse temporal resolution and methodological limitations. To assess responses of phenology and related ecological processes, it is essential to narrow the temporal uncertainty of estimated phenological processes. In this study, we estimated four phenological transitions (greenup, maturity, senescence and dormancy) using twice-daily Moderate Resolution Imaging Spectroradiometer (MODIS) imagery from 2000 to 2015. We then compared them with phenological ground observations to evaluate their performance capturing phenology transitions.
Results/Conclusions
In this study, we developed an improved method to identify LSP transitions dates with change point estimation using twice-daily satellite imagery. Our method also quantified the increasing speed of Enhanced Vegetation Index (EVI) during spring, the decreasing pattern of EVI during summer, and the decreasing speed of EVI during fall time. Comparisons indicated that our change point estimation method had better performance in estimating LSP than the double logistic method, especially for the autumn senescence date. In most cases, the estimated senescence dates from MODIS phenology product were too early (more than 30 days) compared to the visually-observed phenology date in the field. Our final product will improve the temporal resolution of phenological data for investigations on landscape phenology at regional to global scales. The transformative method can be applied to estimate change points from any other relevant time series at any spatial scale.