Assessment of various MODIS products to derive key phenology metrics for a Mediterranean Evergreen Needleleaf forest

Abstract

Monitoring the health of forests has for many years been done using manual observations. More recently with the dissemination of satellite remote sensing platforms, measurements can be continuously taken allowing more complete datasets over larger parts of the globe. With such a comprehensive dataset, assessments on the health of forests over multiple years can be carried out. This study examines changes in phenology on a section of evergreen needleleaf forest located on the Tyrrhenian coast of Tuscany and determines which of a series of MODIS products is the best at tracking these changes. The Flux Tower data at the study site shows variations in annual total productivity and a shortening of the growing season affected mainly by a delay in green up. The phenological timings were extracted from the remotely sensed data using a Savitsky-Golay filter and the 20% threshold method. Of the 6 timeseries created from the 3 MODIS products the MOD17A2H GPP product was the most accurate, precise and robust managing to predict all the dates within the range of validated dates. That being said the correlation between the GPP product, and the validation data was low for the start and end of season R2 = 0.31 and -0.03, respectively. In conclusion, while it is capable of determining the phenology timings it can not depict any changes year on year. This suggests that there may be scope for an optimization of the MOD17A2H radiation conversion efficiency model for evergreen need leaf forests in a Mediterranean climate.