|
Linking interannual phenology from MODIS and Landsat
Eli
Melaas, Boston University, emelaas@bu.edu
(Presenter)
Mark
Friedl, Boston University, friedl@bu.edu
Chris
Holden, Boston University, ceholden@bu.edu
Josh
Gray, Boston University, joshgray@bu.edu
More than a decade of MODIS data is now available for assessing vegetation responses to global climate change. As a result, MODIS has been widely used to study large-scale vegetation dynamics, and a large number of studies have used time series of MODIS EVI and NDVI data to estimate land surface phenology metrics at regional-to-global scales. Despite significant progress in algorithm development, ground-truth data to support validation of year-to-year variability in land surface phenology products derived from MODIS remains a challenge because the moderate resolution of MODIS pixels (≥ 250m) is much larger than the scale at which ground data are collected, and because long-term ground observation datasets (≥ 5 years) are rare. We demonstrate that time series of Landsat TM and ETM+ images have the potential to help resolve this challenge by accurately characterizing long-term average and interannual dynamics in phenology at medium resolution (30m). To do this, we use all available data across six Landsat scenes in the Northeastern United States to generate multi-decadal time series of spring and autumn phenology for deciduous broadleaf forests and compare the results with phenology products derived from MODIS. Our results show that average and year-to-year variation in phenological transition dates estimated from Landsat agree closely with corresponding transition dates from the MODIS Land Cover Dynamics product (MCD12Q2), but that the MODIS product has an early bias of about one week in both spring and autumn. The level of agreement between results from MODIS and Landsat increases for pixels with higher cover of deciduous forest and when the same methodology is used to estimate phenology from MODIS and Landsat. Based on these results, we propose an integrated algorithm for generating phenology time series at 500-m spatial resolution that uses a combination of MODIS and Landsat data.
Presentation Type: Poster
Session: Poster Session 2-A
(Wed 11:00 AM)
Associated Project(s):
Poster Location ID: 63
|