You are here

Harvard Forest >

Harvard Forest REU Symposium Abstract 2012

  • Title: Determining the reliability of MODIS vegetation indices to assess vegetative phenological trends across regional scales
  • Author: Erin A Frick (Mount Holyoke College)
  • Abstract:

    Phenology is the study of recurring biological events such as vegetation growth cycles and animal migration. As plant growth is correlated to land surface temperature, these measures are indicative of climate change trends. The heterogeneity of phenological patterns prevents field observations, which have been implemented to study phenology for centuries, from serving as an optimal method for phenological assessment across regional to continental scales. Alternatively, remote sensing devices such as the Moderate Resolution Imaging Spectroradiometer (MODIS), may be used to estimate phenological trends using measures of vegetation greenness inferred from wavelengths of reflected solar radiation. A compelling question that is frequently addressed using these data is where and how vegetation phenology is shifting in response to climate change.

    Several datasets within the MODIS archive provide time series of vegetation greenness. It is necessary to determine which of these products provides the greatest quantity of reliable data to estimate the timing of phenological events with confidence. We investigated the Nadir BRDF-Adjusted Reflectance (MCD43A4) product which accounts for the effect of sun-illumination geometry and the Vegetation Indices (MOD13A1) product which is not BRDF-adjusted. A key question is whether or not MOD13A1 contains enough high-quality information to accurately calculate phenophase transition dates compared to data provided by MCD43A4. MODIS datasets, including these products, differ as a result of unique processing nuances involving compositing procedures, algorithmic manipulation and incorporation of ancillary data. To assess the differences between, and overall ‘usefulness’ of these products, we produced various plots and maps based on vegetation greenness data spanning a decade. These results include comparisons of the mean date of spring onset and assessment of the length of high-quality EVI time series across MODIS products and forested land cover types.

    Our results suggest that the usefulness of the MOD13A1 product is limited by quality control filtering. If this filtering is too exclusive, insufficient data are available to derive accurate phenophase transition dates. With respect to the reliability of spring onset dates derived from these products, the MOD13A1 product consistently returns substantially later spring onset dates than both MCD products. This result is counterintuitive: we would expect these dates to be similar as they use the same EVI values to determine spring onset. By adjusting this quality threshold to balance the competing needs for quality and quantity, we expect the MOD13A1 product to become a more reliable source of data for phenophase transition date calculations. These estimations may then be used to model future vegetative response to changing climatic conditions on a global scale.

  • Research Category: Ecological Informatics and Modelling