Evaluating the potential of MODIS satellite data to track temporal dynamics of autumn phenology in a temperate mixed forest

Abstract

Autumn phenology plays a critical role in regulating growing season duration and can be estimated from satellite remote sensing. However, to date, little work has been undertaken to evaluate the performance of remotely sensed autumn phenology, mainly due to a lack of spatiotemporally compatible field observations. To address this limitation, we conducted intensive ground observations of leaf coloration and leaf fall from 610 deciduous trees at two 625×625m study areas within a mixed forest in northern Wisconsin, USA during 2010 and 2012. We derived landscape phenology (LP) indices by upscaling these plot-level observations to facilitate spatially compatible comparisons with coarse resolution satellite measures. The satellite-derived land surface phenology (LSP) was based on 250m Moderate Resolution Imaging Spectroradiometer (MODIS) Normalized Difference Vegetation Index (NDVI) and Enhanced Vegetation Index (EVI) data from both standard 16-day composite (MOD13Q1) and daily Nadir Bidirectional Reflectance Distribution Function (BRDF)-Adjusted Reflectance (NBAR) products. The results revealed that LSP dormancy onset differed from the observed date of full leaf coloration by 5.25days on average (ranging from 0 to 12days). Furthermore, progression of the autumn season as determined from LSP and LP showed close agreement as increasing LP leaf coloration corresponded to declining NDVI and EVI values. In addition, the end of the leaf coloring phase was marked by a simultaneous stabilizing of both NDVI and EVI time series whereas the timing of the end of the growing season (leaf fall) closely corresponded to minimum NDVI values. These findings clearly support the use of satellite measurements to effectively monitor temporal dynamics of autumn phenology in a temperate mixed forest.