Evaluating autumn phenology derived from field observations, satellite data, and carbon flux measurements in a northern mixed forest, USA

Common approaches currently used to monitor forest phenology include direct field observation and indirect approaches such as satellite remote sensing and carbon flux measurements. However, differences in both temporal and spatial scales of these methods make direct comparison challenging. In order to evaluate the reliability of indirect measures of autumn phenology in estimating direct observations, we compared the timing of three transition dates and the rate of autumn progression derived from (i) satellite data (MOD13Q1 006 enhanced vegetation index (EVI) and normalized difference vegetation index (NDVI) products, 2000-2017), (ii) carbon flux measurements (net ecosystem exchange (NEE) and gross primary production (GPP), 1997-2016), and (iii) field observation (2010, 2012 for the north site and 2010, 2012, and 2013 for the south site) from a mixed forest in northern Wisconsin, USA. Overall, the transition dates and progression rates derived from NDVI were closest to that of field observations. Furthermore, the start of autumn derived from satellite data was earlier than directly observed leaf coloration (LC), which resulted from species-specific canopy senescence patterns and the sensitivity of the vegetation indices. Even after full leaf fall was reached, EVI continued to detect coloring which was likely due to the presence of understory plant species. Finally, NEE and GPP changes tended to start before LC as a result of tree physiological and environmental changes and continued after full leaf fall possibly due to understory and coniferous activity. These results highlight the need for long-term field observations of both trees and understory species, information on species-specific canopy senescence patterns, and species composition in understanding the efficiency of indirect approaches in estimating autumn forest phenology.