Satellite sensors have long been used to monitor ecosystem phenology. These measurements are important for detecting ecosystem change over time and for projecting ecosystem response to global change. However, satellite-derived measures of phenology often compare poorly to phenological markers observed on the ground. In 2007 and 2008 we collected ground measurements focused on estimating changes in overall ecosystem greenness at Sevilleta National Wildlife Refuge, a Long Term Ecological Research (LTER) site in central New Mexico. Field sampling was timed to correspond to Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) satellite image acquisitions. Two perennial grassland sites were selected, one dominated by blue grama grass (Bouteloua gracilis), the other by black grama grass (Bouteloua eriopoda). Measurements of percent green cover were collected within systematically arrayed 30x30 cm quadrats at two-week intervals during the growing season. In addition to visual estimations of percent greenness, digital photographs were taken of each quadrat. We compared visual percent green cover and a greenness index derived from the digital photos to a Normalized Difference Vegetation Index (NDVI) derived from the ASTER data to determine how well the phenology measured in the field corresponds to phenology measured by the satellite.
Results/Conclusions
We show that visual estimates of green cover are strongly correlated to satellite NDVI (R2 > 0.43), while indices derived from digital photographs are poorly correlated (R2 < .13). The poor relationship between the digital camera results and satellite NDVI is likely due to the high percentage of soil cover, which strongly affects greenness indices derived from digital photos. This poor relationship implies that long-term monitoring of phenology based on digital photos may not compare reliably to satellite estimates, particularly in semi-arid systems. We further show that there is no correlation between visual estimates of green cover and satellite NDVI when green cover is lower than 25%. This lack of relationship suggests that phenological changes in semi-arid systems may be very challenging to measure until a threshold of ecosystem greenness has been crossed. This type of comparative study is an important first step for creating better linkages between remote sensing and community-level measurements of ecosystem phenology.
