Land-surface phenology (LSP) measurements (e.g. MODIS Land Cover Dynamics) are retrieved from time series of spectral vegetation indices (e.g. NDVI) that respond primarily to absorbed photosynthetically-active radiation. Thus, LSP measurements record the presence and amount of healthy, green leaf area, and should correspond to the capacity to do photosynthesis. In contrast, the actual rate of photosynthesis is determined by green leaf area in concert with a variety of factors including weather, soil moisture, and nutrient availability. Therefore, while LSP measurements are consistent indicators of the timing of canopy development and senescence, their function as indicators of gross primary production (GPP) is less clear. Some prior work has investigated the correspondence between LSP and photosynthesis using data from eddy-covariance flux towers, but for a limited number of sites and with prior generation LSP products. Here, we examine the relationship between the MODIS data product MCD12Q2 Collection 6 phenometrics and their equivalent FLUXNET2015-derived GPP, where the MODIS phenometrics are based on EVI2 (Enhanced Vegetation Index) and FLUXNET2015 is a standardized dataset of flux tower measurements from around the world. We analyzed 119 sites comprising about 649 site-years of data and covering 13 different International Geosphere-Biosphere Programme (IGBP) land cover types.
Results/Conclusions
Results show a surprising, significant correlation between the phenometrics and GPP thresholds. Most MCD12Q2 phenometrics explained more than 50% of the variation in their GPP equivalents, with the strongest correlations found between MCD12Q2 mid-green-up / dormancy phenometrics and spring GPP 50% / fall 15% amplitude thresholds, respectively (R^2=0.73). However, despite strong agreement for most phenometrics, MCD12Q2 Peak had the weakest correspondence among phenometrics (0.26 R^2 compared with peak GPP). We hypothesize the mostly strong relationship observed here is due to MCD12Q2’s correlation with photosynthesis, it captures variation in green leaf area and absorbed photosynthetically active radiation (PAR). In addition, the scales of MODIS and the flux tower measurements are closely aligned. These results support the use of LSP more directly in ecosystem models. Given the observed relationships yield a more thorough understanding of phenometrics themselves, the results also allow for future analyses to further parse the effect of phenological change on carbon dynamics. By incorporating site data from across the FLUXNET network, this study represents the most comprehensive analysis of its type to date.