Matching land surface phenology with the phenology of net ecosystem exchange (NEE) in the Kruger National Park, South Africa

Abstract

Vegetation phenology plays an important role in the inter-annual variability of Net Ecosystem Exchange (NEE). However, the relationship between vegetation phenology and NEE is still largely unknown in the southern African region. The relatively recent availability of Eddy-Covariance (EC) flux towers in the region, which allow for measurements of the interactions between the terrestrial biosphere and the atmosphere, have expanded our understanding of Carbon Flux Phenology (CFP), but the towers' spatial coverage and representation is still limited. By contrast, satellite-based measurements of terrestrial environments provide repeat global spatial coverage that offers a complimentary approach to flux tower estimates of phenology and NEE. This study therefore aimed to investigate the relationship between Land Surface Phenology (LSP), calculated from FAPAR (Fraction of Absorbed Photosynthetically Active Radiation) derived from MODIS (Moderate Resolution Imaging Spectroradiometer) and from MISR (Multi-angle Imaging Spectroradiometer), and CFP using 18 years of EC flux data at the Skukuza flux tower site, Kruger National Park, South Africa. Cross-correlation results showed that there is a moderate association between NEE and FAPAR from both sensors (MISR-HR, R2 = 0.54; MODIS, R2 = 0.53) where FAPAR leads NEE by 8-days. Specifically, only the start of carbon uptake (SOSNEE) has a strong correlation with FAPAR derived start of season (SOSFAPAR). In this instance, MISR-HR FAPAR outperformed MODIS at estimating the SOSNEE dates (i.e., R2 = 0.61 and R2 = 0.28, respectively) and overall showed the least difference to CFP (Mean signed difference ∼ 7–12 days). The findings show that MISR-HR FAPAR can be successfully used to approximate the SOSNEE as well as NEE. In summary, this study presents insight into the relationship between LSP and CFP that contributes to our understanding of how LSP can be used to assist monitoring of carbon dynamics in African savannas. Further investigation into the complicated nature of the interactions between environmental variables (e.g. rainfall, soil moisture, temperature, etc.) that influence both vegetation and carbon dynamics will provide important insight into the potential of upscaling NEE estimates.