Integration of carbon fluxes and ground-based broadband vegetation indices (VIs) in a tropical dry forest in Western Mexico

Background/Question/Methods .

Monitoring plant phenology is fundamental to quantify the global carbon cycle and to understand climate-vegetation interactions. Remote sensing of canopy greenness from satellites allows tracking phenological changes over large areas. However, remote estimates of canopy status does not always converge with direct observations of carbon fluxes at ground, partly due to the spatiotemporal mismatch of the measurement scales. In this context, the use of ground-based VIs provides a way to monitor phenology at the same spatial scale of the source area for carbon fluxes, and at comparable temporal frequencies, while being directly scalable to satellite estimates of primary productivity. To understand the dynamics of carbon exchange in a tropical dry forest (TDF), in 2008 we deployed a set of optical sensors at a height of 18 m, in the Chamela-Cuixmala Biosphere Reserve, in Jalisco, Mexico. The optical set consisted of two Photosyntetically Active Radiation sensors and two pyranometers, one installed looking upwards and one looking downwards. From the sensors signals, the broadband Normalized Difference Vegetation Index (NDVI) and the Enhanced Vegetation Index 2 (EVI2) were calculated every 30 min and compared to the eddy covariance Net Ecosystem Exchange of CO₂ (NEE) measured simultaneously.

Results/Conclusions .

The broadband VIs accurately captured the phenology of the TDF. New leaves develop within 10-13 days after the first substantial precipitation, remain active as long as rain falls regularly, and senesce and fall for several months into the dry season. VIs captured two secondary leaf-flushings triggered by rain pulses in the dry season of 2010 and 2013, and the sudden defoliation caused by hurricane Jova in 2011, followed by the quick recovery of canopy over 13 days. The results of relating the VIs to NEE indicate that 1) NDVI shows a greater saturation at high levels of NEE compared to EVI2, 2) Pearson´s correlation coefficients between midday EVI2 and NEE are high and statistically significant (P<0.01) for an entire growing season (r =-0.95 for 2008, r =-0.85 for 2009, r=-0.91 for dry season rains in 2010, y r =-0.88 for the rainy season of 2010), and 3) inclusion of the soil water content as a second predictor for NEE better resolves the saturation of the EVI2 at high values of NEE. In a system highly driven by the dependence of canopy phenology to water availability, the combination of flux measurements with ground-based VIs advances our understanding of the seasonal and annual variability of carbon exchange.