Daytime water and CO2 exchange within and above the Amazon rainforest

Abstract

We studied the sub-hourly variability of water and CO2 fluxes within and above the Amazon tropical forest during the dry season. Our aim was to investigate how forest layers contribute to the net ecosystem exchange (NEE) and evapotranspiration (ET) by accounting for the existing vertical profiles of leaf traits and microclimate, and the presence of clouds. To this end, we estimated NEE and ET with a 3-layer land-surface model driven by vertical profiles observed at the Amazon Tall Tower Observatory (ATTO) site during CloudRoots-Amazon22 field campaign. Our analysis focused on a typical “shallow-convective” day, composed of 6 individual days with similar conditions. The observed vertical profiles characterized microclimatic variables (photosynthetic active radiation, air temperature, humidity, CO2, wind and turbulence) as well as leaf traits (maximum carboxylation rate, Vcmax; maximum electron transport, Jmax; and 13C leaf isotopic composition, δ13Cp) across three layers that represented the understory, the mid-canopy and the forest top-crown. The modeled NEE and ET were subsequently compared with eddy-covariance (EC) fluxes observed above the canopy. We found distinctive vertical profiles of leaf traits and microclimate that were maintained during most of the day. In particular, we observed a persistent inversion of temperature within the canopy which hindered air mixing between the top-crown and the lower layers. Modeled NEE and ET fluxes were comparable in magnitude to the EC-measured fluxes, with flux dynamics dominated by the exchange of the top-crown layer. However, differences between observed and modeled NEE emerged during the morning transition (from 7:30 to 9:00 LT), when CO2 stored within the canopy overnight was released. We conclude that tropical forests exhibit complex, distinctive vertical profiles of microclimate and leaf traits that influence the water and CO2 vegetation exchange and the transport of air within the canopy.