Carbon flux responses to seasonal and annual hydroclimatic variability in a tropical dry forest in South Ecuador

Abstract. Tropical dry forests play an important role in the global carbon cycle, but their responses to climate variability are still not well understood. Using a three-year period (April 2022 to March 2025) of eddy covariance measurements, we studied seasonal and annual controls on carbon balances in a Tumbesian dry forest in Southern Ecuador. During the study period, the forest functioned as a net carbon sink, with a net ecosystem exchange (NEE) of -285 gCm⁻² year⁻¹. The strongest carbon uptake occurred during the wet period (Feb–May) with 173.86 ± 66 gCm⁻² month⁻¹, while it was reduced to 39.80 ± 8.12 gCm⁻² month⁻¹ in the dry season (August–November).

Light use efficiency (LUE) and water use efficiency (WUE) were used to characterize the functional controls on carbon fluxes at both seasonal and annual scales. WUE showed relatively stable water–carbon exchange, whereas LUE displayed clear seasonal variation, reflecting the strong influence of seasonal vegetation growth and greenness. Principle component analysis (PCA) was conducted to further analyze controlling mechanisms in carbon fluxes. Seasonal results showed that gross primary productivity (GPP) was mainly controlled by energy-related factors, while ecosystem respiration (Reco) was primarily driven by a moisture–temperature gradient. Annually, GPP was predominantly influenced by variations in vapor pressure deficit (VPD), soil temperature (Ts), and incoming radiation (Rg), reflecting a strong coupling between surface energy balance and atmospheric moisture demand. These drivers were further modulated by ENSO related climate variability, as reflected by shifts in their PCA loadings across years. Overall, the results reveal a decoupling between photosynthesis and respiration and show that tropical dry forests are highly vulnerable to increasing climate extremes, highlighting the need for improved representation of these processes in Earth system models.