Evaluation of Partitioning Methods to Identify Timescale-Dependent Drivers of Light Use Efficiency in a Wet Tropical Forest

Abstract. Tropical forests play a major role in the global terrestrial carbon cycle; however, substantial uncertainties remain regarding their capacity to continue acting as a carbon sink under changing climate conditions. This uncertainty is driven largely by the limited number of carbon flux observations in tropical regions. Using eddy covariance (EC) measurements over an 11-month period at La Selva, Costa Rica, we quantified net ecosystem exchange (NEE) and evaluated partitioning methods to estimate gross primary production (GPP) and ecosystem respiration. Daytime partitioning using morning flux data introduced substantial variability in respiration estimates due to rapidly changing micrometeorological conditions. In contrast, daytime afternoon, nighttime, and sundown partitioning methods produced strong agreement and physiologically realistic results. We therefore use the afternoon daytime partitioning method for subsequent light use efficiency (LUE) analysis. The La Selva rainforest functioned as a strong temporally stable carbon sink with no detectable seasonal trend, averaging 2.5 µmol CO₂ m^-2 s^-1, corresponding to an annual carbon uptake of 947 g C m^-2 yr^-1. GPP averaged 11.8 µmol CO₂ m^-2 s^-1. Random forest modeling revealed timescale-dependent environmental controls on LUE, with photosynthetically active radiation dominating short-term variability, while temperature and atmospheric moisture increasingly constrained productivity at weekly timescales. VPD and temperature exerted a strong negative effect on half-hourly LUE, indicating vulnerability to increasing atmospheric dryness. Our findings demonstrate that partitioning method choice and temporal scale strongly shape inferred GPP magnitude and drivers in this wet tropical forest, with important implications for interpreting carbon flux estimates from EC measurements.