Carbon allocation to reproduction: implications for predicting tropical forest carbon dynamics

In vegetation demographic models (VDMs), reproductive allocation (RA) is the first step of tree reproduction, whereby a fraction of available carbon (i.e. NPP after respiration and tissue turnover) that is allocated to reproduction (construction of flowers, fruits, seeds). RA is not only necessary to drive regeneration within VDMs, but also represents a key demographic trade-off: surplus carbon not allocated to reproduction is allocated to growth. In part because RA is difficult to directly observe, parameter values (typically a fixed fraction, 0.10-0.30) are not well supported. However, recent empirical work suggests that RA differs across plant functional types (PFTs) and increases with tree size/age (decreasing fractional carbon allocated to growth). In order to better constrain RA within VDMs, we first synthesize tree census data and weekly leaf and reproductive litterfall observations from Barro Colorado Island, Panama to construct ecosystem- and PFT- level RA benchmarks. We then compare these benchmarks to model runs from a VDM, FATES. Additionally, we test the sensitivity of growth rates and aboveground biomass (AGB) accumulation to RA. We hypothesize that a low RA value (0.10) will overestimate growth rates and AGB accumulation, particularly in the largest size classes where the pool of surplus carbon is expected to be largest. We find that light-demanding PFTs allocate slightly more to reproduction than shade-tolerant PFTs per unit leaf production (0.38 vs. 0.19), but do not find significant differences between drought-tolerant and drought-intolerant PFTs. Our sensitivity analysis finds that an increase in RA from 0.10 to 0.40 results in an average decrease of 56% in ecosystem-level growth rates and an 18% decrease in AGB. These results demonstrate the influence of RA on growth rates and AGB, and suggest that PFT-specific RA would drive differential growth rates in VDMs, with implications for competitive dynamics, recruitment, and PFT-composition. Future work will further investigate the empirical relationship between RA and PFTs, and will explore size-class dependent RA parameterization within VDMs.