Competition alters predicted forest carbon cycle responses to nitrogen availability and elevated CO<sub>2</sub>: simulations using an explicitly competitive, game-theoretic vegetation demographic model

Abstract

Abstract. Competition is a major driver of carbon allocation to different plant tissues (e.g., wood, leaves, fine roots), and allocation, in turn, shapes vegetation structure. To improve their modeling of the terrestrial carbon cycle, many Earth system models now incorporate vegetation demographic models (VDMs) that explicitly simulate the processes of individual-based competition for light and soil resources. Here, in order to understand how these competition processes affect predictions of the terrestrial carbon cycle, we simulate forest responses to elevated atmospheric CO2 concentration [CO2] along a nitrogen availability gradient, using a VDM that allows us to compare fixed allocation strategies vs. competitively optimal allocation strategies. Our results show that competitive and fixed strategies predict opposite fractional allocation to fine roots and wood, though they predict similar changes in total net primary production (NPP) along the nitrogen gradient. The competitively optimal allocation strategy predicts decreasing fine root and increasing wood allocation with increasing nitrogen, whereas the fixed strategy predicts the opposite. Although simulated plant biomass at equilibrium increases with nitrogen due to increases in photosynthesis for both allocation strategies, the increase in biomass with nitrogen is much steeper for competitively optimal allocation due to its increased allocation to wood. The qualitatively opposite fractional allocation to fine roots and wood of the two strategies also impacts the effects of elevated [CO2] on plant biomass. Whereas the fixed allocation strategy predicts an increase in plant biomass under elevated [CO2] that is approximately independent of nitrogen availability, competition leads to higher plant biomass response to elevated [CO2] with increasing nitrogen availability. Our results indicate that the VDMs that explicitly include the effects of competition for light and soil resources on allocation may generate significantly different ecosystem-level predictions of carbon storage than those that use fixed strategies.

Towards better representations of carbon allocation in vegetation: a conceptual framework and mathematical tool

“…Huntingford et al 2000), and models that represented the carbon allocation using allometric equations (e.g. Weng et al 2019) or structural equation modelling (e.g. Hofhansl et al 2014).…”

A model-independent data assimilation (MIDA) module and its applications in ecology

“…The first case study is to demonstrate that MIDA can be effective for independent data assimilation with the data assimilation linked ecosystem carbon (DALEC) model (Lu et al, 2017). DALEC has been used for data assimilation in several studies (Bloom et al, 2016;Lu et al, 2017;Richardson et al, 2010;Safta et al, 2015;Williams et al, 2005). Previous studies all incorporated data assimilation algorithms into DALEC, which requires invasive coding.…”

“…This case study is to examine the efficiency of MIDA to integrate remote sensing data into a dynamic vegetation model. The model used in this study is Biome Ecological strategy simulator (BiomeE) (Weng et al, 2019). This model simulates vegetation demographic processes with individualbased competition for light, soil water, and nutrients.…”

“…Data assimilation (DA), a statistically rigorous method to integrate observations and models, is gaining increasing attention for parameter estimation and uncertainty evaluation. It has been successfully applied to many ecological models (Fox et al, 2009;Keenan et al, 2012;Richardson et al, 2010;Safta et al, 2015;Wang et al, 2009;Williams et al, 2005;Zobitz et al, 2011). However, almost all those DA studies require model-dependent, invasive coding (Walls et al, 2005).…”

“…And a two-pool interactive model was selected after DA to best represent SOC decomposition with priming. Additionally, DA can be applied to locate the most informative data to reduce uncertainty, thus guiding the sensor network design (Keenan et al, 2013;Raupach et al, 2005;Shi et al, 2018;Williams et al, 2005). One DA study at Harvard Forest (Keenan et al, 2013) indicated that only a few data sources contributed to the significant reduction in parameter uncertainty.…”