California annual grass phenology and allometry influence ecosystem dynamics and fire regime in a vegetation demography model

Abstract

Introduction

Grasslands cover > 30% of the Earth surface; therefore, accurately representing grassland ecosystems in Earth System Models (ESMs) is important for understanding vegetation–climate–fire feedbacks (Blair et al., 2014). Grasslands also store about one-third of global terrestrial carbon stocks, mostly in the form of soil organic matter, which may be more stable under changing climate and shifting disturbance regimes than living biomass (Bai & Cotrufo, 2022; Wilcox et al., 2023). Grasslands are one of the predominant vegetation types in arid and semiarid regions where tree cover is limited by climate and recurrent disturbances (Anderson, 2006). Persistence of grasses in ecosystems such as grasslands and savannas depends on just-enough precipitation and periodic disturbances to prevent woody plant encroachment and maintain a dynamic equilibrium (Scholes & Archer, 1997; Marañón et al., 2009). However, anticipated changes in the frequency and intensity of precipitation extremes and fire disturbances will likely alter species composition and thus ecosystem structure and carbon dynamics in grasslands (Staver et al., 2011; Yu et al., 2017; D'Onofrio et al., 2019). Yet, representing change in these grassy ecosystems in ESMs remains a modeling challenge due to the complexity introduced by climate–vegetation–fire feedbacks and limited investment in simulating herbaceous communities (Beckage et al., 2009, Dantas et al., 2016, Holdo & Nippert, 2023).

In the last decade, dynamic vegetation demography models (VDMs) that capture size-dependent growth, mortality, and competition for water, nutrients and light have been a focus of development by the ESM community to better predict the role of vegetation dynamics on global carbon cycles (Fisher et al., 2018). They are also useful tools for understanding the local and regional drivers of community structure and ecosystem function. However, most vegetation demographic models (e.g. LPJ-GUESS, ED2, and FATES but see aDGVM) were originally developed for closed-canopy forests with most model applications hitherto focused on tree-dominated systems, resulting in less developed model processes and poorly calibrated model parameters for grass plant functional types (PFTs) and open ecosystems (Sitch et al., 2003; Medvigy et al., 2009; Moncrieff et al., 2014; Koven et al., 2020). One of the fundamental differences between trees and grasses is the size-dependent carbon allocation to different plant structures (Niklas, 2004), which is important for understanding plant–environment interactions and species competition (Shipley & Meziane, 2002; Metcalf et al., 2006; McCarthy & Enquist, 2007). For instance, greater allocation to stem biomass enables greater access to light, while greater allocation to deep root biomass enables access to groundwater to help plants avoid drought and effectively compete for water compared to shallowly rooted plants (Holmes & Rice, 1996). Due to the lack of empirical data on individual grass biomass allocation and plant architecture, parameterization of grass allometry (hereafter referring to the size dependence of biomass, allocation, and canopy architecture) in these VDMs is less constrained or does not distinguish between different grass functional types (e.g. C₃ vs C₄ and annual vs perennial grass) despite known species differences in growth and development (Sitch et al., 2003; Medvigy et al., 2009; Nafus et al., 2009).

However, differences in biomass allocation can potentially influence ecosystem structure, function, and fire regime through partitioning of net primary productivity to aboveground biomass, and thus fuel load, or between photosynthetic and supporting structures to influence carbon assimilation and transport (Li et al., 2018). Variations in plant canopy architecture can also influence community structure: A larger crown area may result in reduced stem density per unit ground area due to stronger competition for space and light (Pretzsch et al., 2012). Interactions between biomass partitioning and canopy architecture and the resulting impacts on community structure and ecosystem functioning are not clear. In addition, some grasses are annual plants that differ from perennials in terms of leaf phenology and plant life span, which both influence the seasonal variation in matter and energy exchange and fire regime (Davies & Nafus, 2013). Yet, most VDMs assume grass PFTs to be perennials (Bart et al., 2017). Misrepresenting grass life history, allometry and phenology in vegetation demographic models can generate unrealistic biomass patterns in grasslands, which in turn may affect how the model simulates community structure, fire behavior, and vegetation–fire feedbacks (Wilcox et al., 2023).

Grasses in Mediterranean regions are mainly annual species that are adapted to seasonal droughts by completing reproduction before the onset of the dry season and persisting as dormant seeds until the first rainfall in early winter (Fernández Ales et al., 1993; Volis et al., 2002; Sherrard & Maherali, 2006). Seasonal dynamics in matter and energy exchange, which are driven by the phenology and life history of the dominant species, thus coincide with the seasonal variation in soil water content in these annual grasslands (Xu & Baldocchi, 2004; Liu et al., 2011). Grasses are also the main components of surface fuels in open ecosystems, influencing the regional fire regime (Vilà et al., 2001; Rahlao et al., 2009; Balch et al., 2013). Change in grass phenological phase therefore directly influences wildfire dynamics: As the curing level of grass fuels increases, fire risk increases due to decreased fuel moisture (Wittich, 2011; Cruz et al., 2015). Capturing annual grass phenology, life span and fuel conditions, and their responses to climate variability is important for reconstructing the historical fire regime and projecting community dynamics into the future for these fire-prone ecosystems.

Using modeling experiments with generalized and species-specific grass allometries and varying plant traits, we addressed the following research questions: (1) how does variation in biomass partitioning between leaf and stem, and in canopy architecture, influence simulated ecosystem structure and function in California annual grasslands? (2) What plant traits and ecological processes are important in controlling the mean state in ecosystem properties and fire behavior, and how does trait importance to ecosystem properties change seasonally? We also use site-optimized parameters to run the model across California annual grasslands, validating model performance and determining the main drivers for grassland annual burned area. We hypothesize that: (1) a higher carbon partitioning to leaf than to stem will result in a more productive ecosystem despite the variations in canopy architecture; (2) traits controlling leaf physiology, plant phenology, and sensitivity to soil moisture are important in controlling ecosystem energy and matter exchange and their seasonal variations; (3) fire behavior is determined by both fuel traits and key ecological processes such as mortality and litter decomposition; and (4) variation in ecosystem productivity drives annual burned area at the regional scale.