Litter accumulation and fire risks show direct and indirect climate-dependence at continental scale

A large share of terrestrial net primary production enters the litter layer via aboveground litterfall, after which litter accumulates, decomposes to CO2, becomes stabilized organic matter, or is transferred in dissolved form. Global biogeochemical and Earth system models largely rely on litterbag-based understandings derived from single-species foliage, despite non-additive interactions in mixtures and the influences of decomposer fauna. Short-term litterbag studies (1–3 years; a few up to a decade) often overestimate long-term decomposition, and mesh size can bias fauna access. The decomposition of complex mixtures (leaves, bark, wood) is under-studied. This study uses a continental-scale dataset from Australian eucalypt forests and woodlands to test the hypothesis that litter accumulation (the outcome of litterfall inputs and decomposition) is regulated by climate (moisture and temperature), litter quality, and time since last fire. Formally, litter mass should relate to elapsed accumulation time and indices capturing moisture/temperature (e.g., aridity index) and litter quality, providing a robust basis to improve carbon and fire risk models.

Prior work shows climate as a primary control on decomposition at global scales, with biota and litter quality becoming dominant at regional and local scales. Litterbag studies reveal non-additive mixture effects, strong soil fauna influences, and that short-term measurements overestimate long-term rates; decomposition is often incomplete. Large programs (LIDET, CIDET) extended to ~10 years but still miss multi-decadal dynamics. Decomposer fauna, microbial community composition, and physical proximity among substrates influence decomposition. Common modeling with single exponential decay constants (k) has been criticized for overestimating decay and ignoring changing rates over time and incomplete decomposition. Studies highlight that decomposition of mixtures differs from single species, and that indirect effects (e.g., climate shaping litter quality and species composition) may be important. Fire models often assume rapid approach to fixed fine-fuel limits, an assumption questioned by empirical observations of continued accumulation over decades in eucalypt systems.

Data: Compiled a continental-scale dataset for Australian eucalypt forests and woodlands including standing litter mass and annual litterfall, with site-level climate and fire records. The full database contains >1812 litterfall and 3887 standing litter datapoints; analysis focused on eucalypt-dominated systems (genera Eucalyptus, Corymbia, Angophora). Exclusions included wet tropical forests with mixed dominance, inland Acacia shrublands, Casuarina/Callitris woodlands, and plantations. Litter and litterfall data were categorized into leaves, wood, bark, fruits; wood/bark particle-size thresholds varied (most commonly <6 mm; some at 10 mm or 26 mm), and reported values were used as-is. C and N concentrations (n>274) indicated higher C:N for wood/bark than leaves, informing quality proxies. Independent variables: Time since last fire (Tsf) per observation; climate via Aridity Index (AI = precipitation / potential evapotranspiration; TerraClimate 1970–2000); litter quality proxies: Qlf = leaf litterfall / total litterfall, and Ql = leaf litter / total standing litter. Qlf reflects input mixture quality; Ql reflects in situ composition influenced by early rapid losses. Species/community aggregation used dominant overstory eucalypt species; additional aggregations by formation (grassy forests/woodlands) and ash forests (genetic grouping) were also considered. Modeling: Evaluated functional forms (linear, exponential, power, polynomial/quadratic) for relationships between litter mass and Tsf within communities and across all eucalypt sites. Compared models using AIC and R2, with heteroscedasticity-consistent errors (lm_robust in R). Addressed multicollinearity using orthogonal polynomials (found similar explained variance; presented raw polynomials for interpretability). Conducted stepwise inclusion of AI and then Q metrics to quantify gains in explained variance. Due to data bias (92% of Tsf observations <40 years), analyses were performed for (a) all Tsf and (b) Tsf ≤ 40 years. Climate–quality relationships were examined (Qlf vs AI). Visualizations and regressions were performed in R; climate data from TerraClimate. Fire record uncertainties beyond ~40 years constrained long-Tsf inference.

- Litter accumulation increases with time since fire across the Australian continent and within individual/aggregated eucalypt communities for at least 40 years. Accumulation tends to slow with time; power and quadratic functions generally fit better than linear or exponential forms (AIC-based). - Direct climate effects: Litter mass shows a non-linear relationship with AI; mass peaks around AI ≈ 1.5 (precipitation 50% greater than PET). - Indirect climate effects via quality: Qlf (leaf fraction of litterfall) varies five-fold (~0.2–1.0) across the continent and is climate-dependent (emergent property). For five representative communities, Q vs AI: Q = 1.009 − 0.387 AI (P < 0.001, R = 0.32). - Multivariate models combining Tsf, AI, and Qlf (modified quadratic: X_Tsf = 1 + aTsf + bTsf^2 + cAI + dQlf) significantly improved fit over time-only models (all P < 0.001). Time remained the dominant driver of explained variance, accounting for up to 90% in some contexts. - Model performance (Table 2): • All eucalypt forests, all Tsf: n = 119; R2 = 0.44; coefficients significant for Tsf (a), Tsf^2 (b), AI (c), Qlf (d) except where noted. • Representative forests, all Tsf: n = 85; R2 = 0.45. • All eucalypt forests, Tsf < 40 years: n = 108; R2 = 0.59. • Representative forests, Tsf < 40 years: n = 78; R2 = 0.65. - Ql (in situ quality) improved models but was inferior to Qlf; models targeting leaf-only litter were weaker than those for total litter mixtures. - For several communities (e.g., E. marginata, E. diversicolor, grassy forests and woodlands), Tsf alone explained most variance in litter mass. - Figure 6 demonstrates model predictions showing how fixed quality across AI gradients or fixed AI across quality gradients affect accumulation trajectories, illustrating direct and indirect climate effects. - Contrary to many fire models, data do not support exponential accumulation with fixed upper limits; fine litter continues to accumulate for at least 40 years, implying implications for fire behavior and management under changing climates.

Findings support the hypothesis that both direct climate (via moisture and temperature captured by AI) and indirect climate effects (via climate-driven variation in litterfall quality) regulate decomposition and hence litter accumulation, compounding through time. Time since fire is the dominant predictor, but inclusion of AI and Qlf significantly increases explained variance at continental and community scales. The climate–quality linkage shows scale dependence (emergent property) in contrast to more scale-invariant direct climate effects. The results challenge common assumptions in decomposition and fire-risk modeling: single-pool exponential decay constants (k) overestimate long-term decomposition and ignore deceleration and incomplete decomposition; assumptions of rapid convergence to fixed fine-fuel limits are not supported. Complex litter mixtures behave differently from single-species foliage observed in litterbags, with biotic interactions and physical proximity facilitating transfers among substrates and buffering decomposition. Microclimate feedbacks from accumulating litter (insulation, moisture retention) likely contribute to the modest quadratic behavior, with potential eventual declines if decomposition outpaces inputs. Incorporating parsimonious, empirically grounded functions of Tsf, AI, and Qlf can improve continental-scale estimates of carbon pools/fluxes and refine fire risk assessments. The demonstrated climate dependence of input quality underscores how climate change (e.g., drought-induced shifts in leaf vs wood/bark inputs, crown dieback) could alter litter dynamics and fire hazards. The work aligns with calls to integrate low-parameter empirical models into complex ecosystem models to enhance robustness.

This continental-scale study provides long-term (up to 40+ years) evidence that litter accumulation in eucalypt forests and woodlands is primarily driven by time since fire, with substantial additional regulation by direct climate (AI) and climate-mediated litterfall quality (Qlf). Parsimonious quadratic models explain 44–65% of variance across datasets and outperform assumptions of exponential decay and fixed fuel limits. The approach offers practical algorithms to improve predictions of carbon and nutrient dynamics and fire risk. Future research should: (1) test these models on other continents and biomes; (2) expand long-term measurements of both litter accumulation and litterfall quantity/quality; (3) refine quality metrics and incorporate biotic drivers (e.g., termites, ecosystem engineers) where scale-appropriate; and (4) update fire risk models to reflect continued accumulation and climate-linked quality effects under changing climates.

- Temporal bias: Tsf recorded for ~50% of litter observations; ~92% of Tsf data are <40 years, limiting inference at longer times due to uncertain fire histories and potential outsized influence of sparse long-Tsf data. - Data completeness: Litterfall datasets are more limited than standing litter; Qlf could not be computed for some communities (e.g., insufficient E. regnans observations). - Methodological variability: Inconsistent particle-size thresholds for wood/bark across sources introduce added variance; potential contamination by incompletely combusted pre-fire litter in early post-fire years. - Multicollinearity among predictors required attention (addressed with orthogonal polynomials in sensitivity analyses). - Scope: Results pertain to Australian eucalypt systems; generalizability to other regions needs confirmation. - Quality metrics: Ql reflects in situ composition and early mass loss processes, complicating interpretation relative to initial input quality; other quality measures (e.g., lignin) had too few observations.