A significant portion of terrestrial net primary production (NPP) contributes to the litter layer. Litter's fate—accumulation (until fire), decomposition, soil incorporation, or dissolved transfer—has global implications due to involved carbon fluxes. Current biogeochemical and Earth system models rely on decomposition understanding primarily from single-species foliage studies, neglecting the non-additive effects observed in mixed-species litter decomposition. Litterbag studies, while dominant in the literature, typically cover 1–3 years, potentially overestimating long-term decomposition rates. These studies often overlook the impact of decomposer fauna and incomplete decomposition. Prior research has highlighted strong interactions between soil fauna and nutrient release in mixed foliage and significant differences in decomposition rates between leaf mixtures and individual species, highlighting the need for long-term studies. Accumulated litter, especially fine fuel, is a key driver of wildfires. The limited applicability of litterbag studies to field conditions necessitates long-term accumulation studies for accurate fire risk assessment. This study uses a continental-scale dataset on litter accumulation in Australian eucalypt forests and woodlands to address knowledge gaps regarding litter decomposition and its consequences for carbon biogeochemistry and fire risk models. The frequent fire regime in these ecosystems facilitates the study of litter accumulation over extended periods. The primary hypothesis is that litter mass is influenced by elapsed time since the last fire, moisture, temperature, and litter quality. Climate indices, frequently used in decomposition studies, are employed, along with a metric for litter quality based on the proportion of different litter types.

Existing research emphasizes the hierarchical control of litter decomposition: climate (moisture, temperature) at the global scale and biota (with biotic interactions with climate) at regional scales. Litter quality, often assessed by species identity, is a major biotic regulator. Short-term litterbag studies often overestimate long-term decomposition rates, and decomposition is seldom complete. Studies of complex litter mixtures (leaves, bark, wood) are limited. Previous work has shown strong interactions between soil fauna and carbon/nitrogen release in mixed-species litter and that differences in decomposition rates between leaf mixtures and single species can be substantial. This gap in knowledge hinders the accuracy of global models.

The study uses a continental-scale dataset on litter accumulation in Australian eucalypt forests and woodlands, facilitated by frequent fire regimes. The data includes litter mass, litterfall, climate data (annual precipitation, mean daily temperature, Aridity Index (AI)), and geographic location. Litter quality was assessed using the proportion of leaves in annual litterfall (Qlf). Five well-described eucalypt communities, representing ~40% of the litter observations, were selected for detailed analysis, with data also aggregated for other vegetation formations and ash forests. A range of models (linear, power, exponential, polynomial) were tested to assess the relationship between litter mass and elapsed time since fire (Tsf), AI, and Qlf. The best-fit models were selected based on AIC and goodness-of-fit (R²). Data limitations, particularly the reduced availability of data for longer elapsed times and incomplete combustion issues, were considered in the analysis. Alternative models and quality metrics (based on litter in situ rather than litterfall) were also tested. Multivariate models were used to evaluate the effects of Tsf, AI, and Qlf on litter accumulation. Heteroscedasticity-consistent errors were calculated using the lm_robust function in R. Collinearity was addressed using orthogonal polynomials. Analyses were conducted using both total litter and leaf litter as target variables. Climate data from the TerraClimate database was used (1970–2000).

Geographic distribution of litter and litterfall data reflected continental patterns of eucalypt-dominated forests and woodlands. Sampling year had no effect on litter mass, but annual litterfall varied considerably. Litter accumulation increased continuously with time (up to 40 years). Power and polynomial functions better described litter accumulation than linear or exponential functions for most communities. Climate (AI) directly affected litter accumulation, with a maximum litter mass at AI = 1.5. Litter quality (Qlf) showed scale-dependence related to climate and large variation within forest types. Multivariate models incorporating Tsf, AI, and Qlf significantly improved the explanation of litter mass variance compared to models with only Tsf. Models including AI and Qlf explained ~45% of the variance in litter mass for over 100 years. Model fit was improved by limiting data to 40 years of elapsed time (up to 65% explained variance). For some individual forest communities, elapsed time alone explained most of the variance in litter accumulation. The robust model for complex litter layers showed that litter accumulation over time is influenced by both constant climate and constant litterfall quality (Fig. 6). Climate influence on litter accumulation was both direct (via AI) and indirect (via Qlf), with the latter representing an emergent ecosystem property. The climate-Qlf relationship was scale-dependent, contrasting with the scale-invariant direct effects of climate.

This continental-scale analysis provides a robust test of hypotheses derived from shorter-term litterbag studies. Climate and litter quality clearly regulate decomposition, and their effects compound over time. The study highlights limitations of litterbag studies, including the assumption of exponential decomposition behavior, which is not supported by the field data. The use of decomposition constants (k) is problematic due to the overestimation of decomposition rates and incomplete decomposition. The study's models, spanning a minimum 40-year period, far exceed the duration of the longest litterbag studies. The models demonstrate the ability to explain variance in litter mass in situ for periods extending beyond 100 years, contrasting with the lower explained variance in long-term litterbag studies. The inclusion of AI and Qlf significantly improved model performance, highlighting the importance of both direct and indirect climate effects on litter accumulation. The climate-dependent Qlf represents an emergent ecosystem property that should be considered in models. Assumptions of exponential litter accumulation behavior and limits to litter mass used in fire risk modeling are not supported by the data, necessitating adjustments in these models.

This study provides a robust continental-scale model for litter accumulation in eucalypt forests, accounting for the influence of time, climate, and litter quality. The model significantly improves the accuracy and reliability of predictions of carbon and nutrient dynamics and fire risk compared to models based solely on shorter-term studies and assumptions of exponential decomposition. Future research should extend this analysis to other continents to validate the findings and incorporate detailed studies of litterfall to better understand the interplay between climate, litter quality, and fire risk.

The dataset's bias towards shorter time periods since fire (Tsf) affects the model's accuracy for longer Tsf. Although orthogonal polynomials addressed collinearity among variables, interpretations might benefit from using orthogonal polynomials to compare covariate importance. The analysis was limited to eucalypt forests and woodlands in Australia, limiting the generalizability to other forest types and geographical regions. Furthermore, data limitations regarding longer Tsf might have outsized influence in some statistical analysis.