Information content in time series of litter decomposition studies and thetransit time of litter in arid lands

Plant litter decomposition stands at the intersection betweencarbon (C) loss and sequestration in terrestrial ecosystems. During thisprocess organic matter experiences chemical and physical transformationsthat affect decomposition rates of distinct components with differenttransformation fates. However, most decomposition studies only fit one-poolmodels that consider organic matter in litter as a single homogenous pooland do not incorporate the dynamics of litter transformations and transfersinto their framework. As an alternative, compartmental dynamical systems aresets of differential equations that serve to represent both theheterogeneity in decomposition rates of organic matter and thetransformations it can undergo. This is achieved by including parameters for the initial proportion of mass in each compartment, their respectivedecomposition rates, and mass transfer coefficients between compartments.The number of compartments as well as their interactions, in turn,determine the model structure. For instance, a one-pool model can be considereda compartmental model with only one compartment. Models with two or moreparameters, in turn, can have different structures, such as a parallel one if eachcompartment decomposes independently or in a series if there is mass transferfrom one compartment to another. However because of these differences inmodel parameters, comparisons in model performance can be complicated. Inthis context we introduce the concept of transit time, a random variabledefined as the age distribution of particles when they are released from asystem, which can be used to compare models with different structures. Inthis study, we first asked what model structures are more appropriate torepresent decomposition from a publicly available database of decompositionstudies in arid lands: aridec. For this purpose, we fit one- and two-pooldecomposition models with parallel and series structures, compared theirperformance using the bias-corrected Akaike information criterion (AICc) andused model averaging as a multi-model inference approach. We then asked whatthe potential ranges of the median transit times of litter mass in arid landsare and what their relationships with environmental variables are. Hence, wecalculated a median transit time for those models and explored patterns in thedata with respect to mean annual temperature and precipitation, solarradiation, and the global aridity index. The median transit time was 1.9 yearsfor the one- and two-pool models with a parallel structure and 5 years forthe two-pool series model. The information in our datasets supported allthree models in a relatively similar way and thus our decision to use amulti-model inference approach. After model averaging, the median transit timehad values of around 3 years for all datasets. Exploring patterns oftransit time in relation to environmental variables yielded weak correlationcoefficients, except for mean annual temperature, which was moderate andnegative. Overall, our analysis suggests that current and historical litterdecomposition studies often do not contain information on how litter qualitychanges over time or do not last long enough for litter to entirelydecompose. This makes fitting accurate mechanistic models very difficult.Nevertheless, the multi-model inference framework proposed here can help toreconcile theoretical expectations with the information content from fieldstudies and can further help to design field experiments that betterrepresent the complexity of the litter decomposition process.

Automated summary

Plant litter decomposition plays a crucial role in the carbon dynamics of terrestrial ecosystems. Most decomposition studies only use one-pool models that do not consider the heterogeneity and dynamics of litter transformation. In this study, we compared the performance of one-pool and two-pool decomposition models with different structures and used a multi-model inference approach to analyze the median transit times of litter mass in arid lands and their relationships with environmental variables.