Simulating thin and patchy soils using an Agent-Based Model of forest dynamics, root growth, and soil production

Abstract:

Thin and patchy soils are ubiquitous features of upland hillslopes where erosion rates are high and/or soil production rates are low. Heterogeneous soil cover and depths are important to runoff generation, ecosystem structure and function, hillslope erodibility, and sediment delivery to river channels. Yet modern landscape evolution theory and models struggle to capture the continuous transition from soil-mantled to bedrock-dominated hillsides within a single framework. This challenge is partially due to changes in sediment transport process (e.g., creeping to mass wasting) and a dearth of mechanistic models for soil production. Here, we show how one important soil production mechanism, physical disaggregation of bedrock by tree roots, can be represented using an Agent-Based Model (ABM) of forest dynamics that includes resource competition, seed production, seed dispersal, tree growth, and tree mortality. The key novelty in this approach is that conversion of bedrock into mobile regolith is directly linked to plants via root growth rates that vary with soil depth, lithology, and fracture density. This allows us to simulate forest dynamics for different plant functional types and assess their role on long-term soil production rates. Using the Landlab 2.0 modeling library, we couple the ABM for soil production to linear, depth-dependent, and nonlinear diffusion models of hillslope sediment transport. In this way, we are able to evaluate how fractional soil cover and soil depth distributions respond to differences in model structure and rates of base level fall. We find that tors, or isolated bedrock outcrops, are emergent features in the model and that partial soil cover is a steady state property of hillslopes that is responsive to rates of base level fall. Even though response timescales for forest dynamics are much shorter (10’s to 100’s of years) than for soil development (1,000’s to 10,000’s of years), long-term soil production rates are directly, albeit nonlinearly, controlled by rules for tree reproduction, growth, and death. This model is also among the first to couple a NetLogo ABM to Python-based Landlab via the pyNetLogo library. As such, we believe this use-case is a template for other landscape evolution studies targeting feedbacks among biological agents, substrate properties, and sediment transport.