Agent-based modelling of the social-ecological nature of poverty traps in southwestern Madagascar

Abstract

CONTEXTIn Madagascar land use pressure on natural resources has increased drastically over the last decades. For subsistence farmers this situation often leads to an accelerated unsustainable exploitation of resources with negative consequences for their livelihood, which can be described as a social-ecological trap (SET). OBJECTIVEOur paper seeks to understand the dynamics of SETs and demonstrates a model-driven scenario on how farmers may escape the trap. For this, we developed a spatially explicit empirical agent-based model (SEALM) to simulate possible future trends of farmers' crop management and the effects of these trends on the environment, household (HH) economy, food self-sufficiency (FSS) and HH coping strategies to food insecurity. METHODSUsing links between social-ecological system (SES) components that depict real-world conditions, SEALM focuses on three village communities for which extensive data were available from transdisciplinary research comprising surveys, remote sensing analysis, and field-based validation. To identify the most significant parameters and to quantify how parameter uncertainty influences model outcomes, local and global sensitivity analyses were performed. Simulations were conducted in annual time steps from 2015 to 2045 to depict two contrasting scenarios: A baseline scenario (A) and an improved crop management scenario to increase FSS (B). For scenario assessments we used direct simulation outputs (FSS, HH income, land cover) and further indices calculated based on model outcomes (coping strategy index, landscape fragmentation, and the Gini coefficient). RESULTS AND CONCLUSIONSScenario A predicted expansion of cropland and fragmentation of remaining forest areas. Simulations at each time step displayed how SET situations occurred due to self-reinforcing ecological and social constraints and affected HHs future coping capacity to food shortages. Here, FSS was insufficient (< 1.0) for most HHs and was <0.5 for years with rainfall deficits. Scenario B demonstrated a way to navigate out of the SET by increasing crop yields reflecting improved management and resulting in significantly higher FSS values and a lower coping strategy index. SEALM considers the heterogeneity in farmers' abilities, constraints, behaviour and decision-making and depicts the dynamic interplay of SES components contributing to SET. SIGNIFICANCESEALM serves as a discussion support with local stakeholders and experts and it allows to better tailor site specific land use recommendations due to its empirical driven and spatially explicit character. However, strategies and interventions that lower the poverty rate in the long term and encourage exit out of trap situations require major policy changes.