RT Dissertation/Thesis
T1 Extreme climate shock and locust infestation impacts in Ethiopia : farm-level agent-based simulation of adaptation and policy options
A1 Ejeta,Alemu Tolemariam
WP 2023/07/05
AB Extreme climate shocks have been a daunting problem for smallholder farmers in Ethiopia for a decade. In recent years, locust invasions in many parts of the country have become another livelihood challenge to the subsistence farming population who already lives in dire livelihood situations. These two compounding shocks can lead to total crop failure at the early crop development stage or any crop growth stage. They are creating a massive economic upheaval in rainfed-dependent countries particularly affecting the well-being of resource-poor subsistence farmers. To reduce the effect of recurring shocks, especially climate risks, farmers have been implementing different risk management strategies. In addition to farmer autonomous adaptation practices, the government has been supporting farmer climate adaptation efforts by designing different policy interventions. In locust-hit areas, government and non-governmental organizations have designed and implemented different locust relief programs aimed at reducing associated welfare losses. Whether farmers can adapt to the effects of climate shocks or not by autonomous adaptation and/or with policy support is an empirical policy question. Moreover, as there are no studies of locust impacts and locust relief programs evaluation, the degree of locust livelihood devastation and the roles of locust relief policy interventions in minimizing the effect of locust shock are policy concerns.
To address these important and key empirical questions, this thesis applied a farm-level agent-based simulation model. MPMAS, a modeling framework developed at the University of Hohenheim for agent-based simulations, was applied to capture inseparable production and consumption decisions of subsistence farming households in the Central Rift Valley of Ethiopia. The modeling framework uses a whole-farm mathematical programming modeling approach to represent complex dynamics of farm household decisions where a set of constraints and their complex relationships are considered. This simulation model enables scenario-based policy analysis by comparing different climate, locust, and policy scenarios which is hardly possible using statistical and other reduced forms of econometrics models. Through establishing scenarios, the model helps to disentangle the pathways through which external shocks may affect the well-being of smallholder farmers.
MPMAS has been extensively applied for policy simulations in different countries including Ethiopia. This thesis extends previous MPMAS applications in Ethiopia by including new features for Central Rift Valley (MPMAS_CRV). MPMAS_CRV was parameterized from the CIMMYT household survey augmented with CSA datasets and own field research. Smallholder farmers ex-ante considerations of risk management strategies for possible climate shock are explicitly captured in MPMAS_CRV to assess their role in climate adaptation and welfare improvements. As part of enhancing the adaptive capacity of farm households to recurring climate shocks, the effect of policy interventions such as better access to credit services and improved agricultural technology are quantified by establishing climate and policy scenarios. Similarly, the thesis quantified the impact of locust invasions on household welfare outcomes and their response to locust relief interventions including food or cash transfers complemented with inputs and livestock provisions. Locust simulation is one of the novelties of this research as it is the first study to explicitly capture the welfare effects of the desert locust and assess the roles of locust relief programs through the application of MPMAS.
To enable climate and locust shock effects quantification and associated policy interventions, different simulation experiments were designed comprised of climate and locust shock frequencies and policy scenarios. The simulation experiments and analysis were performed using the computational resources of bwForCluster within the bwHPC infrastructure in the state of Baden-Württemberg, Germany.
Before using MPMAS_CRV for policy simulations, its reliability was validated using land use, livestock holding, and amount of crop sales by comparing simulated against observed survey values. The validation results suggest that MPMAS_CRV can represent and reflect real-world conditions so that it is reliable to use for impact quantification and policy simulations. In addition to empirical validation, the thesis conducted a global uncertainty analysis to check the robustness of the simulation results under different parameter variations and combinations to minimize erroneous policy formulations. Uncertainty analysis results show that the model converges rapidly at 50 repetitions which implies that these model repetitions are enough to cover the model uncertainty space.
In terms of extreme climate impacts and adaptations, the simulation results suggest that climate shocks affect the welfare of agents adversely to the extent that they face temporary food shortages, loss of discretionary income, and depletion of livestock assets. The welfare losses are similar for both with and without ex-ante measure scenarios which indicates that farm agents cannot adapt to extreme shocks by employing autonomous adaptations. After the shocks are over, the simulation results reveal that agents cannot recover income and livestock losses immediately even when they consider ex-ante measures in the planning for possible risks. This suggests that for resource-poor farm agents, income and assets recovery takes a longer period after perturbation which can lead to a long-term livelihood crisis and a poverty trap. But, according to the simulation results in this thesis, agents can recover from food shortage immediately after the shocks are over, as meeting minimum food requirements are an absolute priority for agents (which is also true with real-world subsistence smallholder farmers) over other competing goals.
Credit and technology policy simulation analysis further depict that welfare losses are partly compensated compared to without policies. Welfare losses of agents are better compensated when credit and technology are used jointly than when they are implemented separately. Similarly, technology policy intervention is better in compensating welfare losses compared to credit policy. Though policy interventions have compensational effects in minimizing the losses, they cannot completely offset the negative effects of extreme climate shocks even when implemented jointly. Disaggregation of simulation results by resource endowments suggests that agents with higher baseline income (without policy) and farm size appeared to be relatively less affected by shocks, and benefit from policy interventions the most.
Locust simulation results also suggest that locust shock leads to agent livelihood crisis and makes slower recovery of income and livestock assets rebuild without any relief intervention programs. Simulation of different locust relief policy interventions reveals that combined relief policy interventions appear to be superior in compensating welfare losses compared to individual relief interventions. When food or cash transfer is combined with inputs and assets the welfare losses are considerably reduced compared to the individual policy intervention. When asset recuperation is combined with other relief programs, livestock losses are substantially reduced which signifies the importance of asset support in building an asset base which has long-term benefits. Strengthening early warning systems by including seasonal weather forecasting has paramount importance to prevent the crisis of desert locust plague.
K1 Klimaänderung
K1 Heuschreckenplage
K1 unsichere Analysen 
PP Hohenheim
PB Kommunikations-, Informations- und Medienzentrum der Universität Hohenheim
UL http://opus.uni-hohenheim.de/volltexte/2023/2154