Drought is Devastating the Horn of Africa. Here’s How Complex Modeling is Aiding in Anticipatory Action.
System Dynamics Modeling is helping humanitarians understand when and where their intervention is needed most.
Strengthening Anticipatory Action
Somalia continues to be battered by severe, climate change-induced drought, raising concerns about how the devastating famine stands to erode the livelihoods of millions of households. To make matters worse, supply chain issues from the COVID-19 pandemic and the war in Ukraine have triggered inflation and a shortage of essential items. The Mercy Corps Somalia team is all too familiar with the on-the-ground reality. During droughts, families are consuming less. Some agropastoralists–who earn their livelihoods with a combination of agriculture and livestock herding–default to feeding their livestock over feeding themselves in hopes of sustaining their livelihood. A circumstance like this begs the question: how can humanitarians strengthen anticipatory action to systemic shocks to livestock systems to prevent the crippling effects of recurrent crises?
Mercy Corps’ Technology for Development team, particularly the Data for Impact (D4I) initiative, is exploring anticipatory action in its development of a livestock systems model to simulate behavior which reflects the observed, historical behaviors during drought to project future outcomes and better inform the timing and precision of aid. The challenge comes in the nature of the problem: the chain of links between rainfall, grain markets, livestock body conditions, and household income has many nonlinear dynamics, where small changes can lead to large impacts, often with substantial delays.
Why Use System Dynamics Modeling?
System Dynamics is a complex systems modeling technique well-matched to simulating such non-linear interactions. It allows a team to map the key elements of a system: the flow of animals from birth to market sale, the inflows and outflows of money in a household, the balance of trade of domestic grain production with the demand signal for grain imports. Then, it creates a mathematical framework to map the interdependencies between these elements: how herd management impacts household expenses (feed, fodder, vaccinations, etc), and how changes to grain markets impact these household economics and herd health.
The method is based on two concepts: 1) the flow of stock from one state to another (e.g., a unit of grain from planting to consumption by an animal); and 2) the feedback loops that mediate the flow of those stocks (rainfall needed for plant growth, transportation costs of grain shipments, etc).
This type of modeling is interdisciplinary, challenging experts across sectors to examine historical trends in each other’s disciplines and probe the interconnections. A meteorologist is aware of historical rainfall patterns and what that may signal for the present and future. A pastoralist has a wealth of knowledge on how much food livestock consume or what the rate of reproduction is depending on the season. And an agropastoralist brings an added focus to agriculture and grazing in caring for livestock. The modeling method connects these dynamics into simultaneous equations, linking climate, agriculture, and herd management into a simulation of potential futures. The objective is to capture the key dynamics of a system–not in its full complexity, but rather in sufficient simplicity that the model can simulate what might happen under specific conditions, such as when a drought shock hits Somalia or when Mercy Corps provides cash programming to households. Khaled Gaafar, a system dynamics expert who built the model being explored at Mercy Corps, emphasizes a key point in making these interconnections: “the point isn’t to keep modeling until we reach some objective truth. The point is to help decision makers choose the policy options that capture the most dynamic aspects of the system to weigh the trade-offs of choosing a certain policy or approach.”
Noting the interdisciplinary nature of the model, Lugard Ogaro, Director of Programs of Mercy Corps Somalia, says the model is meeting a need that was previously unmet for his team by serving as a stakeholder map.
“To be able to graphically show the complexity involved in livestock as a sector is useful in exposing the areas where governments and NGOs and other actors need to look into as they do programming.” — Lugard Ogaro, Director of Programs of Mercy Corps Somalia
From the model, the Somalia team has identified other areas that can be focused on from a programmatic standpoint such as fodder assistance, the timing of cash assistance in different parts of the growing season, and how to restock livestock in a way that takes into account if the restocked animals can withstand the conditions. Prior to the model, the team mostly focused on restocking livestock as a response to the shocks of drought on this sector.
The model is based on another System Dynamics model from the University of Bergen (Norway), which linked changing climate conditions in Guatemala with agriculture and household cash flows. The Data for Impact team extended the model and adapted it to the arid climate of Somalia using data from multiple agencies. It is a first attempt to build a method that connects the elements of the Pathway to Possibility strategy into a quantitative framework.
The quantitative elements of system dynamics are not intended to displace the insights of experts that have historically guided practices and approaches. Instead, system dynamics is a tool to simulate the impact of choices across a complex system.
Engaging the Experts
In late September, D4I held a 2-day workshop with the Mercy Corps Somalia team, a forum for bridging the rich qualitative expertise of agropastoralists with the technical expertise of system dynamics modelers to be able to holistically understand the magnitude and complexity of the crisis, simulate outcomes, and weigh different courses of action in programming. For Gaafar, the workshop functioned as a means to get a “reality check” of the assumptions that guide the model so that the disciplinary experts can confront these assumptions, which are guided by literature and the experiences of smallholder farmers, and refute them, allowing the model to be adjusted accordingly. In real time, this confrontation unraveled to unveil the full story and context of what Somalia and Somalis face as a core livelihood sector continues to be under immense threat.
As the D4I team demonstrated the model, the Somalia team members interjected with questions any actor should be considering. How can conflict be modeled to better understand how it serves as a shock to the livelihood sector? How can a model account for fluctuations and evolutions when it must contain constants in its equations? Ogaro brought his agricultural expertise to the forefront, highlighting the complexity of the matter and how the livestock systems model can capture that complexity. Navigating the model size is a challenge for both modelers and programming actors. On the one hand, a simple and compact model is more easily actionable, allowing constants to become dynamic as they interact and are interconnected through feedback loops. On the other hand, agropastoralists know that the situation on the ground is immensely complex and continuing to change, introducing new dilemmas and new considerations for humanitarian response.
Managing model size is a challenge as it is crucial to stay as true as possible to the available data. The variables and parameters of the model have to be reflective of actual data variables that can be measured and captured by the real systems. Gaafar suggests a rule of thumb: to allow the model to adequately reflect the data and then attempt model simplification once the outcome is strong.
“A modeler is always limited to the certain constraints of the model.” — Khaled Gaafar
The Hidden Potential of the Livestock Systems Model
For now, the model has supported the Somalia team in determining when to stock households. Prior to the introduction of the model, the approach mainly focused on stocking those who had lost their livelihoods due to drought. The model introduces more precise considerations of when is the best time to restock so that the number of restocked cattle can be sustained. The model also opened up other areas of focus such as fodder assistance and cash assistance. When asked how his team plans to further incorporate the model into their programming work, Ogaro spotlights the hidden potential of the tool. While existing gaps in data at these early stages leave room for growth in using the tool for decision making purposes, Ogaro sees the opportunity to use the model to engage those on the ground most deeply impacted. He shares that pastoralists and agropastoralists sometimes have tensions as pastoralists come during drought season to graze their animals in the land of the agropastoralists, degrading the rangeland and negatively impacting animal health. A potential use of the model could inform programming to build the relationship between these two groups to form an enterprise as resilience and adaptation approaches will have to encompass the needs of both groups.
Moving Forward
From the workshops with the Somalia team, D4I has identified a smaller, core group of individuals who are interested in engaging with the model in more depth. This level of engagement ensures that the model remains sensitive to the changes on the ground and reflective of what the country team is witnessing in reality. New policy options and considerations may even be generated from this kind of space. A key challenge of the model remains the gaps in available data. As it remains in its early stages, the next steps will require collecting time series data or proxies for variables of interest, whether qualitative or quantitative. The model continues to unlock new insights to combat the shocks to livelihoods and household food security, demonstrating the utility of combining qualitative expertise and knowledge with quantitative metrics. With deteriorating conditions throughout the Horn, a tool like this gives the sector a better understanding of what is happening so that the policy options can be as impactful and targeted as possible.
