Enhancing the usefulness of weather forecasts and impact data for effective early action by development, humanitarian and community organisations is central to one of the Observatory’s challenge questions. In early May 2023, we brought together a group of people involved in creating, analysing, disseminating and using drought and related data to explore how data-driven products and innovation can better contribute to early warning and early action on food security and drought in Kenya. In this post, Gary Watmough, Philipp Barthelme, John Mutua and Peter Ballantyne reflect on the discussions and their relevance to the wider goals of the Jameel Observatory for Food Security Early Action. It is the third in a series of four posts on recent discussions convened in Kenya by the Observatory.
Data, forecasts and anticipatory actions
Early warning and anticipatory action are fundamental strategies to deliver the early actions that governments and communities can implement to prepare for, respond to, and recover from climatic shocks and droughts. While details vary, the common ‘engine’ in all such systems is forecast data on expected conditions (rainfall, soil moisture, surface temperature) and their likely impacts on availability of water for people, livestock, pastures and crops. Short to medium term rainfall and temperature forecasts are used to calculate likely conditions while additional earth observation data, on vegetation or water bodies for example, also give insights that help assess near-term scenarios.
Different agencies use this data to trigger pre-agreed actions and funding and to alert governments and citizens about upcoming conditions. Often the data is benign: with weather forecasts, farmers, for example, can decide when to plant crops or herders can plan to move their herds. Where the predicted changes reveal anomalies or suggest risks or threats – drought, flood, heatwave, disease or pest emergence, the data are used to guide interventions at different levels.
For decision makers in government and humanitarian agencies, the power of this observation data is revealed when it is combined with socio-economic data to indicate likely impacts of the forecasted conditions on people, infrastructure or their livelihoods. Using known trends and anomalies, we can estimate how different conditions will influence, for example, forage availability for livestock, market prices for livestock and crops, food availability for vulnerable families, likely demands for cash payments, movements of herders, and so on.
This requires that the diverse data exists and is provided in formats that allow it to be effectively combined and communicated to show socio-economic effects on people and communities and the implications of these.
The datathon
Growing from the past year of research and collaboration with our partners to better understand and address data challenges and recognising that there is certainly no ‘data drought’ in this area, the datathon was convened to explore, based on real use cases, promising ways that data on drought and food security can be combined and put to use.
Note: ‘A datathon is an event where participants gather to solve practical problems through the application of data science tools and techniques, by working together in teams to generate insights and potential solutions.’
The datathon brought together eighteen people working with drought/food security data – as customers, owners and wranglers – to explore ways to bring together and combine and visualise predictive/forecast data and indicators and to explore and co-create ideas and products that could become proofs of concept for improved or new products.
Gary Watmough from the University of Edinburgh framed the discussions and objectives, which were to:
- Co-design a set of mini-projects based on use-cases.
- Collectively explore the needs of data users in relation to the available data.
- Collectively understand what is needed to achieve the desired outcomes from various actors engaged in or wanting to engage in drought early action.
- Build trust and understanding across a diverse range of actors.
Drought and food security use cases
Initial discussions were centred around four ‘use cases’ from Save the Children, the Kenya National Drought Management Authority (NDMA), the Kenya State Department of Livestock and Baringo Pastoralist Network. Structured interviews probed four topics:
- What are the organisations’ objectives in using data; what problems are they trying to solve?
- Which data do they use and how?; Which data do they collect?
- Are there decisions that they want to take that suffer due to data issues? What is preventing effective data use in predictive decisions?
- In terms of data to inform future decisions, what is their dream future scenario?
Steve Mutiso explained how Save the Children in Kenya seek data and insights that help them take ‘forecast based’ decisions and actions relevant to upcoming situations and as far ahead of a crisis as possible. The goal is to act early, minimise losses and be more effective. He wants to know where and when different communities will be affected, who in the community is going to be most affected, and how. With this information they can set priorities and allocate resources. They currently rely on seasonal data from weather forecasts, looking two or three months ahead of the rainy season. Anything shorter gives insufficient time to act. They combine the forecasts with socio-economic data – their own household economic analysis, from other sources, from communities or context monitoring – to predict likely future impacts of a forecasted crisis with different confidence levels. Although the socioeconomic data overlay on weather forecast does provide robust predictive analysis over large livelihoods zones and can be downscaled to districts levels to predict food deficits, it does not involve the use of other commonly deployed data sources such as satellite data on rainfall and vegetation. Further, automation of data generation and analysis is insufficient. Harmonization of data generation by different actors is weak as data is collected at different spatial and temporal resolutions with less attention to household socioeconomic data. His vision: A user-friendly predictive model that can take data from different sources, run scenarios based on community input, and provide several months (at least) warning of likely food deficits.
Jeremiah Kithamah of the Kenya Government’s State Department for Livestock explained how his agency coordinates emergency responses and choice of interventions based on predictions and severity of droughts as informed by the drought phases. He uses data from the Met Office and other early warning systems especially from National Drought Management Authority (NDMA) to identify the numbers of livestock at risk of starvation and death in diverse livelihood zones of the country. Using government data on livestock numbers, feed availability and other environmental indicators, estimation of livestock numbers at risk of death can be estimated. This can inform the required intervention for early action and early response. The interventions may involve destocking, feed provision, disease management and or water tracking depending on the severity of the drought. The biggest challenge he faces is collecting sufficient and reliable data. His vision: Digitizing data to analyse trends and make this available through a platform or dashboard for policy and decision making.
Francis Murithi of Kenya’s National Drought Management Authority explained that his agency collects data, compiles, combines and communicates drought early warning information that support drought risk management efforts in the county. He explained how data is collected through a network of sentinel sites in twenty-three counties across Kenya’s arid and semi-arid lands. Spread across livelihood zones, socio-economic and environmental data is routinely collected from communities focusing environmental indicators (water, vegetation condition, pasture and browse availability), production indicators (animal body condition, livestock diseases, milk production, status of crop production), access indicators (livestock prices at the markets, food commodities prices), terms of trade (household purchasing power), milk consumption, utilization indicators (household food consumption,- i.e Food Consumption Score, Coping Strategies Index, malnutrition levels). An annual survey (refered to as SMART Survey) assesses the malnutrition levels at county level to update nutrition management programming. An important aim of this, he said, is to provide prognosis and recommendations to all stakeholders engaged in drought risk management, that is local communities, national government, county government, non-state actors and donor community. He noted that there have been mismatches between the information and recommendations provided, and local actions taken, as the fundamental challenge. The uptake of the recommendations is mainly believed to be hindered by negative cultural practices and beliefs that are superimposed with religion.
Limo Kibet of the Baringo Agro-Pastoral Community Network explained how his group works mainly with pastoral and agro-pastoral communities to improve local understanding of the impacts of drought and to promote interventions or local solutions that can minimise climate risks. The Network works almost like an extension agency, sourcing data and information from different sources and sharing them with other users. He also highlighted new initiatives like the use of video, photographs and social media to document both drought impacts from local perspectives as well as indigenous/traditional forms of predictions and adaptations to changing climate.
Emerging project ideas
After mapping potential data sources (see the wordcloud above), participants worked in three teams to work out how each of three project ideas could be developed:
- Scenario-based livelihood forecasts to inform early and anticipatory actions on drought by humanitarian agencies.
- Integrated livestock vulnerability assessment for coordinated actions to manage the effects of drought on livestock.
- Communicating price information to facilitate early action by pastoralists against drought and other emergencies
Ideas from the first two discussions are introduced here, they will be refined and revisited with champions to explore ways to create enhanced predictive tools in these areas. The accompanying visualisations were created during the meeting to illustrate how data could support such decision processes.
Initial work on the third topic ‘Communicating price information to facilitate early action by pastoralists against drought and other emergencies’, though promising, was inconclusive and many of these ‘last mile’ ideas are being taken up in a new Observatory impact collaboration.
For the fourth topic ‘identifying indigenous knowledge and information sources and indicators for more effective drought management’ it was decided to connect these ideas with the recently-started PhD project of George Tsitati.
The ideas generated were shared to the wider community of practice meeting in May and are being further refined and discussed with potential collaborators and partners.
Scenario-based livelihood forecasts to inform early and anticipatory actions on drought by humanitarian agencies
A critical challenge for humanitarian agencies is that accurate forecasts of droughts are often only available 2-3 months in advance. As decision-making and planning efforts typically require multiple weeks, this leaves almost no time for early action.
To tackle this issue, the team came up with a potential approach that utilises more uncertain 6-9 month weather forecasts which would provide more time for preparation and early action to happen.
To address the higher uncertainty associated with 6-9 month forecasts, the team proposed to derive three scenarios instead of one from the weather forecasts and to prepare livelihood impact forecasts and interventions for each of them. These three scenarios would each fall within the uncertainty bounds of the longer-range forecast. As more accurate 3-month weather forecasts and other data arrive, the scenarios can be updated and implementation of already-prepared actions can begin.
While drought-related information (precipitation, soil moisture, vegetation indices, etc.) lay the foundation of this approach, the team recognized the need to complement these datasets with sub-county level socioeconomic data (census data, market prices, building footprints, etc.). Combining these data sources would allow humanitarian agencies to create more accurate livelihood forecasts and derive effective actions that address the unique challenges faced by each community.
Integrated livestock vulnerability assessment for coordinated actions to manage the effects of drought on livestock
To manage the effects of drought on the livestock sector, policymakers must have a clear understanding of livestock vulnerability. Of particular importance is assessing the balance between available feed resources and livestock requirements. Other important factors include physiological needs, age, and productivity of the animals.
The team explored ways to find or produce accurate evaluations of feed availability, using data on, for example, pasture conditions, forage production, and vegetation dynamics, and leveraging technologies like satellite imagery and remote sensing to produce updated ‘feed balance’ information for locations, livestock types and production systems. This feed balance analysis guides policymaking by comparing feed availability with requirements, identifying imbalances, and recommending how resources should be allocated.
With such information, early warning systems trigger timely actions such as de-stocking, feed supplementation, or re-location, with priority given to vulnerable livestock populations and optimized resource allocation.
Collaboration among stakeholders, including livestock owners, extension services, veterinary services, and government agencies, fosters coordinated decision-making. By prioritizing feed availability and employing coordinated actions based on vulnerability assessments, policymakers can effectively manage the impacts of drought on livestock, ensuring sector resilience and sustainability while safeguarding livelihoods.
Strategic opportunities
Reflecting on the interactions, participants saw a positive outlook for further collaboration in this area.
Strengths to build on include: There is much data available, though not yet always fully accessible. We can call on growing, and diverse, expertise from different stakeholders and fields. | Opportunities that stakeholders ask for include: There is much interest in early warning systems among stakeholders – donors, research, civil society, pastoralists, government. We observe calls for integrated multiple data for better understanding. Strong partnerships in food and nutrition security and early warning systems seek data and tools they can apply. |
We have complementary aspirations, including: To collaboratively leverage on the strengths (data and expertise) to deliver a set of useful and effective tools for drought management. To provide easy access to data.To reduce the vulnerabilities of populations and build resilience. | We see some clear indicators we can use to assess progress, including: The widespread and open availability of data for vulnerability assessment, potentially through a central data sharing platform. Upscaling and replication of [data-assisted] drought management tools |
Read the other posts:
Community of practice meeting / Dialogue on local drought response and resilience / Conversations with pastoralist elders
Contact us for more information or to partner with us:
[/et_pb_text][/et_pb_column] [/et_pb_row] [/et_pb_section]