Remote Sensing in Humanitarian Crises: A Case Study From Sudan
In complex emergencies, humanitarian aid can be hampered by numerous factors. One important factor is the ability to quickly assess on-the-ground conditions without immediate or comprehensive access to the context. As remote sensing satellite technologies become more freely available, Mercy Corps is experimenting with the use of geospatial data science to inform aid and development programs. In early 2023, Mercy Corps developed a methodology to assess the impacts of the ongoing war in Sudan on agricultural production as it relates to food security.
Conflict in Sudan
Since April 2023, increased conflict in Sudan has had a profound impact on food security across the country. Currently, an estimated 42% of people are unable to consume adequate food, putting their lives or livelihoods in danger. Domestic agricultural production is critically important as it supports availability of food — particularly sorghum and millet, staple parts of the diet — and is the main source of livelihood for much of the population.
Mercy Corps in Sudan aims to reduce vulnerability and increase income and food security for farmers. For example, between July 2019 and April 2023, Mercy Corps partnered with ADRA to support 31,000 male and female smallholder farmers, through enhanced gender-equitable market access and agricultural production in Blue Nile and South Kordofan states (via the Strengthening Agricultural Markets and Food Security [SAFE] programme funded by SIDA). Assistance has continued throughout 2023: between the June-September planting season, Mercy Corps provided agricultural support to over 5,800 smallholder farmers, including initiatives like e-voucher seed fairs and providing affordable, climate-resilient seeds.
The outbreak of the conflict created significant challenges for farmers across Sudan. During the planting season, some farmers faced direct security threats, and in all states, it was extremely difficult to obtain agricultural inputs. Whilst a Mercy Corps survey in Blue Nile and South Kordofan found that 90% of farmers had started preparing their field or intended to plant in July 2023, it was difficult to assess the potential impact of the conflict on agricultural inputs such as seed, water, or fertilizer. Further, conflict has significantly increased access challenges across the country, meaning large-scale and regular primary data collection is extremely challenging.
To provide a wider aperture on the status of agriculture across Sudan, Mercy Corps developed an analysis using freely available satellite imagery and other data sources to monitor the Normalized Difference Vegetation Index (NDVI) in agricultural areas across the country.
You can read about the key findings and review the reports here.
Analytical Results
Overall, the August results showed that the average NDVI across the whole of Sudan was generally high compared to the reference period. This was attributed to good early season rainfall and favorable growing conditions. By September, the NDVI had fallen below the reference period average; most notably in Khartoum as well as “hotspots” in east Sudan. Across the east, in September the NDVI anomaly declined in comparison to July. This is despite the FAO Agricultural Stress Index showing smaller areas affected by severe drought in September as compared to July. Our analysts concluded that this could indicate that non-climate related factors may be affecting vegetation, such as the impact of conflict on irrigated agriculture. In comparison, agriculture in the Darfurs and Kordofans is typically rainfed, and vegetation indexes in September remained higher than the reference period (although the positive anomaly was smaller than in July).
Technical Methodology & Opportunities for Scale
For this analysis, we selected NDVI as the vegetation index1. NDVI assesses vegetation health by using the difference between near-infrared and red bands. It is commonly used in agricultural analyses to assess crop growth and health status. Using the free and public imagery product MOD09GA.061 Terra Surface Reflectance Daily Global at 500-meter resolution, we corrected for cloud cover. We created two time series image stacks — one for the years 2013–2022 (reference period) and one for 2023 (current period). We averaged the daily NDVI values to create composite images for the 10-year reference period, then calculated the percentage difference between the reference value and the daily image in the current period. Finally, we averaged the total NDVI anomaly to create a composite image.
While this method provides a view into the current state of agricultural growth as it relates to “normal” (non-conflict) years, there were trade-offs as well as limitations. The mean-of-means method and percent difference measure were selected for this analysis. In some contexts where other externalities (such as drought) have hampered agricultural growth during the reference period, median values may be preferred to decrease the impact of outlier years on the analysis. Cumulative anomalies may be a better approach when long run variables such as climate change are of interest in the analysis. This approach is also limited because it does not provide any indication of the relationship between NDVI agricultural areas and total agricultural production vis a vis food security. This relationship was assumed and interpreted qualitatively in our reporting. An exercise in calculating statistical correlation could be undertaken if we can secure a sufficient sample size of data for annual production totals.
These reports also highlighted the need to ground-truth results and identify driving factors behind changes in vegetation indices. For example, in the case of Sudan, the factors behind a reduction in NDVI in agricultural regions was dependent on the administrative area and ranged from fuel shortages to lack of funding for irrigation to drought to the inability to plant as the result of direct conflict. These nuances have a substantive impact on the planning of possible humanitarian interventions but would not be apparent from a remote sensing analysis alone. A compilation of remote sensing analyses and subject-matter expertise will continue to be the best path forward when using these technologies for humanitarian decision making.
Despite limitations, this methodology is easily scalable and may have applicability in contexts beyond conflict. Assessing the impact of acute natural disasters such as storms, volcano eruptions, earthquakes, and others may be another application of these methods. In principle, the technical requirements to scale this analysis are relatively light. Freely available satellite data enables this analysis on a sufficiently granular scale to identify trends at the subnational level. While the Sudan pilot was conducted in an ad hoc method via Google Earth Engine, plans to replicate and scale will include an explicit parametrization of the analysis so it can be made readily available via API or online platform to our country teams across the globe.
Now what?
Beyond just generating insights, the information created by remote sensing analyses can be used to make decisions on existing humanitarian and development programming. The Mercy Corps Sudan team is currently using this analysis to inform and adapt programming on the ground. Our analysis of staple crops in Gedaref state in particular will be used as evidence to scale-up ongoing agricultural programs in the region.
There is clear value in the use of remote sensing for contexts with humanitarian access constraints; the use of pure remote sensing analyses gives humanitarians the opportunity to identify possible patterns and trends before dedicating financial and human resources toward collecting primary data that may be necessary for more robust program design. The increasing affordability and accessibility of geospatial data, cloud services, and AI/ML tools also make early warning analyses of this type increasingly possible.
Interested in our work? Contact us at dataforimpact@mercycorps.org
