Editors Note: This blog is based on findings from the paper "Geospatial impact evaluation of a low-cost agricultural intervention for enhancing environmental resilience", authored by Ariel BenYishay, Pratap Khattri, Brad Sagara, Rachel Sayers, Kunwar K. Singh, Ryan Slapikas, Chet Bahadur Tamang, and Dinee Tamang. The original study evaluates the effects of sugarcane cultivation in western Nepal using a geospatial impact evaluation that combines satellite imagery, machine learning, and econometric analysis.
Farmers in Nimwabojhi, western Nepal, use bamboo structures to protect young sugarcane crops from river cutting along erosion-prone banks.
In western Nepal, communities residing along the floodplains of the Mohana and Mahakali river systems face a dual challenge: frequent flooding and persistent economic vulnerability. One intervention, the promotion of sugarcane cultivation along riverbanks, is designed to address both challenges simultaneously. Sugarcane, with its deep root structure and dense foliage, offers the potential to both stabilize soil and generate income.
Since 2014, Mercy Corps has implemented the Managing Risk through Economic Development (M-RED) program to reduce disaster risk while promoting climate-resilient livelihoods. Through this program, farmers in high-risk areas were incentivized to cultivate sugarcane along erosion-prone riverbanks.
Yet, a critical question remains: has the strategy been effective in achieving its dual goals? To answer this, the GeoField partnership applied Earth observation (EO) tools by pairing satellite imagery with robust analytical techniques to examine the causal impacts of sugarcane adoption on both agricultural uptake and environmental resilience.
This Nepal case study exemplifies the collaborative mission of GeoField, a joint initiative of AidData, Mercy Corps, and DevGlobal, supported by the Gates Foundation. GeoField is a platform that brings EO tools into the hands of evaluators working on agriculture and climate resilience programs. Think of it as a bridge between satellites and field-level decision-making.
In the M-RED context, the geospatial impact evaluation was led by a joint research team from AidData and Mercy Corps, including Pratap Khattri, Rachel Sayers, Kunwar K. Singh, Ryan Slapikas, and Ariel BenYishay from AidData, together with Chet Bahadur Tamang, Dinee Tamang, and Brad Sagara from Mercy Corps. The team applied geospatial impact evaluation (GIE) methods using a combination of Copernicus Sentinel-2 and PlanetScope satellite imagery, alongside machine learning algorithms.
By combining these data with causal inference techniques such as optimal full matching (a method for creating statistically similar comparison groups) and difference-in-differences (which estimates changes over time between treated and untreated areas), the team could rigorously estimate changes in sugarcane cultivation and landscape conditions across more than 50 communities in Nepal’s Terai region.
Satellite imagery offers a valuable complement to traditional evaluation tools, especially in complex and dynamic landscapes. In this evaluation, Earth observation brought three critical advantages. First, scale: high-resolution imagery allowed researchers to monitor sugarcane adoption and environmental changes across dozens of villages over multiple years. Second, timeliness: multi-temporal observations from Sentinel-2 and PlanetScope enabled year-by-year tracking of land cover, riverbank shifts, and vegetation changes. Third, objectivity: remote sensing provided an independent and consistent lens, which was particularly useful in hard-to-reach areas or where ground data were sparse or costly to collect.
By integrating EO with advanced econometric methods, the AidData team could discern not just where sugarcane was adopted, but also whether it led to meaningful changes in environmental indicators such as soil erosion and river movement.
The evaluation revealed that sugarcane cultivation expanded significantly in M-RED communities, with adoption also spreading to nearby (spillover) villages. This suggests strong appeal and potential for self-sustaining uptake even beyond targeted areas.
However, the environmental impacts were less conclusive. The study found no consistent evidence that sugarcane cultivation alone reduced riverbank erosion or stabilized flood-prone landscapes over the evaluation period.
Rather than signaling a failed intervention, this insight reflects the complexity of environmental systems. Measurable impacts such as riverbank stabilization may take longer to emerge, or may require complementary interventions like bioengineering, community-based erosion control, or upstream water management infrastructure.
The Nepal sugarcane case demonstrates both the promise and the limitations of integrating EO into climate-sensitive agricultural evaluations. While remote sensing helped uncover clear patterns of crop adoption, its ability to detect environmental outcomes highlights the need for long-term monitoring and multi-pronged strategies.
As climate threats become more urgent, tools like Earth Observation and partnerships like GeoField can help us root resilience in both evidence and action.
In upcoming blogs, we will unpack the analytical framework in greater detail and explore what this approach can tell us about designing effective, scalable interventions for climate adaptation.
