Proportion of each feed item in the diet for the selected species and production system.
NA = feed item not used in this system.
Based on: Herrero, M., Havlík, P., Valin, H., Notenbaert, A., Rufino, M. C., Thornton, P. K., Blümmel, M., Weiss, F., Grace, D., & Obersteiner, M. (2013). Biomass use, production, feed efficiencies, and greenhouse gas emissions from global livestock systems. Proceedings of the National Academy of Sciences, 110(52), 20888–2093. https://doi.org/10.1073/pnas.1308149110
Updating map...
Scenarios
| Baseline | Current situation: existing breed and herd size |
| Scenario 1 | Same herd size, improved breed with higher yield per animal |
| Scenario 2 | Same total output, fewer animals with improved breed |
| Scenario 3 | Same total enteric fermentation emissions, improved breed |
| Scenario 4 | Same total output as Scenario 1, original breed (more animals) |
Median household-level outcomes from RHoMIS survey data, including herd size, production, sales, income, feed requirements, and product availability for consumption.
Estimated daily contribution of milk to household nutritional requirements, expressed as a percentage of the recommended dietary allowance (RDA) per person per day, for each nutrient across scenarios.
MERLIN combines regional feed basket data with gridded crop production statistics (SPAM) and livestock survey data (RHoMIS) to estimate the environmental and socioeconomic effects of livestock productivity improvements. The core pipeline runs through five components:
Feed requirements are derived from species- and production-system-specific feed baskets from Herrero et al. (2013), which provide the proportional composition of diets by livestock species, production system, and region. For each scenario, the number of animals, their productivity (e.g. milk yield in kg/head/year), and the efficiency of feed conversion are used to calculate total dry matter intake and the breakdown by feed type (e.g. improved pasture, crop residues, concentrates).
Feed items that require dedicated cropland (e.g. concentrates, oilseeds) are mapped to SPAM crop categories. The area of cropland required to produce each feed item is calculated by dividing the total feed demand by the weighted-average crop yield within the selected region, derived from SPAM 2020 (V2r2) harvest area and production rasters at 5-arcminute resolution. Irrigation water consumption (blue and green) is estimated using crop-specific water use coefficients from Chukalla et al. (2025), spatially weighted by SPAM harvest area within the region.
Crop-related GHG emissions are estimated by applying regional emission factors to the estimated cropland areas. Emission sources include synthetic fertilizer application, manure management, crop residue incorporation, rice paddy CH₄, peatland cultivation, and crop burning, following Cao et al. (2026) global gridded emission layers at 0.5° resolution.
Herd-level environmental indicators are calculated analytically from the feed requirement model outputs for each scenario. These cover total animals, production, enteric fermentation, manure emissions, and direct GHG emissions per unit of product.
Socioeconomic and nutritional outcomes are estimated from RHoMIS survey data aggregated over the selected region. RHoMIS collects standardized farm-level data including herd size, household size, production, sales, and income. Survey medians weighted by number of data points per administrative unit serve as the household baseline.
Household production is computed directly from herd size and per-animal yield for both the baseline and all scenarios. Using the same yield basis (PNAS default or user input) for baseline and scenarios ensures that differences between scenarios reflect only the intervention effect.
On the Analysis tab, use the Country and Administrative Level dropdowns to choose the geographic scope of your analysis. The map will update to show administrative boundaries for the selected country. Click one or more polygons on the map to select the region of interest; selected areas are highlighted in amber. Click a selected polygon again to deselect it.
Select the target species (e.g. Dairy cattle, Beef cattle, Dairy sheep and goats) and the production system (e.g. Mixed Rainfed, Grassland-based) from the dropdowns. These determine which feed basket is used in the analysis.
Enter the current (baseline) yield and a projected improved yield. Yields are expressed as kg of product per head per year (e.g. kg milk/cow/year for dairy cattle). Also enter the number of animals in the herd. The feed basket proportions panel shows the composition of the feed used for the selected species and system.
Click Run Analysis. MERLIN will calculate feed requirements, cropland area, GHG emissions, water consumption, and household-level outcomes for each scenario. Results are displayed automatically in the Results section.
The Results section contains three sub-tabs:
MERLIN (Modeling Environmental Repercussions of Livestock Interventions) is a quantitative tool for assessing the environmental impacts of livestock productivity interventions targeting smallholder farmers. It combines geospatial crop and livestock data with feed requirement modeling to estimate how changes in animal yields affect land use, greenhouse gas emissions, water consumption, and household-level socioeconomic and nutritional outcomes.
Currently MERLIN covers Ethiopia, India, Kenya and Nigeria, at administrative levels 0 to 3.
Four production types are modeled: dairy cattle, beef cattle, dairy sheep and goats, and meat sheep and goats. Pigs and poultry are present in the underlying data but are not yet calibrated.
Feed baskets describe systems circa 2000 and are held constant across scenarios. Therefore, results are best read as comparisons between scenarios rather than absolute predictions. Household indicators depend on RHoMIS survey coverage, which varies by region.
See the Methodology tab for how the model works, and the Help tab for how to run an analysis.
Supported by the Gates Foundation.