A shift from cattle to camel and goat farming can sustain milk production with lower inputs and emissions in north sub-Saharan Africa's drylands

The study addresses whether current cattle-dominated dairy systems in north sub-Saharan Africa (NSSA) are suitable to cope with ongoing and future climatic and biophysical challenges. Dairy is crucial to livelihoods and food security in NSSA and demand for milk is projected to continue growing rapidly. However, climate change is impacting feed availability, water accessibility, and thermal comfort for livestock, with SSA dairy systems already exhibiting high greenhouse gas (GHG) emissions per kg of milk compared to global averages. Evidence from pastoralist communities indicates shifts from cattle to more climate-resilient browsers (goats and camels) in response to drought, feed scarcity, and heat stress. The study aims to (1) identify where environmental changes in NSSA drylands most jeopardize cattle milk production, and (2) quantify the potential impacts on milk output, resource use, and emissions of shifting herd composition from cattle toward goats and camels.

The paper synthesizes evidence that NSSA dairy productivity faces increasing constraints from climate change, including higher heat stress, declining or variable biomass production, and water scarcity. Prior studies document high GHG intensity of milk in SSA and observed shifts by pastoralists (e.g., Borana, Maasai, Fulani) from cattle to goats and camels as adaptive strategies during environmental extremes. Literature indicates browsers’ superior drought tolerance and ability to utilize woody vegetation, making them more resilient under dryland conditions. Studies also show expected increases in heat stress frequency across East and West Africa by late-century. Market and nutrition literature underscores the importance of animal-sourced foods in NSSA and growing demand for camel and goat products, while highlighting constraints such as costs, husbandry knowledge gaps, and supply chain limitations.

Study area: Arid and semi-arid Sudano-Sahel and Greater Horn of Africa (5°–20°N, 20°W–55°E; ~10 million km²). Data: ERA5-Land climate (temperature, relative humidity); Copernicus Global Land Service remote sensing (DMP, NDVI, Small Water Bodies); FAO GLW3 livestock populations (2010) for cattle, goats; modeled camel distribution; FAO Global Perspective Studies milk supply/demand; NDVI-derived length of growing period. Period analyzed: 2001–2020 (remote sensing from 1998 for DMP processing). - Indicators modeled per 0.1° grid: (1) annual dry matter productivity (DMP) via BioGenerator v5.1000 by cumulating decadal DMP across the growing season after NDVI-based filtering; (2) water accessibility index (WAI) via HydroGenerator v4.3 using a Gaussian decay from detected surface water bodies within 30 km and background by aridity zone; (3) heat stress (HS) frequency for dairy cattle using daily temperature-humidity index (THI) derived from daily maximum temperature and relative humidity, counting annual days above moderate/severe thresholds. - Trend detection: Mann–Kendall test and Sen’s slope (significance at 5% and 1%) for 2001–2020 trends in DMP, WAI, HS. - Identifying areas where cattle conditions worsened: Grid cells flagged if any of the following held: (1) simultaneous significant changes (P<0.05): HS increasing, DMP decreasing, WAI decreasing; (2) DMP significant at 1% and decadal ratio (2011–2020)/(2001–2010) < 0.5; (3) HS significant at 1% and ratio > 10%; (4) WAI significant at 1% and ratio < 0.5. - Herd composition and milk contribution: For flagged cells, extracted GLW3 livestock numbers and FAO milk production/demand. Estimated per-species milk yield parameters by solving systems Ax + By + Cz = milk production per grid (x,y,z are species counts) across N grids; fitted distributions yielded average per-head milk yields used subsequently: cattle=171, goat=68, camel=295 (units consistent with dataset). - Resource use and emissions per kg milk: From literature review, averages used: feed dry matter (kg/kg milk): cattle 1.78, goat 1.43, camel 1.17; water (L/kg milk): cattle 5.73, goat 2.73, camel 4.90; GHG (kg CO2e/kg milk): cattle 7.25, goat 6.70, camel 6.00 (enteric CH4 and manure CH4/N2O under extensive/mixed systems in dryland NSSA). - Optimization of herd composition: Maintained constant total tropical livestock units (TLU; 1 TLU=250 kg BW). Iteratively adjusted composition to reduce cattle and increase goats/camels. Objective prioritized: (1) maximize aggregate milk (combinations >75th percentile), then (2) minimize water and feed use (combinations <25th percentile), then (3) minimize GHG emissions (combinations <25th percentile). Constraints: in semi-arid mixed systems, maximum cattle reduction capped at 50% per grid; in arid pastoral systems, up to 100% replacement permitted. Also evaluated three ‘extreme’ scenarios: 100% shift of reducible cattle to camels; 100% to goats; 50/50 to camels and goats. - Aggregation: Computed region-wide changes in milk production, water and feed demand, and GHG emissions for ideal and extreme scenarios; compared to current composition; summarized by subregion.

- Environmental trends: Average annual DMP ~30 t ha−1 yr−1 (range ~0 to ~53); significant DMP decreases in ~9% of study area (greening in Sahel ~19%). WAI averaged ~0.04; significant WAI decrease in ~15% of area. Average detrimental HS frequency 77 days yr−1 (~21%), exceeding 50% of days in coastal Greater Horn; HS frequency significantly increased in ~38% of area (2001–2020). - Areas where cattle milk production conditions worsened: ~17% of the study area (~1.7 million km²) flagged; hosts ~14.2 million cattle, ~18.11 million goats, ~2.99 million camels. Ten independent case study areas reporting shifts from cattle to goats/camels overlap with these zones. - Optimized ‘ideal’ herd composition outcome (region-wide): Decrease cattle by ~24% (~−5.9 million), increase goats by ~14% (~+7.7 million) and camels by ~10% (~+1.2 million) while holding TLU constant yields: +0.14 Mt milk (+5.7% vs current), water use −1,683.6 million m³ (−15.3%), feed use −404.3 Mt (−11.2%), GHG emissions −1,224.6 MtCO2e (−7.9%). - Extreme scenarios: 100% shift to goats lowers feed (−15%), water (−33%), and GHG (−9%) but reduces milk by 26%. Other extremes similarly underperform relative to the optimized multi-species shift. - Demand context: Figure indicates 2030 regional milk demand ~4.40 Mt; production under ideal vs current shares (2010) ~2.42 vs ~2.28 Mt, suggesting composition shifts can help narrow the supply–demand gap. - Species resilience: Results align with observed pastoral shifts favoring browsers (goats, camels) due to drought/heat tolerance and ability to utilize woody vegetation, supporting climate-resilient milk supply.

The analysis shows that climatic and biophysical changes—reduced biomass production, declining surface water accessibility, and increasing heat stress—are already undermining cattle-based milk systems in NSSA drylands. By optimizing herd composition toward goats and camels while maintaining total livestock biomass (TLU), the region can modestly increase aggregate milk production and substantially reduce water and feed demands and GHG emissions. This directly addresses the research question by demonstrating that alternative species compositions improve both productivity and environmental outcomes under current and emerging constraints. The approach recognizes the essential nutritional role of animal-sourced foods in NSSA, where reducing consumption is not generally viable, and instead leverages species better adapted to heat and feed scarcity. Beyond milk, multispecies herding may enhance ecosystem services (vegetation diversity, grass availability), diversify incomes, and improve resilience to climate and market shocks. However, adoption requires attention to social, economic, and infrastructural factors, including producer preferences, market development for goat and camel products, training in husbandry, and overcoming initial investment costs.

Shifting dairy herd composition away from cattle and toward goats and camels in NSSA drylands can sustain and slightly increase milk production while reducing water and feed requirements and lowering GHG emissions. Approximately one-fifth of the region is already experiencing environmental changes that challenge cattle milk production, and documented local shifts corroborate the model results. Policy and investment should support development of goat and camel dairy value chains, including breeding (heat-tolerant cattle and improved goat/camel milk yields), animal health and nutrition services, and processing, transport, and market infrastructure. Future research should refine spatially explicit data on water (including groundwater) and woody vegetation dynamics, improve species-specific milk yield and emission factors under local management, evaluate socio-economic feasibility and adoption barriers, and assess broader ecosystem service impacts of multispecies pastoralism.

- Water accessibility index captures surface water only; wells and groundwater resources are not included due to data limitations, potentially underrepresenting water availability. - Limited, non-uniform data on groundwater use by livestock and infrastructure changes over time. - Woody vegetation dynamics not explicitly integrated despite indications of increased woody cover in many affected areas; lack of spatially explicit, process-based data constrained inclusion. - Assumed similar milk energy values across species; actual milk composition varies by species, breed, diet, and management, which could affect nutritional equivalence assessments. - Species-level parameters (milk yield per head, water/feed use, GHG per kg milk) derived from literature and modeled relationships may not capture local breed/management heterogeneity. - Optimization imposes simplifying constraints (constant TLU; capped cattle reduction in semi-arid mixed systems), which may differ from real-world transitions and path dependencies. - Economic and adoption barriers (animal costs, husbandry knowledge, equipment needs) and policy/institutional factors are discussed but not quantitatively modeled.