Impact of an accelerated melting of Greenland on malaria distribution over Africa

Malaria, particularly caused by Plasmodium falciparum, remains a major public health burden in Africa, with climate playing a critical role in vector and parasite dynamics. Transmission depends on temperature and rainfall patterns that regulate Anopheles mosquito survival, breeding, and parasite development. Prior work shows climate change can alter the length and intensity of the malaria transmission season, with potential decreases in very hot, already endemic regions (e.g., Sahel) and increases in cooler highland areas (e.g., East African plateaus). However, most impact studies use standard RCP scenarios from CMIP frameworks that omit potential non-linear climate system dynamics such as rapid Greenland ice-sheet melting and associated freshwater input that can slow the AMOC, alter temperature and precipitation patterns, and shift the African rain-belt southward. The research question is how an accelerated Greenland ice-sheet melting, superimposed on high-emissions forcing, could additionally impact African climate and consequently malaria transmission. The purpose is to explore mechanisms and spatial patterns of malaria risk under extreme freshwater release scenarios using a multi-model malaria framework driven by IPSL climate simulations, not to make precise forecasts but to assess plausible impacts beyond standard scenarios.

Previous studies (e.g., Martens and colleagues) suggest climate change impacts are pronounced at the fringes of endemic regions and in temperate zones with competent vectors but temperatures too low for current transmission. Rising temperatures can reduce transmission in the hottest regions (semi-arid Sahel) yet increase risk in cooler highlands (East Africa). Ensemble assessments (e.g., Caminade et al.) under RCP8.5 project decreased Length of Transmission Season (LTS) over warm West African plains and increased LTS over East African highlands by the 2080s. Standard CMIP RCP scenarios, however, neglect rapid destabilizing processes like ice-sheet melt and permafrost thaw. Freshwater inputs can weaken the AMOC, cool the North Atlantic, modify pressure gradients, and shift the ITCZ southward, historically linked to Sahel drying. There is substantial inter-model spread in projected African monsoon responses under standard scenarios, but paleo evidence and modeling indicate a robust southward ITCZ shift with freshwater hosing. This motivates examining tipping-point scenarios for health impacts beyond traditional RCP-only analyses.

• Malaria models: Five mathematical malaria models (MMMs) were used. Two are dynamical, daily driven models producing prevalence (%): Liverpool Malaria Model (LMM) and VECTRI. Three are monthly models producing Length of Transmission Season (LTS, months/year): MARA, MIASMA, and a steady-state LMM version (LMM_R0).
  • All are parameterized for P. falciparum and primarily for An. gambiae. LMM simulates mosquito and parasite life cycles, with rainfall influencing eggs/larvae and temperature affecting mosquito mortality, gonotrophic cycle, and parasite sporogony. VECTRI includes precipitation, temperature, simple hydrology (pool model, flushing with extreme rain), population density (urban dilution), and immunity dynamics; inputs are daily rainfall and temperature plus static 2005 population density. MARA and MIASMA use monthly rainfall/temperature thresholds and functions to define months suitable for transmission; LTS sums suitable months. LMM_R0 uses monthly climate to compute R0; months with R0 ≥ 1 are suitable and summed for LTS.
• Validation datasets: (1) Malaria Atlas Project (MAP) prevalence (children 2–10 years) for 2000–2017 to compare with LMM and VECTRI prevalence; regions with prevalence <1% considered malaria-free. (2) Lysenko and Semashko early 1900s endemicity classes (pre-intervention), used to compare MARA, MIASMA, and LMM_R0 after converting LTS to endemicity categories (holoendemic >9 months; hyper 6–9; meso 3–6; hypo 1–3; free <1 month).
• Climate model and scenarios: IPSL-CM5A-LR (CMIP5 ESM) provided temperature and rainfall. Reference RCP8.5 simulation used as baseline high-emissions scenario. Additional freshwater “water-hosing” experiments (ICEXm) superimposed on RCP8.5 to mimic accelerated Greenland melting: ICE0.5m, ICE1m, ICE1.5m, ICE3m corresponding to global SLR equivalents; ICE1m releases 0.22 Sv freshwater to the North Atlantic from 2020–2070. ICE3m represents a large destabilization akin to Heinrich-like events.
• Bias correction: Applied CDF-t method to IPSL outputs for 1950–2099 using WFDEI (ERA-Interim-derived) observations at 0.5° grid; monthly bias-correction preserves long-term trends though not all moments/quantiles.
• Analysis periods and metrics: Reference period 2000–2020 (“2010s”); maximum disturbance period 2040–2050 (“2040s”). Computed difference maps (2040s minus 2010s). Additional ice-melt effect quantified by differences ICE1m minus RCP8.5 for the 2040s. Defined four regions for time series averaging land points with 6-year running mean: West Africa, East Africa, Central African west coasts, and southern Africa.
• Model skill: Compared MMM outputs against MAP prevalence and Lysenko & Semashko endemicity. Reported Pearson correlations (e.g., LMM prevalence 0.527; MARA LTS 0.561), NMAE, and RMSE; LMM and MARA showed better overall skill among models.

• Validation: MMMs reproduce broad malaria-free zones and high-prevalence regions, though VECTRI tends to overestimate prevalence and spatial extent relative to MAP. MARA and LMM_R0 align well with historical endemicity patterns; MIASMA and VECTRI are more expansive.
• RCP8.5 (2040s vs 2010s):
  • East African highlands: All MMMs project increased risk. LMM prevalence increase up to ~18%; MARA LTS lengthening by ~3 months over Ethiopia. Drivers: temperature rise up to +3 °C and precipitation increase ≥40 mm/month.
  • West Africa/Sahel and southern Africa: LMM simulates decreased prevalence over Sahel and southern Africa as already warm regions exceed vector survival thresholds (~+1 °C warming over Sahel; limited rainfall change). Future near-year-round transmission is simulated only in favorable zones (LTS sometimes >10 months in parts of East Africa). Temperatures >32 °C or <20 °C and arid zones (<50 mm/month) remain malaria-limiting.
• Additional Greenland ice-melt (ICE1m vs RCP8.5, 2040s):
  • Climate signal: Cooler over much of Africa (strongest over Sahara and parts of southern Africa) with pronounced drying over Sahel, East, and Central Africa; wetter over southern Africa, indicating a southward rain-belt shift.
  • Sahel/West Africa: Further reduction in malaria risk. Prevalence decreases by ~4–8% (LMM and VECTRI) relative to RCP8.5; signal strengthens in ICE3m.
  • East African highlands: The RCP8.5 increase is moderated. LMM prevalence reduced by ~6–16%; VECTRI reduction up to ~10%. Monthly models show LTS shortened by ~1 month (LMM_R0, MIASMA); MARA shows −1 to +3 months depending on location.
  • Southern Africa and western Central African coasts: Increased risk with ICE1m. Prevalence increases by ~6–14%; LTS increases by ~1–3 months by the late 21st century. Largest changes occur in southern Africa, linked to increased precipitation.
  • Endemicity classes: >80% of grid points retain their endemicity class; where changes occur, shifts are typically by one class.
• Mechanism: Freshwater-induced AMOC slowdown cools the North Atlantic, shifts the ITCZ and subtropical jet southward, leading to Sahel drying and southern Africa moistening, consistent with paleo evidence and hosing experiments.

The study demonstrates that relying solely on standard high-emissions scenarios (RCP8.5) can miss substantial spatial redistributions of malaria risk driven by non-linear climate dynamics. Under RCP8.5, malaria risk increases in cooler highlands of East Africa and decreases in the hottest parts of West Africa due to thermal limits on vectors. Superimposing rapid Greenland melt (ICE1m) cools and dries northern tropical Africa and moistens southern Africa, moderating the projected East African increases, amplifying West African decreases, and promoting emergence or intensification in southern Africa. These patterns are consistent across diverse malaria models and align with a robust southward shift of the African rain-belt associated with AMOC weakening seen in paleo records and climate models. The findings have significant public health relevance, warning of potential emerging risk in southern Africa, a region with limited recent exposure, and emphasizing that tipping-point climate processes can alter disease risk landscapes beyond those suggested by greenhouse-gas forcing alone. While African monsoon projections vary among models under standard scenarios, the direction of rain-belt shift under freshwater hosing is comparatively robust, reinforcing the plausibility of the malaria responses shown here.

This work quantifies how an accelerated Greenland ice-sheet melting, via AMOC weakening and a southward-shifted African rain-belt, could reshape malaria risk in Africa beyond standard RCP8.5 projections: attenuating increases over East African highlands, further decreasing risk over the Sahel/West Africa, and increasing risk over southern Africa. The multi-model malaria framework and validated climate–disease links provide mechanistic insight rather than precise predictions. The study underscores the need to incorporate rapid climate tipping-point scenarios into health, agriculture, and water-resource risk assessments. Future research should: (1) perform multi-model climate hosing experiments across centers to quantify inter-model uncertainty; (2) increase spatial resolution and incorporate hydrology and land–atmosphere–vegetation feedbacks; (3) couple malaria models with dynamic demographics, interventions, and socio-economic trajectories; and (4) explore additional tipping processes (e.g., permafrost thaw) and compounding effects on vector-borne diseases.

• Malaria models consider primarily climatic drivers; socio-economic factors, intervention scale-up, drug/treatment evolution, and dynamic population changes are not represented (VECTRI includes static population effects only). No age stratification or adaptive control measures are modeled.
• Validation relies on MAP (a statistical product with its own biases) and historical endemicity categories; model parameterizations focus on P. falciparum and An. gambiae and may not capture all vector/parasite ecotypes.
• Single climate model (IPSL-CM5A-LR) used for hosing experiments due to complexity and computational demands; inter-model spread of African monsoon responses under standard scenarios is known to be large.
• Climate feedbacks (e.g., vegetation dynamics) and basin-scale hydrology, rivers, lakes, and dams are not fully resolved at current resolution.
• Bias-correction (CDF-t) preserves trends but not all statistical moments/quantiles; potential residual biases remain.