Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2025 Sep 6.
Published in final edited form as: Lancet Planet Health. 2025 Aug 20;9(8):101293. doi: 10.1016/j.lanplh.2025.101293

Climate-smart public health for global health resilience

Christopher D Golden 1, Marissa L Childs 1, Oladimeji E Mudele 1, F Adolphe Andriamizarasoa 1, Timothy A Bouley 1, Giacomo De Nicola 1, Max A Fontaine 1, Peter J Huybers 1, Paubert T Mahatante 1, Rija Rabemananjara 1, Norohasina Rakotoarison 1, H Rakoto Ramambason 1, Herlyne Ramihantaniarivo 1, Hervet J Randriamady 1, Hanitra Randriatsara 1, M Ando Ravelomanantsoa 1, A K Symphonia Razafinimanana 1, Angela J Rigden 1, Joy Shumake-Guillemot 1, Lethicia L Yasmine 1, Francesca Dominici 1
PMCID: PMC12412889  NIHMSID: NIHMS2106502  PMID: 40848733

Abstract

Climate change poses urgent public health risks from rising global temperatures and extreme weather events, including heatwaves, droughts, and floods, which disproportionately affect vulnerable populations. To address the current silos embedded in climate, environmental, and public health monitoring and surveillance systems, climate-smart public health (CSPH) creates an integrated platform for action across these sectors, enabling more rapid and efficient responses to climate-related public health challenges. In this Personal View, we introduce the concept of CSPH, a data-driven framework designed to monitor, assess, and adapt to climate-related health impacts. CSPH incorporates surveillance, risk assessment, early warning systems, and resilient health-care infrastructure to address the evolving challenges of climate change. The framework adopts an iterative, community-centred model that responds to local needs and incorporates feedback from health-care providers and policy makers. CSPH also leverages data science and artificial intelligence to address a wide range of health concerns, including infectious diseases, non-communicable diseases, nutrition, and mental health. We applied this framework in Madagascar, a region highly vulnerable to climate impacts, where poverty, malnutrition, and frequent extreme weather events make climate adaptation particularly urgent. Early data analysis has shown strong climate sensitivity in important diseases such as malaria and diarrhoea, which could enable preparedness efforts to target some regions more efficiently. CSPH provides a pathway to enhance resilience in such settings by improving the capacity of public health systems to withstand and respond to climate-related stressors.

Introduction

Climate change poses substantial and multifaceted risks to human health, representing a complex and interconnected challenge that necessitates urgent attention.1,2 Rising global temperatures contribute to extreme weather events, such as heatwaves, droughts, storms, wind intensification, and floods. Such events can harm individuals and disrupt the functioning of health-care systems.3 Furthermore, these impacts are not distributed equally; vulnerable populations, including low-income and minoritised communities and those without adequate access to health care, are disproportionately affected, as evidenced by global trends in increased infant mortality and reduced life expectancy corresponding with increased temperatures across 170 countries between 1960 and 2016.4 Addressing the health consequences of climate change and ecosystem degradation requires interdisciplinary collaboration, policy interventions, and public health initiatives to enhance adaptive capacities against environmental risks. Such efforts can ensure intergenerational equity and safeguard the health of present and future generations. Although climate mitigation remains a global priority, continued warming is expected even under the most optimistic scenarios. Existing public health information systems do not include primary exposure data related to climate and environmental change, inhibiting an understanding of critically important drivers of ill health. Public health systems must therefore adapt urgently to face this challenge.

Climate-smart public health (CSPH): what it is and why we need it

In this Personal View, we introduce an integrated operational framework called CSPH for a data-driven approach to detect, quantify, and adapt to climate-related health impacts at a global scale. CSPH builds on and expands upon the elements of WHO’s operational framework for building climate-resilient health systems.5 CSPH also aims to support intergenerational equity by protecting present and future communities, reducing health disparities, and improving public health system resilience to climate and environmental change. The panel (page 2) summarises the components of CSPH.

The four components of CSPH are not necessarily sequential; CSPH uses an iterative structure with feedback loops to integrate inputs from local communities, health practitioners, and implementation organisations into the system to understand vulnerabilities via a bottom-up approach. This adaptive and iterative approach facilitates a CSPH system that can address evolving climate hazards.

Our vision for CSPH incorporates climate change adaptation into public health practice, leveraging data science and artificial intelligence (AI) to guide decision making across health domains, including infectious diseases, non-communicable diseases, nutrition, and mental health. Given the global open data landscape and advancements in remote sensing technologies, the volume of accessible data on climate stressors, socioeconomic variables, and health information is increasing. Although this massive data availability is a major opportunity, the complexity and heterogeneity of data in this field make the use of AI necessary for the seamless ingestion, integration, and analysis of deep-learning models.6 However, AI has only recently been made available to the public health community. Public health interventions have typically focused on data-rich evidence, yet some social groups are often under-researched because of their vulnerability.7 Harnessing big data (including remote sensing information) and the use of AI can help to compensate for these missing data and adopt approaches to target vulnerable populations in the absence of primary survey data.

The proposed CSPH system intends to empower policy makers and health-care professionals with data-driven decision making to adapt to climate-related health stressors. Our framework builds on previous work developed with a focus on climate-sensitive infectious diseases8,9 and leverages international platforms, such as WHO’s Early Warning and Response Systems. For instance, in Madagascar, innovative tools, such as malaria early warning systems, were conceived at local scales.10,11 Similar initiatives in the Caribbean were aimed to bridge environmental, climate, and health data with mainstream public health informatics using planetary health and data science approaches.12

CSPH in Madagascar

Madagascar presents an ideal setting to implement CSPH because of its high vulnerability to climate change and ongoing preparedness efforts. The country is prone to frequent cyclonic events and associated floods,13,14 major heatwaves,15 and drought.16,17 Furthermore, more than 80% of the population lives in poverty.18 Child mortality (age <5 years) remains high at 66 deaths per 1000 livebirths,19 and malnutrition results in stunted growth of nearly half the population.20 The maternal mortality rate in Madagascar is among the highest in the world, with 392 deaths per 100 000 livebirths.21 These health and socioeconomic challenges are expected to intensify with future climate and environmental changes. Our previous research linked climate-related stressors to decreased nutrition and food security, through ocean warming that alters fisheries22 and changes in atmospheric circulation that increase agricultural water stress.23 In addition, deforestation is associated with increased incidences of diarrhoeal24 and vector-borne diseases.25

Operationalising CSPH in Madagascar

To illustrate the application of CSPH, we describe two ongoing programmes in Madagascar: drought monitoring to assess the risk of crop failure and associated malnutrition, and harmful algal bloom (HAB) detection and monitoring to assess the risk of marine food intoxication and diarrhoeal illness.

Part 1: Surveillance and monitoring

The first component of CSPH involves developing integrated surveillance of climate and ecological exposure data in relation to relevant health outcomes. Madagascar has a centralised system to monitor public health and disease, collating data from more than 2770 public clinics. Between 2010 and 2019, these data were managed by local platforms; in 2019, the data were migrated to District Health Information Software 2 (DHIS2), a free and open-source health management data platform used in more than 100 countries.26

Nevertheless, active monitoring of spatial and temporal health trends influenced by climate stressors remains limited because of infrastructure and capacity constraints. Moreover, the health surveillance platform is not directly integrated with any environmental, climate, or agricultural data source; thus, how changes in these systems might affect human health remains unclear. To address this challenge, we developed a georeferenced and temporally explicit clinical health database integrated with in situ climate data, remote sensing observations, and reanalysis products related to climate and environmental exposures. This database enables rapid health research, prediction, and public health planning.

The integration of climate and environmental exposure data into the health database constitutes climate services for health.27 Our health database includes monthly incidence data for more than 60 health outcomes (infectious, non-communicable, and nutritional conditions) sourced from more than 2770 geographically distributed and geospatially referenced public health-care clinics between 2010 and 2025 (figure 1). This database is currently housed at the Harvard TH Chan School of Public Health, where it was cleaned, geocoded, and coherently linked across data collection waves before and after DHIS2 transition. In partnership with the Direction des Études et de la Planification et du Système d’Information (DEPSI) at the Madagascar Ministry of Public Health, we are developing a platform to automate climate-health analyses, perform data visualisations, and enable public health planning. DHIS2 had recently launched a climate-health portal that focuses on temperature, precipitation, and humidity. However, our version differs from it by focusing on broader climate and environmental exposures and includes health data from 2010 to present, adding approximately a decade of health data predating DHIS2, which was established in 2019.

Figure 1:

Figure 1:

Open in a new tab

Research data platform to operationalise climate-smart public health

To address disparities that are pronounced in settings with limited data infrastructure, such as Madagascar, our system exploits low-bandwidth ground-based measurements, such as low-cost air quality sensors, for enhanced local accuracy, in addition to leveraging DHIS2 and freely available remote sensing and reanalysis data. Selecting Madagascar as our pilot location ensures that CSPH is battle-tested for similar settings.

Data governance is central to our approach, with clear data use agreements allowing local governments to define access and terms of use. Institutional Review Boards provide ethical oversight, especially for sensitive health data. We also implement data anonymisation protocols and prioritise the development of community-centred tools to ensure local engagement and relevance.

Drought and malnutrition

Madagascar is vulnerable to drought, which affects major staple crops such as rice, maize, and cassava. In recent decades, drought incidence in Southern Madagascar has been exacerbated by climate change17 and contributed to famine. In some regions, increasing temperatures and declining soil moisture during growing seasons reduced crop yields, resulting in poor harvests and economic losses for farmers.28,29 Previous research has linked water stress to changes in crop production at regional scales across Madagascar.23 The next step of this analysis is to estimate crop production at fine spatial resolutions across the country using remote sensing and AI approaches. We intend to identify the relationships between satellite-based observations, including temperature and soil moisture, and granular crop production data, focusing on rice, cassava, maize, and sweet potato.

Our workflow for estimating the annual production of major staple crops in Madagascar leverages high spatial and temporal resolution data from products such as the Sentinel-1 and Sentinel-2 satellite missions of the European Space Agency. The high temporal resolution of these missions enables high fidelity of crop phenological information, which is an essential component to link crop development to water availability.30 Once the sensitivity of all crops to different climatic conditions is quantified, we might be able to attribute the reduction in crop productivity to temperature and soil moisture changes and estimate agricultural vulnerability to future drought events.30

In the proposed framework, AI involves training deep-learning models to map rice paddies across Madagascar for every growing season over multiple years. The models, trained on time-series data that capture crop phenological patterns, will be able to distinguish rice from other crops based on unique seasonal growth patterns.31 We aim to collect field samples across diverse agro-ecological zones to serve as training and validation points, tagging key stages such as transplanting, tillering, flowering, and harvest. These ground observations will be aligned with satellite-derived vegetation indices and Sentinel-1 data to develop reliable phenological signatures of rice. Once trained, the models will produce high-resolution, season-by-season maps of rice extent, which can be used to estimate the planted area, monitor changes over time, and compute the yield.32 The effects of climate stress on crops can then be linked to our health systems database using the resulting time series of acreage and yield estimates. The different variables can be aggregated to administrative boundaries, and statistical causal inference tools can be applied to estimate how changes in food production and security might be associated with an increasing incidence of malnutrition (figure 2).

Figure 2: Research data platform to evaluate the associations among climate stressors, agricultural production, and malnutrition.

Figure 2:

Open in a new tab

Climate and environmental data note the highest currently available resolution temporally and spatially from different products. VPD, Vapour pressure deficit; NDVI, Normalised difference vegetation index.

Harmful algal blooms and health

Harnessing a methodology similar to that described above, we can also retrospectively detect and prospectively monitor environmental phenomena such as HABs.33 Increasing sea and freshwater surface temperatures, possibly in combination with changes in nutrient levels and thermal stratification, create conditions conducive for HABs.34 HABs can produce toxins that get into aquatic foods, resulting in marine food contamination and associated poisonings and diarrhoeal disease (figure 3). HABs are caused by different phytoplankton species in marine and freshwater environments. Although satellite-based assessment of water colour can identify blue-green algae in freshwater and red tides in seawater,35 these approaches are yet to be consistently applied in Madagascar. Thus, a CSPH approach will require translating such detection systems to consistently identify HAB occurrences in salt and freshwater ecosystems over time. The relationships between water surface temperatures, HABs, and health effects such as diarrhoea and marine food intoxication can then be evaluated. Machine-learning models could be trained to detect HABs using known occurrences of algal blooms and marine food poisonings in Madagascar and other regions—eg, from the Harmful Algal Event Database (HAEDAT)—with features from satellite and reanalysis products such as the multisensor Global Ocean Color Plankton Chlorophyll-a36 as well as weather variables from the European Centre for Medium-Range Weather Forecasts reanalysis data (ERA5).37

Figure 3: Research data platform to evaluate the associations among climate stressors, harmful algal blooms (HABs), and food-borne illness.

Figure 3:

Open in a new tab

Climate and environmental data note the highest currently available resolution temporally and spatially from different products. SST, Sea surface temperature.

To execute these monitoring and surveillance activities, expertise from numerous collaborators must be harnessed, and different data sources from various organisations should be analysed. CSPH relies on ERA-5 data for air temperature, wind speed, and precipitation;37 MODIS/Terra vegetation indices for vegetation condition measurement;38 and CHIRPS precipitation products for rainfall and drought.39 These datasets are openly accessible with global coverage, making them invaluable for scalable workflows. Consequently, the workflows that will be developed using CSPH can be extended to other countries, particularly in resource-limited settings, by leveraging consistent, high-quality satellite-derived environmental data. To minimise energy consumption and carbon footprint during AI model training, lightweight deep-learning architectures—specifically tailored to remote sensing tasks—shall be used. This choice also avoids the unnecessary use of resource-intensive systems such as large language models.

Part 2: Risk assessment

Risk assessment in climate-health relationships involves a systematic process to evaluate the potential health impacts of climate-related stressors on ecological and health outcomes. After identifying a climate-related stressor or hazard (eg, extreme heat, drought, floods, and cyclones), the short-term and long-term consequences of its exposure to health outcomes should be estimated. Risk assessment should therefore leverage the best available statistical and causal inference methods to causally link impacts of climate and environment with health outcomes. Risk assessment also involves assessing exposure–response relationships to identify how the risk of adverse health outcomes increases with the intensity or frequency of climate hazards (eg, temperature thresholds). Furthermore, as factors such as socioeconomic status and other temporal and geographical variables might influence the impact of these climate hazards, vulnerable populations must be identified. For example, identifying locations that are sensitive to drought and crop failure or economically poor and distant from food markets might improve the identification of populations vulnerable to malnutrition caused by climate-driven food system shocks. Ultimately, the goal of climate-health risk assessment is to provide evidence-based recommendations for public health policies and adaptation strategies to address the health impacts of climate change. Risk assessment also offers a starting point for dialogue with policy makers; by clearly and causally identifying climate-health linkages and vulnerable populations, risk assessments can motivate the discussion for appropriate protective actions.

Part 3: Climate-smart early warning and response systems

Once the first two components of CSPH (integrated surveillance and risk assessment) are established, modelling approaches can facilitate the third step: early warning and response systems. We are developing early warning systems to empower policy makers and practitioners to efficiently act on evidence, supporting public health specialists to prevent disease onset and strengthen climate resilience. Partnership with local stakeholders can facilitate the identification of appropriate lead times and spatial scales for intervention, ensuring that early warning systems are tailored to support resilience efforts and meet local needs.

When applied to climate threats such as droughts, early warning systems can enhance forecasts and outline the expected impacts on agriculture and health, thus promoting climate adaptation and increasing human resilience. Drought prediction helps identify areas of agricultural vulnerability, enabling targeted interventions in food systems and guiding humanitarian responses to upcoming crop failures. The relationships linking food systems and weather can be characterised using historical observations, offering potential for predictability at various time scales. For example, skilful prediction of La Niña and El Niño–Southern Oscillation events was demonstrated up to 18 months in advance,40 with important implications for forecasting the risks posed by tropical storms and droughts in Madagascar. Similarly, using satellite-derived sea surface temperature data, Funk and colleagues41 and Shukla and colleagues42 predicted spring drought in several African countries experiencing food insecurity. Such predictions can support regional forecasting efforts and land surface data assimilation to facilitate early warning and guide climate outlooks. On short time scales, novel technologies, including satellite-based direct monitoring of crop development and end-of-season harvest, can provide a lead time of several months for anticipating food shortages. These advances can reduce reliance on ad-hoc, last-minute preparations that are currently used. Further, timely warnings could lead to interventions that foster climate adaptation and overall system resilience, mitigating impacts on malnutrition and downstream effects on mental health.

By combining remote sensing-derived variables (such as mean chlorophyll concentration and sea surface temperature) with long-term, multisensor ocean-colour time series and ground-truth data, machine-learning models can be trained to predict and anticipate the occurrence of HABs.43 By detecting the environmental precursors, such as elevated sea surface temperatures or increased nutrient loads from rain-fed runoff, the models can warn public health officials, who could then intervene before exposure occurs.

To execute the early warning and response system, we plan to develop systems that feature vulnerability mapping to identify communities at risk for HABs, automated alerts for HAB detection, and decision-support tools to guide targeted interventions. By integrating climate and environmental data, we can improve spatial and temporal precision of predicted issues associated with climate-related marine food systems to support rapid decision making. Historically, HAB responses in Madagascar have relied on blanket seasonal warnings or alerts after marine intoxications were detected; a more targeted forewarning should allow for better and more nuanced responses. The Ministry of Fisheries and the Blue Economy has prioritised this research area, with plans to issue both social media (eg, Facebook) and radio alerts when HABs are detected. Early warning and response systems can become less reactive and more formalised into health care and public health interventions, laying the groundwork for the fourth component of CSPH, which mainstreams climate-health understandings into health-care infrastructure and provider capacity.

Improving estimates of health impacts from climate-related stressors is essential for informing climate adaptation strategies. However, translating such estimates into actionable recommendations and intervention strategies considering real-world constraints and limited resources is equally crucial,44 representing another scenario in which AI, particularly, reinforcement learning, can be revolutionary and used to recommend optimal actions from data. In reinforcement learning, algorithms learn optimal actions using real-world conditions and trade-offs through an agentic perspective in a decision-making environment with a reward or cost signal.45 In the climate-health context, reinforcement learning can help to identify policies for reducing hospitalisations and mortality (over multiple time scales) during early warnings and guide interventions addressing extreme heat, wildfires, flooding, air quality, and infectious diseases.46 These examples emphasise the need to harness advances in open data technology with innovations in AI and machine learning to enable state-of-the-art applications in climate-health practice.

Part 4: Climate-smart health infrastructure and provider capacity

Climate-smart health infrastructure and provider capacity are essential for strengthening health-care systems to withstand the direct and indirect impacts of climate-related stressors while ensuring long-term community resilience and continuity of care. The three described components of CSPH should enable a better understanding of the types of diseases, populations, and geographical areas most vulnerable to climate hazards and susceptible to climate-sensitive health outcomes. The output should be then used by public health systems to develop climate-smart built infrastructure, medical records and patient tracking, and educational tools for patients and practitioners. This process involves designing and constructing health-care facilities that are energy-efficient, environmentally sustainable, and resilient to extreme weather events such as floods, heatwaves, and hurricanes. Hospitals, clinics, and other health infrastructure should be located and built by considering climate risks and incorporating climate-resistant materials and energy systems such as solar power to minimise dependency on vulnerable grids. In addition, climate-smart infrastructure must include adaptive features such as temperature regulation systems, clean water and sanitation solutions, and backup power systems to ensure uninterrupted service during extreme weather events.

Equally important to developing technologies is empowering health-care providers to recognise and respond to evolving climate-related health challenges. This task requires integrating climate knowledge into medical training and continuing education, enabling health-care professionals to recognise symptoms and diseases exacerbated by climate-related stressors, such as heat-related illnesses, respiratory issues from air pollution, or vector-borne diseases such as malaria and dengue. Health-care systems must also develop robust surveillance and early warning systems for climate-related health risks, allowing for timely responses to emerging threats described previously in this article. Training health workers in these areas will ensure that they are capable of providing immediate care during climate emergencies and are equipped to engage in preventive health care and public health strategies that reduce the long-term health impacts of climate-related stressors. By enhancing both infrastructure and provider expertise, society can better mitigate the health risks posed by a changing climate and ensure that communities remain resilient in the face of increasing environmental challenges.

Policy infrastructure for CSPH in Madagascar

Globally, an iterative process of vulnerability and adaptation assessment (VAA) followed by a health national adaptation plan (HNAP; or PNASS as a French acronym), typically revised every 5 years, exists. The VAA and HNAP process mirrors our definition of CSPH. The VAA relies on integrated climate and health surveillance systems and detailed risk assessment to examine regional and population-level vulnerabilities, whereas the HNAP focuses on the development of early warning and response systems and the mainstreaming of climate into health-care infrastructure. A key element highlighted in both the VAA and HNAP is the need for robust integrated surveillance and climate-informed early warning systems. Data-driven risk assessment to examine the historical relationships of climate and environmental variation with human health is the first step towards developing such early warning systems.

The Government of Madagascar is committed to taking actions to adapt to climate change. The locus of activity within Madagascar for more than 15 years has been an interagency working group dedicated to addressing climate and health.47 This climate-health interagency working group is housed within the Ministry of Public Health and coordinates with WHO and the World Meteorological Organization. The first WHO-led VAA for Madagascar was conducted in 2011,48 establishing a basis for in-country and international programme building and iteration. In 2015, WHO published a climate and health country profile,49 jointly produced by partners in the country and at WHO, which intended to serve as a quick reference for often hard-to-find data regarding health and climate. In 2016, the World Bank conducted a climate-health stress test and diagnostic in Madagascar, the first of any country, setting a precedent for a global programme by establishing new methodology and pathways for health and climate partnership among a funding agency, the UN, and governments. A primary outcome of this stress test was the observation that the country lacked a strong, unified information system that could enable climate-health analyses.47 The first HNAP for Madagascar was also published in 2016;9 the plan identified the vulnerabilities and strengths of the current health system in the context of climate change. Subsequent years were marked by limited international climate and health engagement in Madagascar, owing partly to the resource and bandwidth prioritisation in the wake of the COVID-19 pandemic. In 2024, the World Bank published an internally prepared vulnerability assessment report.50 All assessments to date have identified the need for a strong integrated climate and health surveillance and monitoring system that relies on interoperable, spatio-temporally explicit datasets. However, the execution of these activities faces obstacles such as insufficient funding, limited rural capacity at the district and communal levels, a shortage of health professionals, and challenges in cross-sectoral collaboration on issues such as food security and nutrition.

Discussion

Governments will need to increasingly focus on preparedness efforts to improve health system resilience to ensure simplistic responses in the face of climate change. Therefore, research and policy must create an infrastructure for public health systems to act under increasing pressure. At the UN Climate Change Conference in Glasgow (COP26), a coalition of 50 countries, including Madagascar, pledged to create climate-resilient and low-carbon health systems, acknowledging the increasing evidence of the sensitivity of the public health sector to climate stressors. The Alliance for Transformative Action on Climate Change and Health (ATACH) has committed to supporting these efforts globally by providing a collaborative ecosystem, and the number of participating countries has now increased to 85 or more. The commitment of Madagascar towards the objectives formulated at COP26 reinforced its past efforts to centre public health on climate resilience.

CSPH must be co-produced to ensure its applicability and ultimate adoption. Such co-production will capitalise on the years of climate-health initiatives across countries. Building coalitions across sectors to address human health impacts of climate change requires a collaborative approach grounded in shared goals and interdependence. These cross-sectoral strategies must incorporate the core elements of WHO’s conceptualisation of a resilient health system, including sound governance, effective leadership, political commitment, human resources, and organisational capacity development, as well as diligent communication and awareness. The policy frameworks (eg, National Adaptation Plans) establish a shared set of objectives that highlight the impacts of climate change on various sectors, such as agriculture, health, and terrestrial and marine environments, thereby linking sectoral efforts to a common challenge. Concerted efforts by all stakeholders will drive the success of CSPH. This success will be assessed through three primary mechanisms: 1) validation of environmental or climate exposure indicators included in the CSPH using state-of-the-art data validation methods; 2) adoption of the CSPH platform by the Ministry of Public Health in its monthly and annual reporting, intervention planning, and VAA documentation; and 3) estimation of changes in the incidence of climate-sensitive diseases attributable to efficient case management resulting from improved preparation efforts.

The development of the CSPH data platform and conceptualisation of early warning systems are shaped through consultation with local partners from the Ministries of Public Health, Environment and Sustainable Development, and Fisheries and the Blue Economy, as reflected in the authorship of this manuscript. Although incomplete at the time of publication, this interdisciplinary and cross-sectoral partnership will underpin all outputs of the CSPH programme. This framework currently does not include a formal mechanism to incorporate indigenous traditional knowledge (ITK). However, ITK could be woven into all four elements of CSPH, especially in surveillance, early warning systems, and in developing adaptation and resilience capacity to address future risks. Evidence suggests that indigenous groups have historically monitored and predicted local climate conditions.51 Such practices are effective at predicting health events and developing integrated early warning systems.52,53 Although integrating ITK systematically into mainstream national public health systems is challenging, attention should be paid to the unique and important health implications for indigenous populations.54

The CSPH framework relies on a wide range of data on health, climate, and environmental conditions. Understanding and contextualising the limitations of data are essential to avoid reinforcing or creating disparities in public health responses. For example, global environmental and climate datasets might be better validated in and more representative of locations that have more well-established ground monitoring infrastructure (eg, meteorological stations with air temperature records). Both monitoring and risk assessment components must contribute to understanding environmental datasets in the relevant context using ground-truth data, thereby preventing inaccurate data from biasing the estimates. A growing body of literature highlights the importance of using predicted or remotely sensed estimates of climate and environmental conditions.55 As with the environmental data, understanding the representativeness of the health data is crucial for ensuring that CSPH does not reinforce existing inequalities in health infrastructure and access to care by ignoring under-represented populations. Health data collected from clinics and hospitals can be compared to data from community health surveys or cohort studies to identify well-captured populations and health conditions. Both these data should be considered when expanding CSPH to new settings; further, CSPH broadening may necessitate ground data collection to better leverage large-scale datasets from satellite imagery or national health systems. Such data-intensive functions and applications pose challenges to high-performance computing in settings such as Madagascar, where cloud-based workflows are constrained by limited cyberinfrastructure. This necessitates global north–south partnerships until sustainable computing solutions are established.

To conclude, we propose our CSPH research data platform that relies on DHIS2, the world’s most widely adopted health data management system. With an existing blueprint and by developing analytical methods that link climate and environmental data with DHIS2-structured health system data, the project is designed to extend beyond Madagascar. More than 100 countries use DHIS2 for collecting and analysing health data, covering 3·2 billion people (40% of the global population). Although the raw health data for Madagascar cannot be released publicly, all algorithms and data integration methods developed through this work will be adaptable to any country using DHIS2. Furthermore, partnerships have already been established with the Government of Nepal to enable the transfer of learning from one context to the next, with the goal of ensuring that these research innovations contribute to easing the health burden of future climate impacts on health.

Panel: Four components of CSPH.

  • Surveillance and monitoring to estimate spatiotemporal variations in ecological and climate exposure data, alongside climate-sensitive health outcomes

  • Risk assessment to identify and evaluate health risks posed by climate stressors, including the characterisation of vulnerable populations and geographies

  • Climate-smart early warning and response systems to deploy data-driven alerts using state-of-the-art highly localised forecasts (eg, for drought and extreme heat predictions)

  • Climate-smart health infrastructure and provider capacity to strengthen health-care systems to withstand climate-related impacts (eg, cyclonic wind and flood damage) on health-care infrastructure and to respond to long-term needs of populations in a changing environment to ensure resilience and continuity of care.

Acknowledgments

This study was funded by the Ren Che Foundation to CDG, Amazon Web Services’ support of the Harvard Data Science Initiative (awarded to CDG, OM, and MLC), and the National Science Foundation’s AI Institute for Societal Decision Making (grant number IIS2229881 awarded to CDG), and the National Institutes of Health Fogarty International Center (grant number 1P20TW013028-01 awarded to CDG). We also thank Alex Boersma for generating figures 2 and 3 for this article.

Footnotes

Declaration of interests

We declare no competing interests.

For more about the HAEDAT, see https://haedat.iode.org/

Editorial note: The Lancet Group takes a neutral position with respect to territorial claims in published maps and institutional affiliations.

References

  • 1.Romanello M, di Napoli CD, Green C, et al. The 2023 report of the Lancet Countdown on health and climate change: the imperative for a health-centred response in a world facing irreversible harms. Lancet 2023; 402: 2346–94. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2.Mahon MB, Sack A, Aleuy OA, et al. A meta-analysis on global change drivers and the risk of infectious disease. Nature 2024; 629: 830–36. [DOI] [PubMed] [Google Scholar]
  • 3.Intergovernmental Panel on Climate Change. Climate Change 2023: Synthesis report, summary for policymakers. Intergovernmental Panel on Climate Change, 2023. [Google Scholar]
  • 4.Meierrieks D. Weather shocks, climate change and human health. World Dev 2021; 138: 105228. [Google Scholar]
  • 5.WHO. Operational framework for building climate resilient health systems. WHO, 2015. [Google Scholar]
  • 6.Zeng D, Cao Z, Neil DB. Chapter 22: Artificial intelligence–enabled public health surveillance—from local detection to global epidemic monitoring and control. In: Xing L, Giger ML, Min JK, eds. Artificial intelligence in medicine. Academic Press, 2021: 437–53. [Google Scholar]
  • 7.Ellard-Gray A, Jeffrey NK, Choubak M, Crann SE. Finding the hidden participant: solutions for recruiting hidden, hard-to-reach, and vulnerable populations. Int J Qual Methods 2015; 14: 1609406915621420. [Google Scholar]
  • 8.Phung D, Colón-González FJ, Weinberger DM, et al. Advancing adoptability and sustainability of digital prediction tools for climate-sensitive infectious disease prevention and control. Nat Commun 2025; 16: 1644. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Stewart-Ibarra AM, Rollock L, Best S, et al. Co-learning during the co-creation of a dengue early warning system for the health sector in Barbados. BMJ Glob Health 2022; 7: e007842. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Rakotoarison N, Raholijao N, Razafindramavo LM, et al. Assessment of risk, vulnerability and adaptation to climate change by the health sector in Madagascar. Int J Environ Res Public Health 2018; 15: 2643. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Evans MV, Ihantamalala FA, Randriamihaja M, et al. Increasing the resolution of malaria early warning systems for use by local health actors. Malar J 2025; 24: 30. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Díaz AR, Rollock L, Boodram LLG, et al. A demand-driven climate services for health implementation framework: A case study for climate-sensitive diseases in Caribbean Small Island Developing States. PLoS Clim 2024; 3: e0000282. [Google Scholar]
  • 13.Weiskopf SR, Cushing JA, Morelli TL, Myers BJE. Climate change risks and adaptation options for Madagascar. Ecol Soc 2021; 26: 36. [Google Scholar]
  • 14.Fontaine I, Garabedian S, Jammes M. Short-term impact of tropical cyclones in Madagascar: evidence from nightlight data. Appl Econ 2024; 56: 5124–45. [Google Scholar]
  • 15.World weather attribution. Extreme poverty renders Madagascar highly vulnerable to underreported extreme heat that would not have occurred without human-induced climate change. Nov 23, 2023. https://www.worldweatherattribution.org/extreme-poverty-rendering-madagascar-highly-vulnerable-to-underreported-extreme-heat-that-would-not-have-occurred-without-human-induced-climate-change/ (accessed July 15, 2025).
  • 16.Barimalala R, Raholijao N, Pokam W, Reason CJC. Potential impacts of 1.5 °C, 2 °C global warming levels on temperature and rainfall over Madagascar. Environ Res Lett 2021; 16: 044019. [Google Scholar]
  • 17.Rigden A, Golden C, Chan D, Huybers P. Climate change linked to drought in Southern Madagascar. npj Clim Atmos Sci 2024; 7: 41. [Google Scholar]
  • 18.International Organization for Migration. Madagascar crisis response plan 2023. https://crisisresponse.iom.int/response/madagascar-crisis-response-plan-2023 (accessed Nov 13, 2023).
  • 19.World Bank Group. Mortality rate, under-5 (per 1,000 live births) - Madagascar. https://databank.worldbank.org/source/world-development-indicators/Series/SH.DYN.MORT (accessed July 15, 2025).
  • 20.UNICEF. Malnutrition data. 2023. https://data.unicef.org/resources/dataset/malnutrition-data (accessed July 15, 2025).
  • 21.WHO, UNICEF, UNFPA, World Bank Group, UNDESA/Population Division. Trends in maternal mortality 2000 to 2020: estimates by WHO, UNICEF, UNFPA, World Bank Group and UNDESA/Population Division. Feb 23, 2023. https://www.who.int/publications/i/item/9789240068759 (accessed July 15, 2025).
  • 22.Golden CD, Allison EH, Cheung WWL, et al. Nutrition: fall in fish catch threatens human health. Nature 2016; 534: 317–20. [DOI] [PubMed] [Google Scholar]
  • 23.Rigden AJ, Golden C, Huybers P. Retrospective predictions of rice and other crop production in Madagascar using soil moisture and an NDVI-based calendar from 2010–2017. Remote Sens 2022; 14: 1223. [Google Scholar]
  • 24.Herrera D, Ellis A, Fisher B, et al. Upstream watershed condition predicts rural children’s health across 35 developing countries. Nat Commun 2017; 8: 811. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.Arisco NJ, Rice BL, Tantely LM, et al. Variation in Anopheles distribution and predictors of malaria infection risk across regions of Madagascar. Malar J 2020; 19: 348. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26.Dehnavieh R, Haghdoost A, Khosravi A, et al. The District Health Information System (DHIS2): A literature review and meta-synthesis of its strengths and operational challenges based on the experiences of 11 countries. Health Inf Manag 2019; 48: 62–75. [DOI] [PubMed] [Google Scholar]
  • 27.World Meteorological Organization. 2023 State of climate services for health. https://wmo.int/publication-series/2023-state-of-climate-services-health (accessed July 15, 2025).
  • 28.Anderson WB, Seager R, Baethgen W, Cane M, You L. Synchronous crop failures and climate-forced production variability. Sci Adv 2019; 5: eaaw1976. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.Caparas M, Zobel Z, Castanho ADA, Schwalm CR. Increasing risks of crop failure and water scarcity in global breadbaskets by 2030. Environ Res Lett 2021; 16: 104013. [Google Scholar]
  • 30.Heft-Neal S, Lobell DB, Burke M. Using remotely sensed temperature to estimate climate response functions. Environ Res Lett 2017; 12: 014013. [Google Scholar]
  • 31.Zhao Z, Dong J, Zhang G, et al. Improved phenology-based rice mapping algorithm by integrating optical and radar data. Remote Sens Environ 2024; 315: 114460. [Google Scholar]
  • 32.Thorp KR, Drajat D. Deep machine learning with Sentinel satellite data to map paddy rice production stages across west Java, Indonesia. Remote Sens Environ 2021; 265: 112679. [Google Scholar]
  • 33.Hill PR, Kumar A, Temimi M, Bull DR. HABNet: machine learning, remote sensing-based detection of harmful algal blooms. IEEE J Sel Top Appl Earth Obs Remote Sens 2020; 13: 3229–39. [Google Scholar]
  • 34.Moore SK, Trainer VL, Mantua NJ, et al. Impacts of climate variability and future climate change on harmful algal blooms and human health. Environ Health 2008; 7 (suppl 2): S4. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35.Shen L, Xu H, Guo X. Satellite remote sensing of harmful algal blooms (HABs) and a potential synthesized framework. Sensors (Basel) 2012; 12: 7778–803. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36.Copernicus. Global Ocean Colour Plankton MY L4 monthly observations. Nov 30, 2023. https://data.marine.copernicus.eu/product/OCEANCOLOUR_GLO_BGC_L4_MY_009_108/description (accessed March 27, 2025).
  • 37.Hersbach H, Bell B, Berrisford P, et al. The ERA5 global reanalysis. Q J R Meteorol Soc 2020; 146: 1999–2049. [Google Scholar]
  • 38.Didan K, Munoz AB, Solano R, Huete A. MODIS vegetation index user’s guide (MOD13 series). University of Arizona: vegetation index and Phenology Lab 2015; 35: 2–33. [Google Scholar]
  • 39.Funk C, Peterson P, Landsfeld M, et al. The climate hazards infrared precipitation with stations—a new environmental record for monitoring extremes. Sci Data 2015; 2: 150066. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40.Ham YG, Kim JH, Luo JJ. Deep learning for multi-year ENSO forecasts. Nature 2019; 573: 568–72. [DOI] [PubMed] [Google Scholar]
  • 41.Funk C, Hoell A, Shukla S, et al. Predicting East African spring droughts using Pacific and Indian Ocean sea surface temperature indices. Hydrol Earth Syst Sci 2014; 18: 4965–78. [Google Scholar]
  • 42.Shukla S, McNally A, Husak G, Funk C. A seasonal agricultural drought forecast system for food-insecure regions of East Africa. Hydrol Earth Syst Sci 2014; 18: 3907–21. [Google Scholar]
  • 43.Sathyendranath S, Brewin RJW, Brockmann C, et al. An ocean-colour time series for use in climate studies: the experience of the ocean-colour climate change initiative (OC-CCI). Sensors (Basel) 2019; 19: 4285. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 44.Smit B, Pilifosova O. Adaptation to climate change in the context of sustainable development and equity. Sustain Dev 2003; 8: 9. [Google Scholar]
  • 45.Sutton RS, Barto AG. Reinforcement learning: an introduction. The MIT Press, 2018. [Google Scholar]
  • 46.Considine EM, Nethery RC, Wellenius GA, Dominici F, Tec M. Optimizing heat alert issuance for public health in the United States with reinforcement learning. arXiv 2024; published online Dec 19, 2024. 10.48550/arXiv.2312.14196 (preprint). [DOI] [Google Scholar]
  • 47.Bouley T, Midgley A, Shumake-Guillemot J, Golden CD, Ebi K. Madagascar - climate change and health diagnostic: risks and opportunities for climate-smart health and nutrition investment (English). International Bank for Reconstruction and Development/The World Bank, 2018. [Google Scholar]
  • 48.Word Bank. Madagascar—climate and health vulnerability assessment. 2024. https://hdl.handle.net/10986/41846 (accessed Oct 27, 2024).
  • 49.WHO. Climate and health country profile - 2015: Madagascar. WHO, 2016. [Google Scholar]
  • 50.Iglesias Luzardo MG, Flannery JLW, Rabary M. Madagascar - climate and health vulnerability assessment (English). International Bank for Reconstruction and Development/The World Bank, 2024. [Google Scholar]
  • 51.Kalanda-Joshua M, Ngongondo C, Chipeta L, Mpembeka F. Integrating indigenous knowledge with conventional science: enhancing localised climate and weather forecasts in Nessa, Mulanje, Malawi. Phys Chem Earth (Pt A B, C) 2011; 36: 996–1003. [Google Scholar]
  • 52.Macherera M, Chimbari MJ, Mukaratirwa S. Indigenous environmental indicators for malaria: A district study in Zimbabwe. Acta Trop 2017; 175: 50–59. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 53.Chakravarty P, Gattupalli S. Integration of indigenous traditional knowledge and AI in Hurricane Resilience and adaptation. In: Collins J, Done J, Zhu YJ, Wilson P, eds. Advances in Hurricane Risk in a Changing Climate. Springer, 2024: 125–58. [Google Scholar]
  • 54.Redvers N, Aubrey P, Celidwen Y, Hill K. Indigenous Peoples: traditional knowledges, climate change, and health. PLoS Glob Public Health 2023; 3: e0002474. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 55.Jain M. The benefits and pitfalls of using satellite data for causal inference. Rev Environ Econ Policy 2020; 14: 157–69. [Google Scholar]

RESOURCES