Tropical oceans drive variability in soil moisture and malaria in Malawi Authors: M. T. Elling, K. B. Karnauskas, M. Kowalcyk, D. Mategula, J. Chirombo, B. Livneh, R. S. McCann, A. G. Buchwald Published: 2026 ABSTRACT BackgroundTransmission of malaria, one of the world’s deadliest infectious diseases, is highly sensitive to environmental conditions. Understanding the large-scale climate patterns that influence these conditions is crucial for developing forecasting tools, which could be especially valuable for prevention in low-resource nations, like Malawi. Previous research has often focused on statistical correlations between local weather and disease trends, but has rarely explored the underlying physical climate mechanisms.MethodsWe analyze malaria incidence alongside observational, reanalysis, and modeled climate data to identify the mechanistic link between climate and malaria variability in Malawi.ResultsHere, we show that two distinct ocean-based climate patterns drive interannual malaria variability in Malawi via their influence on local climatic and surface hydrologic conditions. A warm tropical Atlantic leads to wet conditions in Malawi and increased malaria cases. In contrast, a warm Indian Ocean drives hot, dry conditions and reduces malaria cases. We find that soil moisture is the crucial link between these remote climate drivers and local disease dynamics, and looking ahead, future climate change is expected to reduce soil moisture levels in the country by 2100 (magnitude uncertain), which could reshape transmission patterns.ConclusionsBy identifying these climate drivers and the physical processes that link them to disease outbreaks, our work provides a foundation for building physically grounded, reliable early warning systems. Introduction Malaria is one of the deadliest infectious diseases globally, with an estimated 250 million cases and 600,000 deaths occurring annually1. Low resource countries are disproportionately impacted by malaria2 and Malawi, where half of the population lives below the country’s poverty line3, is among the hardest hit. The entire population of Malawi, estimated to exceed 20 million people4 is considered at risk of infection with Plasmodium falciparum5, although this risk varies considerably by location6. The impacts are substantial—malaria is a leading cause of mortality and morbidity in the country and contributes to a high child mortality rate7. Malawi has seen important reductions in malaria prevalence (from 44% in 2000 to 19% in 2022) coinciding with the scale-up of malaria interventions, including artemisinin combination therapies, insecticide-treated bed nets, and indoor residual spraying6, like other countries in the region8. Nonetheless, disease rates remain high, and new tools are needed to further reduce malaria burden, especially in hotspots6. One promising tool is climate-based disease forecasting, which leverages the well-established relationships between climate conditions and mosquito populations—the vectors for malaria parasites9–12. Mosquito population dynamics and parasite transmission intensity are sensitive to species-specific weather conditions, such as temperature13, humidity14, and rainfall15–17. For example, temperature impacts mosquito growth, development, and survival rates, and modifies behavioral factors like biting rates. Adult emergence of Anopheles gambiae, one of the primary malaria vectors in Malawi, is estimated to occur between 18 and 34 °C, with optimal temperature ranges for adult emergence at 22–26 °C, and optimal temperatures for adult development between 28 and 32 °C18. Increased rainfall can increase mosquito abundance via the creation of larval habitats, though extreme rain can wash out larvae15,19. The contribution of rainfall to these dynamics is governed by two key factors: first, the hydrological and geomorphological effects on the availability of standing surface water, which serves as the larval habitat for Anopheles funestus, Anopheles gambiae, and Anopheles arabiensis20, the primary malaria vectors in Malawi; and second, the types of aquatic habitats used by the immature stages of each vector species (e.g., temporary or permanent water bodies). To build useful forecasts with these environmental variables, we need a robust understanding of the climate drivers that shape these conditions. The aim of this paper is to uncover and quantitatively describe links between large-scale climate drivers, such as tropical sea surface temperature (SST) patterns, weather and environmental conditions, and malaria in Malawi, which may then be leveraged by climate-based forecast and early warning systems. By predicting changes in transmission intensity, these forecasts can help guide and strengthen intervention strategies21–23. Previous efforts to understand climate-malaria relationships within Malawi have often been built on statistical models that correlate historical weather and environmental variables with disease incidence24–26. These models have identified strong climate-malaria relationships, but they are commonly limited by a focus on correlation of environmental variables rather than the large-scale drivers of these conditions. These statistical approaches can show that a particular temperature or rainfall pattern is associated with an increase in malaria cases, but they often lack the mechanistic detail to explain why that relationship exists and to pinpoint the leading, more predictable climate drivers that influence those weather patterns. These studies are additionally limited by reliance on data from limited periods, sometimes only a few years, which makes it difficult to capture long-term trends and the full range of climate variability. To build useful forecasts with these environmental variables, we focus on SST patterns as our primary climate driver. Tropical modes of coupled climate variability, such as the Subtropical Indian Ocean Dipole (SIOD), El Niño Southern Oscillation (ENSO), and Indian Ocean Dipole (IOD), are known to impact seasonal weather patterns in southern Africa27,28. SSTs provide practical advantages for forecasting: SST anomalies are less noisy than their atmospheric counterparts, evolve gradually and can persist for months, offering predictability at seasonal timescales that local weather variables cannot provide. Although physical mechanistic ties between SSTs and malaria have been studied in neighboring countries29,30, there is a surprising lack of similar research in Malawi. Malawi is an ideal setting for such climate-malaria studies. It’s small size and relatively uniform climate minimize confounding factors like major differences in land use, which can otherwise obscure climate-malaria relationships. Our extensive, relatively long-term dataset (2005–2023) of nationwide, district-level malaria incidence data presents an opportunity to close this critical research gap. To understand how such large-scale climate phenomena influence malaria in Malawi, we investigate how fluctuations in SSTs in the tropical oceans drive regional atmospheric shifts that shape Malawi’s hydroclimate and influence malaria transmission. We find that SSTs in the tropical Atlantic and Indian oceans drive malaria incidence in Malawi, and soil moisture, having good memory of variations in climate, serves as the key link between climate and malaria. Our focus is on uncovering robust climate-malaria links to build a foundation for reliable, mechanistically grounded, climate-based malaria forecasting. Causal analyses have shown that SST teleconnection patterns typically precede soil moisture changes on seasonal timescales (~1–12 months)31,32. This lag is consequential for malaria transmission and adds a potential seasonal timescale of predictability. Given Malawi’s record of acting on malaria data and adopting new interventions6, the climate information presented in our study is well positioned to inform decision-making. Methods To uncover the links between climate variability and malaria trends in Malawi, we use a combination of temporal filtering, statistical decomposition, and regression techniques, focused on isolating the low-frequency signals most relevant to disease dynamics, then investigate the underlying mechanisms by examining how atmospheric changes from remote climate drivers impact weather systems and environmental conditions conducive to malaria vectors and parasite transmission. Malaria incidence data We compile monthly malaria case data reported at the district level across Malawi from three sources: District Health Information Software (DHIS-2; spanning 2018 through 2023), Malawi National Malaria Control Program (NMCP; spanning 2015 through 2023), and a published dataset from Chirombo et al. 202026 (spanning mid-2004 through 2017). DHIS-2 data consists of outpatient confirmed cases reported by health facilities to the district using RDT or microscopy. NMCP data is an aggregated and processed version of DHIS-2 data, while the Chirombo dataset, also outpatient, is directly received from NMCP in a preprocessed form. IRB approval is not required as this analysis uses aggregated data collected as part of routine public health surveillance; no individual-level data was accessed. Where datasets overlap in time, we average the available values. If only one source has data for a given month and district, we use that value directly. Cross-source agreement is high, suggesting that this aggregation had minimal impact on trends (see Supplementary Fig. 4). We use full-year data (2005–2023) so that seasonal cycles are complete before applying low-frequency filtering. Of Malawi’s 28 districts, 27 are analyzed. Likoma District is excluded from the analysis due to its small population size. Malaria incidence data includes confirmed outpatient malaria cases among all ages, diagnosed either by microscopy or Rapid Diagnostic Test. Incidence trends among all ages did not differ from trends among children under 5. In 2012-2013, changes in the NMCP reporting system, including a migration to the web-based DHIS-2 platform, coincided with an abrupt drop in reported cases, likely suppressing the usual rainy-season peak in cases. To account for this discontinuity, we correct the data leveraging malaria incidence data from Mozambique—a neighboring country that borders more than half of Malawi and experiences similar seasonal malaria patterns. The largest disagreement between incidence datasets occurs after the reporting system change in 2013 and 2014. To correct for underreporting, the 2013–2014 data in Malawi are adjusted following the procedure outlined in Supplementary Text 1, which leverages district-level relationships between the two countries to adjust Malawi’s 2013-2014 malaria incidence patterns (see Supplementary Fig. 5). Climate data We use publicly available observed, reanalysis, and modeled climate datasets to evaluate climate conditions relevant to malaria transmission. The NOAA OI SST V2 High Resolution dataset is an observational product that combines records of sea surface temperature from satellites, ships, buoys, and Argo floats and uses statistical methods to fill data gaps and adjust for bias (0.25° × 0.25° resolution), provided by the NOAA Physical Sciences Laboratory, Boulder, Colorado, USA, from their website at https://psl.noaa.gov33,34. European Centre for Medium-Range Weather Forecasts Reanalysis v5 (ERA5) is a reanalysis product, which combines weather observations with models to estimate past weather conditions (0.25° × 0.25° resolution)35. We use ERA5 fields of geopotential height, horizontal wind (u and v components), surface temperature, and vertically integrated moisture flux, accessed via the Copernicus Climate Change Service, Climate Data Store36. A benefit of ERA5 is that, due to its modeling component, many different variables are available at high spatial and temporal resolutions on a global scale. However, some variables in ERA5 are not constrained by observations, such as precipitation. While ERA5 is a reanalysis product, it has been shown to correlate well with station observations of variables like temperature and precipitation in Malawi37. In this case, we use the Climate Hazards Group InfraRed Precipitation with Station data (CHIRPS), which combines in-situ measurements with satellite data to produce high-resolution precipitation data (0.05° × 0.05° resolution)38. Data were accessed via the CHIRPS online archive (https://www.chc.ucsb.edu/data/chirps). To tighten the connection between our climatic analysis and malaria, we analyze the wetness state of the land surface via soil moisture, which we find to be the environmental factor most closely aligned with malaria incidence. Soil moisture reflects the integrated effects of rainfall, evaporation, and infiltration, making it a more direct measure of potential surface water availability, and, therefore, it may capture malaria vector habitability better than precipitation alone. The GRACE-DA-DM dataset offers unique insights into near-surface moisture conditions (0.25° × 0.25° resolution). Data were accessed via the Goddard Earth Sciences Data and Information Services Center online archive (https://disc.gsfc.nasa.gov). GRACE satellites measure small changes in Earth’s gravitational field, which are highly sensitive to changes in the amount of water stored at the surface39,40. The data are expressed as a percentile relative to a base period of 1948–2012. We analyze future projections of surface soil moisture from a suite of global climate models participating in the Coupled Model Intercomparison Project Phase 6 (CMIP6). CMIP6 is an internationally coordinated effort where climate models developed in different locations around the world are run using standardized scenarios to simulate the past, present, and future climate, as well as specialized experiments under specific conditions. We evaluate Shared Socioeconomic Pathway (SSP) experiments, which simulate climate under standardized greenhouse gas emissions trajectories41 (Supplementary Table 2). All CMIP6 outputs are regridded to 0.95° × 1.25° resolution using bilinear interpolation. Supplementary Table 1 summarizes key variables and sources. Spatiotemporal analysis of malaria Empirical Orthogonal Function (EOF) analysis is a method to objectively identify and summarize spatiotemporal patterns by decomposing the data into orthogonal functions based on its covariance matrix, capturing the main modes of variability and their strengths. We apply EOF analysis separately to monthly malaria incidence and precipitation, temperature, and surface soil moisture datasets (2005–2023). We extract spatial patterns (hereafter EOFs) and corresponding time series (hereafter principal components, PCs), reflecting the dominant modes of variability in the data. For more information on EOF analysis, see Lorenz, 195642 and Davis, 197643. All EOFs are calculated using the eofs Python library (v2.0.0)44. While EOF analysis cannot capture potential nonlinear interactions between climate modes, these methods are appropriate for identifying the dominant, orthogonal modes of variability and establishing first-order physical relationships between climate drivers and malaria transmission. The mechanistic consistency between our statistical findings and physical climate responses (see Sections “SSTs drive large-scale dynamical responses”–“Thermodynamic responses vary regionally”) supports the validity of this linear framework for the primary modes identified. Linking SST and malaria We focus on SST as our primary climate predictor because it provides an observable, slowly evolving metric of large-scale climate modes that influence regional weather patterns. While the SST patterns we identify may be affected by other known climate modes, they serve as a practical basis for establishing physically plausible mechanistic links between remote climate forcing and local malaria-relevant environmental conditions. To isolate longer-term variability, all malaria PCs are low-pass filtered using a 24-month Butterworth filter. This allows us to investigate how deviations from normal climate conditions, rather than just seasonal patterns, relate to malaria outbreaks. Low-pass filtering of the data, rather than focusing on a specific season, also allows us to properly track interannual variability by accounting for subtle shifts in the seasonality of malaria in some districts. SSTs are filtered using a 60-month cutoff to focus on the climate signal driving atmospheric circulation. We then calculate the Pearson correlation coefficient between every SST grid cell and PC1cases/PC2cases. Based on this result, two ocean regions are selected for further analysis (see Fig. 2). These regions correspond to the strongest spatially-coherent SST-malaria correlations. All correlations reported in the text meet or exceed the threshold for statistical significance at the 95% confidence level, with coefficients greater than 0.7 and effective degrees of freedom equal to or greater than 6 (assessed following low-pass filtering). Investigating the SST-malaria mechanism To investigate the atmospheric mechanisms linking SSTs to malaria variability, we conduct spatially-resolved regression analyses, using SSTs as the predictor. We calculate anomalies for gridded climate variables (geopotential height, precipitation, soil moisture, temperature, vertically integrated moisture flux divergence, and wind) and regress them onto area-weighted average SST anomalies from the regions identified through analysis described in 2.4: tropical Atlantic (8°S-2°N, 40°W-10°W) and Indian Ocean (26°S-2°N, 64°E-92°E) regions. Predictor (SST) time series are normalized to unit variance, meaning the regression coefficients can be interpreted as changes in atmospheric response per standard deviation change in SST. Regressions are performed over 2005–2023. This approach allows us to understand how regional SST variability causes atmospheric circulation shifts that result in climatic changes over Malawi. Soil moisture-malaria relationship We hypothesize that soil moisture would be the most important predictive variable for identifying mosquito habitats, given that it reflects the combined effects of temperature, humidity, and precipitation in a single state variable. Additionally, soil moisture acts as a form of ‘time integration’ for these factors, meaning it exhibits a distinctly different temporal pattern than the individual factors it is influenced by (e.g., precipitation), particularly showing much greater temporal autocorrelation. We assess the relationship at local and national scales using a combination of spatial correlation, the construction of composite indices, and comparative analysis with precipitation and temperature. We compare malaria and soil moisture PCs to assess national-scale relationships. We create composite indices combining the dominant PCs from malaria and soil moisture EOF analyses to quantify these relationships. For this analysis, soil moisture PCs are filtered using a 36-month window to account for the longer timescale of variability of surface soil moisture. The malaria composite index is defined as the subtraction difference between PC1cases and PC2cases, where PC1cases represents the pattern of variability associated with tropical Atlantic SST and PC2cases represents the pattern associated with Indian Ocean SST. The soil moisture composite index is defined as the difference between PC2soil and PC1soil, where, inverse of malaria PCs, PC1soil represents the pattern of variability associated with Indian Ocean SST, and PC2soil represents the pattern associated with tropical Atlantic SST. The composite approach accounts for the opposing effects of the main two modes of SST variability on malaria cases. We analyze local soil moisture-malaria relationships by examining correlations between lowpass-filtered district-level malaria variability and soil moisture. We compare these results to precipitation- and temperature-malaria correlations. This provides insight into how local conditions influence disease dynamics and how related environmental factors differ in their associations with malaria, including in relation to the timescales for which they vary. Future soil moisture projections CMIP projections of temperature and precipitation have been used to model future malaria risk45–47. While the focus of this study is on the past two decades, our results lead to a clear interest in the future of soil moisture in southern Africa, which is assessed briefly here as groundwork for future studies. To develop projections of changes in suitable environmental conditions for malaria in Malawi under climate change, we analyze surface soil moisture outputs from an ensemble of global climate simulations from the Coupled Model Intercomparison Project Phase 6 (CMIP6), spanning from preindustrial conditions to the end of the 21st century. Model outputs are regridded with bilinear interpolation to place all output onto a common grid. Data are accessed via the Pangeo CMIP6 cloud-optimized catalog48 to enable efficient and reproducible analysis, and all models with surface soil moisture diagnostics available for the relevant experiments are included in the analysis. All data analyses are performed using Python, with xarray49 as a critical tool for handling data. Models are selected based on the availability of surface soil moisture output (CMIP6 variable mrsos) across four experiments: historical, SSP2-4.5, SSP3-7.0, and SSP5-8.5. The SSP scenarios range in greenhouse gas emissions trajectories, from moderate (SSP2-4.5) to extremely high (SSP5-8.5), and emissions generally increase with SSP number. All available model variants are included in the analysis. In total, there are 9 contributing models making up 96 ensemble members for the historical simulations, 69 for SSP2-4.5, 75 for SSP3-7.0, and 64 for SSP5-8.5. Soil moisture changes are evaluated both spatially and temporally. Spatial ensemble means are calculated for the late 21st century (2066–2100) and compared to modern baseline levels (1980–2014). Nationwide ensemble averages over Malawi are calculated to show historical and future soil moisture changes relative to preindustrial conditions (1850–1899). Model specifications and corresponding DOIs are provided in Supplementary Table 2. Statistics and reproducibility Analyses are performed in Python, and results are directly reproducible via scripts hosted in a public repository. Please see the Code Availability Statement for more information. Results Climate in Malawi Malawi has a tropical climate with a distinct wet season from November through March, with an average rainfall of 181 mm/month, and a dry season from June through September (average rainfall 5 mm/month) based on CHIRPS data over the study period. Temperatures are relatively high year-round but hottest in November (average 25.2 °C) and coolest in July (average 17.6 °C), and a warming trend of approximately 0.2 °C per decade since 1960 based on ERA5 data. Malaria in Malawi Malaria incidence closely follows the seasonal climate cycle, with case counts typically peaking from January to May and dropping to annual lows from July to September. The highest malaria rates (per 1000 residents) are concentrated along the northwestern shores of Lake Malawi (Fig. 1b). Nkhata Bay District, located on the lake’s border, has the highest case rate with an average of ~60 monthly cases/1000 residents per month.Fig. 1Overview of malaria incidence in Malawi from 2005 to 2023.a Monthly average number of malaria cases in each district. b Monthly malaria case rate per 1000 residents in each district. c, d First and second empirical orthogonal functions (EOFs) of malaria incidence across districts. e Time series of the first (dark red) and second (light red) principal components (PC1 and PC2) on the left y-axis, along with the total monthly malaria cases reported across the country (grey) on the right y-axis. Overview of malaria incidence in Malawi from 2005 to 2023. a Monthly average number of malaria cases in each district. b Monthly malaria case rate per 1000 residents in each district. c, d First and second empirical orthogonal functions (EOFs) of malaria incidence across districts. e Time series of the first (dark red) and second (light red) principal components (PC1 and PC2) on the left y-axis, along with the total monthly malaria cases reported across the country (grey) on the right y-axis. We focus on two dominant modes of variability resulting from the Empirical Orthogonal Function (EOF) analysis (see Section “Spatiotemporal analysis of malaria” for details), which together account for 73.1% of the total variance in malaria cases, with the first EOF (EOF1cases) accounting for 62.6% of the observed variability and EOF2cases 10.5%. The leading mode (EOF1cases) represents the coherent, nationwide oscillations in malaria incidence that affect the entire country simultaneously. The amplitude of the leading mode is variable across the nation, with maximum impact concentrated in the central-western region (Fig. 1c). EOF2cases captures a north-south seesaw pattern of malaria variability (Fig. 1d). Unlike the leading mode, which describes an oscillation where the entire country is in phase, the second mode describes transmission increasing in one region (e.g., the north) while transmission decreases in the other. The southern component of this EOF2cases dipole is stronger than the northern component, meaning the southern region sees larger swings in malaria transmission when this dipole pattern is active. The principal components (PCs) represent the time series associated with each EOF pattern. PC1cases captures a similar trend to nationwide case counts, while PC2cases, which is orthogonal to PC1cases by construction, captures the next most prominent signal in the malaria case dataset (Fig. 1e). In the following sections, we investigate the extent to which these empirically derived signals (via EOF analysis) relate to physical, remote climate influences, and examine the specific physical mechanisms linking those influences to local environmental drivers of malaria transmission. Malaria fluctuations align with tropical SSTs To explore the specific large-scale drivers that influenced malaria in Malawi from 2005-2023, we conduct a systematic, global SST-malaria correlation analysis focusing on low-frequency variability in both datasets, without assuming a priori which oceanic basins influence malaria dynamics in Malawi, and identify two different ocean regions with prominent, coherent correlations with malaria incidence in Malawi (Fig. 2): the tropical Atlantic, which shows a strong association with PC1cases (r = 0.746), and the central Indian Ocean region, which is strongly associated with PC2cases (r = 0.748). These oceanic regions have coherent and strong correlations with malaria and possess the physical significance necessary to drive regional weather patterns, making them the climatic focal point of this study. These ocean regions provide not only a basis for investigating the physical pathways that link large-scale climate variability to malaria transmission, but also a source of potential interannual predictability, given the relatively slow thermal inertia of oceans.Fig. 2Relationship between malaria incidence in Malawi and SST.a Correlation between SSTs and PC1cases, with a black box surrounding the tropical Atlantic region [8°S, 2°N, 40°W, 10°W] selected for comparisons with PC1cases. b Low-pass filtered and normalized time series of PC1cases and tropical Atlantic SST. c Correlation between SSTs and PC2cases, with a black box around the Indian Ocean region [26°S, 2°N, 64°E, 92°E] selected for comparisons with PC2cases. d Corresponding low-pass filtered and normalized time series of PC2cases and Indian Ocean SST. Relationship between malaria incidence in Malawi and SST. a Correlation between SSTs and PC1cases, with a black box surrounding the tropical Atlantic region [8°S, 2°N, 40°W, 10°W] selected for comparisons with PC1cases. b Low-pass filtered and normalized time series of PC1cases and tropical Atlantic SST. c Correlation between SSTs and PC2cases, with a black box around the Indian Ocean region [26°S, 2°N, 64°E, 92°E] selected for comparisons with PC2cases. d Corresponding low-pass filtered and normalized time series of PC2cases and Indian Ocean SST. SSTs drive large-scale dynamical responses To investigate potential causality underlying these statistical relationships, we examine how SST anomalies in these regions influence regional climate factors that shape mosquito population dynamics and ultimately drive malaria transmission. SST anomalies in the identified regions generate large-scale surface pressure redistributions that propagate through well-defined atmospheric wave sequences, directly influencing environmental conditions for malaria in Malawi via downstream impacts on weather patterns and soil moisture. Here, we diagnose those observed patterns with a focus on the impacts in southeastern Africa, which is examined in more detail in terms of Malawian hydrometeorology in Section “Thermodynamic responses vary regionally”. To understand how these large-scale atmospheric anomalies translate to regional impacts, we turn to the upper tropospheric (300 mb) geopotential height field. This level is ideal for diagnosis of large-scale, wave-like atmospheric responses to tropical SST forcing due to lack of frictional effects from the surface. Regression of 300-mb geopotential height anomalies onto the index of tropical Atlantic SST reveals the presence of a Rossby wave train propagating across the South Pacific, South Atlantic, and Southern Oceans (Supplementary Fig. 1). This wave train is indicative of the kind of far-reaching atmospheric response that can influence climate conditions in Malawi through geostrophic advection of temperature and moisture. The Rossby wave prevails throughout the troposphere, with similar wave patterns near the surface (Supplementary Fig. 1b). Despite being closer to the surface, the 850 mb wind response still closely reflects geostrophic balance. Focusing on southeastern Africa, the wave train drives an anticyclonic (high pressure) system in the lower troposphere over the Mozambique Channel southeast of Malawi, bringing easterly (westward flowing) winds over Malawi (Supplementary Fig. 1c). These winds flow over Lake Malawi and southeastern Malawi with enhanced moisture flux convergence over the mountainous eastern highlands, contributing to increased precipitation. There is also a distinct wave-like atmospheric response to Indian Ocean warming in the upper troposphere, characterized by a strong high-pressure center in the Southern Ocean directly southwest of Chile and a corresponding strong low-pressure center to the southeast of Argentina. Near the surface, a weaker sequence of adjacent anomalous pressure centers stretches horizontally across the Indian Ocean (Supplementary Fig. 1d). This wave train originating in the central Indian Ocean results in a strengthened 850 mb subtropical ridge (high pressure center) positioned south of the Mozambique Channel. A weaker high extends northward over Malawi and the Mozambique Channel, driving southerly winds that curve eastward over the center of the country due to the pressure gradient orientation around the ridge (Supplementary Fig. 1e). Additionally, northwesterly wind anomalies flow over Malawi’s mountainous northern regions, where there are notable increases to moisture flux convergence and precipitation. Thermodynamic responses vary regionally The large-scale wave patterns described above show that distant SSTs influence atmospheric circulation over Malawi. These circulation changes are reflected in moisture and temperature conditions, which directly affect mosquito population dynamics and, consequently, malaria transmission. Tropical Atlantic warming induces negative vertically integrated moisture flux divergence anomalies across the country, indicating areas of moisture convergence, which are favorable for increased rainfall via the atmospheric moisture budget50. The response is strongest over the mountainous regions, suggesting that topographic uplift of moist air from the nearby ocean is increasing convergence (Fig. 3a). The moisture-convergent environment coincides with widespread precipitation increases up to +55 mm/month per standard deviation (SD) increase in SST. On average, precipitation increases by about +23.8 mm per SD increase in SST, over 13% of the typical wet season peak rainfall in the country (Fig. 3b).Fig. 3Hydroclimate variable regressions onto tropical Atlantic and Indian Ocean SST anomalies.Climate variables onto tropical Atlantic (TA) SST anomalies: a vertically integrated moisture flux divergence (vimfd), b precipitation, c 2 m air temperature, and d surface soil moisture percentile. e–h Corresponding regressions onto Indian Ocean (IO) SST anomalies for the same climate variables. Ocean basin bounding boxes (TA and IO) are defined in Section “Malaria fluctuations align with tropical SSTs”. All regressions are performed using anomalies regressed onto anomalies, with SST predictor time series normalized to unit variance. Regression coefficients, therefore, represent climate response changes per standard deviation of SST. Hydroclimate variable regressions onto tropical Atlantic and Indian Ocean SST anomalies. Climate variables onto tropical Atlantic (TA) SST anomalies: a vertically integrated moisture flux divergence (vimfd), b precipitation, c 2 m air temperature, and d surface soil moisture percentile. e–h Corresponding regressions onto Indian Ocean (IO) SST anomalies for the same climate variables. Ocean basin bounding boxes (TA and IO) are defined in Section “Malaria fluctuations align with tropical SSTs”. All regressions are performed using anomalies regressed onto anomalies, with SST predictor time series normalized to unit variance. Regression coefficients, therefore, represent climate response changes per standard deviation of SST. Surface temperatures respond by rising slightly but uniformly, with an average increase of about +0.23 °C per SD increase in SST (Fig. 3c). Soil moisture responds with positive anomalies throughout the country, with maximum values along the western coast of Lake Malawi. Overall, anomalously warm tropical Atlantic SSTs drive increased precipitation and slightly warmer conditions in Malawi. The net effect on surface water, which is typically favorable for larval Anopheles mosquito habitat and determined by the balance between precipitation, evaporation, runoff, and other factors, is captured by soil moisture anomalies (Fig. 3d), showing wetter soils. Indian Ocean SST anomalies produce a distinctly different regional hydroclimate response, highlighting a more even balance between precipitation and evaporation. The precipitation response to warming Indian Ocean SSTs varies spatially, with decreased precipitation west of Malawi, moderate increases in precipitation over central Malawi (average of approximately +7.7 mm per SD increase in SST), and substantial increases over the northern districts, Chipita and Karonga (up to +73.7 mm per SD increase in SST), influenced by increased convergence driving rainfall (Fig. 3e, f). The Indian Ocean exerts a stronger temperature influence over Malawi than the tropical Atlantic, with an average increase in temperature of about +0.66 °C per SD increase in SST, nearly three times the temperature change observed under tropical Atlantic influence. This surface air temperature response is primarily driven by horizontal temperature advection. When the Indian Ocean SSTs are warm, there is a pronounced temperature gradient off the eastern coast of southern Africa, with warmer conditions southeast of Malawi and cooler conditions over the country itself (Supplementary Fig. 2). The wind response to Indian Ocean warming (highlighted in Supplementary Fig. 1f and superimposed in Fig. 3e) therefore draws warmer air toward Malawi from the south, effectively raising surface air temperatures across the country. Although Indian Ocean SST warming is associated with increased precipitation over Malawi, the effect on soil moisture is largely negative (Fig. 3h). This is likely the result of elevated temperatures drying out soils through evaporation that outpaces precipitation increases. These drier, hot conditions are typically unfavorable for mosquitoes. The exception is in the northern region, where increases in precipitation associated with warm Indian Ocean SSTs dominate soil moisture changes, highlighting important spatial variability in the remotely driven surface climate signal within Malawi. These results reveal fundamentally different responses to two oceanic forcing regions: warm tropical Atlantic SST anomalies drive warm, moist conditions favorable for mosquitoes, while warm Indian Ocean SST anomalies drive hot, dry conditions that suppress mosquito breeding and malaria transmission. Given that the two ocean basins’ empirical relationships with malaria are shown to be separable through EOF analysis, their combination may prove a powerful tool for understanding and predicting malaria dynamics in Malawi. Cases reflect local soil moisture levels To understand how large-scale climate drivers influence local malaria dynamics, we examined key environmental variables that affect mosquito breeding and survival: soil moisture, precipitation, and temperature. Among these, soil moisture stands out as the most consistent and spatially coherent correlate of malaria. Soil moisture levels within a district correspond well to malaria incidence within that district, with correlation coefficients reaching up to 0.8 between the low-frequency variability of malaria incidence and that of soil moisture in several districts. The strongest relationships are in the central districts along Lake Malawi and throughout the southern region (Fig. 4a).Fig. 4District-level relationships between soil moisture, precipitation, temperature, and malaria incidence.a Pearson correlation coefficients between soil moisture and malaria incidence for each district, where soil moisture values are spatially averaged within district boundaries. b and c Same as a but for precipitation and temperature. Soil moisture, precipitation, and temperature data are low-pass filtered with a 36-month window, and malaria data are low-pass filtered with a 24-month window before correlation analysis. District-level relationships between soil moisture, precipitation, temperature, and malaria incidence. a Pearson correlation coefficients between soil moisture and malaria incidence for each district, where soil moisture values are spatially averaged within district boundaries. b and c Same as a but for precipitation and temperature. Soil moisture, precipitation, and temperature data are low-pass filtered with a 36-month window, and malaria data are low-pass filtered with a 24-month window before correlation analysis. While SST-driven atmospheric circulation patterns produce coherent precipitation and temperature responses across Malawi (Fig. 3b, f), the relationships between these factors and malaria incidence are weak and spatially inconsistent (Fig. 4b) because they do not account for the land-surface hydrology, whereas soil moisture effectively integrates rainfall, evaporation, and runoff to better represent the actual environmental suitability for vector breeding. Correlations between precipitation and malaria incidence are positive in the central region but tend to be negative in the northernmost and southernmost districts (Fig. 4b). Temperature-malaria correlations do not show a coherent spatial pattern across the country (Fig. 4c). The lack of coherence in these relationships suggests that temperature and precipitation effects may be impacted by additional local factors. Soil type is likely one such mediating factor. While precipitation determines the input of water into a system, it does not determine how much moisture is retained at the surface—soil texture plays a critical role in this process51. Malawi has a diverse terrain consisting of several different major soil textures (Supplementary Fig. 3), which impact how precipitation water is stored on the surfaces or becomes runoff. Soil types respond differently to heavy rainfall, affecting how much water stays on the surface or runs off. Cambisols, for example, experience high soil erosivity during rainfall events with an advanced intensity pattern, where peak rainfall occurs early, making them prone to surface runoff52. These soils dominate the central lakeshore and southern regions, notably along mountainous areas where steep slopes increase susceptibility to erosion. High, advanced intensity precipitation events might reduce transmission in Cambisol-dominated regions as surface runoff flushes Anopheles mosquito larvae from aquatic habitats, contributing to the negative correlations observed between precipitation and malaria incidence in the southern part of Malawi, where Cambisols are common. Soil moisture and malaria PCs align While the strength of soil moisture-malaria relationships varies across districts, aggregating to the national level reduces local disparities and reveals clear overarching patterns. Again, at the national scale, soil moisture has the strongest relationship with malaria incidence in this study, with the soil PC composite index (see Section “Soil moisture-malaria relationship”) explaining 58% of the variance in the malaria composite index (Fig. 5e). PC1cases, which is connected to tropical Atlantic SSTs, is positively related to soil moisture variability across all of Malawi, with notably high correlations along the mountainous regions bordering Lake Malawi (Fig. 5a). This pattern corresponds to tropical Atlantic Ocean warming driving wet conditions in Malawi that coincide with increases in malaria transmission (see Fig. 3a–d for comparison). Conversely, soil moisture is inversely correlated to PC2cases across most of Malawi, with the northern tip of the country having the opposite (positive) relationship with PC2cases (Fig. 5b). This pattern corresponds to Indian Ocean SST warming drying soils over central and southern Malawi and increasing soil moisture in northern Malawi (Fig. 3h). Overall, Malawi’s malaria variability appears to be driven by oceanic forcing, where tropical Atlantic warming drives wetter soils and increases in malaria incidence, and Indian Ocean warming drives drier soils and reductions to incidence.Fig. 5Relationships between soil moisture and malaria variability.a Spatial correlation between 36-month low-pass filtered soil moisture and PC1cases. b Same as a but correlated with PC2cases. c Relationship between PC1cases and PC2soil, both associated with tropical Atlantic SST variability. d Relationship between PC2cases and PC1soil, both associated with Indian Ocean SST variability. e Composite indices comparing (PC1-PC2) malaria and -(PC1-PC2) soil, representing tropical Atlantic variability minus Indian Ocean variability for malaria and soil moisture, respectively. Relationships between soil moisture and malaria variability. a Spatial correlation between 36-month low-pass filtered soil moisture and PC1cases. b Same as a but correlated with PC2cases. c Relationship between PC1cases and PC2soil, both associated with tropical Atlantic SST variability. d Relationship between PC2cases and PC1soil, both associated with Indian Ocean SST variability. e Composite indices comparing (PC1-PC2) malaria and -(PC1-PC2) soil, representing tropical Atlantic variability minus Indian Ocean variability for malaria and soil moisture, respectively. As anticipated by the close relationships between soil moisture and malaria across Malawi, as shown above (Fig. 5a, b), their PCs are also closely correlated. We focus on the first two PCs of malaria and soil moisture to capture the two largest, orthogonal patterns of variation. PC1soil captures the Indian Ocean-driven variability, while PC2soil captures the tropical Atlantic-driven patterns. The combined influence of both oceanic regions is revealed through the difference between these cross-correlated components (Fig. 5e). The composite approach accounts for the opposing effects of the tropical Atlantic and Indian Oceans on malaria variability. Subtraction of the Indian Ocean-aligned components reflects the negative influence of the Indian Ocean on malaria incidence and soil moisture. These composite indices are highly correlated, demonstrating that the combined effect of tropical Atlantic and Indian Ocean forcing on malaria is consistently detectable in soil moisture patterns. A decrease in soil moisture projected While future projections of malaria in Africa often focus on changes in temperature and precipitation45,53,54, the strong, regionally distinct relationships between soil moisture and malaria incidence highlight the value of assessing future soil moisture projections as an indicator of changes in transmission risk under changing climate conditions. Temperatures in southern Africa (including Malawi) are expected to increase throughout the century, with summer precipitation decreasing and winter precipitation showing variable or inconsistent changes in this area55. Drought intensity is also projected to increase in southern regions56. However, there is considerable uncertainty in the magnitude and spatial patterns of the hydroclimatic changes, with uncertainty increasing over time57. To provide preliminary insight into Malawi’s projected hydroclimate from a soil moisture perspective, we analyze a multi-model ensemble of CMIP6 soil moisture projections, which simulate historical soil moisture conditions that are generally consistent with reanalysis data58, across three emissions scenarios. Models show soil moisture reductions across Malawi under all three emissions trajectories evaluated in this study (SSP2-4.5, SSP3-7.0, and SSP5-8.5), consistent with previously described increases in temperature and decreases in precipitation. The drying progressively intensifies with increasing greenhouse gas emissions, from SSP2-4.5 through SSP5-8.5 scenarios. The projected drying trend in soil moisture is likely driven primarily by greenhouse gas forcing rather than the ocean basin variability identified in this study, which operates on the timescales of interannual climate variability. By the end of the century, ensemble means project soil moisture decreasing by ~6%, 9%, and 11% relative to 1850–1899 baseline levels in the SSP2-4.5, SSP3-7.0, and SSP5-8.5 experiments, respectively (Fig. 6a–c). Soil moisture levels depart from pre-industrial levels beginning in the early 2000s and continue to decrease through the end of the century. Individual model projections span a wide range, with some models forecasting soil moisture decreases exceeding 30% by 2100, while a minority simulate changes within the range of natural variability. Nearly all models agree that future soil moisture conditions will be drier than pre-industrial conditions in the southern half of Malawi over all SSP scenarios, and all but three models project a decrease in soil moisture in the northern half. The three dissenting models show insignificant changes to soil moisture in the northern region. This widespread drying trend suggests a shift in Malawi’s hydroclimate that could substantially reduce environmental suitability for malaria transmission in currently warm, moderately wet environments and increase environmental suitability in cooler, wet environments. Meanwhile, the role of internal climate variability (including that driven by the adjacent ocean basins) will remain crucial for predictions on the time scales of years to a decade. This is because short- to mid-term climate can be strongly influenced by naturally occurring shifts in ocean temperatures, which can influence weather patterns independently of long-term climate trends.Fig. 6Future changes in soil moisture in Malawi under three greenhouse gas emissions scenarios.a–c Projected soil moisture change (2066–2100 average minus 1980–2014 historical simulation average) from CMIP6 variable mrsos for a SSP2-4.5, b SSP3-7.0, and c SSP5-8.5 scenarios. Hatched grid cells indicate where 8 out of 9 models agree on the sign of change. d Ensemble mean of projected soil moisture changes in all three scenarios and historical period (1980–2014) versus preindustrial levels (1850–1899). Future changes in soil moisture in Malawi under three greenhouse gas emissions scenarios. a–c Projected soil moisture change (2066–2100 average minus 1980–2014 historical simulation average) from CMIP6 variable mrsos for a SSP2-4.5, b SSP3-7.0, and c SSP5-8.5 scenarios. Hatched grid cells indicate where 8 out of 9 models agree on the sign of change. d Ensemble mean of projected soil moisture changes in all three scenarios and historical period (1980–2014) versus preindustrial levels (1850–1899). Discussion As we show here, climate has strong ties to malaria, but it is only part of the story. Even the most advanced malaria forecasts will have a ceiling on their impact without synchronous improvements in housing, access to healthcare facilities, and economic opportunities59. Evidence shows that improved living conditions reduce malaria transmission60,61, while scaled-up interventions bring economic benefits62,63. A comprehensive malaria early warning system would therefore integrate both climatic and socioeconomic factors64. This is the subject of future work. These social, economic, and infrastructural challenges are compounded by the projected impacts of climate change. Our results indicate substantial reductions to soil moisture across Malawi by 2100 under moderate to high emissions scenarios. In some districts, these conditions may lower malaria risk, but the same climate changes driving these malaria benefits can negatively impact other aspects of life. With much of Malawi’s population employed in primarily rain-fed agriculture65, the country is highly vulnerable to climate change. In Malawi, droughts, which are projected to intensify in southern Africa throughout the century56, can reduce yields of certain crops, like maize66, in specific regions, deepen poverty, and are associated with a range of social and public health consequences, including increased malnutrition, gender-based violence, suicide rates, and sexually transmitted infections67,68. Malaria early warning systems should therefore be a part of broader development efforts, focused on reducing vulnerability to adverse health impacts of climate variability and change. Conclusion Our findings demonstrate a clear connection between natural, large-scale climate drivers and malaria incidence in Malawi, providing an essential foundation for developing climate-informed forecasting tools. We establish relationships between tropical SSTs and malaria incidence, while avoiding a usual approach of making a priori assumptions about relevant climatic drivers. This approach is particularly important given Malawi’s unique topography, including its mountainous terrain and the presence of Lake Malawi, which can modify the local impacts of large-scale climate drivers on the environmental factors relevant to mosquitoes (i.e., moisture and temperature), as well as its geographic location in relation to influential climate teleconnections. For example, Malawi is situated between ENSO wet and dry response regions, while bordering Mozambique falls within a conventional ENSO teleconnection pattern69, meaning neighboring countries within the broad region of southern Africa can be impacted differently by major modes of climate variability. While prior studies have linked malaria in other African countries to ocean-atmosphere modes such as the El Niño-Southern Oscillation and the Indian Ocean Dipole12,30,70–73, we find that malaria incidence in Malawi is driven by interannual SST variability in the tropical Atlantic and the central Indian Ocean through their downstream impacts on local climate. Although there are many complex, nonlinear relationships in the climate system (e.g., between the Indian Ocean and El Niño-Southern Oscillation)74–76, our findings indicate that SST variability in the tropical Atlantic and central Indian Ocean captures the influences most relevant to malaria variability in Malawi. Identifying these Malawi-specific climate drivers lays the groundwork for developing a national malaria early warning system. SST-based malaria early warning systems developed for other African countries have shown predictive skill at seasonal-scale lead times29,77. While the timescales over which these two SST patterns could support operational forecasting will be investigated in future work, the slowly evolving nature of SST anomalies in both regions suggests that seasonal to multi-year predictability may be achievable. At a seasonal lead time, such forecasts could plausibly inform vector control strategies like bed net distribution and indoor residual spraying, which operate on similar planning timelines and have been identified as key response actions triggered by early warning systems78. Such a tool tailored for Malawi has strong potential for uptake, given the country’s track record of adopting new malaria intervention strategies. Beyond large-scale climate drivers, a key finding of our study is that soil moisture is the key environmental determinant of malaria incidence in Malawi. This strong relationship with malaria incidence is likely due to soil moisture being a more comprehensive indicator of surface-level hydrological conditions relevant to mosquito larval habitats, as it accounts for the balance of multiple hydrological processes, including precipitation, evapotranspiration, and land surface properties. Although links between soil moisture and malaria have been established79–81, soil moisture is often overlooked due to challenges in acquiring reliable soil moisture data (e.g., sparse direct observations, challenges in models, etc.). However, the existence of satellite-based soil moisture monitoring (e.g., GRACE-DA-DM and NASA Soil Moisture Active Passive (SMAP)82) presents a unique opportunity to increase predictability of malaria on an intra-annual timescale83. Predictability of soil moisture is substantially better than precipitation on seasonal to multi-year timescales84,85, and our results demonstrate that soil moisture can be a more important factor to consider than temperature and precipitation, which are most commonly used in future projections of malaria45,53,54. Supplementary information Transparent Peer Review file Supplementary Material Transparent Peer Review file Supplementary Material