Rethinking malaria seasonality: humidity-driven transmission shifts and emerging hotspots in Zambia (2009–2023)

Introduction

Climate change is increasingly recognized as one of the greatest threats to global health security. Its effects disproportionately impact countries that have contributed least to its acceleration (1–3). Health impacts include increased heat-related mortality, food and water insecurity, altered distribution of infectious diseases, and disruption of health systems (2). A large proportion of climate-related mortality is attributable to increased transmission of climate-sensitive infectious diseases (CSID), including malaria (4).

Malaria remains one of the deadliest infectious diseases worldwide, with an estimated 263 million cases and 597,000 deaths in 2023, an increase of about 11 million cases from 2022 (5). About 94% of global cases and 95% of deaths occurred in sub-Saharan Africa (6). Since 2000, malaria interventions have averted an estimated 2.2 billion cases and 12.7 million deaths. Yet, progress toward the WHO Global Technical Strategy milestones has stalled, and incidence and mortality remain more than double the 2023 targets (5, 6).

Despite significant global and national investments, endemic countries remain far from achieving elimination, highlighting the need for climate-informed strategies that address both biological and environmental drivers of transmission (7–9). Climate-driven changes in precipitation, temperature and humidity can extend periods of environmental suitability for malaria transmission (10, 11). High-burden, climate-vulnerable settings are underrepresented in the literature exploring the link between climatic shifts and disease transmission (12).

Zambia, for example, is among 21 African countries that have achieved reductions since 2015 but remains off track for the 2025 goal of a 75% reduction (5). In 2019, malaria incidence reached 340 malaria cases per 1,000 population, and parasite prevalence among children under five years rose from 24% in 2018 to 29% in 2019, signaling an upward trend (13), and deviating from the national malaria elimination target of 5 cases per 1,000 population by 2026 (8). These trends reflect overlapping global challenges, including health and humanitarian emergencies, financial crises, and climate change, all of which compound malaria transmission risks and complicate the delivery of core interventions (3, 5).

Whereas climate change reflects long-term shifts with occasional catastrophic events such as hurricanes and floods, climate variability captures short- to medium-term departures that more directly influence seasonal malaria risk, provide early warning signals for outbreak detection and response planning (14, 15). Earlier modeling studies across Zambia have shown malaria incidence peaks during the rainy season, and is influenced by short-term warming and cooling periods, expanding breeding habitats in previously low-risk areas (13, 16). Thus, climate variability can increase transmission risk and strain under-resourced health systems (3, 7).

However, there is limited district-level evidence quantifying how specific climate parameters, particularly relative humidity, rainfall and temperature, shape spatial and temporal malaria trends in Zambia. A clearer understanding of these relationships is essential for designing climate-informed malaria strategies and improving outbreak preparedness. Therefore, this study assessed 15 years of climate and malaria surveillance data to determine how climate variability influences transmission patterns across Zambia. The study also sought to identify adaptation measures that can be integrated into national malaria control efforts.

Methodology

Study design

This study constitutes the quantitative component of a mixed-methods, cross-sectional study examining the impact of climate variability on malaria incidence in Zambia. An ecological time-series design was used to investigate temporal and spatial associations between malaria incidence and key climatic variables, including rainfall, temperature and relative humidity, over a 15-year period (2009-2023).

Study setting and context

Zambia is a landlocked country in Southern Africa with a population of approximately 22 million people (17) and ranks among the 20 high-burden countries globally (5). Malaria is endemic, with varying incidence rates across provinces due to geographic, climatic, and socio-economic differences (18). Zambia has a tropical climate, with higher rainfall generally occurring in the northern parts of the country. The rainy season is from October to April. In recent years, Zambia has experienced increased flooding associated with El Niño, followed by severe droughts, including the worst drought on record in 2024 (19). These climatic extremes underscore the importance of understanding how short- and long- term variability influence malaria transmission.

Data sources

Malaria surveillance data

Monthly malaria surveillance data were obtained from the National Malaria Elimination Centre (NMEC) through the District Health Information System Version 2 (DHIS2). Data included confirmed and clinical malaria cases, malaria-related deaths, and district population estimates for the period from January 2009 to December 2023. Monthly malaria incidence was calculated as cases per 1000 population using district population estimates obtained from the Zambia Statistical Agency (ZAMSTAT).

Climate data

Climate data were obtained from the Zambia Meteorological Department at 42 manual observation stations across the country, supplemented with satellite-based datasets available online at https://www.ncei.noaa.gov/cdo-web/. Supplementary products included satellite-based rainfall products from the Climate Hazards Group Infra-Red Precipitation (CHIRPS) dataset and Japanese 55-year atmospheric reanalysis (JRA-55) data. Station data were quality-controlled and merged with satellite products using the Climate Data Tool (CDT) to improve spatial coverage for all 116 districts.

Meteorological and satellite datasets were validated at 90% confidence level using the Climate Data Tool (CDT) and aggregated to monthly district-level estimates of rainfall (mm), minimum and maximum temperatures (°C), mean temperature and relative humidity (%). Five-year rolling averages were calculated for temperature and relative humidity; while rainfall totals were aggregated over time to characterize long-term climatic trends.

Geospatial data

District boundary shapefiles were obtained from the Zambia Statistics Agency. Administrative boundaries changed from 75 to 116 districts during the study period; therefore, harmonized shapefile was used to ensure consistent spatial analysis and hotspot detection throughout the study period.

Data management and processing

Malaria and climate datasets were cleaned and managed in RStudio. Climate and malaria records were linked by district and month-year. Districts were geo-referenced using centroid coordinates to facilitate spatial analyses.

Statistical and spatial analysis

The analysis comprised four stages. First, temporal trends in malaria incidence and climatic variables were described using descriptive statistics and time-series plots. Second, spatial hotspot analyses were undertaken to identify districts with persistently high malaria incidence. Third, district level associations between malaria incidence and climatic variability were explored using Spearman correlation analyses. Finally, time series modeling was performed using Vector Autoregressive (VAR) models, Granger causality testing, Forecast Error Variance Decomposition (FEVD), and Seasonal Autoaggressive Integrated Moving Average (SARIMA) models to examine temporal relationships and forecast malaria incidence.

Trend analysis and hotspot detection

Geo-spatial visualizations were created by overlaying malaria incidence on district shapefiles in QGIS. Three-year moving averages of malaria incidence were calculated to smooth annual fluctuations in malaria incidence. Persistent hotspots were identified using SATSCAN and compared with mapped incidence patterns to confirm spatial consistency.

Climate trend analysis

Climate trends were assessed using rainfall, temperature, and relative humidity data extracted for each district. Rainfall extremes were characterized using the Expert Team on Climate Change Detection and Indices (ETCCDI) metrics implemented in the Climate Data Tool (CDT) (Table 1). Monotonic trends were assessed using the non-parametric Mann-Kendall test.

Table 1. List of expert teams on climate change detection and indices (ETCCDI) extreme precipitation indices.

PCP, Precipitation; RR, Rainfall.

Correlation analysis

Spearman’s rank correlation coefficients were calculated separately for three time periods (2009 – 2013, 2014 – 2018, and 2019 – 2023) to evaluate associations between climatic variables and malaria incidence. Correlation categories were mapped in QGIS to identify districts with consistent climatic influence on malaria incidence overtime. Districts with correlation coefficients above 0.3 across multiple time periods were classified as climate-sensitive hotspots. This threshold is consistent with commonly accepted interpretations of correlation strength in medical research, where coefficients around 0.30 indicate a moderate association (19).

Time-series modeling

Monthly malaria incidence data and climatic variables were analyzed using time-series methods. Descriptive statistics were used to summarize the distribution of malaria incidence and climatic variables. Continuous variables were summarized using means, standard deviations, medians, interquartile ranges (IQRs), minimums, and maximums, while seasonal patterns were explored using monthly autocorrelation boxplots and time-series plots. Variance stationarity was evaluated using the Box–Cox transformation, while stationarity in the mean was assessed using the Augmented Dickey–Fuller (ADF) unit root test. Where non-stationarity was detected, seasonal differencing with a periodicity of 12 months was applied to remove seasonality and achieve stationarity. Seasonal Autoregressive Integrated Moving Average (SARIMA) models were developed following the Box–Jenkins approach. Candidate models were identified through examination of the autocorrelation function (ACF) and partial autocorrelation function (PACF) plots of the stationary series. Competing models were compared using the Akaike Information Criterion (AIC), and parameter estimation was performed using maximum likelihood methods. Model adequacy was assessed through residual diagnostics, including examination of residual ACF and PACF plots, quantile–quantile plots, and the Ljung–Box test to confirm the absence of residual autocorrelation.

Dynamic relationships between malaria incidence and climatic variables were examined using Vector Autoregressive (VAR) models. Lag length was selected using the Final Prediction Error criterion together with the Akaike Information Criterion, Schwarz Bayesian Criterion, Hannan–Quinn Criteria. Prior to VAR modeling, all time-series were tested for stationarity using the ADF test and transformed where necessary. Pre-whitening procedures were applied to remove autocorrelation within individual series before evaluating cross-correlations between malaria incidence and climatic variables.

Twelve lags were added to each variable in each equation to create a VAR model that included the monthly malaria and climate variables (Supplementary Table 1). Lag-length selection was guided by the Final Prediction Error (FPE) criterion, which indicated that a 12-month lag sufficiently captured climatic and epidemiological dynamics (20, 21). The VAR approach mitigated spurious correlations driven by autocorrelation or shared time trends.

Granger causality tests were performed to assess whether historical values of rainfall, temperature, and relative humidity improved prediction of malaria incidence. The dynamic effects of climatic shocks on malaria incidence were further examined using orthogonalized impulse response functions (IRFs), while Forecast Error Variance Decomposition (FEVD) was used to quantify the contribution of each climatic variable to variation in malaria incidence over time, while orthogonalized impulse response functions were used to examine the effects of climatic shocks on malaria transmission (22).

All statistical tests were two-sided, and a p-value <0.05 was considered statistically significant. Analyses were conducted using R statistical software (R Foundation for Statistical Computing, Vienna, Austria).

Results

Temporal trends of malaria incidence and climate variability

Between January 2009 and June 2023, malaria incidence in Zambia showed a strong and predictable seasonal pattern closely aligned with rainfall and humidity (Figure 1). Rainfall exhibited distinct annual peaks that coincided with increases in relative humidity, while malaria incidence consistently increased shortly after these peaks. Temperature showed far less seasonal variability, remaining comparatively stable throughout the study period. Although the seasonal pattern persisted, malaria incidence peaks became more pronounced from 2016 onward, with particularly elevated levels between 2021 and 2022, despite continued implementation of malaria control interventions.

Cumulative malaria incidence ranged from 185,892 cases in 2016 to 1,326,442 cases in 2020, with a mean annual incidence of 497,694 cases. Rainfall ranged from 0.08 mm to 296.62 mm (mean 89.21 mm), temperatures ranged from 16.77 °C to 28.35 °C (mean 22.26 °C), and relative humidity ranged from 22.87% to 81.82% (mean 54.32%). All-time series were stationary following transformation and differencing, as confirmed by the Augmented Dickey-Fuller (ADF) unit root test (p < 0.05).

Malaria incidence was highest between February and April, and lowest during the dry season (July to October) (Figure 2). Rainfall and relative humidity peaked from January to March, with median rainfall exceeding 150 mm and humidity exceeding 65%, preceding increases in malaria incidence. Temperatures showed relatively limited seasonal variation, ranging between approximately 20 °C and 25 °C throughout the year.

Spatial distribution of malaria hotspots (2009–2023)

District-level mapping identified persistent malaria hotspots in northern and north-eastern Zambia, particularly in Luapula, Muchinga, Northern, and parts of Eastern provinces (Figure 3). In contrast, Lusaka, Southern, and Western Provinces consistently reported lower incidence Over the 15-year period, high burden districts intensified and expanded geographically, with several districts in Luapula and Muchinga exceeding 1,000 cases per 10,000 population between 2018 and 2023.

Nationally, incidence increased from a maximum of 118 cases per 10000 persons during 2009–2011 to 132 per 10,000 cases during 2021-2023. Kasempa District (Northwestern Province) had the highest burden, rising from 505.2 per 1000 in 2017 to 1,317 per 1000 in 2023. Meanwhile, Monze (Southern Province) remained the lowest-burden district, with the incidence increasing from 51.4 per 1000 in 2017 to 101.1 per 1000 in 2023.

Correlation between climatic variables and malaria incidence (2009–2023)

Rainfall

Rainfall showed positive correlations with malaria incidence in districts with consistently high precipitation, particularly in northern and north-western Zambia (Figure 4). However, correlations varied across districts with some high-rainfall areas demonstrating weaker associations.

Relative humidity

Relative humidity demonstrated the strongest and most consistent statistical associations with malaria incidence (Figure 5). Correlation coefficients greater than or equal to 0.3 were observed in 60 districts across all three study periods, with coefficients above 0.9 in Serenje and Shiwang’andu.

Temperature

Temperature demonstrated weaker and more localized associations than rainfall and relative humidity. The strongest correlation (r = 0.661) was observed in Chama District (Eastern Province). Sixteen districts showed significant correlations across all timeframes, and 32 showed significant correlations in at least two periods (Figures 6, 7). These districts were predominantly located within valley systems in Eastern, Muchinga and Southern Provinces.

Composite climate correlations

A composite analysis identified nine districts, (Kalabo, Kaoma, Zambezi, Kabompo, Kitwe, Luanshya, Lufwanyama, Masaiti, and Milenge), with correlation coefficients exceeding 0.3 with at least two climatic variables across multiple time frames (Figure 8). A broader band of 70 districts in northern Zambia showed significant correlations between relative humidity and rainfall, indicating extensive spatial overlap between these climatic variables (Figure 9).

Time-series modeling of climate-malaria relationships (VAR and Granger)

Granger causality tests were performed to evaluate whether past values of climatic variables could significantly improve the prediction of malaria incidence. The results indicated statistically significant predictive relationships between all climatic variables and malaria incidence (Table 2).

Table 2. Granger causality tests between climatic variables and malaria incidence in Zambia, 2009–2023.

Relative humidity showed the strongest predictive relationship (F = 4.545, p < 0.001), followed by temperature (F = 4.275, p < 0.001) and precipitation (F = 3.080, p < 0.001). These findings indicate that historical climatic conditions significantly improved prediction of future malaria incidence.

A Vector Autoregressive (VAR) model incorporating monthly malaria incidence, rainfall, temperature and relative humidity was fitted using a 12-month lag structure. Lag selection was guided by the Akaike Information Criterion, Schwarz Criterion, Hannan–Quinn Criterion and Final Prediction Error, all of which supported a VAR (12) model (Supplementary Table 1).

Forecast error variance decomposition and impulse response

The Forecast Error Variance Decomposition (FEVD) analysis showed that previous malaria incidence explained 74.9% of future variation in malaria (Figure 10). Among the climatic variables, relative humidity accounted for 29.3% of forecast variance, followed by rainfall (13.3%) and temperature (12.8%).

Impulse response analysis demonstrated delayed responses of malaria incidence following climatic shocks (Figure 11). Temperature shocks produced the largest positive response nine months after the initial change (around October) while rainfall and relative humidity demonstrated delayed positive effects occurring later in the annual transmission cycle.

Forecasting malaria incidence (SARIMA)

The final SARIMA (1,0,1)(0,1,2) model adequately captured the temporal structure of the malaria time-series. Residual diagnostics indicated good model fit, with no significant residual autocorrelation (Ljung–Box Q = 8.37, df=21, p=0.993). Model performance was further supported by a Mean Absolute Percentage Error (MAPE) of 6.84%, Mean Absolute Error (MAE) of 240.44, Root Mean Square Error (RMSE) of 339.21 and minimal residual autocorrelation (ACF1 -0.003), indicating good forecast accuracy. Forecasting analysis projected that the characteristic seasonal pattern of malaria transmission would persist through 2030, with incidence consistently peaking between March and May, and declining during dry season (July – September) (Figure 12). Forecast uncertainty increased progressively over time, as reflected by widening 95% prediction intervals, although the seasonal transmission pattern remained stable throughout the forecast period. Model diagnostics indicated that the residuals approximated white noise, supporting the adequacy of the fitted SARIMA model for forecasting future malaria incidence.

Discussion

The study provides new national-level evidence linking long-term climate variability to malaria transmission dynamics in Zambia. Over 15 years of surveillance, malaria incidence showed strong seasonal alignment with rainfall and humidity, with both variables consistently preceding increases in transmission. Temperature emerged as a more localized but important driver, particularly in valley and low-lying districts, indicating that multiple interacting factors shape climatic suitability for malaria. By integrating temporal, spatial, correlation and time series analyses, the study demonstrated that climate influences on malaria are both time-sensitive and geographically specific, underscoring the need for climate-informed elimination strategies.

Contrary to anticipated declines projected under Zambia’s Malaria Elimination Strategic plan (8), malaria incidence has not declined steadily. Instead, peaks have been reported, and overall incidence has increased toward 2023 compared to 2009. This reversal suggests a complex interplay between climatic and non-climatic pressures, including insecticide resistance and possible antimalarial resistance, rapid population growth, urbanization, and changes in rainfall and humidity that sustain mosquito breeding. Health system disruptions, particularly during the COVID-19 pandemic, may also have undermined consistent delivery of interventions. Together, these findings highlight the need to strengthen climate-informed surveillance, optimize vector control strategies and enhance program adaptability to protect gains made over the last decade. However, although climatic variables explained important temporal and spatial variation in malaria incidence, the observed patterns almost certainly reflect interactions between climate and health system, demographic and ecological factors that were not explicitly modelled in this study.

Among all climatic indicators, relative humidity emerged as the most consistent and strongest predictor of malaria incidence across regions and time frames. Predictive modeling suggests that humidity and minimum temperature, rather than rainfall alone, are likely to play an important role in future malaria transmission, with projected increases in malaria cases through 2030 even under current intervention levels, if the current relationship persists. Relative humidity influences mosquito survival, host-seeking behavior and parasite development, allowing transmission to persist even where rainfall has ceased (3). This biological mechanism may explain why humidity demonstrated stronger associations than rainfall alone. Although the 2024 drought occurred after the study period and was not included in the analysis, the persistence of malaria transmission reported during this event is consistent with our finding that humidity and temperature, rather than rainfall alone, may sustain transmission under a wider range of climatic conditions.

The geographic distribution of malaria risk is also evolving. Historically high-burden areas were reported in Luapula, Northern, Muchinga, and Eastern provinces and have remained deeply affected, while new hotspots have emerged in Northwestern, Copperbelt, and Western provinces. These expanding patterns correspond with changes in rainfall distribution, extended humidity peaks and extreme weather events (7, 13). Granger causality and variance decomposition further confirm that climate variability is a key driver of seasonal extension, emphasizing the need to move beyond rainfall-based early warning systems towards multi-indicator climate surveillance (19, 23, 24).

Implications for malaria control and elimination strategies

The findings have direct implications for Zambia’s goal of malaria elimination by 2026. The observed extension of the transmission season from December to June, suggests that current interventions, such as indoor residual spraying (IRS), distribution of long-lasting insecticide-treated nets (LLIN), and seasonal malaria chemoprevention (SMC), may need to begin earlier, last longer, or be stratified geographically. Districts showing persistent climate–malaria correlations provide a strong evidence base for adaptive subnational targeting, a strategy proven effective in other endemic settings like Burkina Faso (25) and Madagascar (26).

The marked geographical heterogeneity observed across districts suggests that climate-informed subnational targeting should complement existing malaria stratification approaches (14, 27). Districts experiencing prolonged humidity and favorable ecological conditions, such as Nchelenge, may require different intervention timing and intensity from lower-risk areas (28). These comparisons emphasize that climate-driven risk must be considered alongside program performance and reflect sub-national geographical peculiarities.

Emerging malaria vaccines such as RTS,S and R21 may represent an additional tool for climate-adaptive malaria control. Evidence from Madagascar suggests that strategically timed vaccination could mitigate climate-driven disruptions to malaria programs, particularly where climatic variability alters the timing and duration of transmission seasons (26). Growing evidence suggests that strategically timed vaccination can reduce clinical malaria, severe disease and malaria-related deaths, particularly in settings where climatic variability alters the timing and duration of transmission seasons (25, 29). The present study found that humidity and other climatic factors may influence malaria transmission beyond traditional rainfall-driven patterns, highlighting the need for locally tailored intervention strategies. As malaria vaccines are introduced, aligning vaccination schedules with climate-informed seasonal risk patterns and local epidemiological conditions may help maximize their impact (24, 26, 30, 31). Integrating climate surveillance with health information systems could further support more responsive (26) and targeted deployment of malaria vaccines in areas experiencing shifting transmission dynamics.

Strengths and limitations

Several limitations should be considered when interpreting the findings. First, the study focused primarily on climatic drivers and was unable to account for several non-climatic factors known to influence malaria transmission, including intervention coverage, insecticide resistance, antimalarial resistance, urbanization, population movement, demographic change and disruptions associated with the COVID-19 pandemic. Consequently, observed associations should not be interpreted as reflecting the effects of climate variability alone. Additionally, the analyses were conducted using aggregated district-level data and therefore cannot determine how climatic factors influenced malaria risk at the individual level. Moreover, it was not possible to model the effects of discrete climate extremes such as flash floods and heat waves, as the analysis relied primarily on routinely available meteorological and satellite-derived climate data. Although satellite products were validated against meteorological station observations, some measurement uncertainty may remain in areas with limited ground-based data. Finally, the forecasting analyses assume that historical relationships between climate variables and malaria incidence remain relatively stable over time. The projections should therefore be interpreted as indicative future scenarios rather than precise predictions of future malaria burden.

However, the strengths of this study include the use of a long-term national dataset, integration of multiple climate data sources validated against ground stations, and the application of complementary analytical methods. The combination of hotspot detection, correlation analysis, time-series forecasting, causality testing and forecast error variance decomposition provided a comprehensive assessment of climate-malaria relationships across Zambia. Spatial analysis across harmonized district shapefiles versions also enabled consistent mapping despite boundary changes over time.

Conclusions and recommendations

Relative humidity emerged as a stronger and more consistent predictor of malaria incidence in Zambia than rainfall alone, while minimum temperature shifts also influenced transmission patterns, particularly in valley districts. These findings suggest that malaria risk is becoming increasingly shaped by a broader range of climatic factors than traditionally recognized, contributing to shifts in both the geography and seasonality of transmission.

As climatic variability continues to alter malaria risk, these findings support integrating climate adaptation as a core component of malaria elimination strategies rather than a parallel activity (31–33). Strengthening climate-informed surveillance, improving the geographical targeting of interventions, and integrating emerging tools such as malaria vaccination into climate-responsive planning may help maintain progress towards elimination. The findings also highlight the importance of moving beyond rainfall-based assumptions and incorporating humidity and other climatic indicators into malaria early warning and decision-support systems.

Future research should evaluate how climatic variability interacts with mosquito ecology, intervention coverage and emerging malaria control tools to inform adaptive elimination strategies under changing climatic conditions.

Generative AI statement

The author(s) declared that generative AI was used in the creation of this manuscript. The tool was used for language refinement and copy-editing to improve clarity and readability. The tool was not used for data analysis, interpretation, or the generation of scientific findings. All analyses, interpretations, and conclusions were developed independently by the authors.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fitd.2026.1842716/full#supplementary-material

Residual diagnostic plots for the fitted SARIMA model. The residuals show no significant autocorrelation, approximate normality, and Ljung–Box test p-values above 0.05 across all lags, indicating that the model adequately captures the temporal structure of the malaria incidence series and is suitable for forecasting. Supplementary material at the source

Supplementary material Supplementary material at the source

Supplementary material Supplementary material at the source

Footnotes

Edited by: Emanuele Nicastri, National Institute for Infectious Diseases Lazzaro Spallanzani (IRCCS), Italy

Reviewed by: Enrico Perugini, University of Rome Tor Vergata, Italy

Mohammed Baba-Adamu, Yobe State University, Nigeria

Footnotes

CDT, Climate Data Tool; ETCCDI, Expert Team on Climate Change Detection and Indices; FPE, Final Prediction Error Test; FEVD, Forecast Error Variance Decomposition; IRS, Indoor Residual Spraying; ISDR, Integrated Disease Surveillance and Response; LLIN, Long Lasting Insecticide-treated Nets; NMEC, National Malaria Control Centre; SMC, Seasonal Malaria Chemoprophylaxis; VAR, Vector Auto Regressions; WHO, World Health Organization; ZNPHI, Zambia National Public Health Institute; ZamSTAT, Zambia Statistics Agency.