Climate change significantly impacts the transmission dynamics of vector-borne diseases like malaria and dengue, causing substantial interannual variability in epidemic sizes. While temperature and rainfall have been extensively studied, the role of relative humidity (RH) remains largely unexplored. This is particularly true in urban settings, where the unique environment and presence of vectors like *Anopheles stephensi* – a mosquito that thrives in human-built environments and doesn't rely directly on rainfall for breeding – complicates existing models developed for rural areas. *An. stephensi* poses a significant challenge to malaria control, acting as a reservoir for transmission and hindering elimination efforts. Its reported expansion across the Arabian Peninsula into the Horn of Africa further emphasizes the need to understand its population dynamics. Experimental studies have shown that temperature and RH significantly influence mosquito survival and activity, with RH levels below 60% potentially too low for effective malaria transmission. Despite this, most mathematical models of malaria transmission focus primarily on temperature, neglecting the potential impact of humidity, especially for *An. stephensi* with its relatively short lifespan. This study aims to address this gap by analyzing extensive surveillance data from Surat and Ahmedabad, two Indian cities with differing climates, to determine the relative importance of humidity, temperature, and rainfall in driving interannual variations in urban malaria transmission.

Existing literature highlights the significant interannual variability in climate-sensitive infectious diseases, emphasizing the need to identify key climatic drivers for effective disease prediction and control. While numerous studies examine the role of temperature and rainfall on vector-borne diseases like malaria and dengue, the influence of relative humidity (RH) remains understudied. Early entomological research suggests that humidity is crucial for mosquito survival and activity, impacting malaria transmission. In South Asian cities, *Anopheles stephensi*, an urban vector that breeds in artificial containers, poses a unique challenge to malaria control. Urbanization enhances disease persistence, counteracting global elimination plans. Studies have shown that specific temperature and RH ranges are optimal for *Anopheles* mosquito survival and malaria parasite development, but dynamical effects on disease incidence have not been rigorously tested, particularly for humidity. Most mathematical models of malaria transmission rely heavily on temperature data from studies with *Anopheles gambiae*, neglecting the nuances of other vectors like *An. stephensi*. This study aims to fill this knowledge gap by investigating the role of humidity in shaping urban malaria dynamics.

This study utilized extensive surveillance records of malaria cases from Surat and Ahmedabad, India, spanning from 1997 to 2014. These data, collected by the respective municipal corporations through active and passive surveillance methods, were complemented by monthly climate data (relative humidity, rainfall, and temperature) obtained from local weather stations and verified using gridded climate products. The analysis began by characterizing the temporal scales of interannual variability in malaria cases and their relationship to humidity using wavelet coherence analysis. This approach identified significant correlations between malaria cases and humidity at specific periods, suggesting a coherent relationship between the two variables. A stochastic malaria transmission model, described by a system of stochastic differential equations, was then developed to assess the relative importance of humidity, temperature, and rainfall as drivers of interannual malaria variability. The model incorporated a gamma-distributed time delay to account for the parasite development time within the mosquito vector. The model's parameters were estimated using maximum likelihood iterated filtering (MIF), a simulation-based algorithm that accounts for both measurement and process noise. Model comparisons were performed using likelihood ratio tests and the deviance information criterion (DIC). To evaluate model performance, 1000 simulations were run from each of the models, including the one incorporating the humidity, for comparison with the observed malaria data. The model’s ability to predict malaria cases for years not used in parameter estimation was tested by performing out-of-fit predictions. The MIF algorithm, with the associated filtering procedure, allowed for the continuous assimilation of new data each year, updating parameter estimates and initial conditions. This methodology allowed the researchers to assess the model’s predictive power in real-world forecasting scenarios. A permutation test was also conducted to verify that the strong correlation between malaria transmission and humidity was not simply due to seasonal confounding.

Wavelet coherence analysis revealed significant correlations between malaria cases and relative humidity (RH) at periods of approximately 2 and 4 years, predominantly in Surat and Ahmedabad respectively. A strong positive correlation (R = 0.72 for Ahmedabad, p = 0.0002; R = 0.69 for Surat, p = 0.004) was found between total malaria cases during the transmission season (August-November) and average RH in a critical pre-transmission window. Model comparisons showed that the model incorporating RH as a covariate performed significantly better than models with temperature or rainfall (p < 0.001). Simulations of the best model (incorporating RH) accurately captured the observed interannual variation in the size of seasonal outbreaks, including the pattern of large outbreaks followed by smaller ones, and the main seasonal pattern of reported cases. The model revealed differences in transmission intensity between cities, with higher transmission rates and greater variability in Surat (more humid) compared to Ahmedabad. The humidity coefficient was higher for Ahmedabad, indicating a more pronounced response to RH changes in drier conditions. A permutation test confirmed that the strong correlation between malaria transmission and humidity is unlikely to be due to seasonal confounding. Parameter estimates from the model, including the average dynamical delay between the latent and current force of infection, were consistent with empirical values of parasite developmental time within the mosquito. Out-of-fit predictions for the period 2009-2014, generated using data up to 2008, showed that the median predictions accurately captured interannual variability and that the observations fell within the 10-90% prediction intervals for most months.

The findings strongly support the hypothesis that relative humidity is a crucial driver of interannual variations in urban malaria transmission, exceeding the explanatory power of temperature in these specific Indian cities. This contrasts with many existing models, which primarily focus on temperature, highlighting the importance of considering RH in malaria forecasting and control efforts. The superior predictive ability of the RH-driven model supports its use in developing climate-based early warning systems and evaluating the effectiveness of public health interventions. The differing effects of humidity in the two cities suggest that the relationship between climate and malaria transmission can be complex and context-specific. Further research is needed to fully understand the mechanistic interactions between humidity, temperature, vector ecology, and parasite development. The study’s findings have implications for predicting the effects of climate change on malaria transmission, particularly given the projected increase in humidity in Northwest India. The persistence of *An. stephensi* in urban environments, independent of rainfall, underscores the significant role of RH as a transmission driver, challenging existing assumptions and strategies for malaria elimination.

This study demonstrates the critical and previously underappreciated role of relative humidity in driving interannual variability of urban malaria transmission in semi-arid Indian cities. The developed model accurately captured observed variability and successfully predicted future incidence, highlighting the importance of including humidity in malaria prediction and control efforts. Further research should focus on a more detailed mechanistic understanding of the interaction between humidity, temperature, vector behavior, and parasite development, and on the broader applicability of this model across diverse urban environments and disease vectors.

The study focuses on two specific cities in India and its generalizability to other geographic regions or urban environments needs further investigation. The model simplifies vector dynamics by using a distributed lag, which may not fully capture the complexities of vector population dynamics and their interaction with climate variables. The availability of data on other relevant factors, such as socioeconomic conditions and interventions, could strengthen the model and its interpretation. The findings are limited to *P. falciparum* malaria and might not fully reflect the dynamics of other malaria parasites.