Accounting for spatial variation in climatic factors predicts spatial variations in mosquito abundance in the desert southwest

Accounting for spatial variation in climatic factors predicts spatial variations in mosquito abundance in the desert southwest | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Accounting for spatial variation in climatic factors predicts spatial variations in mosquito abundance in the desert southwest Kayode Oshinubi, Crystal M. Hepp, Ye Chen, Eck Doerry, Nicole Busser, and 5 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7595227/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 11 Mar, 2026 Read the published version in Parasites & Vectors → Version 1 posted 9 You are reading this latest preprint version Abstract Background : Mosquitoes are vectors for diseases globally, making development of models that better explain mosquito abundances imperative. Mosquito population dynamics are particularly sensitive to local weather conditions, and mosquito-borne disease outbreaks can be spatially concentrated. There is a need for improved modeling studies to address whether spatial variation in disease outbreaks is driven by spatial variation in weather conditions, especially in dry and hot environments. In this study, we build a climate-driven model of mosquito population dynamics and compare whether predictions of mosquito abundance at the county scale are improved by accounting for sub-county climate variation. Methods: Using a 5-year time series of weekly mosquito abundance data collected for each zip code in Maricopa County, USA, we assess how local variation in climate can explain and predict mosquito population dynamics. We built a mechanistic model of mosquito population dynamics influenced by daily temperature and 30-day accumulated precipitation. We grouped zip codes based on similar patterns of temperature and precipitation using functional clustering. We compared two approaches: one using county-level average climate and another using data from the identified climate clusters. We use MCMC to fit the mechanistic model using averaged climate data in each cluster, then compare the modeling fit to observed data of the county-level model to the model based on climate-based clusters. Results : Simple, climate-forced modeling accurately estimates detailed mosquito abundance trajectories throughout a five-year period. Modeling mosquito abundances in the sub-county spatial clusters demonstrated that the same effects of temperature and precipitation on population growth rates could explain small-scale changes in mosquito populations. However, when we aggregate the sub-county model fits to the county-scale, the resulting fits are more precise but are sometimes overly confident, leading to lower overall accuracy and predictive performance. Conclusions : Our study demonstrates the importance of collecting fine-scale mosquito abundance data to improve our understanding and the predictability of mosquito population dynamics. The strong performance of both the cluster-based and county-level models illustrates the value of spatially sensitive modeling in this application. We anticipate that such modeling efforts will also aid in using weather forecasts to predict mosquito populations, aiding in efforts to control the spread of infectious disease. West Nile virus Culex Clustering Mosquitoes Climate Figures Figure 1 Figure 2 Figure 3 Figure 4 Background Mosquitoes are carriers of many pathogens that have global impacts, including West Nile virus (WNV), Yellow Fever Virus, Rift Valley Virus, Zika Virus, Chikungunya Virus, Dengue, and Malaria, which cause more than 700,000 deaths annually [ 1 ]. For example, as of 2022, 249 million cases of malaria were identified worldwide [ 2 ]. Dengue virus is widespread in over 100 countries, with annual global infections of 390 million persons [ 3 ]. Furthermore, one in every 150 people infected with WNV is diagnosed with neuroinvasive WNV disease, with many suffering from long-lasting secondary sequelae (e.g., depression, memory loss, and motor dysfunction). As ectotherms and complex-life-cycle organisms, mosquito population dynamics are invariably linked to climate and local environmental conditions [ 4 , 5 ], such as temperature, precipitation, humidity, and the availability of standing water for breeding. It is well documented from laboratory studies that temperature drives the rates of many key demographic traits across different mosquito life stages and across many mosquito species [ 6 ]. Interestingly, humidity can alter the shapes of these thermal performance relationships, indicating that interactions between climate factors will also impact population dynamics [ 7 ]. For some mosquito species, the availability of standing water for breeding helps explain trends in population sizes, such that cumulative measures of precipitation are often used as proxies to model mosquito populations. A recent study in [ 8 ] analyzed the relationships between mosquito population size, irrigation, and WNV cases in humans, finding an association between mosquito population and irrigation activities. Accordingly, climate factors show statistical relationships with the burdens of many mosquito-borne diseases. For instance, statistical modeling studies often use climate factors to explain and predict the cases of WNV in humans [ 9 – 14 ]. Based on the relationships between climate, mosquito population demography, and disease burden, more recent studies attempt to project the effects of climate change on the future global distributions of mosquitoes and the diseases they harbor [ 15 – 19 ]. Given the complicated nature of climate’s effects on mosquito population dynamics, flexible modeling approaches are needed to more mechanistically explain how multiple climate variables influence variation in mosquito abundance patterns over space and time. Mosquito population dynamics can fluctuate at relatively small spatial scales, likely because local weather variation leads to spatial and temporal variation in mosquito abundances [ 20 – 22 ]. For instance, even across a few hundred squared kilometers, mosquito population sizes can vary spatially and seasonally, potentially due to differences in local precipitation and temperature, and the effects of weather may even vary among co-occurring mosquito species [ 23 ]. An important challenge is therefore not only to build mechanistic models that can explain temporal trends in mosquito population fluctuations, but that also help us interrogate the factors that explain spatial heterogeneities in these abundance patterns. Many methods are used to model the spatial and temporal fluctuations in mosquito population abundances, ranging from purely mathematical models (e.g., compartmental differential equation models) to purely statistical models (e.g., linear regression frameworks). Developing an adequate mathematical or statistical model to explain or predict mosquito population dynamics in relation to climate is a challenging endeavor with various trade-offs. Purely statistical models, including statistically-based machine learning algorithms, can do a good job of explaining temporal and spatial trends in mosquito abundances [ 8 , 9 , 24 – 28 ]. A challenge with statistical modeling approaches is that they typically are inadequate representations of the well-known non-linear effects of climate on complex population dynamics. Additionally statistical models are generally correlational, meaning that at least some mechanistic understanding of the system is lost. Classic [ 29 – 31 ] and contemporary mathematical models often introduce large sets of differential equations that represent the various life-stages of mosquitoes, from egg to larvae to adult, and they can include the effects of temperature on certain demographic rates [ 32 – 36 ]. Such models have the strong benefit of inherently including non-linear effects of climate on demographic patterns, but they also have some challenges. It is difficult to derive realistic estimates for the many parameters in such large models (i.e., a dimensionality problem). Parameterizing such models therefore often relies on laboratory experiments that measure thermal performance of mosquitoes, but such parameter estimates are largely untested in field settings (although see [ 37 ]). Fortunately, recent mechanistic modeling and model-fitting approaches show promise in incorporating both temperature and precipitation to explain dynamic mosquito population abundance patterns [ 15 ]. In our study, we seek to develop a mechanistic yet simplified way of incorporating temperature and precipitation into a generic and dynamic modeling approach. This modeling approach allows us to explore if spatiotemporal variation in weather explains the non-linear dynamics of mosquito populations in natural landscapes. We test our modeling approach using the study system of Cx. quinquefasciatus in Maricopa County, AZ, which transmits multiple infections to humans, and is the most abundant vector for West Nile and St. Louis Encephalitis (SLEV) viruses in the county. According to the CDC, Maricopa County is consistently in the top 10 counties yearly in the nation for West Nile virus cases [ 1 ]. From 1999 to 2022, Maricopa County has had the most human cases in the nation, with a total of 3006 infections. Hundreds of traps are set weekly across the county, and mosquito abundances can vary dramatically in magnitude across subdivisions of the county. Hence, this system is ideal for testing how a combination of weather factors may influence spatiotemporal heterogeneity in mosquito population dynamics. We therefore hypothesized that accounting for microclimate variation—by grouping areas with similar weather patterns—would improve our model’s ability to predict mosquito abundance across the county. This modeling study is important because it helps us understand how spatial variation in climate affects spatial variation in mosquito population sizes. Moreover, in future work, the simplified modeling approach we develop here could be incorporated into epidemiological models to account for important climate and spatial variation in disease transmission. Methods Approach Overview First, we develop a semi-mechanistic model of the temporal dynamics of adult mosquito abundance, which incorporates non-linear statistical effects of temperature and accumulated precipitation. We then use Bayesian Markov chain Monte Carlo (MCMC) to fit our model to three years of weekly, adult female Cx. quinquefasciatus abundance observations across Maricopa County (i.e., the abundance data aggregated across all traps established in the county). To test whether spatial heterogeneity in climate explains variation in the patterns of mosquito abundance across Maricopa County, we cluster the county’s zip codes by local similarities in climate, and aggregate the mosquito abundance across traps within these clusters. We develop a method to hierarchically re-fit our model to abundance data in each cluster. Next, we compare whether fitting the data per cluster and then aggregating these fits to the county-level improves the overall fit of the county-level data. In this way, we see if heterogeneity in climate-based spatial clusters can help us better explain the county-level aggregated data. Finally, we also evaluate the out-of-sample prediction accuracy of our models, at both county and cluster scales. Below, we describe each step of the process in detail. Climate-forced mathematical model We developed a mathematical model that explains temporal patterns of adult mosquito abundance, which is driven by two exogenous climatic variables: temperature and precipitation. The general form of the ordinary differential equation is $$\:\frac{dX\left(t\right)}{dt}=\nu\:f\left(\text{Temp}\right(t),\text{Prcp}(t\left)\right)-\mu\:X\left(t\right).$$ \(\:X\left(t\right)\) is the mosquito abundance over time, with model time steps discretized to daily as \(\:X\left({t}_{i}\right)\) where \(\:i=1,2,3,\dots\:\) . The model assumes that there is a baseline population growth rate of mosquitoes, \(\:\nu\:\) , which is modulated by a statistical function of daily measures of temperature \(\:\text{Temp}\left(t\right)\) and precipitation \(\:\text{Prcp}\left(t\right)\) . Specifically, \(\:\text{Temp}\left(t\right)\) is the daily maximum temperature in degrees Celsius, and \(\:\text{Prcp}\left(t\right)\) is the daily measure of 30-day accumulated millimeters of precipitation, which we scale for statistical convenience by dividing by a constant 50. The form of the statistical function is: $$\:f\left(\text{Temp}\right(t),\text{Prcp}(t\left)\right)=-\left(\text{Temp}\right(t)-Tmin)\left(\text{Temp}\right(t)-Tmax))\frac{1}{(1+{\text{e}}^{\alpha\:-\varphi\:\text{Prcp}\left(t\right)})}$$ The temperature effect is modeled as a quadratic equation, representing a thermal performance curve of population growth rate, controlled by a minimum \(\:Tmin\) and a maximum \(\:Tmax\) growth temperature. This structure is agnostic to the fact that many traits of mosquito individuals across their life stages have unique thermal performance curves [ 6 ]. Our strategy is instead to test a simplification by estimating a parsimonious representation of thermal effects on the overall growth rate of the adult mosquito population, integrating across the thermal effects of specific, life-stage specific traits. The precipitation effect is modeled as a logistic, saturating function, whose shape is controlled by constants \(\:\alpha\:\) and \(\:\varphi\:\) . Biologically, this means that as accumulated precipitation increases, the growth rate increases, but the model could estimate how quickly this increase occurs across the range of precipitation and whether there is a saturating effect, above which more accumulated precipitation has negligible effects. Illustrative plots of the temperature and precipitation functions and their corresponding seasonal time series are shown in Additional file 1: Fig. S1 . We fix the per‑capita mortality rate \(\:\mu\:\) (at 0.12) instead of estimating it as the ratio between the growth term and the mortality term are the only factors that govern the net population change; estimating both would create an identifiability issue. Consequently, fluctuations in the fitted growth term implicitly capture both births and deaths. In Table 1 , we described each parameter in our mathematical model. Table 1 Parameters of the model Parameter Signification of parameters Unit Source ν Baseline population growth rate \(\:\frac{\text{mosquitoes}}{{}^{\circ\:}C\cdot\:\text{day}}\) Estimated Tmin Minimum temperature that constrains population growth \(\:{}^{\circ\:}C\) Estimated Tmax Maximum temperature that constrains population growth \(\:{}^{\circ\:}C\) Estimated α Contributes to the inflection point of the precipitation effect curve Unitless Estimated φ Steepness of the precipitation effect curve 1/mm Estimated µ Mosquito death rate 1/day Fixed Table 2 Model evaluation comparison at the county level Model Fitted RMSE (2014–2016) Predicted RMSE (2013) Predicted RMSE (2017) County without ’outliers’ ZCTAs 252.6900 306.0828 494.7659 5 clusters sum 265.8615 325.5769 507.0085 9 clusters sum 269.4774 349.3141 511.3981 Mosquito abundance data The mosquito abundance data used in this article was collected by the Maricopa County Environmental Services Vector Control Division [ 38 ]. The data is collected weekly from \(\:\sim\:800\) \(\:{\text{CO}}_{2}\) traps located in an approximate 1km \(\:{}^{2}\) grid across most of the county. The longitude and latitude of the traps locations is documented for easy spatial analysis. Each trap is placed for a 12-h collection period once per week for 50 weeks of the year. Collections are subsequently sorted by mosquito species and sex. Data used in this study were restricted to Cx. quiquefasciatus , the most frequently trapped mosquito vector in Maricopa County, and is known to carry avian malaria, WNV, SLEV, and other pathogens. This species constitutes about 80% of the total infected mosquitoes in Maricopa County. In our analysis, we used a training data set spanning three years, January 1, 2014 to December 31, 2016, and we used 2013 and 2017 data to validate our model as an out-of-sample prediction data set. We chose 2014–2016 for the training set, because the data includes unique patterns in abundance that can help validate the effects of climate. For instance, in 2014 there was a large monsoon season, and this effect can be visually observed in the abundance data, while years 2015 and 2016 had relatively average abundance trends. For the county-level data set, mosquito abundance per week was aggregated across all traps. We used data for adult female mosquitoes only since they were blood-seeking and attracted to \(\:{\text{CO}}_{2}\) traps. Each trap is identified by GPS coordinates (longitude and latitude), such that for the cluster-level data sets, we first aggregated trap data to the ZIP code-level (i.e., ZIP Code Tabulation Areas, ZCTA), and then clustered ZCTA by climate similarity and aggregated trap data accordingly (see Fig. 1 a). Temperature and precipitation data Daily estimates of maximum temperature in Celsius and precipitation in millimeters were downloaded from PRISM via the prism R package [ 39 ]. We downloaded the Prism daily data in a 4 km by 4 km raster. For the county-level model, we averaged daily values across all the Prism data in the raster within the Maricopa County polygon defined by the 2010 Census Bureau delineations (tigris package [ 40 ]) in Maricopa County. For the cluster-level model, we averaged the Prism data across the grid within each ZCTA polygon, again using 2010 Census Bureau delineations. Then, we averaged these data across all ZCTAs included in a given cluster. In the model, we used a 30-day accumulated precipitation, scaled by dividing by the constant 50. In Fig. 2 , we present the time series plot for temperature and precipitation for 2014–2016 which is averaged across ZCTA in Maricopa County. We explored the spatial distribution of the mosquito abundance data present within the 109 non-outlier ZCTAs across Maricopa County (see details of outlier ZCTAs in Additional file 1: Text S1). Our goal was to cluster these ZCTAs based on similarities in climatic data during the three-year window of the training data set. We used only the precipitation time-series data, as the temperature time-series did not vary significantly across the county (see Fig. 2 ). Using the precipitation data, we clustered the ZCTAs based on a functional time series clustering technique implemented in the FunFEM package in R [ 41 ]. This clustering technique requires the user to specify the number of clusters desired. Therefore, we tested 5 and 10 clusters. Briefly, to apply functional time-series clustering, first, a basis function must be fit to the time-series data to smooth it. In our case, we used a Fourier basis, which is the most appropriate for periodic patterns, such as those observed in the precipitation data [ 42 ]. The functional clustering algorithm compares time-series data by transforming them into functional forms, often using basis expansions like splines or Fourier series, to capture their continuous-time dynamics [ 42 ]. It then measures similarity based on shared features such as trends, seasonal patterns, or amplitude variations using a model-based clustering approach in a discriminative functional subspace, where differences between curves are captured via latent variable modeling and Gaussian mixture models, primarily relying on functional Principal Component Analysis (PCA) and Fisher-like discriminative analysis. Time series with similar functional properties are grouped into contiguous clusters, allowing for the identification of common behaviors and trends across the data while accounting for variability in shape, scale, and temporal alignment. Note that when we specified 10 clusters, one cluster only included four ZCTA, each with a maximum mosquito abundance per week of less than ten. Therefore, we manually moved those to the most similar other cluster, leaving us with a comparison of 5 versus 9 clusters (see middle and last panel of Fig. 1 b). Fitting the model to the time-series data We used a Metropolis-Hastings Markov chain Monte Carlo (MCMC) algorithm to sample from the joint posterior distribution of model parameters. To calculate the likelihood, we compare the solution of our ordinary differential equation model to the weekly observed abundance data. We solve the differential equation model with the deSolve [ 43 ] package using the method lsoda [ 44 ] in R, incorporating daily fluctuations in temperature and precipitation. We compare the model to the data every seven days; therefore, while the model ingests daily data on climate, the model is only compared to the observed mosquito data once per week. The likelihood is defined as a negative binomial probability distribution with an estimated over-dispersion parameter (see Additional file 1: Text S2-S3 for more details). For each spatial data set, we ran four MCMC chains, each with 5,000 iterations, in parallel using the futures R package [ 45 ]. We used the \(\:\widehat{R}\) statistic [ 46 ], also known as the Gelman-Rubin diagnostic, to assess the convergence of the chains. We also performed posterior predictive checks, described below, to ensure the joint posterior defined a reasonable parameter estimation space. We fit the model to the following 5 data partitions: county-level with outliers, county-level without outliers, 2-cluster (outlier ZCTA versus all other ZCTA), 5-cluster without outliers, and 9-cluster without outliers. For the two county-level partitions, we run a standard MCMC process. For the cluster-level partitions, we ran a hierarchical inference approach. Model validation We used within-sample and out-of-sample predictive measures to compare model performance across the different data partitions. We are ultimately interested in understanding if accounting for spatial variation in climate and baseline growth rates might improve the model’s explanation of county-wide, total mosquito abundances. Therefore, in calculating goodness-of-fit metrics, for the cluster-level data partitions, we summed metrics across clusters to a county-level performance metric. The method we employed for goodness-of-fit measurement was Root Mean Square Error (RMSE). To generate this measurement for each data partition and model-fit, we first generate posterior predictive model runs. For each data partition and MCMC analysis, we randomly sampled 100 parameter sets from the joint posterior. For each set, we substitute the ODE to generate the corresponding numerical solution. Across these 100 model runs, we calculated the median model expectation per day for the model. The RMSE was then calculated by comparing the data observations to the median model expectation, and summing across all observations, and across clusters, when appropriate. Results Variation in precipitation and temperature The clustering routine captured logical spatial groupings of zip codes (ZCTA) across Maricopa County based on spatial and temporal variation in 30-day accumulated precipitation (Fig. 2 and Fig. 3 ). When we look at the PRISM-derived precipitation metric, we see that clusters are mostly differentiated by precipitation magnitude and more minimally by precipitation timing. For instance, in the 5-cluster case, all clusters experienced relatively high abundance in the monsoon season of 2014, with minor variation in total magnitude, whereas in 2015, there is some variability in the timing of precipitation, especially in the second half of the year (see Fig. 3 ). Comparing county- and cluster-level modeling In general, our mechanistic model captures detailed patterns of mosquito abundance changing through time across multiple spatial scales for the three years of training data, particularly when we remove the six outlier ZCTA (Fig. 4 and Additional file 1: Fig. S6-S7). For example, the model generally captures the effects of high summer temperatures, where we see corresponding declines in mosquito abundances, and also the effects of monsoonal activity, where we can see dramatic increases in mosquito abundances due to heavy rainfall (e.g., in the second half of 2014). Accounting for heterogeneity in precipitation across clustered ZCTA lead to higher precision in model fits we aggregated model expectations from cluster to county-scales (i.e., narrower error bands); however, these aggregated model fits were sometimes overly confident, such that the error bands did not always capture the true observations (see Fig. 4 ). Thus, the county-level model, which used daily values for temperature and precipitation that were averaged across the ZCTA, had a relatively better fit compared to the aggregated cluster-level models, based on the Root Mean Squared Error (RMSE) metric (Additional file 1: Table S3-S4). Furthermore, there were not clear differences in the aggregated county-level model fit when we compare the 5-cluster to the 9-cluster situations (Fig. 4 ). When we examine model fits per cluster, however, we do see that the model captures variation in mosquito abundances across clusters, explained mostly by differences in baseline population growth rates (Additional file 1: Fig. S6-S7). Also, we observed that the RMSE values vary across clusters depending on the number of clusters (Additional file 1: Table S3-S4). This result demonstrates that our method of hierarchical estimation of cluster-specific growth rates successfully characterized key differences among clusters. Collectively, the results suggest that the aggregate, county-level patterns in mosquito abundance can be most parsimoniously explained by a spatial average of temperature and precipitation data. Yet the cluster-level models provide accurate, finer-scale inference of how mosquito population varies across space and time, though at this finer scale the model fits may be more prone to small prediction errors. Additional details about outliers and county level analysis can be found in Additional file 1: Text S4-S5. Out of sample performance When we predicted mosquito abundance for two additional years of withheld data (2013 and 2017), our model still captures the fundamental characteristics of these mosquito abundance time-series (see Additional file 1: Fig. S6 (first and last panel), S8-S9). Indeed, the key results from the training data still hold true with the out-of-sample data. Particularly, the county-level model performs nearly equally as well as the cluster-level models. Notably, however, across spatial scales, the dynamic model does a poor job at explaining the early-year increases in mosquito abundance observed in 2017, and under-predicts peaks in abundances observed in the later half of both 2013 and 2017. We observed these failures of the model in both the county- and Cluster-level out-of-sample model fits. Discussion This research sheds light on the complex relationships between climate, spatial heterogeneities, and mosquito populations. By implementing a novel mechanistic model with a data-driven clustering strategy, we show that a relatively simple mathematical model that incorporates temperature and precipitation effectively captures the dynamical patterns of mosquito abundance observed over multiple years in a large urban setting. When we fit our model to aggregate data for all of Maricopa County, our climate-driven model accurately explains the seasonal increase in mosquito abundances in the spring, a decline in the hottest part of the summer, and another increase in the warm, wet monsoon season. When we subdivided the county into clusters based on similar climate patterns, we observe that average mosquito abundances vary substantially across the county. However, the cluster-level model still explains a substantial proportion of variance in the these downscaled data, indicating that mosquito populations across the region exhibit consistent numerical responses to temperature and precipitation. When we aggregate these smaller-scale model fits, we improve precision around the median model prediction, but we do not improve county-level model accuracy—likely due to error accumulation across clusters and the sparse data within each group. What this tells us is that the same climate-forced model can be applied at sub-county spatial scales and will explain the key dynamical patterns of the mosquito population dynamics; yet, at these smaller spatial scales, the model will be more prone to observational error, suggesting a trade-off between spatial resolution and model accuracy. Our results underscore the importance of modeling spatial structure when interpreting mosquito dynamics, particularly in regions like the desert Southwest where precipitation and temperature vary dramatically over small spatial scales. Previous studies have demonstrated that local weather conditions influence mosquito abundance and distribution [ 15 , 20 , 47 , 48 ]. Research on Cx. quinquefasciatus , the focal species in our study, has highlighted how temperature and precipitation shape its life cycle. For example, a study in Hawaii [ 49 ] found that Cx. quinquefasciatus abundance increased with temperature, peaking at mid-elevation sites, and showed a non-linear relationship with precipitation—highlighting temperature as a key driver of mosquito distribution across varying terrains. Precipitation also played a significant, albeit complex, role. Interestingly, the relationship between precipitation and abundance indicated negative effects of rainfall, while lagged precipitation showed positive associations. This complexity likely reflects seasonal rainfall patterns and their influence on larval habitat availability. Similarly, Morin and Comrie [ 50 ] used the Dynamic Mosquito Simulation Model (DyMSiM) model to show that the interaction between temperature and precipitation drives mosquito dynamics differently in California and Florida. Valdez et al. [ 51 ] further revealed that not only total rainfall, but also the number of rainy days and daily variability, strongly influence mosquito abundance. Our findings corroborate these earlier studies and extend them by explicitly testing whether accounting for spatial climate heterogeneity improves the performance of a mechanistic model. Although the cluster-level model explains key features of the mosquito population dynamics at smaller spatial scales, aggregating cluster-level predictions to the county scale did not improve inference. A possible contributing reason for this contradiction is that the data we used per mosquito trap (PRISM), interpolates weather patterns across space based on sparse station networks and provides 4x4km grid raster data. In Arizona, localized monsoon events produce intense, short-duration rainfall that varies significantly over short distances. PRISM's data resolution may smooth over these extremes, missing key signals needed for accurate prediction [ 52 , 53 ]. Future studies could consider deploying ground-based weather stations near the mosquito traps to obtain high-resolution climate data. This could enhance the accuracy of fine-scale model predictions. Moreover, model error may accumulate during the aggregation process, particularly when fit to clusters with small sample sizes. While the cluster-aggregated model did not improve county-wide predictions, it remains valuable for local inference and public health planning. Local-level fits can inform targeted mosquito control strategies and enhance situational awareness. In our framework, we estimated a shared climate-response function for the mosquito population growth rate across all the clusters – the quadratic effect of temperature and the exponentially saturating effect of accumulated precipitation. Given that this same functional response explains the data across clusters, this provides evidence that mosquito sub-population growth rates are responding similarly to weather despite differing average abundances. However, the current approach does not account for mechanisms that may explain the average differences abundance across clusters. Future research should explore how landscape features, water availability, or urban infrastructure might explain baseline variation in mosquito abundances. We also encountered six zip codes with unusually high mosquito counts, especially in spring, which the model failed to capture even when fit separately. These outlier patterns were not explained by temperature or 30-day precipitation trends, suggesting local landscape effects such as standing water accumulation may be at play. Investigating these features—e.g., water-retaining infrastructure, wetlands, or irrigation systems—could yield important insights. Future work could incorporate additional covariates to explain baseline mosquito growth rates, such as proxies for standing water, land use, and hydrological features. High-resolution, daily updated data on these variables could better capture early-season surges in abundance. Additionally, spatial movement of mosquitoes—whether natural or human-facilitated—should be considered in spatially explicit models. Spatial processes are known to structure ecological and epidemiological dynamics, influencing everything from vector distribution to disease transmission and species interactions [ 47 , 54 ]. Invasive species spread, habitat fragmentation, and local resource competition all highlight the importance of spatial structure in population persistence and disease risk [ 55 ]. By explicitly incorporating spatial heterogeneity, our model represents a step forward in mosquito modeling efforts. As weather forecasting capabilities continue to improve, spatially structured models will be critical for translating environmental changes into actionable public health responses. Climate change is reshaping ecosystems and disease risk globally [ 16 – 18 ]. Rising temperatures and altered precipitation patterns create more favorable conditions for mosquito proliferation and disease transmission [ 56 , 57 ]. Accurate models that incorporate both temporal and spatial heterogeneity are essential to predict these dynamics and guide interventions. Conclusions In sum, our study demonstrates the value of spatially resolved climate-driven models for understanding mosquito population dynamics. Although challenges remain in scaling predictions and capturing outlier behavior, this work lays a foundation for more refined, localized modeling efforts that can enhance mosquito control and public health preparedness in a changing climate. Declarations Acknowledgement The authors would like to especially acknowledge the manuscript review provided by Brad J. Biggerstaff, Ph.D., of the Division of Vector-Borne Infectious Diseases, National Center for Zoonotic, Vector-Borne, and Enteric Diseases, Centers for Disease Control and Prevention (CDC), Fort Collins, CO. The contents are solely the responsibility of the authors and do not necessarily represent the official views of the Centers for Disease Control and Prevention or National Institutes of Health. Funding Research reported in this publication was supported by the National Institute of Allergy and Infectious Diseases of the National Institutes of Health under award number R01AI168144 to J.M and a subaward from the Pacific Southwest Regional Center of Excellence for Vector-Borne Diseases funded by the U.S. Centers for Disease Control and Prevention Cooperative Agreement 1U01CK000649 to C.M.H. Conflict of interest/Competing interests The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. Ethics approval and consent to participate There is no ethical issue related to this research. Code availability Code used for analyzing the data for this research is available upon request. Author contribution Conceptualization, K.O. and J.M.; methodology, K.O. and J.M.; software, K.O.; validation, K.O., Y.C., E.D., N.B., C.M.H., J.T., J.W., I.R., M.K., and J.M.; formal analysis, K.O.; investigation, K.O., Y.C., E.D., N.B., C.M.H., J.T., J.W., I.R., M.K., and J.M.; resources, J.M.; data curation, K.O. and J.M.; writing and original draft preparation, K.O. and J.M.; writing—review and editing, K.O., Y.C., E.D., N.B., C.M.H., J.T., J.W., I.R., M.K., and J.M.; visualization, K.O. and J.M.; supervision, J.M.; project administration, J.M. All authors have read and agreed to the published version of the manuscript. References Disease Control C for, Prevention, et al. Centers for disease control and prevention (CDC). 2020. World health organization. “Malaria.” (Published March 29, 2023. accessed September 5, 2024). World health organization. “Dengue.” (Published March 17, 2023. accessed September 5, 2024). Mordecai EA, Cohen JM, Evans MV, Gudapati P, Johnson LR, Lippi CA, et al. Detecting the impact of temperature on transmission of zika, dengue, and chikungunya using mechanistic models. PLoS neglected tropical diseases. 2017;11:e0005568. Van Wyk H, Eisenberg JN, Brouwer AF. Long-term projections of the impacts of warming temperatures on zika and dengue risk in four brazilian cities using a temperature-dependent basic reproduction number. PLOS Neglected Tropical Diseases. 2023;17:e0010839. Mordecai EA, Caldwell JM, Grossman MK, Lippi CA, Johnson LR, Neira M, et al. Thermal biology of mosquito-borne disease. Ecology letters. 2019;22:1690–708. Brown JJ, Pascual M, Wimberly MC, Johnson LR, Murdock CC. Humidity–the overlooked variable in the thermal biology of mosquito-borne disease. Ecology letters. 2023;26:1029–49. Kovach TJ, Kilpatrick AM. Irrigation increases and stabilizes mosquito populations and increases west nile virus incidence. Scientific Reports. 2024;14:19913. Farooq Z, Rocklöv J, Wallin J, Abiri N, Sewe MO, Sjödin H, et al. Artificial intelligence to predict West Nile virus outbreaks with eco-climatic drivers. The Lancet Regional Health – Europe. 2022 [cited 2024 Apr 26];17. Available from: https://www.thelancet.com/journals/lanepe/article/PIIS2666-7762(22)00063-1/fulltext Smith KH, Tyre AJ, Hamik J, Hayes MJ, Zhou Y, Dai L. Using climate to explain and predict west nile virus risk in nebraska. GeoHealth. 2020;4:e2020GH000244. Ruiz MO, Chaves LF, Hamer GL, Sun T, Brown WM, Walker ED, et al. Local impact of temperature and precipitation on west nile virus infection in culex species mosquitoes in northeast illinois, USA. Parasites & vectors. 2010;3:1–16. Baker RE, Mahmud AS, Miller IF, Rajeev M, Rasambainarivo F, Rice BL, et al. Infectious disease in an era of global change. Nature Reviews Microbiology. 2022;20:193–205. Parker N. Exploring the role of temperature and other environmental factors in west nile virus incidence and prediction in california counties from 2017–2022 using a zero-inflated model. PLOS Neglected Tropical Diseases. 2024;18:e0012051. Albrecht L, Kaufeld KA. Investigating the impact of environmental factors on West Nile virus human case prediction in Ontario, Canada. Frontiers in Public Health. 2023 [cited 2024 Apr 26];11. Available from: https://www.frontiersin.org/journals/public-health/articles/10.3389/fpubh.2023.1100543/full Caldwell JM, LaBeaud AD, Lambin EF, Stewart-Ibarra AM, Ndenga BA, Mutuku FM, et al. Climate predicts geographic and temporal variation in mosquito-borne disease dynamics on two continents. Nature communications. 2021;12:1233. Aliaga-Samanez A, Romero D, Murray K, Segura M, Real R, Olivero J. Potential climate change effects on the distribution of urban and sylvatic dengue and yellow fever vectors. Pathogens and Global Health. 2024;118:397–407. Medlock JM, Leach SA. Effect of climate change on vector-borne disease risk in the UK. The Lancet Infectious Diseases. 2015;15:721–30. Carlson CJ, Albery GF, Merow C, Trisos CH, Zipfel CM, Eskew EA, et al. Climate change increases cross-species viral transmission risk. Nature. 2022;607:555–62. Ryan S J. Lippi A, Stewart-Ibarra AM. Mapping geographic and demographic shifts for container breeding mosquito-borne disease transmission suitability in central and south america in a warming world. PLOS Climate. 2024;3:1–21. Available from: https://doi.org/10.1371/journal.pclm.0000312 Reisen WK, Cayan D, Tyree M, Barker CM, Eldridge B, Dettinger M. Impact of climate variation on mosquito abundance in california. Journal of Vector Ecology. 2008;33:89–98. Althouse BM, Hanley KA, Diallo M, Sall AA, Ba Y, Faye O, et al. Impact of climate and mosquito vector abundance on sylvatic arbovirus circulation dynamics in senegal. The American journal of tropical medicine and hygiene. 2015;92:88. Ferraccioli F, Riccetti N, Fasano A, Mourelatos S, Kioutsioukis I, Stilianakis NI. Effects of climatic and environmental factors on mosquito population inferred from West Nile virus surveillance in Greece. Scientific Reports. 2023 [cited 2024 Apr 26];13:18803. Available from: https://www.nature.com/articles/s41598-023-45666-3 Baril C, Pilling BG, Mikkelsen MJ, Sparrow JM, Duncan CA, Koloski CW, et al. The influence of weather on the population dynamics of common mosquito vector species in the canadian prairies. Parasites & Vectors. 2023;16:153. Li B, Fu Q, Huang Y, Sun Q, Zhao C, Ma X, et al. Application of circular statistics in temporal distribution of adult mosquitoes in qingdao, shandong province, china, 2021–2023. Parasites & Vectors. 2024;17:325. Rodrı́guez-González S, Izquierdo-Suzán M, Rocha-Ortega M, Córdoba-Aguilar A. Vector mosquito distribution and richness are predicted by socio-economic, and ecological variables. Acta Tropica. 2024;254:107179. Villena OC, McClure KM, Camp RJ, LaPointe DA, Atkinson CT, Sofaer HR, et al. Environmental and geographical factors influence the occurrence and abundance of the southern house mosquito, culex quinquefasciatus, in hawai ‘i. Scientific Reports. 2024;14:604. Gervasi R, Azzali I, Bisanzio D, Mosca A, Bertolotti L, Giacobini M. A genetic programming approach to predict mosquitoes abundance. European conference on genetic programming. Springer; 2019. p. 35–48. Culler LE, Ayres MP, Virginia RA. Spatial heterogeneity in the abundance and fecundity of arctic mosquitoes. Ecosphere. 2018;9:e02345. Available from: https://esajournals.onlinelibrary.wiley.com/doi/abs/10.1002/ecs2.2345 Ross R. The prevention of malaria 2nd edn (london: murray). 1911. M’Kendrick AG. Applications of mathematics to medical problems. Proceedings of the Edinburgh Mathematical Society. 1925;44:98–130. Macdonald. G MG. The epidemiology and control of malaria. 1957. Diabaté AB, Sangaré B, Koutou O. Mathematical modeling of the dynamics of vector-borne diseases transmitted by mosquitoes: Taking into account aquatic stages and gonotrophic cycle. Nonautonomous Dynamical Systems. 2022;9:205–36. Dembele B, Friedman A, Yakubu A-A. Malaria model with periodic mosquito birth and death rates. Journal of biological dynamics. 2009;3:430–45. Jia P, Lu L, Chen X, Chen J, Guo L, Yu X, et al. A climate-driven mechanistic population model of aedes albopictus with diapause. Parasites & vectors. 2016;9:1–15. Ezanno P, Balenghien T, Cailly P, L’Ambert G, Toty C, Tran A-L, et al. A generic climate-driven model to predict mosquito population dynamics. 2014. Shutt D, Goodsman DW, Martinez K, Hemez ZJL, Conrad JR, Xu C, et al. A process-based model with temperature, water, and lab-derived data improves predictions of daily culex pipiens/restuans mosquito density. Journal of Medical Entomology. 2022;59:1947–59. Fay RL, Cruz-Loya M, Keyel AC, Price DC, Zink SD, Mordecai EA, et al. Population-specific thermal responses contribute to regional variability in arbovirus transmission with changing climates. Iscience. 2024;27. Hepp CM, Cocking JH, Valentine M, Young SJ, Damian D, et al. Phylogenetic analysis of west nile virus in maricopa county, arizona: Evidence for dynamic behavior of strains in two major lineages in the american southwest. PLOS ONE. 2018;13:1–12. Available from: https://doi.org/10.1371/journal.pone.0205801 Hart EM, Bell K. Prism: Download data from the oregon prism project [Internet]. 2015. Available from: https://github.com/ropensci/prism Walker KE. Tigris: An r package to access and work with geographic data from the US census bureau. R Project; 2016. Bouveyron C, Jacques J. funFEM: An r package for functional data clustering. Quatrième rencontres r. 2015. Ramsay JO, Hooker G, Graves S. Introduction to functional data analysis. Functional Data Analysis with R and MATLAB. 2009;1–19. Soetaert KE, Petzoldt T, Setzer RW. Solving differential equations in r: Package deSolve. Journal of statistical software. 2010;33. Hindmarsh AC, Petzold LR. LSODA, ordinary differential equation solver for stiff or non-stiff system. 2005. Bengtsson H. A unifying framework for parallel and distributed processing in r using futures. The R Journal. 2021;13:208–27. Available from: https://doi.org/10.32614/RJ-2021-048 Gelman A, Carlin JB, Stern HS, Dunson DB, Vehtari A, Rubin DB. Bayesian Data Analysis. 3rd ed. New York: Chapman; Hall/CRC; 2015. Zeng Q, Yu X, Ni H, Xiao L, Xu T, Wu H, et al. Dengue transmission dynamics prediction by combining metapopulation networks and kalman filter algorithm. PLOS Neglected Tropical Diseases. 2023;17:e0011418. Ciota AT, Bialosuknia SM, Zink SD, Brecher SM, Ehrbar DJ, Morrissette MN, et al. Effects of zika virus strain and aedes mosquito species on vector competence. Emerging infectious diseases. 2014;20:1577–84. Villena O, McClure KM, Camp RJ, LaPointe D, Atkinson CT, Sofaer H, et al. Environmental and geographical factors influence the occurrence and abundance of the southern house mosquito, culex quinquefasciatus, in hawai‘i. Scientific Reports. 2023;13:20652. Morin CW, Comrie AC. Modeled response of the west nile virus vector culex quinquefasciatus to changing climate using the dynamic mosquito simulation model. International Journal of Biometeorology. 2010;54:517–29. Available from: https://doi.org/10.1007/s00484-010-0349-6 Valdez LD, Sibona GJ, Diaz LA, Contigiani MS, Condat CA. Effects of rainfall on culex mosquito population dynamics. Journal of Theoretical Biology. 2017;421:28–38. Available from: https://doi.org/10.1016/j.jtbi.2017.03.024 Sheppard PR, Comrie AC, Packin GD, Angersbach K, Hughes MK. The climate of the US southwest. Climate Research. 2002;21:219–38. Comrie AC, Broyles BB. Variability and spatial modeling of fine-scale precipitation in the southwest united states. Journal of Arid Environments. 2002;50:573–92. Giménez-Mujica U, Velázquez-Castro J, Anzo-Hernández A, Herrera-Ramı́rez T, Barradas I. Estimating the final size of vector-borne epidemics in metapopulation networks: Methodologies and comparisons. Boletı́n de la Sociedad Matemática Mexicana. 2025;31:35. Clark NJ, Ernest SKM, Senyondo H, Simonis J, White EP, Yenni GM, et al. Beyond single-species models: Leveraging multispecies forecasts to navigate the dynamics of ecological predictability. PeerJ. 2025;13:e18929. Jibowu M, Nolan MS, Ramphul R, Essigmann HT, Oluyomi AO, Brown EL, et al. Spatial dynamics of culex quinquefasciatus abundance: Geostatistical insights from harris county, texas. International Journal of Health Geographics. 2024;23:1–12. Sambado S, Sipin TJ, Rennie Z, Larsen A, Cunningham J, Quandt A, et al. The paradoxical impact of drought on west nile virus risk: Insights from long-term ecological data. bioRxiv. 2025;2025–01. Additional Declarations No competing interests reported. Supplementary Files MosqAbundsupp.pdf Supplementary information Additional file 1: Text S1. Some details about outlier ZCTAs. Text S2. Hierarchical model inference across clusters. Text S3. Formulation of likelihood function using negative binomial probability distribution. Text S4. Interpretation of the outliers' ZCTAs results. Text S5. Model evaluation for 2 clusters. Fig. S1. Temperature and precipitation response functions used in the model. We used 2014 temperature and precipitation data from PRISM. Fig. S2. Comparison of clustering prediction, real data, and county prediction for 2 clusters. Fig. S3. A spatio-temporal plot of mosquito abundance per ZCTA. Fig. S4. Four cluster result. Fig. S5. Relationship between estimated baseline mosquito population growth rate and average mosquito abundance across clusters. Fig. S6. 5 Cluster prediction. Fig. S7. 9 Cluster prediction. Fig. S8. 9 Cluster 2013 prediction. Fig. S9. 9 Cluster 2017 prediction. Fig. S10. 2 Cluster prediction. Fig. S10-S15. Histograms of marginal posterior draws of parameters. Fig. S16-S20. Pairs plot displaying joint posterior draws of parameters. Fig. S21. Predicted versus actual for all clusters. Table S1. Parameter estimated value and its prior for one cluster. Table S2. Parameter estimated value and its prior for more than one cluster. Table S3. Model evaluation at cluster level for five clusters. Table S4. Model evaluation at cluster level for nine clusters. abstactfigure.tiff Cite Share Download PDF Status: Published Journal Publication published 11 Mar, 2026 Read the published version in Parasites & Vectors → Version 1 posted Editorial decision: Revision requested 25 Nov, 2025 Reviews received at journal 24 Nov, 2025 Reviews received at journal 22 Oct, 2025 Reviewers agreed at journal 20 Oct, 2025 Reviewers agreed at journal 13 Oct, 2025 Reviewers invited by journal 05 Oct, 2025 Editor assigned by journal 05 Oct, 2025 Submission checks completed at journal 30 Sep, 2025 First submitted to journal 22 Sep, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7595227","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":530278168,"identity":"d585ca08-316e-4b07-9cb3-00f3efbf3d01","order_by":0,"name":"Kayode Oshinubi","email":"","orcid":"","institution":"Northern Arizona University","correspondingAuthor":false,"prefix":"","firstName":"Kayode","middleName":"","lastName":"Oshinubi","suffix":""},{"id":530278170,"identity":"1b2df142-d8f0-4048-a51a-21765e16fe41","order_by":1,"name":"Crystal M. Hepp","email":"","orcid":"","institution":"Northern Arizona University","correspondingAuthor":false,"prefix":"","firstName":"Crystal","middleName":"M.","lastName":"Hepp","suffix":""},{"id":530278172,"identity":"2287d405-4cd6-4e85-a111-d4bd7bf18268","order_by":2,"name":"Ye Chen","email":"","orcid":"","institution":"Northern Arizona University","correspondingAuthor":false,"prefix":"","firstName":"Ye","middleName":"","lastName":"Chen","suffix":""},{"id":530278174,"identity":"e5cf0174-7116-4ad2-8131-872e1f569639","order_by":3,"name":"Eck Doerry","email":"","orcid":"","institution":"Northern Arizona University","correspondingAuthor":false,"prefix":"","firstName":"Eck","middleName":"","lastName":"Doerry","suffix":""},{"id":530278176,"identity":"5230159b-e57f-48d3-852d-d77c60b21ec6","order_by":4,"name":"Nicole Busser","email":"","orcid":"","institution":"Vector Control Division","correspondingAuthor":false,"prefix":"","firstName":"Nicole","middleName":"","lastName":"Busser","suffix":""},{"id":530278177,"identity":"c8975308-1fa8-4bbd-a867-e1c1e59452c1","order_by":5,"name":"John Townsend","email":"","orcid":"","institution":"Vector Control Division","correspondingAuthor":false,"prefix":"","firstName":"John","middleName":"","lastName":"Townsend","suffix":""},{"id":530278178,"identity":"13906306-e36e-41c5-a531-a8191670759b","order_by":6,"name":"James Will","email":"","orcid":"","institution":"Vector Control Division","correspondingAuthor":false,"prefix":"","firstName":"James","middleName":"","lastName":"Will","suffix":""},{"id":530278179,"identity":"c2f0c51c-7105-4ac4-988b-03bc23b278e3","order_by":7,"name":"Irene Ruberto","email":"","orcid":"","institution":"Arizona Department of Health Services","correspondingAuthor":false,"prefix":"","firstName":"Irene","middleName":"","lastName":"Ruberto","suffix":""},{"id":530278180,"identity":"1a8a3542-45c9-41c8-aeaf-dd018f18824c","order_by":8,"name":"Melissa Kretschmer","email":"","orcid":"","institution":"Maricopa County Department of Public Health","correspondingAuthor":false,"prefix":"","firstName":"Melissa","middleName":"","lastName":"Kretschmer","suffix":""},{"id":530278181,"identity":"f5bb859d-3392-4c55-87a9-7cfeacd8aaf8","order_by":9,"name":"Joseph Mihaljevic","email":"data:image/png;base64,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","orcid":"","institution":"Northern Arizona University","correspondingAuthor":true,"prefix":"","firstName":"Joseph","middleName":"","lastName":"Mihaljevic","suffix":""}],"badges":[],"createdAt":"2025-09-11 23:53:05","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7595227/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7595227/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1186/s13071-026-07326-z","type":"published","date":"2026-03-11T15:57:57+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":93697601,"identity":"78fc34ed-a85f-45af-8c7a-298c95c3447a","added_by":"auto","created_at":"2025-10-16 14:59:12","extension":"tiff","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1452846,"visible":true,"origin":"","legend":"","description":"","filename":"Fig.1.tiff","url":"https://assets-eu.researchsquare.com/files/rs-7595227/v1/4b6bd89804d220dc5f0e6cb8.tiff"},{"id":93698179,"identity":"6b144bd0-626d-44f3-8ea5-6730b20b16e4","added_by":"auto","created_at":"2025-10-16 15:07:12","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":58619,"visible":true,"origin":"","legend":"","description":"","filename":"FinalMosqAbundV8.docx","url":"https://assets-eu.researchsquare.com/files/rs-7595227/v1/3b04699354f4f23a5241728c.docx"},{"id":93697609,"identity":"160aae30-306d-4faf-a751-9f8b024295c0","added_by":"auto","created_at":"2025-10-16 14:59:12","extension":"tiff","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1461486,"visible":true,"origin":"","legend":"","description":"","filename":"Fig.2.tiff","url":"https://assets-eu.researchsquare.com/files/rs-7595227/v1/f1bd2c669ba3c332e4f6f990.tiff"},{"id":93699204,"identity":"c0e85fa3-fba9-45cd-b45e-fb99bded5059","added_by":"auto","created_at":"2025-10-16 15:15:12","extension":"tiff","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":718398,"visible":true,"origin":"","legend":"","description":"","filename":"Fig.3.tiff","url":"https://assets-eu.researchsquare.com/files/rs-7595227/v1/6da998c657e191b07c11ff98.tiff"},{"id":93698184,"identity":"bf19dd67-2cfc-4e90-9cbd-b6d345687287","added_by":"auto","created_at":"2025-10-16 15:07:12","extension":"tiff","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":878246,"visible":true,"origin":"","legend":"","description":"","filename":"Fig.4.tiff","url":"https://assets-eu.researchsquare.com/files/rs-7595227/v1/c5de2666bd6c169a4ea2ca7b.tiff"},{"id":93697606,"identity":"b1d80b62-cf9c-4baa-8566-b9e230dbc4cf","added_by":"auto","created_at":"2025-10-16 14:59:12","extension":"tiff","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":878246,"visible":true,"origin":"","legend":"","description":"","filename":"abstactfigure.tiff","url":"https://assets-eu.researchsquare.com/files/rs-7595227/v1/a38dd72ae02bb683a9affe45.tiff"},{"id":93698183,"identity":"09edc7c8-1bd7-4c8f-81f0-7f5bc417ef24","added_by":"auto","created_at":"2025-10-16 15:07:12","extension":"json","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":12399,"visible":true,"origin":"","legend":"","description":"","filename":"4d456721a6934f80afd6797d30e3215d.json","url":"https://assets-eu.researchsquare.com/files/rs-7595227/v1/0642d0714f4c81eb6ec95f4b.json"},{"id":93697626,"identity":"fbd3da00-2e5c-44c8-893c-b661bba6b1ae","added_by":"auto","created_at":"2025-10-16 14:59:13","extension":"pdf","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":84750516,"visible":true,"origin":"","legend":"","description":"","filename":"MosqAbundsupp.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7595227/v1/2c23cfe0cff38fdc67537f6c.pdf"},{"id":93699586,"identity":"4dfa0bac-7d03-442f-b597-d3c090858de3","added_by":"auto","created_at":"2025-10-16 15:23:12","extension":"xml","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":133169,"visible":true,"origin":"","legend":"","description":"","filename":"4d456721a6934f80afd6797d30e3215d1enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-7595227/v1/e9575a08927149945cc7abd2.xml"},{"id":93697614,"identity":"319144b6-68d6-4743-aef2-767d9175d3ee","added_by":"auto","created_at":"2025-10-16 14:59:12","extension":"tiff","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1452846,"visible":true,"origin":"","legend":"","description":"","filename":"Fig.1.tiff","url":"https://assets-eu.researchsquare.com/files/rs-7595227/v1/27e67c181987a8124f71a32c.tiff"},{"id":93697620,"identity":"e396893e-4d53-41e3-a542-225eab3f3a60","added_by":"auto","created_at":"2025-10-16 14:59:12","extension":"tiff","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1461486,"visible":true,"origin":"","legend":"","description":"","filename":"Fig.2.tiff","url":"https://assets-eu.researchsquare.com/files/rs-7595227/v1/8f6735450b143cd262c450cc.tiff"},{"id":93698190,"identity":"c54d6020-9f0c-4122-92d2-bffb231e5d7c","added_by":"auto","created_at":"2025-10-16 15:07:12","extension":"tiff","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":718398,"visible":true,"origin":"","legend":"","description":"","filename":"Fig.3.tiff","url":"https://assets-eu.researchsquare.com/files/rs-7595227/v1/11c11ee0702c2253241b634d.tiff"},{"id":93697621,"identity":"416b05f1-b9cd-4097-b6a3-61472f709558","added_by":"auto","created_at":"2025-10-16 14:59:12","extension":"tiff","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":878246,"visible":true,"origin":"","legend":"","description":"","filename":"Fig.4.tiff","url":"https://assets-eu.researchsquare.com/files/rs-7595227/v1/17e47fc2f696fc203e6fde0f.tiff"},{"id":93699208,"identity":"25255c96-e75f-4a72-b967-8dd24398de6f","added_by":"auto","created_at":"2025-10-16 15:15:12","extension":"tiff","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":878246,"visible":true,"origin":"","legend":"","description":"","filename":"abstactfigure.tiff","url":"https://assets-eu.researchsquare.com/files/rs-7595227/v1/adb271b97868cd9dc8625251.tiff"},{"id":93698186,"identity":"e58b9204-6c15-4675-ac27-3cff2a1c9cb5","added_by":"auto","created_at":"2025-10-16 15:07:12","extension":"png","order_by":14,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":37101,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFig.1.png","url":"https://assets-eu.researchsquare.com/files/rs-7595227/v1/fa0260cb4a04d6aa4b1534f9.png"},{"id":93697608,"identity":"54f358d8-0f61-46d4-b29d-474054eca57f","added_by":"auto","created_at":"2025-10-16 14:59:12","extension":"png","order_by":15,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":39784,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFig.2.png","url":"https://assets-eu.researchsquare.com/files/rs-7595227/v1/4e2aa15db21499a2a06026a1.png"},{"id":93700700,"identity":"80ffa7a7-a231-47a8-b739-4f2e4b5f9220","added_by":"auto","created_at":"2025-10-16 15:31:12","extension":"png","order_by":16,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":18192,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFig.3.png","url":"https://assets-eu.researchsquare.com/files/rs-7595227/v1/5e6b143c9620bbe57c7ab915.png"},{"id":93699205,"identity":"c9ec5cbe-9566-4607-bc03-20867659a3ea","added_by":"auto","created_at":"2025-10-16 15:15:12","extension":"png","order_by":17,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":20682,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFig.4.png","url":"https://assets-eu.researchsquare.com/files/rs-7595227/v1/1d0c192d8de92a02a210ca9a.png"},{"id":93697610,"identity":"9d6c3402-6857-49f9-ba9c-b534620ba05e","added_by":"auto","created_at":"2025-10-16 14:59:12","extension":"png","order_by":18,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":24217,"visible":true,"origin":"","legend":"","description":"","filename":"Onlineabstactfigure.png","url":"https://assets-eu.researchsquare.com/files/rs-7595227/v1/cb71fad136cb1d981100e00f.png"},{"id":93697618,"identity":"782afc1b-eb09-4055-a95e-b68ddabd079b","added_by":"auto","created_at":"2025-10-16 14:59:12","extension":"xml","order_by":19,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":131770,"visible":true,"origin":"","legend":"","description":"","filename":"4d456721a6934f80afd6797d30e3215d1structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-7595227/v1/3f64e4e7ae439b8c61b3248b.xml"},{"id":93697612,"identity":"d0b05138-62de-4816-b02c-f9956d4156de","added_by":"auto","created_at":"2025-10-16 14:59:12","extension":"html","order_by":20,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":146518,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-7595227/v1/99d4fa567f6a41311b111308.html"},{"id":93697599,"identity":"b4fa7265-ac9e-4c33-8077-75ebcf57c702","added_by":"auto","created_at":"2025-10-16 14:59:12","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":348179,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003e(a)\u003c/strong\u003e Map of mosquito trap locations and ZIP Code Tabulation Areas (ZCTAs) in Maricopa County. Each point represents a CO₂-baited mosquito trap monitored weekly by the Maricopa County Environmental Services Vector Control Division. Traps are spaced approximately 1 km² apart and record species- and sex-specific mosquito counts during 12-hour collections. For this study, we aggregated adult female Cx. quinquefasciatus counts at the ZCTA level to analyze spatial patterns in mosquito abundance and perform climate-based clustering of ZCTAs for model training and inference. \u003cstrong\u003e(b) \u003c/strong\u003eClustering result. 2 clusters represent six outlier ZCTAs (red) versus all others (purple), whereas the 5 and 9 clusters were clustered based on precipitation data of all other ZCTAs.\u003c/p\u003e","description":"","filename":"Fig.1.png","url":"https://assets-eu.researchsquare.com/files/rs-7595227/v1/00ce191817dce7d2aea9a5d8.png"},{"id":93700699,"identity":"65e90d1d-9770-4e96-a98e-b8907691311a","added_by":"auto","created_at":"2025-10-16 15:31:12","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":270842,"visible":true,"origin":"","legend":"\u003cp\u003eDaily climate variables in Maricopa County from 2014 to 2016. \u003cstrong\u003e(a)\u003c/strong\u003e Daily maximum temperature (°C) and \u003cstrong\u003e(b)\u003c/strong\u003e daily precipitation (mm), both averaged across all ZCTAs within Maricopa County using 4 km × 4 km PRISM gridded data. These aggregated time series were used as climate inputs for both county- and cluster-level mosquito abundance models. The different colors represent temperature and precipitation for each ZCTA.\u003c/p\u003e","description":"","filename":"Fig.2.png","url":"https://assets-eu.researchsquare.com/files/rs-7595227/v1/8f1fdd53d2db8efa630f0a36.png"},{"id":93699582,"identity":"1db43949-cd1c-403d-9de4-8191f43f5c0c","added_by":"auto","created_at":"2025-10-16 15:23:12","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":171873,"visible":true,"origin":"","legend":"\u003cp\u003eSpatio-temporal patterns in precipitation and mosquito abundance across five climate-based clusters of ZCTAs in Maricopa County. \u003cstrong\u003e(a)\u003c/strong\u003e 30-day accumulated precipitation (mm), showing differences in precipitation magnitude and timing among clusters. \u003cstrong\u003e(b)\u003c/strong\u003eWeekly mosquito abundance, highlighting cluster-level variation in response to climatic conditions, particularly during the 2014 monsoon season and mid-to-late 2015.\u003c/p\u003e","description":"","filename":"Fig.3.png","url":"https://assets-eu.researchsquare.com/files/rs-7595227/v1/cdf90da2c1a0c3f104c3422e.png"},{"id":93697605,"identity":"3e732a36-dea1-43e5-ad86-81f696f5f58f","added_by":"auto","created_at":"2025-10-16 14:59:12","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":216663,"visible":true,"origin":"","legend":"\u003cp\u003eComparison of clustering prediction, real data, and county prediction for 5 (top) and 9 (bottom) clusters. Shaded ribbons in blue indicate 95% credible intervals for county prediction, while shaded ribbons in red indicate 95% credible intervals for cluster sum prediction. The black dotted lines represent observed mosquito abundance data over time.\u003c/p\u003e","description":"","filename":"Fig.4.png","url":"https://assets-eu.researchsquare.com/files/rs-7595227/v1/d2b1967193e430874fdf76cc.png"},{"id":104739319,"identity":"6d3c4904-f6dc-49d9-9cc0-0f0300bbb352","added_by":"auto","created_at":"2026-03-16 16:01:53","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1781197,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7595227/v1/17dad0cf-e9f2-4250-b6f5-13e8bb4a0ef4.pdf"},{"id":93697627,"identity":"e0f0c4f0-3630-48f7-a341-0384072ba321","added_by":"auto","created_at":"2025-10-16 14:59:15","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":84750516,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSupplementary information\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAdditional file 1: Text S1.\u003c/strong\u003e Some details about outlier ZCTAs. \u003cstrong\u003eText S2.\u003c/strong\u003e Hierarchical model inference across clusters. \u003cstrong\u003eText S3.\u003c/strong\u003e Formulation of likelihood function using negative binomial probability distribution. \u003cstrong\u003eText S4.\u003c/strong\u003e Interpretation of the outliers' ZCTAs results. \u003cstrong\u003eText S5.\u003c/strong\u003e Model evaluation for 2 clusters. \u003cstrong\u003eFig. S1.\u003c/strong\u003e Temperature and precipitation response functions used in the model. We used 2014 temperature and precipitation data from PRISM. \u003cstrong\u003eFig. S2.\u003c/strong\u003e Comparison of clustering prediction, real data, and county prediction for 2 clusters. \u003cstrong\u003eFig. S3.\u003c/strong\u003e A spatio-temporal plot of mosquito abundance per ZCTA. \u003cstrong\u003eFig. S4.\u003c/strong\u003e Four cluster result. \u003cstrong\u003eFig. S5.\u003c/strong\u003e Relationship between estimated baseline mosquito population growth rate and average mosquito abundance across clusters. \u003cstrong\u003eFig. S6.\u003c/strong\u003e 5 Cluster prediction. \u003cstrong\u003eFig. S7.\u003c/strong\u003e 9 Cluster prediction. \u003cstrong\u003eFig. S8\u003c/strong\u003e. 9 Cluster 2013 prediction. \u003cstrong\u003eFig. S9.\u003c/strong\u003e 9 Cluster 2017 prediction. \u003cstrong\u003eFig. S10.\u003c/strong\u003e 2 Cluster prediction. \u003cstrong\u003eFig. S10-S15.\u003c/strong\u003e Histograms of marginal posterior draws of parameters. \u003cstrong\u003eFig. S16-S20. \u003c/strong\u003ePairs plot displaying joint posterior draws of parameters. \u003cstrong\u003eFig. S21.\u003c/strong\u003e Predicted versus actual for all clusters. \u003cstrong\u003eTable S1.\u003c/strong\u003e Parameter estimated value and its prior for one cluster. \u003cstrong\u003eTable S2.\u003c/strong\u003e Parameter estimated value and its prior for more than one cluster. \u003cstrong\u003eTable S3.\u003c/strong\u003e Model evaluation at cluster level for five clusters. \u003cstrong\u003eTable S4.\u003c/strong\u003e Model evaluation at cluster level for nine clusters.\u003c/p\u003e","description":"","filename":"MosqAbundsupp.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7595227/v1/d2b46eddef774bfc538a2e2a.pdf"},{"id":93699201,"identity":"ac1274ef-efa4-4885-b089-73555da499ed","added_by":"auto","created_at":"2025-10-16 15:15:12","extension":"tiff","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":95515,"visible":true,"origin":"","legend":"","description":"","filename":"abstactfigure.tiff","url":"https://assets-eu.researchsquare.com/files/rs-7595227/v1/83719d92cafbab6b293d33fa.tiff"}],"financialInterests":"No competing interests reported.","formattedTitle":"Accounting for spatial variation in climatic factors predicts spatial variations in mosquito abundance in the desert southwest","fulltext":[{"header":"Background","content":"\u003cp\u003eMosquitoes are carriers of many pathogens that have global impacts, including West Nile virus (WNV), Yellow Fever Virus, Rift Valley Virus, Zika Virus, Chikungunya Virus, Dengue, and Malaria, which cause more than 700,000 deaths annually [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]. For example, as of 2022, 249\u0026nbsp;million cases of malaria were identified worldwide [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]. Dengue virus is widespread in over 100 countries, with annual global infections of 390\u0026nbsp;million persons [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. Furthermore, one in every 150 people infected with WNV is diagnosed with neuroinvasive WNV disease, with many suffering from long-lasting secondary sequelae (e.g., depression, memory loss, and motor dysfunction). As ectotherms and complex-life-cycle organisms, mosquito population dynamics are invariably linked to climate and local environmental conditions [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e, \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e], such as temperature, precipitation, humidity, and the availability of standing water for breeding. It is well documented from laboratory studies that temperature drives the rates of many key demographic traits across different mosquito life stages and across many mosquito species [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]. Interestingly, humidity can alter the shapes of these thermal performance relationships, indicating that interactions between climate factors will also impact population dynamics [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. For some mosquito species, the availability of standing water for breeding helps explain trends in population sizes, such that cumulative measures of precipitation are often used as proxies to model mosquito populations. A recent study in [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e] analyzed the relationships between mosquito population size, irrigation, and WNV cases in humans, finding an association between mosquito population and irrigation activities. Accordingly, climate factors show statistical relationships with the burdens of many mosquito-borne diseases. For instance, statistical modeling studies often use climate factors to explain and predict the cases of WNV in humans [\u003cspan additionalcitationids=\"CR10 CR11 CR12 CR13\" citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]. Based on the relationships between climate, mosquito population demography, and disease burden, more recent studies attempt to project the effects of climate change on the future global distributions of mosquitoes and the diseases they harbor [\u003cspan additionalcitationids=\"CR16 CR17 CR18\" citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]. Given the complicated nature of climate\u0026rsquo;s effects on mosquito population dynamics, flexible modeling approaches are needed to more mechanistically explain how multiple climate variables influence variation in mosquito abundance patterns over space and time.\u003c/p\u003e\u003cp\u003eMosquito population dynamics can fluctuate at relatively small spatial scales, likely because local weather variation leads to spatial and temporal variation in mosquito abundances [\u003cspan additionalcitationids=\"CR21\" citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]. For instance, even across a few hundred squared kilometers, mosquito population sizes can vary spatially and seasonally, potentially due to differences in local precipitation and temperature, and the effects of weather may even vary among co-occurring mosquito species [\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e]. An important challenge is therefore not only to build mechanistic models that can explain temporal trends in mosquito population fluctuations, but that also help us interrogate the factors that explain spatial heterogeneities in these abundance patterns.\u003c/p\u003e\u003cp\u003eMany methods are used to model the spatial and temporal fluctuations in mosquito population abundances, ranging from purely mathematical models (e.g., compartmental differential equation models) to purely statistical models (e.g., linear regression frameworks). Developing an adequate mathematical or statistical model to explain or predict mosquito population dynamics in relation to climate is a challenging endeavor with various trade-offs. Purely statistical models, including statistically-based machine learning algorithms, can do a good job of explaining temporal and spatial trends in mosquito abundances [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e, \u003cspan additionalcitationids=\"CR25 CR26 CR27\" citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e]. A challenge with statistical modeling approaches is that they typically are inadequate representations of the well-known non-linear effects of climate on complex population dynamics. Additionally statistical models are generally correlational, meaning that at least some mechanistic understanding of the system is lost. Classic [\u003cspan additionalcitationids=\"CR30\" citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e] and contemporary mathematical models often introduce large sets of differential equations that represent the various life-stages of mosquitoes, from egg to larvae to adult, and they can include the effects of temperature on certain demographic rates [\u003cspan additionalcitationids=\"CR33 CR34 CR35\" citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e]. Such models have the strong benefit of inherently including non-linear effects of climate on demographic patterns, but they also have some challenges. It is difficult to derive realistic estimates for the many parameters in such large models (i.e., a dimensionality problem). Parameterizing such models therefore often relies on laboratory experiments that measure thermal performance of mosquitoes, but such parameter estimates are largely untested in field settings (although see [\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e]). Fortunately, recent mechanistic modeling and model-fitting approaches show promise in incorporating both temperature and precipitation to explain dynamic mosquito population abundance patterns [\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eIn our study, we seek to develop a mechanistic yet simplified way of incorporating temperature and precipitation into a generic and dynamic modeling approach. This modeling approach allows us to explore if spatiotemporal variation in weather explains the non-linear dynamics of mosquito populations in natural landscapes. We test our modeling approach using the study system of \u003cem\u003eCx. quinquefasciatus\u003c/em\u003e in Maricopa County, AZ, which transmits multiple infections to humans, and is the most abundant vector for West Nile and St. Louis Encephalitis (SLEV) viruses in the county. According to the CDC, Maricopa County is consistently in the top 10 counties yearly in the nation for West Nile virus cases [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]. From 1999 to 2022, Maricopa County has had the most human cases in the nation, with a total of 3006 infections. Hundreds of traps are set weekly across the county, and mosquito abundances can vary dramatically in magnitude across subdivisions of the county. Hence, this system is ideal for testing how a combination of weather factors may influence spatiotemporal heterogeneity in mosquito population dynamics. We therefore hypothesized that accounting for microclimate variation\u0026mdash;by grouping areas with similar weather patterns\u0026mdash;would improve our model\u0026rsquo;s ability to predict mosquito abundance across the county. This modeling study is important because it helps us understand how spatial variation in climate affects spatial variation in mosquito population sizes. Moreover, in future work, the simplified modeling approach we develop here could be incorporated into epidemiological models to account for important climate and spatial variation in disease transmission.\u003c/p\u003e"},{"header":"Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003eApproach Overview\u003c/h2\u003e\u003cp\u003eFirst, we develop a semi-mechanistic model of the temporal dynamics of adult mosquito abundance, which incorporates non-linear statistical effects of temperature and accumulated precipitation. We then use Bayesian Markov chain Monte Carlo (MCMC) to fit our model to three years of weekly, adult female \u003cem\u003eCx. quinquefasciatus\u003c/em\u003e abundance observations across Maricopa County (i.e., the abundance data aggregated across all traps established in the county). To test whether spatial heterogeneity in climate explains variation in the patterns of mosquito abundance across Maricopa County, we cluster the county\u0026rsquo;s zip codes by local similarities in climate, and aggregate the mosquito abundance across traps within these clusters. We develop a method to hierarchically re-fit our model to abundance data in each cluster. Next, we compare whether fitting the data per cluster and then aggregating these fits to the county-level improves the overall fit of the county-level data. In this way, we see if heterogeneity in climate-based spatial clusters can help us better explain the county-level aggregated data. Finally, we also evaluate the out-of-sample prediction accuracy of our models, at both county and cluster scales. Below, we describe each step of the process in detail.\u003c/p\u003e\u003c/div\u003e\n\u003ch3\u003eClimate-forced mathematical model\u003c/h3\u003e\n\u003cp\u003eWe developed a mathematical model that explains temporal patterns of adult mosquito abundance, which is driven by two exogenous climatic variables: temperature and precipitation. The general form of the ordinary differential equation is\u003cdiv id=\"Equa\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equa\" name=\"EquationSource\"\u003e\n$$\\:\\frac{dX\\left(t\\right)}{dt}=\\nu\\:f\\left(\\text{Temp}\\right(t),\\text{Prcp}(t\\left)\\right)-\\mu\\:X\\left(t\\right).$$\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003e\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:X\\left(t\\right)\\)\u003c/span\u003e\u003c/span\u003e is the mosquito abundance over time, with model time steps discretized to daily as \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:X\\left({t}_{i}\\right)\\)\u003c/span\u003e\u003c/span\u003e where \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:i=1,2,3,\\dots\\:\\)\u003c/span\u003e\u003c/span\u003e. The model assumes that there is a baseline population growth rate of mosquitoes, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\nu\\:\\)\u003c/span\u003e\u003c/span\u003e, which is modulated by a statistical function of daily measures of temperature \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\text{Temp}\\left(t\\right)\\)\u003c/span\u003e\u003c/span\u003e and precipitation \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\text{Prcp}\\left(t\\right)\\)\u003c/span\u003e\u003c/span\u003e. Specifically, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\text{Temp}\\left(t\\right)\\)\u003c/span\u003e\u003c/span\u003e is the daily maximum temperature in degrees Celsius, and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\text{Prcp}\\left(t\\right)\\)\u003c/span\u003e\u003c/span\u003e is the daily measure of 30-day accumulated millimeters of precipitation, which we scale for statistical convenience by dividing by a constant 50.\u003c/p\u003e\u003cp\u003eThe form of the statistical function is:\u003cdiv id=\"Equb\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equb\" name=\"EquationSource\"\u003e\n$$\\:f\\left(\\text{Temp}\\right(t),\\text{Prcp}(t\\left)\\right)=-\\left(\\text{Temp}\\right(t)-Tmin)\\left(\\text{Temp}\\right(t)-Tmax))\\frac{1}{(1+{\\text{e}}^{\\alpha\\:-\\varphi\\:\\text{Prcp}\\left(t\\right)})}$$\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eThe temperature effect is modeled as a quadratic equation, representing a thermal performance curve of population growth rate, controlled by a minimum \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:Tmin\\)\u003c/span\u003e\u003c/span\u003e and a maximum \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:Tmax\\)\u003c/span\u003e\u003c/span\u003e growth temperature. This structure is agnostic to the fact that many traits of mosquito individuals across their life stages have unique thermal performance curves [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]. Our strategy is instead to test a simplification by estimating a parsimonious representation of thermal effects on the overall growth rate of the adult mosquito population, integrating across the thermal effects of specific, life-stage specific traits. The precipitation effect is modeled as a logistic, saturating function, whose shape is controlled by constants \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\alpha\\:\\)\u003c/span\u003e\u003c/span\u003e and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\varphi\\:\\)\u003c/span\u003e\u003c/span\u003e. Biologically, this means that as accumulated precipitation increases, the growth rate increases, but the model could estimate how quickly this increase occurs across the range of precipitation and whether there is a saturating effect, above which more accumulated precipitation has negligible effects. Illustrative plots of the temperature and precipitation functions and their corresponding seasonal time series are shown in Additional file 1: Fig. \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e.\u003c/p\u003e\u003cp\u003eWe fix the per‑capita mortality rate \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\mu\\:\\)\u003c/span\u003e\u003c/span\u003e (at 0.12) instead of estimating it as the ratio between the growth term and the mortality term are the only factors that govern the net population change; estimating both would create an identifiability issue. Consequently, fluctuations in the fitted growth term implicitly capture both births and deaths.\u003c/p\u003e\u003cp\u003eIn Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e, we described each parameter in our mathematical model.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eParameters of the model\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"4\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eParameter\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eSignification of parameters\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eUnit\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eSource\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cem\u003eν\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eBaseline population growth rate\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\frac{\\text{mosquitoes}}{{}^{\\circ\\:}C\\cdot\\:\\text{day}}\\)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eEstimated\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cem\u003eTmin\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eMinimum temperature that constrains population growth\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{}^{\\circ\\:}C\\)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eEstimated\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cem\u003eTmax\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eMaximum temperature that constrains population growth\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{}^{\\circ\\:}C\\)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eEstimated\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cem\u003eα\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eContributes to the inflection point of the precipitation effect curve\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eUnitless\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eEstimated\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cem\u003eφ\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eSteepness of the precipitation effect curve\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1/mm\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eEstimated\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cem\u003e\u0026micro;\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eMosquito death rate\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1/day\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eFixed\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eModel evaluation comparison at the county level\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"4\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eModel\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eFitted RMSE (2014\u0026ndash;2016)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003ePredicted RMSE (2013)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003ePredicted RMSE (2017)\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCounty without \u0026rsquo;outliers\u0026rsquo; ZCTAs\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e252.6900\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e306.0828\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e494.7659\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e5 clusters sum\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e265.8615\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e325.5769\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e507.0085\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e9 clusters sum\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e269.4774\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e349.3141\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e511.3981\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\n\u003ch3\u003eMosquito abundance data\u003c/h3\u003e\n\u003cp\u003eThe mosquito abundance data used in this article was collected by the Maricopa County Environmental Services Vector Control Division [\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e]. The data is collected weekly from \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\sim\\:800\\)\u003c/span\u003e\u003c/span\u003e \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\text{CO}}_{2}\\)\u003c/span\u003e\u003c/span\u003e traps located in an approximate 1km\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{}^{2}\\)\u003c/span\u003e\u003c/span\u003e grid across most of the county. The longitude and latitude of the traps locations is documented for easy spatial analysis. Each trap is placed for a 12-h collection period once per week for 50 weeks of the year. Collections are subsequently sorted by mosquito species and sex. Data used in this study were restricted to \u003cem\u003eCx. quiquefasciatus\u003c/em\u003e, the most frequently trapped mosquito vector in Maricopa County, and is known to carry avian malaria, WNV, SLEV, and other pathogens. This species constitutes about 80% of the total infected mosquitoes in Maricopa County.\u003c/p\u003e\u003cp\u003eIn our analysis, we used a training data set spanning three years, January 1, 2014 to December 31, 2016, and we used 2013 and 2017 data to validate our model as an out-of-sample prediction data set. We chose 2014\u0026ndash;2016 for the training set, because the data includes unique patterns in abundance that can help validate the effects of climate. For instance, in 2014 there was a large monsoon season, and this effect can be visually observed in the abundance data, while years 2015 and 2016 had relatively average abundance trends. For the county-level data set, mosquito abundance per week was aggregated across all traps. We used data for adult female mosquitoes only since they were blood-seeking and attracted to \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\text{CO}}_{2}\\)\u003c/span\u003e\u003c/span\u003e traps. Each trap is identified by GPS coordinates (longitude and latitude), such that for the cluster-level data sets, we first aggregated trap data to the ZIP code-level (i.e., ZIP Code Tabulation Areas, ZCTA), and then clustered ZCTA by climate similarity and aggregated trap data accordingly (see Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ea).\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\n\u003ch3\u003eTemperature and precipitation data\u003c/h3\u003e\n\u003cp\u003eDaily estimates of maximum temperature in Celsius and precipitation in millimeters were downloaded from PRISM via the prism R package [\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e]. We downloaded the Prism daily data in a 4 km by 4 km raster. For the county-level model, we averaged daily values across all the Prism data in the raster within the Maricopa County polygon defined by the 2010 Census Bureau delineations (tigris package [\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e]) in Maricopa County. For the cluster-level model, we averaged the Prism data across the grid within each ZCTA polygon, again using 2010 Census Bureau delineations. Then, we averaged these data across all ZCTAs included in a given cluster. In the model, we used a 30-day accumulated precipitation, scaled by dividing by the constant 50. In Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e, we present the time series plot for temperature and precipitation for 2014\u0026ndash;2016 which is averaged across ZCTA in Maricopa County.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eWe explored the spatial distribution of the mosquito abundance data present within the 109 non-outlier ZCTAs across Maricopa County (see details of outlier ZCTAs in Additional file 1: Text S1). Our goal was to cluster these ZCTAs based on similarities in climatic data during the three-year window of the training data set. We used only the precipitation time-series data, as the temperature time-series did not vary significantly across the county (see Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). Using the precipitation data, we clustered the ZCTAs based on a functional time series clustering technique implemented in the FunFEM package in R [\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e]. This clustering technique requires the user to specify the number of clusters desired. Therefore, we tested 5 and 10 clusters. Briefly, to apply functional time-series clustering, first, a basis function must be fit to the time-series data to smooth it. In our case, we used a Fourier basis, which is the most appropriate for periodic patterns, such as those observed in the precipitation data [\u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eThe functional clustering algorithm compares time-series data by transforming them into functional forms, often using basis expansions like splines or Fourier series, to capture their continuous-time dynamics [\u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e]. It then measures similarity based on shared features such as trends, seasonal patterns, or amplitude variations using a model-based clustering approach in a discriminative functional subspace, where differences between curves are captured via latent variable modeling and Gaussian mixture models, primarily relying on functional Principal Component Analysis (PCA) and Fisher-like discriminative analysis. Time series with similar functional properties are grouped into contiguous clusters, allowing for the identification of common behaviors and trends across the data while accounting for variability in shape, scale, and temporal alignment.\u003c/p\u003e\u003cp\u003eNote that when we specified 10 clusters, one cluster only included four ZCTA, each with a maximum mosquito abundance per week of less than ten. Therefore, we manually moved those to the most similar other cluster, leaving us with a comparison of 5 versus 9 clusters (see middle and last panel of Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eb).\u003c/p\u003e\n\u003ch3\u003eFitting the model to the time-series data\u003c/h3\u003e\n\u003cp\u003eWe used a Metropolis-Hastings Markov chain Monte Carlo (MCMC) algorithm to sample from the joint posterior distribution of model parameters. To calculate the likelihood, we compare the solution of our ordinary differential equation model to the weekly observed abundance data. We solve the differential equation model with the deSolve [\u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e] package using the method lsoda [\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e] in R, incorporating daily fluctuations in temperature and precipitation. We compare the model to the data every seven days; therefore, while the model ingests daily data on climate, the model is only compared to the observed mosquito data once per week. The likelihood is defined as a negative binomial probability distribution with an estimated over-dispersion parameter (see Additional file 1: Text S2-S3 for more details). For each spatial data set, we ran four MCMC chains, each with 5,000 iterations, in parallel using the futures R package [\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e]. We used the \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\widehat{R}\\)\u003c/span\u003e\u003c/span\u003e statistic [\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e], also known as the Gelman-Rubin diagnostic, to assess the convergence of the chains. We also performed posterior predictive checks, described below, to ensure the joint posterior defined a reasonable parameter estimation space.\u003c/p\u003e\u003cp\u003eWe fit the model to the following 5 data partitions: county-level with outliers, county-level without outliers, 2-cluster (outlier ZCTA versus all other ZCTA), 5-cluster without outliers, and 9-cluster without outliers. For the two county-level partitions, we run a standard MCMC process. For the cluster-level partitions, we ran a hierarchical inference approach.\u003c/p\u003e\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\u003ch2\u003eModel validation\u003c/h2\u003e\u003cp\u003eWe used within-sample and out-of-sample predictive measures to compare model performance across the different data partitions. We are ultimately interested in understanding if accounting for spatial variation in climate and baseline growth rates might improve the model\u0026rsquo;s explanation of county-wide, total mosquito abundances. Therefore, in calculating goodness-of-fit metrics, for the cluster-level data partitions, we summed metrics across clusters to a county-level performance metric.\u003c/p\u003e\u003cp\u003eThe method we employed for goodness-of-fit measurement was Root Mean Square Error (RMSE). To generate this measurement for each data partition and model-fit, we first generate posterior predictive model runs. For each data partition and MCMC analysis, we randomly sampled 100 parameter sets from the joint posterior. For each set, we substitute the ODE to generate the corresponding numerical solution. Across these 100 model runs, we calculated the median model expectation per day for the model. The RMSE was then calculated by comparing the data observations to the median model expectation, and summing across all observations, and across clusters, when appropriate.\u003c/p\u003e\u003c/div\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec10\" class=\"Section2\"\u003e\u003ch2\u003eVariation in precipitation and temperature\u003c/h2\u003e\u003cp\u003eThe clustering routine captured logical spatial groupings of zip codes (ZCTA) across Maricopa County based on spatial and temporal variation in 30-day accumulated precipitation (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e and Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). When we look at the PRISM-derived precipitation metric, we see that clusters are mostly differentiated by precipitation magnitude and more minimally by precipitation timing. For instance, in the 5-cluster case, all clusters experienced relatively high abundance in the monsoon season of 2014, with minor variation in total magnitude, whereas in 2015, there is some variability in the timing of precipitation, especially in the second half of the year (see Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e).\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\u003ch2\u003eComparing county- and cluster-level modeling\u003c/h2\u003e\u003cp\u003eIn general, our mechanistic model captures detailed patterns of mosquito abundance changing through time across multiple spatial scales for the three years of training data, particularly when we remove the six outlier ZCTA (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e and Additional file 1: Fig. S6-S7). For example, the model generally captures the effects of high summer temperatures, where we see corresponding declines in mosquito abundances, and also the effects of monsoonal activity, where we can see dramatic increases in mosquito abundances due to heavy rainfall (e.g., in the second half of 2014).\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eAccounting for heterogeneity in precipitation across clustered ZCTA lead to higher precision in model fits we aggregated model expectations from cluster to county-scales (i.e., narrower error bands); however, these aggregated model fits were sometimes overly confident, such that the error bands did not always capture the true observations (see Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e). Thus, the county-level model, which used daily values for temperature and precipitation that were averaged across the ZCTA, had a relatively better fit compared to the aggregated cluster-level models, based on the Root Mean Squared Error (RMSE) metric (Additional file 1: Table S3-S4). Furthermore, there were not clear differences in the aggregated county-level model fit when we compare the 5-cluster to the 9-cluster situations (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e). When we examine model fits per cluster, however, we do see that the model captures variation in mosquito abundances across clusters, explained mostly by differences in baseline population growth rates (Additional file 1: Fig. S6-S7). Also, we observed that the RMSE values vary across clusters depending on the number of clusters (Additional file 1: Table S3-S4). This result demonstrates that our method of hierarchical estimation of cluster-specific growth rates successfully characterized key differences among clusters. Collectively, the results suggest that the aggregate, county-level patterns in mosquito abundance can be most parsimoniously explained by a spatial average of temperature and precipitation data. Yet the cluster-level models provide accurate, finer-scale inference of how mosquito population varies across space and time, though at this finer scale the model fits may be more prone to small prediction errors. Additional details about outliers and county level analysis can be found in Additional file 1: Text S4-S5.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\u003ch2\u003eOut of sample performance\u003c/h2\u003e\u003cp\u003eWhen we predicted mosquito abundance for two additional years of withheld data (2013 and 2017), our model still captures the fundamental characteristics of these mosquito abundance time-series (see Additional file 1: Fig. S6 (first and last panel), S8-S9). Indeed, the key results from the training data still hold true with the out-of-sample data. Particularly, the county-level model performs nearly equally as well as the cluster-level models. Notably, however, across spatial scales, the dynamic model does a poor job at explaining the early-year increases in mosquito abundance observed in 2017, and under-predicts peaks in abundances observed in the later half of both 2013 and 2017. We observed these failures of the model in both the county- and Cluster-level out-of-sample model fits.\u003c/p\u003e\u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eThis research sheds light on the complex relationships between climate, spatial heterogeneities, and mosquito populations. By implementing a novel mechanistic model with a data-driven clustering strategy, we show that a relatively simple mathematical model that incorporates temperature and precipitation effectively captures the dynamical patterns of mosquito abundance observed over multiple years in a large urban setting. When we fit our model to aggregate data for all of Maricopa County, our climate-driven model accurately explains the seasonal increase in mosquito abundances in the spring, a decline in the hottest part of the summer, and another increase in the warm, wet monsoon season. When we subdivided the county into clusters based on similar climate patterns, we observe that average mosquito abundances vary substantially across the county. However, the cluster-level model still explains a substantial proportion of variance in the these downscaled data, indicating that mosquito populations across the region exhibit consistent numerical responses to temperature and precipitation. When we aggregate these smaller-scale model fits, we improve precision around the median model prediction, but we do not improve county-level model accuracy\u0026mdash;likely due to error accumulation across clusters and the sparse data within each group. What this tells us is that the same climate-forced model can be applied at sub-county spatial scales and will explain the key dynamical patterns of the mosquito population dynamics; yet, at these smaller spatial scales, the model will be more prone to observational error, suggesting a trade-off between spatial resolution and model accuracy. Our results underscore the importance of modeling spatial structure when interpreting mosquito dynamics, particularly in regions like the desert Southwest where precipitation and temperature vary dramatically over small spatial scales.\u003c/p\u003e\u003cp\u003ePrevious studies have demonstrated that local weather conditions influence mosquito abundance and distribution [\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e, \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e, \u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e, \u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e]. Research on \u003cem\u003eCx. quinquefasciatus\u003c/em\u003e, the focal species in our study, has highlighted how temperature and precipitation shape its life cycle. For example, a study in Hawaii [\u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e49\u003c/span\u003e] found that \u003cem\u003eCx. quinquefasciatus\u003c/em\u003e abundance increased with temperature, peaking at mid-elevation sites, and showed a non-linear relationship with precipitation\u0026mdash;highlighting temperature as a key driver of mosquito distribution across varying terrains. Precipitation also played a significant, albeit complex, role. Interestingly, the relationship between precipitation and abundance indicated negative effects of rainfall, while lagged precipitation showed positive associations. This complexity likely reflects seasonal rainfall patterns and their influence on larval habitat availability. Similarly, Morin and Comrie [\u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e50\u003c/span\u003e] used the Dynamic Mosquito Simulation Model (DyMSiM) model to show that the interaction between temperature and precipitation drives mosquito dynamics differently in California and Florida. Valdez et al. [\u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e51\u003c/span\u003e] further revealed that not only total rainfall, but also the number of rainy days and daily variability, strongly influence mosquito abundance. Our findings corroborate these earlier studies and extend them by explicitly testing whether accounting for spatial climate heterogeneity improves the performance of a mechanistic model.\u003c/p\u003e\u003cp\u003eAlthough the cluster-level model explains key features of the mosquito population dynamics at smaller spatial scales, aggregating cluster-level predictions to the county scale did not improve inference. A possible contributing reason for this contradiction is that the data we used per mosquito trap (PRISM), interpolates weather patterns across space based on sparse station networks and provides 4x4km grid raster data. In Arizona, localized monsoon events produce intense, short-duration rainfall that varies significantly over short distances. PRISM's data resolution may smooth over these extremes, missing key signals needed for accurate prediction [\u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e52\u003c/span\u003e, \u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e53\u003c/span\u003e]. Future studies could consider deploying ground-based weather stations near the mosquito traps to obtain high-resolution climate data. This could enhance the accuracy of fine-scale model predictions. Moreover, model error may accumulate during the aggregation process, particularly when fit to clusters with small sample sizes. While the cluster-aggregated model did not improve county-wide predictions, it remains valuable for local inference and public health planning. Local-level fits can inform targeted mosquito control strategies and enhance situational awareness.\u003c/p\u003e\u003cp\u003eIn our framework, we estimated a shared climate-response function for the mosquito population growth rate across all the clusters \u0026ndash; the quadratic effect of temperature and the exponentially saturating effect of accumulated precipitation. Given that this same functional response explains the data across clusters, this provides evidence that mosquito sub-population growth rates are responding similarly to weather despite differing average abundances. However, the current approach does not account for mechanisms that may explain the average differences abundance across clusters. Future research should explore how landscape features, water availability, or urban infrastructure might explain baseline variation in mosquito abundances.\u003c/p\u003e\u003cp\u003eWe also encountered six zip codes with unusually high mosquito counts, especially in spring, which the model failed to capture even when fit separately. These outlier patterns were not explained by temperature or 30-day precipitation trends, suggesting local landscape effects such as standing water accumulation may be at play. Investigating these features\u0026mdash;e.g., water-retaining infrastructure, wetlands, or irrigation systems\u0026mdash;could yield important insights. Future work could incorporate additional covariates to explain baseline mosquito growth rates, such as proxies for standing water, land use, and hydrological features. High-resolution, daily updated data on these variables could better capture early-season surges in abundance. Additionally, spatial movement of mosquitoes\u0026mdash;whether natural or human-facilitated\u0026mdash;should be considered in spatially explicit models.\u003c/p\u003e\u003cp\u003eSpatial processes are known to structure ecological and epidemiological dynamics, influencing everything from vector distribution to disease transmission and species interactions [\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e, \u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e54\u003c/span\u003e]. Invasive species spread, habitat fragmentation, and local resource competition all highlight the importance of spatial structure in population persistence and disease risk [\u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e55\u003c/span\u003e]. By explicitly incorporating spatial heterogeneity, our model represents a step forward in mosquito modeling efforts. As weather forecasting capabilities continue to improve, spatially structured models will be critical for translating environmental changes into actionable public health responses.\u003c/p\u003e\u003cp\u003eClimate change is reshaping ecosystems and disease risk globally [\u003cspan additionalcitationids=\"CR17\" citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]. Rising temperatures and altered precipitation patterns create more favorable conditions for mosquito proliferation and disease transmission [\u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e56\u003c/span\u003e, \u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e57\u003c/span\u003e]. Accurate models that incorporate both temporal and spatial heterogeneity are essential to predict these dynamics and guide interventions.\u003c/p\u003e"},{"header":"Conclusions","content":"\u003cp\u003eIn sum, our study demonstrates the value of spatially resolved climate-driven models for understanding mosquito population dynamics. Although challenges remain in scaling predictions and capturing outlier behavior, this work lays a foundation for more refined, localized modeling efforts that can enhance mosquito control and public health preparedness in a changing climate.\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003e\u003cstrong\u003eAcknowledgement\u003c/strong\u003e\u003c/h2\u003e\n\u003cp\u003eThe authors would like to especially acknowledge the manuscript review provided by Brad J. Biggerstaff, Ph.D., of the Division of Vector-Borne Infectious Diseases, National Center for Zoonotic, Vector-Borne, and Enteric Diseases, Centers for Disease Control and Prevention (CDC), Fort Collins, CO. The contents are solely the responsibility of the authors and do not necessarily represent the official views of the Centers for Disease Control and Prevention or National Institutes of Health.\u003c/p\u003e\n\u003ch2\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/h2\u003e\n\u003cp\u003eResearch reported in this publication was supported by the National Institute of Allergy and Infectious Diseases of the National Institutes of Health under award number R01AI168144 to J.M and a subaward from the Pacific Southwest Regional Center of Excellence for Vector-Borne Diseases funded by the U.S. Centers for Disease Control and Prevention Cooperative Agreement 1U01CK000649 to C.M.H.\u003c/p\u003e\n\u003ch2\u003e\u003cstrong\u003eConflict of interest/Competing interests\u003c/strong\u003e\u003c/h2\u003e\n\u003cp\u003eThe authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.\u003c/p\u003e\n\u003ch2\u003e\u003cstrong\u003eEthics approval and consent to participate\u003c/strong\u003e\u003c/h2\u003e\n\u003cp\u003eThere is no ethical issue related to this research.\u003c/p\u003e\n\u003ch2\u003e\u003cstrong\u003eCode availability\u003c/strong\u003e\u003c/h2\u003e\n\u003cp\u003eCode used for analyzing the data for this research is available upon request.\u003c/p\u003e\n\u003ch2\u003e\u003cstrong\u003eAuthor contribution\u003c/strong\u003e\u003c/h2\u003e\n\u003cp\u003eConceptualization, K.O. and J.M.; methodology, K.O. and J.M.; software, K.O.; validation, K.O., Y.C., E.D., N.B., C.M.H., J.T., J.W., I.R., M.K., and J.M.; formal analysis, K.O.; investigation, K.O., Y.C., E.D., N.B., C.M.H., J.T., J.W., I.R., M.K., and J.M.; resources, J.M.; data curation, K.O. and J.M.; writing and original draft preparation, K.O. and J.M.; writing—review and editing, K.O., Y.C., E.D., N.B., C.M.H., J.T., J.W., I.R., M.K., and J.M.; visualization, K.O. and J.M.; supervision, J.M.; project administration, J.M. All authors have read and agreed to the published version of the manuscript.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eDisease Control C for, Prevention, et al. Centers for disease control and prevention (CDC). 2020.\u003c/li\u003e\n\u003cli\u003eWorld health organization. \u0026ldquo;Malaria.\u0026rdquo; (Published March 29, 2023. accessed September 5, 2024).\u003c/li\u003e\n\u003cli\u003eWorld health organization. \u0026ldquo;Dengue.\u0026rdquo; (Published March 17, 2023. accessed September 5, 2024).\u003c/li\u003e\n\u003cli\u003eMordecai EA, Cohen JM, Evans MV, Gudapati P, Johnson LR, Lippi CA, et al. Detecting the impact of temperature on transmission of zika, dengue, and chikungunya using mechanistic models. PLoS neglected tropical diseases. 2017;11:e0005568.\u003c/li\u003e\n\u003cli\u003eVan Wyk H, Eisenberg JN, Brouwer AF. Long-term projections of the impacts of warming temperatures on zika and dengue risk in four brazilian cities using a temperature-dependent basic reproduction number. PLOS Neglected Tropical Diseases. 2023;17:e0010839.\u003c/li\u003e\n\u003cli\u003eMordecai EA, Caldwell JM, Grossman MK, Lippi CA, Johnson LR, Neira M, et al. Thermal biology of mosquito-borne disease. Ecology letters. 2019;22:1690\u0026ndash;708.\u003c/li\u003e\n\u003cli\u003eBrown JJ, Pascual M, Wimberly MC, Johnson LR, Murdock CC. Humidity\u0026ndash;the overlooked variable in the thermal biology of mosquito-borne disease. Ecology letters. 2023;26:1029\u0026ndash;49.\u003c/li\u003e\n\u003cli\u003eKovach TJ, Kilpatrick AM. Irrigation increases and stabilizes mosquito populations and increases west nile virus incidence. Scientific Reports. 2024;14:19913.\u003c/li\u003e\n\u003cli\u003eFarooq Z, Rockl\u0026ouml;v J, Wallin J, Abiri N, Sewe MO, Sj\u0026ouml;din H, et al. Artificial intelligence to predict West Nile virus outbreaks with eco-climatic drivers. The Lancet Regional Health \u0026ndash; Europe. 2022 [cited 2024 Apr 26];17. Available from: https://www.thelancet.com/journals/lanepe/article/PIIS2666-7762(22)00063-1/fulltext\u003c/li\u003e\n\u003cli\u003eSmith KH, Tyre AJ, Hamik J, Hayes MJ, Zhou Y, Dai L. Using climate to explain and predict west nile virus risk in nebraska. GeoHealth. 2020;4:e2020GH000244.\u003c/li\u003e\n\u003cli\u003eRuiz MO, Chaves LF, Hamer GL, Sun T, Brown WM, Walker ED, et al. Local impact of temperature and precipitation on west nile virus infection in culex species mosquitoes in northeast illinois, USA. Parasites \u0026amp; vectors. 2010;3:1\u0026ndash;16.\u003c/li\u003e\n\u003cli\u003eBaker RE, Mahmud AS, Miller IF, Rajeev M, Rasambainarivo F, Rice BL, et al. Infectious disease in an era of global change. Nature Reviews Microbiology. 2022;20:193\u0026ndash;205.\u003c/li\u003e\n\u003cli\u003eParker N. Exploring the role of temperature and other environmental factors in west nile virus incidence and prediction in california counties from 2017\u0026ndash;2022 using a zero-inflated model. PLOS Neglected Tropical Diseases. 2024;18:e0012051.\u003c/li\u003e\n\u003cli\u003eAlbrecht L, Kaufeld KA. Investigating the impact of environmental factors on West Nile virus human case prediction in Ontario, Canada. Frontiers in Public Health. 2023 [cited 2024 Apr 26];11. Available from: https://www.frontiersin.org/journals/public-health/articles/10.3389/fpubh.2023.1100543/full\u003c/li\u003e\n\u003cli\u003eCaldwell JM, LaBeaud AD, Lambin EF, Stewart-Ibarra AM, Ndenga BA, Mutuku FM, et al. Climate predicts geographic and temporal variation in mosquito-borne disease dynamics on two continents. Nature communications. 2021;12:1233.\u003c/li\u003e\n\u003cli\u003eAliaga-Samanez A, Romero D, Murray K, Segura M, Real R, Olivero J. Potential climate change effects on the distribution of urban and sylvatic dengue and yellow fever vectors. Pathogens and Global Health. 2024;118:397\u0026ndash;407.\u003c/li\u003e\n\u003cli\u003eMedlock JM, Leach SA. Effect of climate change on vector-borne disease risk in the UK. The Lancet Infectious Diseases. 2015;15:721\u0026ndash;30.\u003c/li\u003e\n\u003cli\u003eCarlson CJ, Albery GF, Merow C, Trisos CH, Zipfel CM, Eskew EA, et al. Climate change increases cross-species viral transmission risk. Nature. 2022;607:555\u0026ndash;62.\u003c/li\u003e\n\u003cli\u003eRyan S J. Lippi A, Stewart-Ibarra AM. Mapping geographic and demographic shifts for container breeding mosquito-borne disease transmission suitability in central and south america in a warming world. PLOS Climate. 2024;3:1\u0026ndash;21. Available from: https://doi.org/10.1371/journal.pclm.0000312\u003c/li\u003e\n\u003cli\u003eReisen WK, Cayan D, Tyree M, Barker CM, Eldridge B, Dettinger M. Impact of climate variation on mosquito abundance in california. Journal of Vector Ecology. 2008;33:89\u0026ndash;98.\u003c/li\u003e\n\u003cli\u003eAlthouse BM, Hanley KA, Diallo M, Sall AA, Ba Y, Faye O, et al. Impact of climate and mosquito vector abundance on sylvatic arbovirus circulation dynamics in senegal. The American journal of tropical medicine and hygiene. 2015;92:88.\u003c/li\u003e\n\u003cli\u003eFerraccioli F, Riccetti N, Fasano A, Mourelatos S, Kioutsioukis I, Stilianakis NI. Effects of climatic and environmental factors on mosquito population inferred from West Nile virus surveillance in Greece. Scientific Reports. 2023 [cited 2024 Apr 26];13:18803. Available from: https://www.nature.com/articles/s41598-023-45666-3\u003c/li\u003e\n\u003cli\u003eBaril C, Pilling BG, Mikkelsen MJ, Sparrow JM, Duncan CA, Koloski CW, et al. The influence of weather on the population dynamics of common mosquito vector species in the canadian prairies. Parasites \u0026amp; Vectors. 2023;16:153.\u003c/li\u003e\n\u003cli\u003eLi B, Fu Q, Huang Y, Sun Q, Zhao C, Ma X, et al. Application of circular statistics in temporal distribution of adult mosquitoes in qingdao, shandong province, china, 2021\u0026ndash;2023. Parasites \u0026amp; Vectors. 2024;17:325.\u003c/li\u003e\n\u003cli\u003eRodrı́guez-Gonz\u0026aacute;lez S, Izquierdo-Suz\u0026aacute;n M, Rocha-Ortega M, C\u0026oacute;rdoba-Aguilar A. Vector mosquito distribution and richness are predicted by socio-economic, and ecological variables. Acta Tropica. 2024;254:107179.\u003c/li\u003e\n\u003cli\u003eVillena OC, McClure KM, Camp RJ, LaPointe DA, Atkinson CT, Sofaer HR, et al. Environmental and geographical factors influence the occurrence and abundance of the southern house mosquito, culex quinquefasciatus, in hawai \u0026lsquo;i. Scientific Reports. 2024;14:604.\u003c/li\u003e\n\u003cli\u003eGervasi R, Azzali I, Bisanzio D, Mosca A, Bertolotti L, Giacobini M. A genetic programming approach to predict mosquitoes abundance. European conference on genetic programming. Springer; 2019. p. 35\u0026ndash;48.\u003c/li\u003e\n\u003cli\u003eCuller LE, Ayres MP, Virginia RA. Spatial heterogeneity in the abundance and fecundity of arctic mosquitoes. Ecosphere. 2018;9:e02345. Available from: https://esajournals.onlinelibrary.wiley.com/doi/abs/10.1002/ecs2.2345\u003c/li\u003e\n\u003cli\u003eRoss R. The prevention of malaria 2nd edn (london: murray). 1911.\u003c/li\u003e\n\u003cli\u003eM\u0026rsquo;Kendrick AG. Applications of mathematics to medical problems. Proceedings of the Edinburgh Mathematical Society. 1925;44:98\u0026ndash;130.\u003c/li\u003e\n\u003cli\u003eMacdonald. G MG. The epidemiology and control of malaria. 1957.\u003c/li\u003e\n\u003cli\u003eDiabat\u0026eacute; AB, Sangar\u0026eacute; B, Koutou O. Mathematical modeling of the dynamics of vector-borne diseases transmitted by mosquitoes: Taking into account aquatic stages and gonotrophic cycle. Nonautonomous Dynamical Systems. 2022;9:205\u0026ndash;36.\u003c/li\u003e\n\u003cli\u003eDembele B, Friedman A, Yakubu A-A. Malaria model with periodic mosquito birth and death rates. Journal of biological dynamics. 2009;3:430\u0026ndash;45.\u003c/li\u003e\n\u003cli\u003eJia P, Lu L, Chen X, Chen J, Guo L, Yu X, et al. A climate-driven mechanistic population model of aedes albopictus with diapause. Parasites \u0026amp; vectors. 2016;9:1\u0026ndash;15.\u003c/li\u003e\n\u003cli\u003eEzanno P, Balenghien T, Cailly P, L\u0026rsquo;Ambert G, Toty C, Tran A-L, et al. A generic climate-driven model to predict mosquito population dynamics. 2014.\u003c/li\u003e\n\u003cli\u003eShutt D, Goodsman DW, Martinez K, Hemez ZJL, Conrad JR, Xu C, et al. A process-based model with temperature, water, and lab-derived data improves predictions of daily culex pipiens/restuans mosquito density. Journal of Medical Entomology. 2022;59:1947\u0026ndash;59.\u003c/li\u003e\n\u003cli\u003eFay RL, Cruz-Loya M, Keyel AC, Price DC, Zink SD, Mordecai EA, et al. Population-specific thermal responses contribute to regional variability in arbovirus transmission with changing climates. Iscience. 2024;27.\u003c/li\u003e\n\u003cli\u003eHepp CM, Cocking JH, Valentine M, Young SJ, Damian D, et al. Phylogenetic analysis of west nile virus in maricopa county, arizona: Evidence for dynamic behavior of strains in two major lineages in the american southwest. PLOS ONE. 2018;13:1\u0026ndash;12. Available from: https://doi.org/10.1371/journal.pone.0205801\u003c/li\u003e\n\u003cli\u003eHart EM, Bell K. Prism: Download data from the oregon prism project [Internet]. 2015. Available from: https://github.com/ropensci/prism\u003c/li\u003e\n\u003cli\u003eWalker KE. Tigris: An r package to access and work with geographic data from the US census bureau. R Project; 2016.\u003c/li\u003e\n\u003cli\u003eBouveyron C, Jacques J. funFEM: An r package for functional data clustering. Quatri\u0026egrave;me rencontres r. 2015.\u003c/li\u003e\n\u003cli\u003eRamsay JO, Hooker G, Graves S. Introduction to functional data analysis. Functional Data Analysis with R and MATLAB. 2009;1\u0026ndash;19.\u003c/li\u003e\n\u003cli\u003eSoetaert KE, Petzoldt T, Setzer RW. Solving differential equations in r: Package deSolve. Journal of statistical software. 2010;33.\u003c/li\u003e\n\u003cli\u003eHindmarsh AC, Petzold LR. LSODA, ordinary differential equation solver for stiff or non-stiff system. 2005.\u003c/li\u003e\n\u003cli\u003eBengtsson H. A unifying framework for parallel and distributed processing in r using futures. The R Journal. 2021;13:208\u0026ndash;27. Available from: https://doi.org/10.32614/RJ-2021-048\u003c/li\u003e\n\u003cli\u003eGelman A, Carlin JB, Stern HS, Dunson DB, Vehtari A, Rubin DB. Bayesian Data Analysis. 3rd ed. New York: Chapman; Hall/CRC; 2015.\u003c/li\u003e\n\u003cli\u003eZeng Q, Yu X, Ni H, Xiao L, Xu T, Wu H, et al. Dengue transmission dynamics prediction by combining metapopulation networks and kalman filter algorithm. PLOS Neglected Tropical Diseases. 2023;17:e0011418.\u003c/li\u003e\n\u003cli\u003eCiota AT, Bialosuknia SM, Zink SD, Brecher SM, Ehrbar DJ, Morrissette MN, et al. Effects of zika virus strain and aedes mosquito species on vector competence. Emerging infectious diseases. 2014;20:1577\u0026ndash;84.\u003c/li\u003e\n\u003cli\u003eVillena O, McClure KM, Camp RJ, LaPointe D, Atkinson CT, Sofaer H, et al. Environmental and geographical factors influence the occurrence and abundance of the southern house mosquito, culex quinquefasciatus, in hawai\u0026lsquo;i. Scientific Reports. 2023;13:20652.\u003c/li\u003e\n\u003cli\u003eMorin CW, Comrie AC. Modeled response of the west nile virus vector culex quinquefasciatus to changing climate using the dynamic mosquito simulation model. International Journal of Biometeorology. 2010;54:517\u0026ndash;29. Available from: https://doi.org/10.1007/s00484-010-0349-6\u003c/li\u003e\n\u003cli\u003eValdez LD, Sibona GJ, Diaz LA, Contigiani MS, Condat CA. Effects of rainfall on culex mosquito population dynamics. Journal of Theoretical Biology. 2017;421:28\u0026ndash;38. Available from: https://doi.org/10.1016/j.jtbi.2017.03.024\u003c/li\u003e\n\u003cli\u003eSheppard PR, Comrie AC, Packin GD, Angersbach K, Hughes MK. The climate of the US southwest. Climate Research. 2002;21:219\u0026ndash;38.\u003c/li\u003e\n\u003cli\u003eComrie AC, Broyles BB. Variability and spatial modeling of fine-scale precipitation in the southwest united states. Journal of Arid Environments. 2002;50:573\u0026ndash;92.\u003c/li\u003e\n\u003cli\u003eGim\u0026eacute;nez-Mujica U, Vel\u0026aacute;zquez-Castro J, Anzo-Hern\u0026aacute;ndez A, Herrera-Ramı́rez T, Barradas I. Estimating the final size of vector-borne epidemics in metapopulation networks: Methodologies and comparisons. Boletı́n de la Sociedad Matem\u0026aacute;tica Mexicana. 2025;31:35.\u003c/li\u003e\n\u003cli\u003eClark NJ, Ernest SKM, Senyondo H, Simonis J, White EP, Yenni GM, et al. Beyond single-species models: Leveraging multispecies forecasts to navigate the dynamics of ecological predictability. PeerJ. 2025;13:e18929.\u003c/li\u003e\n\u003cli\u003eJibowu M, Nolan MS, Ramphul R, Essigmann HT, Oluyomi AO, Brown EL, et al. Spatial dynamics of culex quinquefasciatus abundance: Geostatistical insights from harris county, texas. International Journal of Health Geographics. 2024;23:1\u0026ndash;12.\u003c/li\u003e\n\u003cli\u003eSambado S, Sipin TJ, Rennie Z, Larsen A, Cunningham J, Quandt A, et al. The paradoxical impact of drought on west nile virus risk: Insights from long-term ecological data. bioRxiv. 2025;2025\u0026ndash;01.\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"parasites-and-vectors","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"parv","sideBox":"Learn more about [Parasites \u0026 Vectors](http://parasitesandvectors.biomedcentral.com/)","snPcode":"13071","submissionUrl":"https://submission.nature.com/new-submission/13071/3","title":"Parasites \u0026 Vectors","twitterHandle":"@bugbittentweets","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"West Nile virus, Culex, Clustering, Mosquitoes, Climate","lastPublishedDoi":"10.21203/rs.3.rs-7595227/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7595227/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cstrong\u003eBackground\u003c/strong\u003e: Mosquitoes are vectors for diseases globally, making development of models that better explain mosquito abundances imperative. Mosquito population dynamics are particularly sensitive to local weather conditions, and mosquito-borne disease outbreaks can be spatially concentrated. There is a need for improved modeling studies to address whether spatial variation in disease outbreaks is driven by spatial variation in weather conditions, especially in dry and hot environments. In this study, we build a climate-driven model of mosquito population dynamics and compare whether predictions of mosquito abundance at the county scale are improved by accounting for sub-county climate variation.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMethods: \u003c/strong\u003eUsing a 5-year time series of weekly mosquito abundance data collected for each zip code in Maricopa County, USA, we assess how local variation in climate can explain and predict mosquito population dynamics. We built a mechanistic model of mosquito population dynamics influenced by daily temperature and 30-day accumulated precipitation. We grouped zip codes based on similar patterns of temperature and precipitation using functional clustering. We compared two approaches: one using county-level average climate and another using data from the identified climate clusters. We use MCMC to fit the mechanistic model using averaged climate data in each cluster, then compare the modeling fit to observed data of the county-level model to the model based on climate-based clusters.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eResults\u003c/strong\u003e: Simple, climate-forced modeling accurately estimates detailed mosquito abundance trajectories throughout a five-year period. Modeling mosquito abundances in the sub-county spatial clusters demonstrated that the same effects of temperature and precipitation on population growth rates could explain small-scale changes in mosquito populations. However, when we aggregate the sub-county model fits to the county-scale, the resulting fits are more precise but are sometimes overly confident, leading to lower overall accuracy and predictive performance.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConclusions\u003c/strong\u003e: Our study demonstrates the importance of collecting fine-scale mosquito abundance data to improve our understanding and the predictability of mosquito population dynamics. The strong performance of both the cluster-based and county-level models illustrates the value of spatially sensitive modeling in this application. We anticipate that such modeling efforts will also aid in using weather forecasts to predict mosquito populations, aiding in efforts to control the spread of infectious disease.\u003c/p\u003e","manuscriptTitle":"Accounting for spatial variation in climatic factors predicts spatial variations in mosquito abundance in the desert southwest","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-10-16 14:59:07","doi":"10.21203/rs.3.rs-7595227/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-11-25T13:59:09+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-11-24T20:57:43+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-10-22T21:31:51+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"223715833141247205681841072701885185401","date":"2025-10-20T14:45:24+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"76500686618057030503820401838918310514","date":"2025-10-13T16:05:16+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-10-05T21:10:13+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-10-05T21:08:57+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-09-30T14:39:12+00:00","index":"","fulltext":""},{"type":"submitted","content":"Parasites \u0026 Vectors","date":"2025-09-22T19:31:36+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"parasites-and-vectors","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"parv","sideBox":"Learn more about [Parasites \u0026 Vectors](http://parasitesandvectors.biomedcentral.com/)","snPcode":"13071","submissionUrl":"https://submission.nature.com/new-submission/13071/3","title":"Parasites \u0026 Vectors","twitterHandle":"@bugbittentweets","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"473fd42b-f4e9-4a15-9752-e177712eb601","owner":[],"postedDate":"October 16th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[],"tags":[],"updatedAt":"2026-03-16T16:01:01+00:00","versionOfRecord":{"articleIdentity":"rs-7595227","link":"https://doi.org/10.1186/s13071-026-07326-z","journal":{"identity":"parasites-and-vectors","isVorOnly":false,"title":"Parasites \u0026 Vectors"},"publishedOn":"2026-03-11 15:57:57","publishedOnDateReadable":"March 11th, 2026"},"versionCreatedAt":"2025-10-16 14:59:07","video":"","vorDoi":"10.1186/s13071-026-07326-z","vorDoiUrl":"https://doi.org/10.1186/s13071-026-07326-z","workflowStages":[]},"version":"v1","identity":"rs-7595227","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7595227","identity":"rs-7595227","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}