Skip to main content
Download PDF

Ecological drivers of dog heartworm transmission in California

Parasites & Vectors volume 15, Article number: 388 (2022) Cite this article

Abstract

Background

Effectively controlling heartworm disease—a major parasitic disease threatening animal health in the US and globally—requires understanding the local ecology of mosquito vectors involved in transmission. However, the key vector species in a given region are often unknown and challenging to identify. Here we investigate (i) the key vector species associated with transmission of the parasite, Dirofilaria immitis, in California and (ii) the climate and land cover drivers of vector presence.

Methods

To identify key mosquito vectors involved in transmission, we incorporated long-term, finely resolved mosquito surveillance data and dog heartworm case data in a statistical modeling approach (fixed-effects regression) that rigorously controls for other unobserved drivers of heartworm cases. We then used a flexible machine learning approach (gradient boosted machines) to identify the climate and land cover variables associated with the presence of each species.

Results

We found significant, regionally specific, positive associations between dog heartworm cases and the abundance of four vector species: Aedes aegypti (Central California), Ae. albopictus (Southern California), Ae. sierrensis (Central California), and Culiseta incidens (Northern and Central California). The proportion of developed land cover was one of the most important ecological variables predicting the presence or absence of the putative vector species.

Conclusion

Our results implicate three previously under-recognized vectors of dog heartworm transmission in California and indicate the land cover types in which each putative vector species is commonly found. Efforts to target these species could prioritize surveillance in these land cover types (e.g. near human dwellings in less urbanized settings for Ae. albopictus and Cs. incidens) but further investigation on the natural infection prevalence and host-biting rates of these species, as well as the other local vectors, is needed.

Graphical Abstract

Background

Heartworm is a globally distributed, mosquito-borne disease of high veterinary and economic importance. The parasitic roundworm, Dirofilaria immitis, can cause severe organ damage and death in a range of companion and wild animals including canids (e.g. dogs, coyotes), felids (e.g. cats, tigers), mustelids (e.g. ferrets, otters), and pinnipeds (e.g. sea lions, seals) [1]. Dirofilaria immitis can also cause pulmonary dirofilariasis in humans, making heartworm a zoonotic disease of concern to public health [2]. Heartworm preventatives are highly effective, yet they are prohibitively expensive for many pet owners and drug-resistant pathogen strains have recently emerged [1, 3, 4]. As a result, roughly 1–12.5% of domestic dogs in the US are estimated to be infected [5, 6]. More effective disease prevention and control require a better understanding of the drivers of disease transmission.

Heartworm transmission involves several developmental stages of D. immitis in both the mosquito vector and definitive mammalian host [7, 8]. Mosquitoes ingest microfilariae (L1) during blood-feeding on an infected host, which develop within the mosquito to reach the infective third larval stage (L3). This process takes approximately 8–30 days depending on temperature, with faster parasite development at higher temperatures [9]. During the next mosquito blood meal, the infective stage enters the bite wound of a susceptible host where it completes development. Between 6 to 9 months after host infection, mature D. immitis release microfilariae, which can be detected in the host blood stream.

Mosquito vector presence was identified as the greatest risk factor of dog heartworm prevalence in the US [10], making vector control necessary for disease prevention. At least 25 species from five genera (Aedes, Anopheles, Culex, Culiseta, and Psorophora) are potential vectors based on natural infections with L3 filariids in collected individuals, and many additional species have demonstrated competence in laboratory assessments [11]. However, determining which mosquito species are important drivers of transmission in a given region is challenging as this depends not only on physiological competence, but also vector abundance and propensity for feeding on host species [12, 13]. Furthermore, within-species variation in vector competence and host-biting rates have been observed, leading to variation in the key vector species involved in transmission between different regions and vector communities in the US [14,15,16,17]. As the most effective vector control measures can vary by species, preventing transmission requires understanding the local mosquito ecology and disease transmission cycle [11].

Climate and land cover factors have also been identified as important drivers of dog heartworm transmission, primarily given their impact on vector distributions and abundance [10]. Temperature impacts mosquito development rates, survival, and reproduction, as well as pathogen development rates within the mosquito, while moisture conditions govern the availability of mosquito breeding habitat [18, 19]. However, identifying the specific aspects of temperature and precipitation that most strongly impact the abundance of a particular species remains challenging [19,20,21,22]. Accordingly, several temperature variables such as maximum, minimum, mean, and diurnal temperature and precipitation variables at various lags have been proposed as drivers of vector presence and dog heartworm transmission [10, 24,25,26,27]. Similarly, numerous land cover variables such as the percent of forest, wetland, or impervious cover in the region have been proposed given their influence on mosquito and host abundance and mosquito-host contact rates [10, 16, 17, 25, 28]. Furthermore, each of the potential vector species may respond differently to these ecological features. For example, Ae. vexans are typically found at higher abundance in partially shaded floodwaters and woodlands, while Cx. quinquefasciatus are found at higher abundance in more developed landscapes [11]. Identifying which aspects of climate and land cover are most strongly associated with the abundance of key vector species can aid in determining when and where to target vector control and heartworm preventative efforts.

The complexity of dog heartworm transmission, including the many interacting and nonlinear biotic and abiotic factors, and lags in infection and transmission, as well as the lack of pathogen surveillance in mosquitoes, make determining the drivers of transmission challenging. Here, we investigate the ecological drivers of dog heartworm transmission in California—an ideal region for this investigation as the state harbors nine putative vector species (Table 1), contains high ecological diversity (Fig. 1a), and has experienced recent increases in dog heartworm cases in several regions (Fig. 1b). Furthermore, long-term mosquito surveillance records have been collected by vector control districts across the state, and dog heartworm case counts have been collected for the past decade, making it possible to infer relationships between vector populations and transmission. We leverage these rich data sources to investigate the biotic and abiotic drivers of dog heartworm transmission in the state. We analyze Northern, Central, and Southern California separately as variations in vector competence and dog-biting rates have been found between these regions (Table 1).

Table 1 Prior information on putative vector species
Full size table
Fig. 1
figure 1

a California bioregions as developed by the Interagency Natural Areas Coordinating Committee, based on distinct physiographic properties and ecological communities. Counties not included here (shown in white) were those without reported mosquito surveillance data. b Trends in dog heartworm cases, shown here as the total number of reported cases in a given year, from 2012 to 2020 for each California bioregion

Full size image

Specifically, we ask: (i) What are the primary vector species associated with dog heartworm transmission in Northern, Central, and Southern California? (ii) What are the key ecological drivers of vector presence for these species? For the first investigation, we use a statistical modeling approach (fixed-effects regression) that explicitly accounts for unobserved variation in dog heartworm cases between different biological regions and years to identify causal relationships between vector abundance and cases. For the second investigation, we use a flexible machine learning approach (gradient boosted machines) that can handle complex and collinear relationships between predictors and response variables and incorporate numerous climate and land cover variables at varying lags and scales to comprehensively assess the impact of climate and land cover on mosquito presence.

Methods

Dog heartworm case data

We used publicly available dog heartworm case data compiled by the Companion Animal Parasite Council [29]. These data are based on reports from IDEXX Laboratories, ANTECH Diagnostics, and Banfield Pet Hospital. Data include the number of dog heartworm antigen tests performed and the number of positive test results at the county and monthly scale from 2012 to the present. Antigen tests are typically performed at annual veterinary visits, but the CAPC notes the actual number of tests conducted is likely greater than their reports. The antigen tests have nearly 100% specificity and can detect sexually mature, adult female D. immitis from the blood of infected dogs starting at 5 months post-infection [30]. Infected dogs are typically treated and will test negative for antigens in subsequent years, meaning positive test results likely capture infection within the prior year. The reported county for a given test is based on the location of the veterinary clinic and thus could differ from the county in which the dog was infected.

Mosquito data

We considered all mosquito species in California that are potential vectors of dog heartworm [11]: Aedes aegypti, Ae. albopictus, Ae. sierrensis, Ae. vexans, Anopheles freeborni, Culiseta incidens, Cu. inornata, Culex quinquefasciatus, and Cx. tarsalis. Surveillance data for these species were provided by the California Vectorborne Disease Surveillance Program (CalSurv), which represents over 60 California mosquito and vector control agencies, the California Department of Public Health, and the Davis Arbovirus Research and Training Lab at the University of California, Davis. We subset the available data to include only abundance records for adult females collected in a single trap night using gravid, oviposition, \({CO}_{2}\)-baited, light, or BG sentinel traps or resting boxes (i.e. we excluded estimates from multiple collection nights or rare trap types). Furthermore, we removed any observation from vector control districts contributing fewer than five total records to minimize the effect of variation in surveillance methods and species identification. From these data, we calculated the average mosquito species abundance in a given county and year.

Climate data

We used publicly available, modeled climate data from the NOAA Physical Sciences Laboratory (https://psl.noaa.gov). We obtained daily maximum and minimum land surface temperature and precipitation data for the period of 1979–2021 at a 0.5° × 0.5° spatial resolution. We use the climate data associated with the closest latitude and longitude for each mosquito surveillance site (n = 31,389 unique surveillance sites). From these data, we calculated several climate variables capturing temperature and precipitation conditions including the minimum, maximum, and mean (the average of the maximum and minimum) temperature, the diurnal temperature range (the difference in the maximum and minimum temperature), and total precipitation. We calculated these variables at varying time periods and lags, as in Skaff et al. [31]. This included daily, weekly, monthly, and quarterly variables at one, two, or three time steps prior to the date of mosquito collection (e.g. one, two, or three days prior to the surveillance record date for the daily climatic variables).

Land cover data

We used the publicly available National Land Cover Database (NLCD) from the Multi-Resolution Land Characteristics consortium (https://www.mrlc.gov/data). Land cover data are available at 30-m resolution for the years: 2001, 2004, 2006, 2008, 2011, 2013, 2016, and 2019. From these data, we calculated the proportion coverage of the following classes: forest (evergreen, deciduous, or mixed), wetland, herbaceous, shrubland, low intensity developed (a mixture of constructed and vegetative materials), and high intensity developed (80–100% impervious cover). We calculated each of these at both 100- and 1000-m buffers around each surveillance site. Variables were calculated using the raster package in R v 4.0.2 [32]. For each surveillance record, we used the NLCD data from the given surveillance year or the closest preceding year.

Dog density and human socioeconomic data

We used publicly available, annual, county-level data from the US Census Bureau (USCB) to estimate dog density (https://www.census.gov/programs-surveys/acs/data.html). Specifically, we used the estimated number of housing units, multiplied by 0.614—the average number of dogs per household [33]—divided by the county area \({(mi}^{2})\). While the densities of other hosts such as coyotes and other wild canids are putative drivers of transmission, there are no reliable, comprehensive, spatially and temporally resolved data on these species. We also obtained annual, county-level median household income data from the USCB, as prior studies have found lower dog heartworm cases associated with higher household income because of increased heartworm preventative compliance [10, 34].

California bioregions

To delineate different regions of California based on ecological similarity, we used the 10 bioregions developed by the Interagency Natural Areas Coordinating Committee. These bioregions were determined based on distinctions in physiographic properties and ecological communities [35]. For counties that overlapped multiple bioregions, we selected the bioregion in which the largest portion of the human population resided, as this likely represented where dog heartworm cases were reported (Fig. 1a).

Data analysis: identifying key vector species

To identify key vector species associated with dog heartworm transmission, we used a least squares dummy variable (or fixed effects) regression approach—a statistical modeling approach used to isolate potential causal relationships in settings where randomized experiments are infeasible [36]. We used both bioregion and year as dummy variables to control for any unobserved heterogeneity that might influence dog heartworm cases in a particular bioregion across all years (e.g. geographic features, number of veterinary clinics) or influence cases in all bioregions in a given year (e.g. an influx of shelter dogs to the state due to natural disaster [37], higher case reporting). Using these models, we included the abundance of each putative vector species as a predictor. Specifically, for each species, we used adult female abundance at a 1-year lag (to account for the lag between transmission and potential case detection) averaged across all traps within the county. We assessed the multicollinearity of these predictors by calculating the variance inflation factor (VIF). None of the mosquito predictors had a VIF value > 3; thus, we did not exclude any species from the model (Additional file 1: Figure S1). As host population density and socioeconomic status can impact case reporting [10, 25], we also included county-level dog population density and median household income as predictors, each at a 1-year lag. As broad spatial variation in vector abundance, competence, and dog-biting rates has been observed in California (Table 1), we ran separate models for Northern, Central, and Southern California. We classified the Bay Delta, Klamath, Sierra, and Sacramento Valley bioregions as Northern California, the Central Coast and San Joaquin Valley bioregions as Central California, and the South Coast and Colorado Desert bioregions as Southern California (Fig. 1a).

Data analysis: ecological drivers of vector presence

To identify the key climate and land cover predictors of each of the nine putative vector species, we used a gradient boosted machine (GBM) approach implemented using the XGBoost package in R [78]. Briefly, gradient boosting is a supervised machine learning approach in which regression or classification trees is sequentially built from the prediction errors of the prior tree. GBM algorithms allow for complex, nonlinear relationships among predictor and outcome variables and collinearity between predictors, making them well suited for this analysis. Extreme gradient boosting is a scalable and efficient GBM implementation that minimizes overfitting and has been shown to achieve high predictive accuracy [79]. Here, we developed an XGboost classification model predicting the presence or absence of each putative vector species (specifically an adult female of that species in a single night at a single trap). Absences here refer to a specific surveillance site and day where a mosquito was found, but not of the given species (i.e. ‘true’ absences rather than pseudo-absences). All climate and land cover predictors described above (n = 78; see “Methods”: “Climate data,” and “Land cover data”) were initially considered as predictors for the model. However, as including many collinear variables can minimize interpretability, we removed any predictor with a > 0.90 pairwise correlation with another predictor (n = 31) and preserved those with the greatest biological relevance (n = 47; Additional file 3: Table S1). Furthermore, we log-transformed any highly skewed ecological predictors to reduce the influence of extreme outliers on model accuracy (Additional file 3: Table S1). To account for unmeasured spatial and temporal variation in mosquito presence, we also included collection latitude and longitude and the week, month, and year of each surveillance record as predictors. Furthermore, we included vector control agency and trap type as predictors to account for variation in surveillance methods.

Model fitting

For each model, we randomly split observations into a training (80%) and testing (20%) dataset to evaluate model accuracy, where each observation was a surveillance record from a single collection date and location. Using the training data, we tuned model hyperparameters using fivefold cross validation and Bayesian optimization implemented with rBayesianOptimization. Specifically, we tuned the maximum tree depth, learning rate, and gamma (which controls regularization), as these parameters typically have the largest impact on model performance [80]. We provided initial hyperparameter values aimed at reducing overfitting (i.e. shallow trees, low learning rates, and high regularization). For cross validation, we used log loss as the model learning objective and ended training on the validation set after 10 rounds if no reductions in log loss were made. Additionally, we corrected for unbalanced classes (i.e. an unequal number of presence and absence records) in the less common species by setting the model parameter ‘scale_pos_weight’ equal to the square root of the total number of absence records divided by the total number of presence records for each species ([37]; Additional file 4: Table S2). We then fit an XGboost classification model using these optimal hyperparameters (Additional file 4: Table S2). As model output can vary based on the subsample of the data used, we conducted 100 iterations of the final model fitting, bootstrapping with a random 80% subset of the full dataset each time. This enabled us to generate confidence intervals for the model evaluation and predictor importance metrics (described below).

Model evaluation and predictor importance

To evaluate the model performance, we used the withheld test data set and calculated AUC, which captures the model’s ability to accurately distinguish between classes (here, vector presence vs, absence), using pROC. To identify which predictors were most important for model performance, we estimated variable importance using XGboost. Specifically, we used ‘gain’—the relative improvement in model performance when adding a split in the tree on a given variable. We calculated these metrics for each of the 100 final model fits.

Results

Identifying key vector species

The vector species that were significantly, positively associated with dog heartworm cases varied regionally and included four species in total: Ae. aegypti (Central California; p = 0.010), Ae. albopictus (Southern California; p = 0.008), Ae. sierrensis (Central California; p = 0.045), and Cs. incidens (Northern and Central California; p = 0.005, 0.006) (Table 2). The effect sizes of these associations varied from an additional 8–35 cases in the year following a one standard deviation increase in mosquito abundance (Table 2). Of these key species, only Cs. incidens is consistently found at relatively high abundance in the given region, while the other identified vectors are relatively less common (Fig. 2, Additional file 2: Figure S2). Several mosquito species were significantly negatively associated with dog heartworm cases: Culex tarsalis in Northern California and Ae. sierrensis and Ae. vexans in Southern California (Table 2). Cases were also negatively associated with Ae. albopictus in Northern California, but this species was only found here in one bioregion and 1 year (Klamath in 2020) so this estimated relationship may be biased. The annual abundance of other putative vector species—An. freeborni, Cs. inornata, and Cx. quinquefasciatus—was not significantly associated with variation in dog heartworm cases in any region.

Table 2 Effect of mosquito predictors on dog heartworm cases
Full size table
Fig. 2
figure 2

Annual trends in abundance of putative vector species for each California bioregion from 2012 to 2020. Abundance here denotes the total count of trapped adult female mosquitoes of a given species. Note: y-axes scales vary between bioregions

Full size image

Despite the associations described above, most of the variation in dog heartworm cases in all regions was explained by the bioregion and year dummy variables (together explaining 40.9%, 49.9%, and 59.8% of the variation in Northern, Central, and Southern California, respectively; Table 2), indicating that there is a large amount of unobserved heterogeneity at this level impacting cases. We found that estimated dog density had a significant, positive association with dog heartworm cases in Southern California, but a negative association in Central California (Table 2). However, changes in the density of other host species such as coyotes and feral dogs, which were not included in our model, may be driving additional bioregion- or year-level variation. Furthermore, median household income was only significantly associated with dog heartworm cases in Southern California, but other factors related to income such as the use of prophylactics or pet relinquishment rates, which were not included here, may also be contributing to the spatial and temporal variation.

Ecological drivers of vector presence

Classification model accuracy

The classification models predicting the presence or absence of each vector species in a given trap location and date had high performance with a mean out-of-sample AUC > 0.93 for all species (Additional file 5: Table S3). The classification model sensitivity—the ability to accurately predict vector presence—varied from 0.674 to 0.993 and was highest for the most abundant species, Cx. quinquefasciatus and Cx. tarsalis. Model specificity—the ability to accurately predict vector absence—was typically higher, ranging from 0.686 to 0.995, and was highest for the less abundant species, all Aedes spp., An. freeborni, and Cs. inornata (Additional file 5: Table S3).

Predictor importance

For the four species identified as positively associated with dog heartworm cases in at least one region (Ae. aegypti, Ae. albopictus, Ae. sierrensis, and Cs. incidens), the proportions of developed land cover in the region near the trap, as well as maximum and minimum temperature in the preceding seasons, were consistently among the top 10 predictors associated with their presence or absence (Fig. 3, Table 3). For Ae. aegypti, presence was associated with a greater proportion of developed cover (of both low and high intensity) in the region surrounding the trap as well higher minimum daily temperatures in the preceding winter. Conversely, Ae. sierrensis presence was associated with less developed and more forested land cover as well as cooler, wetter conditions in the preceding winter. For Ae. albopictus and Cs. incidens, we found a likely association with neighborhoods and human dwellings in less urbanized settings: presence was associated with higher proportions of low-intensity developed cover (e.g. areas with a mixture of constructed materials and vegetation, most commonly single-family neighborhoods) in the immediate surroundings (100 m), but less developed cover (of both low and high intensity) in the surrounding kilometer. Furthermore, Cs. incidens presence was associated with lower maximum daily temperatures in the preceding winter, while Ae. albopictus presence was associated with higher maximum temperatures in the preceding summer. In general, the proportions of low and high intensity developed land cover surrounding the trap were frequently included among the most important in predicting presence or absence for the nine putative vector species, while diurnal temperature range was not an important predictor for any species.

Fig. 3
figure 3

Variable importance in predicting vector species presence/absence. For each species, predictors are ranked based on their mean gain across the 100 model iterations

Full size image
Table 3 Mean values of ecological predictors for vector presence and absence
Full size table

Discussion

Identifying the ecological drivers of dog heartworm transmission is critical for disease prevention. Vector presence, climate, and land cover conditions are known to be key drivers, but the specific species and abiotic factors influencing transmission in a particular setting are often unknown and challenging to identify [10, 25]. Here, we leverage a comprehensive dog heartworm case data set as well as long-term, finely resolved vector surveillance data to investigate the vector species, climate, and land cover features associated with dog heartworm transmission in California. Using a statistical modeling approach that controls for unobserved spatial and temporal variation in cases, we identify four vectors of dog heartworm transmission in different regions of California: Culiseta incidens in Northern and Central California, Aedes aegypti and Ae. sierrensis in Central California only, and Aedes albopictus in Southern California only. Using flexible machine learning models, we find that the proportions of developed cover, and minimum and maximum daily temperature in preceding seasons, are the strongest drivers of these species’ presence or absence.

Over 25 mosquito species have been identified as potential vectors of dog heartworm in the US, including nine species in California [11]. While Aedes sierrensis is often described as the state’s principal vector [59, 68, 81] because of its widespread distribution, we found its abundance was only significantly associated with dog heartworm cases in Central California. We found that, despite being broadly distributed, Ae. sierrensis is not present at high abundance in any bioregion. Furthermore, prior studies have found low infection prevalence in field-collected Ae. sierrensis adults in California [56, 65] and low rates of dog-biting in some regions [60]. Thus, discrepancies in the relative importance of Ae. sierrensis in dog heartworm transmission could be due to the type of evidence being considered (e.g. physiological competence, abundance, distribution, infection prevalence, host-biting rates) and/or regional variation in these factors.

Ae. aegypti, Ae. albopictus, and Cs. incidens have not been previously been considered major vectors of dog heartworm in California. For each of these species, their physiological vector competence (e.g. ability to acquire and maintain D. immitis) has been established by prior laboratory studies in other regions [39, 41, 42, 46, 48,49,50,51, 57, 71, 82]. However, their rates of vector efficiency have varied based on the mosquito strain tested [49, 51, 57] and were generally low in the case of Ae. aegypti [38, 39, 41, 42]. For Cs. incidens, natural infection with D. immitis [65, 68] and a willingness to bite dogs [57, 61, 65] have been observed in California populations specifically. Furthermore, the species is widely distributed across the state and has been previously identified as a potentially important secondary vector of dog heartworm [57, 65, 71]. Our finding that Cs. incidens abundance is significantly positively associated with dog heartworm cases in Northern and Central California suggests that this species may play an under-recognized role in transmission in these regions. For Ae. albopictus, prior studies in the Eastern and Midwestern US found moderate rates of infection in field-collected adults [16, 17, 46, 47] and high rates of dog-biting [52, 53]. Similarly, natural infection and high rates of dog-biting have been observed in Ae. aegypti in the Southern US [43, 45] as well as Mexico and Argentina [44, 83]. Aedes albopictus and Ae. aegypti were rarely detected in California prior to their establishment in 2011 and 2015, respectively [84]. Since then, both species have rapidly increased in abundance and have become established in Coastal, Central, and Southern California [85]. However, to our knowledge, no studies have yet evaluated the vector competence, field infection rates, or dog-biting rates of these species in California. Our finding that Ae. aegypti and Ae. albopictus are significantly positively associated with dog heartworm cases in Central and Southern California, respectively, highlights their potential role in dog heartworm transmission in these regions and the need for further investigation of these species’ vector potential.

While our analysis of the vector species associated with dog heartworm included nearly a decade of mosquito surveillance and case data, we cannot definitively claim that Ae. aegypti, Ae. albopictus, Ae. sierrensis, and Cs. incidens are the key species driving transmission in California. In particular, the available heartworm case data include only antigen test results from domestic animals, thus not capturing heartworm transmission intensity in wild animals. Furthermore, as cases are assigned to the county from which the test was reported, local and imported cases cannot be distinguished. Our statistical modeling approach, which examined differences in cases over time for a particular bioregion and controlled for annual variation in cases, could address some of these limitations. However, an ideal investigation would incorporate information on field infection prevalences and host-biting rates in addition to vector abundances and distributions. This more rigorous investigation is hampered by the lack of systematic parasite surveillance in mosquitoes in California, and the US as a whole, likely due to the high cost of testing many different species and individuals. However, California has a uniquely comprehensive vector surveillance system (CalSurv), which includes finely resolved spatial and temporal data on mosquito abundances across much of the state, dating back decades for many locations and species. Leveraging this rich data source, we provide a critical first step towards understanding the vectors involved in dog heartworm transmission in the Western US, a region that has broadly experienced an increase in dog heartworm cases in the past decade [29].

In investigating the ecological drivers of vector presence, we found that land cover feature, namely the proportion of developed land cover in the region surrounding the surveillance site, was consistently one of the strongest predictors of a given vector species’ presence. Of the four species associated with higher heartworm prevalence, we found Ae. aegypti was more common in highly developed areas (e.g. apartment complexes or commercial settings), while Ae. sierrensis was more common in less developed, more forested areas—findings that match prior habitat associations of these two species [86,87,88,89]. Culiseta incidens and Ae. albopictus were both more likely in areas with some human development in the immediate surroundings, but with less high intensity development in the broader environment. This is also consistent with prior findings that these species are typically found in peridomestic settings [62, 88, 90]. Given these species-specific land cover associations, strategies for targeted vector surveillance may vary by region (e.g. prioritizing placing traps directly within areas of low intensity development in Northern and Southern California to target Cs. incidens and Ae. albopictus, respectively), while efforts to survey a wider range of potential vectors would aim to place traps in a mixture of land cover types.

Conclusions

As cases of heartworm disease are increasing across much of the US, including California, a better understanding of the drivers of transmission is needed to protect companion and wild animals. Mosquito vector presence has been identified as the main risk factor for transmission, but identifying the primary vector species in a specific region is challenging given the large number of putative vector species and spatial variation in their ecologies and vector status. Here, we leveraged long-term mosquito surveillance, climate, and land cover data to identify four regionally specific vector species—Ae. aegypti (Central California), Ae. albopictus (Southern California), Ae. sierrensis (Central California), and Cs. incidens (Northern and Central California)—and their local habitat associations. Investigating natural infection prevalence and host biting rates in these species, as well as other local vectors, is an important next step in understanding the local transmission ecology. Doing so will enable more targeted and effective vector control and disease prevention.

Availability of data and materials

All data and code used in this study are available on GitHub in the following repository: https://github.com/lcouper/DogHeartworm.

References

  1. Bowman DD, Atkins CE. Heartworm biology, treatment, and control. Vet Clin North Am Small Anim Pract. 2009;39:1127–58.

    Article  PubMed  Google Scholar 

  2. Simón F, Siles-Lucas M, Morchón R, González-Miguel J, Mellado I, Carretón E, et al. Human and animal dirofilariasis: the emergence of a zoonotic mosaic. Clin Microbiol Rev. 2012;25:507–44.

    Article  PubMed  PubMed Central  Google Scholar 

  3. Wolstenholme AJ, Evans CC, Jimenez PD, Moorhead AR. The emergence of macrocyclic lactone resistance in the canine heartworm. Dirofilaria immitis Parasitology. 2015;142:1249–59.

    Article  CAS  PubMed  Google Scholar 

  4. Hampshire VA. Evaluation of efficacy of heartworm preventive products at the FDA. Vet Parasitol. 2005;133:191–5.

    Article  PubMed  Google Scholar 

  5. Bowman D, Little SE, Lorentzen L, Shields J, Sullivan MP, Carlin EP. Prevalence and geographic distribution of Dirofilaria immitis, Borrelia burgdorferi, Ehrlichia canis, and Anaplasma phagocytophilum in dogs in the United States: results of a national clinic-based serologic survey. Vet Parasitol. 2009;160:138–48.

    Article  PubMed  Google Scholar 

  6. Lee ACY, Montgomery SP, Theis JH, Blagburn BL, Eberhard ML. Public health issues concerning the widespread distribution of canine heartworm disease. Trends Parasitol. 2010;26:168–73.

    Article  PubMed  Google Scholar 

  7. Otto GF. Geographical distribution, vectors, and life cycle of Dirofilaria immitis. J Am Vet Med Assoc. 1969;154:370–3.

    PubMed  Google Scholar 

  8. Grieve RB, Lok JB, Gllickman LT. Epidemiology of canine heartworm infection. Epidemiol Rev. 1983;5:220–46.

    Article  CAS  PubMed  Google Scholar 

  9. Christensen BM, Hollander AL. Effect of temperature on vector-parasite relationships of Aedes trivittatus and Dirofilaria immitis. Proc Helminthol Soc Wash. 1978;45:5.

    Google Scholar 

  10. Brown HE, Harrington LC, Kaufman PE, McKay T, Bowman DD, Nelson C, et al. Key factors influencing canine heartworm, Dirofilaria immitis, in the United States. Parasit Vectors. 2012;5:245.

    Article  PubMed  PubMed Central  Google Scholar 

  11. Ledesma N, Harrington L. Mosquito vectors of dog heartworm in the United States: vector status and factors influencing transmission efficiency. Top Companion Anim Med. 2011;26:178–85.

    Article  PubMed  Google Scholar 

  12. Kain MP, Skinner EB, van den Hurk AF, McCallum H, Mordecai EA. Physiology and ecology combine to determine host and vector importance for Ross River virus. eLife. 2021;10:e67018.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  13. Plowright RK, Parrish CR, McCallum H, Hudson PJ, Ko AI, Graham AL, et al. Pathways to zoonotic spillover. Nat Rev Microbiol. 2017;15:502–10.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  14. Ahid SMM, Vasconcelos PSS, Lourenço-de-Oliveira R. Vector competence of Culex quinquefasciatus say from different regions of Brazil to Dirofilaria immitis. Mem Inst Oswaldo Cruz. 2000;95:769–75.

    Article  CAS  PubMed  Google Scholar 

  15. Thiemann TC, Lemenager DA, Kluh S, Carroll BD, Lothrop HD, Reisen WK. Spatial Variation in Host Feeding Patterns of Culex tarsalis and the Culex pipiens complex (Diptera: Culicidae) in California. J Med Entomol. 2012;49:903–16.

    Article  CAS  PubMed  Google Scholar 

  16. Paras KL, O’brien VA, Reiskind MH. Comparison of the vector potential of different mosquito species for the transmission of heartworm, Dirofilaria immitis, in rural and urban areas in and surrounding Stillwater Oklahoma USA. Med Vet Entomol. 2014;28:60–7.

    Article  PubMed  Google Scholar 

  17. Spence Beaulieu MR, Federico JL, Reiskind MH. Mosquito diversity and dog heartworm prevalence in suburban areas. Parasit Vectors. 2020;13:12.

    Article  PubMed  PubMed Central  Google Scholar 

  18. Shocket MS, Anderson CB, Caldwell JM, Childs ML, Couper LI, Han S, et al. Environmental drivers of vector-borne disease. In: Shocket MS, Anderson CB, Caldwell JM, Childs ML, Couper LI, Han S, et al., editors. Population biology of vector-borne diseases. Oxford: Oxford University Press; 2021. p. 85–118.

    Google Scholar 

  19. Mordecai EA, Caldwell JM, Grossman MK, Lippi CA, Johnson LR, Neira M, et al. Thermal biology of mosquito-borne disease. Ecol Lett. 2019;22:1690–708.

    Article  PubMed  PubMed Central  Google Scholar 

  20. Beck-Johnson LM, Nelson WA, Paaijmans KP, Read AF, Thomas MB, Bjørnstad ON. The importance of temperature fluctuations in understanding mosquito population dynamics and malaria risk. R Soc Open Sci. 2017;4:160969.

    Article  PubMed  PubMed Central  Google Scholar 

  21. Gilioli G, Mariani L. Sensitivity of Anopheles gambiae population dynamics to meteo-hydrological variability: a mechanistic approach. Malar J. 2011;10:294.

    Article  PubMed  PubMed Central  Google Scholar 

  22. Roiz D, Ruiz S, Soriguer R, Figuerola J. Climatic effects on mosquito abundance in Mediterranean wetlands. Parasit Vectors. 2014;7:333.

    Article  PubMed  PubMed Central  Google Scholar 

  23. Reisen WK, Cayan D, Tyree M, Barker CM, Eldridge B, Dettinger M. Impact of climate variation on mosquito abundance in California. J Vector Ecol. 2008;33:89–98.

    Article  PubMed  Google Scholar 

  24. Genchi C, Mortarino M, Rinaldi L, Cringoli G, Traldi G, Genchi M. Changing climate and changing vector-borne disease distribution: the example of Dirofilaria in Europe. Vet Parasitol. 2011;176:295–9.

    Article  PubMed  Google Scholar 

  25. Wang D, Bowman DD, Brown HE, Harrington LC, Kaufman PE, May T, et al. Factors influencing US canine heartworm (Dirofilaria immitis) prevalence. Parasit Vectors. 2014;7:264.

    Article  PubMed  PubMed Central  Google Scholar 

  26. McGill E, Berke O, Peregrine AS, Weese JS. Epidemiology of canine heartworm (Dirofilaria immitis) infection in domestic dogs in Ontario, Canada: Geographic distribution, risk factors and effects of climate. Geospatial Health. 2019. https://doi.org/10.4081/gh.2019.741.

    Article  PubMed  Google Scholar 

  27. Montoya-Alonso JA, Carretón E, Morchón R, Silveira-Viera L, Falcón Y, Simón F. The impact of the climate on the epidemiology of Dirofilaria immitis in the pet population of the Canary Islands. Vet Parasitol. 2016;30:66–71.

    Article  Google Scholar 

  28. Čabanová V, Miterpáková M, Valentová D, Blažejová H, Rudolf I, Stloukal E, et al. Urbanization impact on mosquito community and the transmission potential of filarial infection in central Europe. Parasit Vectors. 2018;11:261.

    Article  PubMed  PubMed Central  Google Scholar 

  29. Companion Animal Parasite Council. Parasite Prevalence Maps. Companion Animal Parasite Council. 2022. https://capcvet.org/maps. Accessed 2022 Sep 3

  30. Little S, Saleh M, Wohltjen M, Nagamori Y. Prime detection of Dirofilaria immitis: understanding the influence of blocked antigen on heartworm test performance. Parasit Vectors. 2018;11:186.

    Article  PubMed  PubMed Central  Google Scholar 

  31. Skaff NK, Cheng Q, Clemesha RES, Collender PA, Gershunov A, Head JR, et al. Thermal thresholds heighten sensitivity of West Nile virus transmission to changing temperatures in coastal California. Proc R Soc B Biol Sci. 1932;2020:20201065.

    Google Scholar 

  32. Hijmans RJ, van Etten J. raster: Geographic analysis and modeling with raster data. 2012. Available from: http://CRAN.R-project.org/package=raster

  33. AVMA 2017–2018. U.S. pet ownership & demographics sourcebook. Schaumburg: American Veterinary Medical Association; 2018.

    Google Scholar 

  34. Gates MC, Nolan TJ. Factors influencing heartworm, flea, and tick preventative use in patients presenting to a veterinary teaching hospital. Prev Vet Med. 2010;93:193–200.

    Article  PubMed  Google Scholar 

  35. Thomas CW. Bureaucratic landscapes: interagency cooperation and the preservation of biodiversity. Cambridge: MIT Press; 2003. p. 382.

    Google Scholar 

  36. Larsen AE, Meng K, Kendall BE. Causal analysis in control-impact ecological studies with observational data. Methods Ecol Evol. 2019. https://doi.org/10.1111/2041-210X.13190.

    Article  Google Scholar 

  37. Levy JK, Lappin MR, Glaser AL, Birkenheuer AJ, Anderson TC, Edinboro CH. Prevalence of infectious diseases in cats and dogs rescued following Hurricane Katrina. J Am Vet Med Assoc. 2011;238:311.

    Article  PubMed  Google Scholar 

  38. McGreevy PB, McClelland GAH, Lavoipierre MMJ. Inheritance of susceptibility to Dirofilaria immitis infection in Aedes aegypti. Ann Trop Med Parasitol. 1974;68:97–109.

    Article  CAS  PubMed  Google Scholar 

  39. Apperson CS, Engber B, Levine JF. Relative suitability of Aedes albopictus and Aedes aegypti in North Carolina to support development of Dirofilaria immitis. J Am Mosq Control Assoc. 1989;5:377–82.

    CAS  PubMed  Google Scholar 

  40. Nayar JK, Knight JW. Isoenzyme variation in Aedes aegypti correlated with Dirofilaria immitis infectability. Med Vet Entomol. 2002;16:424–9.

    Article  CAS  PubMed  Google Scholar 

  41. Tiawsirisup S, Nithiuthai S. Vector competence of Aedes aegypti (L.) and Culex quinquefasciatus (Say) for Dirofilaria immitis (Leidy). Southeast Asian J Trop Med Public Health. 2006;37:110–4.

    PubMed  Google Scholar 

  42. Serrão ML, Labarthe N, Lourenço-de-Oliveira R. Vectorial competence of Aedes aegypti (Linnaeus 1762) Rio de Janeiro strain, to Dirofilaria immitis (Leidy 1856). Mem Inst Oswaldo Cruz. 2001;96:593–8.

    Article  PubMed  Google Scholar 

  43. Ledesma NA, Kaufman PE, Xue RD, Leyen C, Macapagal MJ, Winokur OC, et al. Entomological and sociobehavioral components of heartworm (Dirofilaria immitis) infection in two Florida communities with a high or low prevalence of dogs with heartworm infection. J Am Vet Med Assoc. 2019;254:93–103.

    Article  PubMed  Google Scholar 

  44. Vezzani D, Mesplet M, Eiras DF, Fontanarrosa MF, Schnittger L. PCR detection of Dirofilaria immitis in Aedes aegypti and Culex pipiens from urban temperate Argentina. Parasitol Res. 2011;108:985–9.

    Article  PubMed  Google Scholar 

  45. Olson MF, Ndeffo-Mbah ML, Juarez JG, Garcia-Luna S, Martin E, Borucki MK, et al. High rate of non-human feeding by Aedes aegypti reduces Zika virus transmission in south Texas. Viruses. 2020;12:453.

    Article  PubMed Central  Google Scholar 

  46. Nayar JK, Knight JW. Aedes albopictus (Diptera: Culicidae): an Experimental and Natural Host of Dirofilaria immitis (Filarioidea: Onchocercidae) in Florida, USA. J Med Entomol. 1999;36:441–8.

    Article  CAS  PubMed  Google Scholar 

  47. Licitra B, Chambers EW, Kelly R, Burkot TR. Detection of Dirofilaria immitis (Nematoda: Filarioidea) by polymerase chain reaction in Aedes albopictus, Anopheles punctipennis, and Anopheles crucians (Diptera: Culicidae) From Georgia, USA. J Med Entomol. 2010;47:634–8.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  48. Scoles GA, Dickson SL. New foci of canine heartworm associated with introductions of new vector species: Aedes albopictus in New Orleans and Aedes sierrensis in Utah. In: Proceedings of the Heartworm Symposium ’95. Auburn, Alabama; 1995. p. 27–35.

  49. Scoles GA, Craig GBJ. Variation in susceptibility to Dirofilaria immitis among U.S. strains of Aedes albopictus. Vector Control Bull. 1993;2:98–103.

    Google Scholar 

  50. Lai CH, Tung KC, Ooi HK, Wang JS. Competence of Aedes albopictus and Culex quinquefasciatus as vector of Dirofilaria immitis after blood meal with different microfilarial density. Vet Parasitol. 2000;90:231–7.

    Article  CAS  PubMed  Google Scholar 

  51. Tiawsirisup S, Kaewthamasorn M. The potential for Aedes albopictus (Skuse) (Diptera: Culicidae) to be a competent vector for canine heartworm. Southeast Asian J Trop Med Public Health. 2007;38:208.

    Google Scholar 

  52. Savage HM, Niebylski ML, Smith GC, Mitchell CJ, Craig GB. Host-Feeding Patterns of Aedes albopictus (Diptera: Culicidae) at a Temperate North American Site. J Med Entomol. 1993;30:27–34.

    Article  CAS  PubMed  Google Scholar 

  53. Faraji A, Egizi A, Fonseca DM, Unlu I, Crepeau T, Healy SP, et al. Comparative Host Feeding Patterns of the Asian Tiger Mosquito, Aedes albopictus, in Urban and Suburban Northeastern USA and Implications for Disease Transmission. PLoS Negl Trop Dis. 2014;8:e3037.

    Article  PubMed  PubMed Central  Google Scholar 

  54. Richards SL, Ponnusamy L, Unnasch TR, Hassan HK, Apperson CS. Host-Feeding Patterns of Aedes albopictus (Diptera: Culicidae) in Relation to Availability of Human and Domestic Animals in Suburban Landscapes of Central North Carolina. J Med Entomol. 2006;43:543–51.

    Article  PubMed  Google Scholar 

  55. Niebylski ML, Savage HM, Roger N, George CJr. Blood hosts of Aedes albopictus in the United States. J Am Mosq Control Assoc. 1994;10(3):447–50.

  56. Walters LL, Lavoipierre MMJ. Aedes Vexans and Aedes Sierrensis (Diptera: Culicidae): Potential Vectors of Dirofilaria Immitis in Tehama County, Northern California, USA1. J Med Entomol. 1982;19:15–23.

    Article  CAS  PubMed  Google Scholar 

  57. Theis JH, Kovaltchouk JG, Fujioka KK, Saviskas B. Vector Competence of Two Species of Mosquitoes (Diptera: Culicidae) from Southern California for Dirofilaria immitis (Filariidea: Onchocercidae). J Med Entomol. 2000;37:295–7.

    Article  CAS  PubMed  Google Scholar 

  58. Tran TD, Nelms BM, Koschik ML, Scott JJ, Thiemann T. Prevalence of filarial parasites in field-caught mosquitoes in northwestern California. J Vector Ecol. 2022;47:61–8.

    Article  Google Scholar 

  59. Sacks BN, Chomel BB, Kasten RW. Modeling the Distribution and Abundance of the Non-Native Parasite, Canine Heartworm. California Coyotes Oikos. 2004;105:415–25.

    Article  Google Scholar 

  60. Thiemann TC, Woodward DL, Fang Y, Ryan BM, Nelms BM, Scott JJ, et al. Abundance and Bloodfeeding Patterns of Mosquitoes (Diptera: Culicidae) in an Oak Woodland on the Eastern Slope of the Northern Coast Range of California. J Med Entomol. 2017;54:1344–53.

    Article  PubMed  Google Scholar 

  61. Tempelis GH, Washino RK. Host-Feeding Patterns of Culex Tarsalis in the Sacramento Valley, California, With Notes on Other Species1. J Med Entomol. 1967;4:315–8.

    Article  CAS  PubMed  Google Scholar 

  62. Reisen WK, Reeves WC. Bionomics and ecology of Culex tarsalis and other potential mosquito vector species. In: Epidemiology and control of mosquito borne arboviruses in California, 1943–1987. Sacramento, California: Proceedings of the California Mosquito Control Association; 1990.

  63. Mckay T, Bianco T, Rhodes L, Barnett S. Prevalence of Dirofilaria immitis (Nematoda: Filarioidea) in Mosquitoes From Northeast Arkansas, the United States. J Med Entomol. 2013;50:871–8.

    Article  PubMed  Google Scholar 

  64. Bemrick WJ, Sandholm HA. Aedes vexans and Other Potential Mosquito Vectors of Dirofilaria immitis in Minnesota. J Parasitol. 1966;52:762–7.

    Article  CAS  PubMed  Google Scholar 

  65. Huang S, Smith DJ, Molaei G, Andreadis TG, Larsen SE, Lucchesi EF. Prevalence of Dirofilaria immitis (Spirurida: Onchocercidae) Infection in Aedes, Culex, and Culiseta Mosquitoes From North San Joaquin Valley. CA J Med Entomol. 2013;50:1315–23.

    Article  PubMed  Google Scholar 

  66. Yen CH. Studies on Dirofilaria immitis Leidy, with Special Reference to the Susceptibility of Some Minnesota Species of Mosquitoes to the Infection. J Parasitol. 1938;24:189–205.

    Article  Google Scholar 

  67. Burkot TR, DeFoliart GR. Bloodmeal sources of Aedes triseriatus and Aedes vexans in a southern Wisconsin forest endemic for La Crosse encephalitis virus. Am J Trop Med Hyg. 1982;31:376–81.

    Article  CAS  PubMed  Google Scholar 

  68. Walters LL. Risk factors for heartworm infection in northern California. In: Proceedings of the Heartworm Symposium. Batavia, IL: American Heartworm Society; 1995.

  69. McGreevy PB, Conrad RD, Bulgin MS, Stitzel KA. Canine filariasis in northern California. Am J Vet Res. 1970;31:1325–8.

    CAS  PubMed  Google Scholar 

  70. Kartman L. Factors influencing infection of the mosquito with Dirofilaria immitis (Leidy, 1856). Exp Parasitol. 1953;2:27–78.

    Article  Google Scholar 

  71. Acevedo RA. Potential vectors of Dirofilaria immitis in San Mateo County California. 1982 p. 272–4. (Mosquito News). Report No.: 42.

  72. Villavaso EJ, Steelman CD. Laboratory and Field Studies of the Southern House Mosquito, Culex Pipiens Quinquefasciatus Say, Infected with the Dog Heartworm, Dirofilaria Immitis (Leidy), In Louisiana1. J Med Entomol. 1970;7:471–6.

    Article  CAS  PubMed  Google Scholar 

  73. Tolbert RH, Johnson WJ. Potential vectors of Dirofilaria immitis in Macon County. Alabama Am J Vet Res. 1982;43:2054–6.

    CAS  PubMed  Google Scholar 

  74. Lowrie. Poor vector efficiency of Culex quinquefasciatus following infection with Dirofilaria immitis. J Am Mosq Control Assoc. 1991;7(1):30–6.

  75. Brito AC, Fontes G, Rocha EM da, Rocha DA, Regis L. Development of Dirofilaria immitis (Leidy) in Aedes aegypti (L.) and Culex quinquefasciatus (Say) from Maceió, Alagoas, Brazil. Mem Inst Oswaldo Cruz. 1999 Jul;94(4):575–6.

  76. Thiemann TC, Lemenager DA, Kluh S, Carroll BD, Lothrop HD, Reisen WK. Spatial Variation in Host Feeding Patterns of Culex tarsalis and the Culex pipiens complex (Diptera: Culicidae) in California. J Med Entomol. 2012;49:903–16.

    Article  CAS  PubMed  Google Scholar 

  77. Molaei G, Cummings RF, Su T, Armstrong PM, Williams GA, Cheng ML, et al. Vector-Host Interactions Governing Epidemiology of West Nile Virus in Southern California. Am J Trop Med Hyg. 2010;83:1269–82.

    Article  PubMed  PubMed Central  Google Scholar 

  78. Chen T, Guestrin C. XGBoost: A Scalable Tree Boosting System [Internet]. 2016. Available from: https://dl.acm.org/doi/https://doi.org/10.1145/2939672.2939785

  79. Asselman A, Khaldi M, Aammou S. Enhancing the prediction of student performance based on the machine learning XGBoost algorithm. Interact Learn Environ. 2021 May 19;0(0):1–20.

  80. Van Doren BM, Horton KG. A continental system for forecasting bird migration. Science. 2018;361:1115–8.

    Article  PubMed  Google Scholar 

  81. Weinmann CJ, Garcia R. Canine heartworm in California, with observations on Aedes sierrensis as a potential vector. 1974 p. 45–50. (California Vector Views). Report No.: 21.

  82. Spence Beaulieu MR, Reiskind MH. Comparative Vector Efficiency of Two Prevalent Mosquito Species for Dog Heartworm in North Carolina. Faraji A, editor. J Med Entomol. 2019 Nov 5;tjz190.

  83. Estrada-Franco JG, Fernández-Santos NA, Adebiyi AA, López-López M de J, Aguilar-Durán JA, Hernández-Triana LM, et al. Vertebrate-Aedes aegypti and Culex quinquefasciatus (Diptera)-arbovirus transmission networks: Non-human feeding revealed by meta-barcoding and next-generation sequencing. PLoS Negl Trop Dis. 2020 Dec 31;14(12):e0008867.

  84. Porse CC, Kramer V, Yoshimizu MH, Metzger M, Hu R, Padgett K, et al. Public Health Response to Aedes aegypti and Ae. albopictus Mosquitoes Invading California, USA. Emerg Infect Dis. 2015 Oct;21(10):1827–9.

  85. Metzger ME, Hardstone Yoshimizu M, Padgett KA, Hu R, Kramer VL. Detection and Establishment of Aedes aegypti and Aedes albopictus (Diptera: Culicidae) Mosquitoes in California, 2011–2015. J Med Entomol. 2017;54:533–43.

    Article  PubMed  Google Scholar 

  86. Landau KI, van Leeuwen WJD. Fine scale spatial urban land cover factors associated with adult mosquito abundance and risk in Tucson. Arizona J Vector Ecol. 2012;37:407–18.

    Article  PubMed  Google Scholar 

  87. Kamgang B, Happi JY, Boisier P, Njiokou F, Hervé JP, Simard F, et al. Geographic and ecological distribution of the dengue and chikungunya virus vectors Aedes aegypti and Aedes albopictus in three major Cameroonian towns. Med Vet Entomol. 2010;24:132–41.

    Article  CAS  PubMed  Google Scholar 

  88. Bohart RM, Washino RK. Mosquitoes of California. Agriculture and Natural Resources: University of California; 1978. p. 164.

    Google Scholar 

  89. Pless E, Saarman NP, Powell JR, Caccone A, Amatulli G. A machine-learning approach to map landscape connectivity in Aedes aegypti with genetic and environmental data. Proc Natl Acad Sci USA. 2021;118:e2003201118.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  90. Kache PA, Eastwood G, Collins-Palmer K, Katz M, Falco RC, Bajwa WI, et al. Environmental DETERMINANTS of Aedes albopictus abundance at a Northern Limit of Its Range in the United States. Am J Trop Med Hyg. 2020;102:436–47.

    Article  PubMed  Google Scholar 

Download references

Acknowledgements

We thank the California vector control districts and personnel who collected, maintained and kindly shared the mosquito surveillance data that made this work possible. We are grateful for the publicly available data sources including those from the CAPC, NOAA, and the USCB.

Funding

L.I.C. was funded by the Stanford Center for Computational, Evolutionary, and Human Genomics. E.A.M. was funded by the National Science Foundation (DEB-2011147), the National Institutes of Health (R35GM133439, R01AI168097, and R01AI102918), the Stanford King Center on Global Development, the Stanford Woods Institute for the Environment, and the Terman Award.

Author information

Authors and Affiliations

  1. Department of Biology, Stanford University, Stanford, CA, 94305, USA

    Lisa I. Couper & Erin A. Mordecai

Authors
  1. Lisa I. Couper
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Erin A. Mordecai
    View author publications

    You can also search for this author in PubMed Google Scholar

Contributions

LIC and EAM conceived of the project and designed the analyses. LIC obtained the data, performed the analyses, and drafted the manuscript. Both authors read and approved the final manuscript.

Corresponding author

Correspondence to Lisa I. Couper.

Ethics declarations

Competing interests

The authors declare that they have no competing interests.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

Additional file 1: Figure S1.

Correlation matrix indicating correlations between the annual, county-level abundance of different mosquito species.

Additional file 2: Figure S2.

Relative abundance of each mosquito species by bioregion. Abundance here reflects the average number of trapped adult females of a given species between 2010–2020.

Additional file 3: Table S1.

Ecological predictors retained for analysis (n = 47). Predictors listed in the Methods that are not included in this list were removed because of high collinearity (pairwise correlations > 0.90 with an included predictor). These ecological predictors were included in models along with the year, month, week, latitude, and longitude of surveillance to account for spatiotemporal variation and trap type and vector control agency to account for variation in surveillance methods.

Additional file 4: Table S2.

Optimal hyperparameters for models predicting the presence/absence of each vector species, identified using Bayesian optimization.

Additional file 5: Table S3.

Performance metrics for models predicting species presence/absence in a given trap night and location. Values shown here for AUC, sensitivity, specificity, accuracy, and balanced accuracy are the mean values from the 100 model iterations. Values are shown for the full model, which contains all the ecological and spatiotemporal predictors for a given species, as well as for null models predicting all ‘present’ or all ‘absent.’

Additional file 6: Table S4.

Variable importance for each predictor in models predicting the presence/absence of each vector species. The left column under each species indicates the mean gain from the 100 model iterations. The right column indicates the predictor rank (i.e. 1–54) based on mean gain.

Additional file 7: Table S5.

Mean values of the ecological predictors for when a given vector species is absent (0) or present (1). Predictors that were ranked as among the top 10 most important, based on mean gain, for that species are denoted with a *.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Couper, L.I., Mordecai, E.A. Ecological drivers of dog heartworm transmission in California. Parasites Vectors 15, 388 (2022). https://doi.org/10.1186/s13071-022-05526-x

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doi.org/10.1186/s13071-022-05526-x

Keywords

Parasites & Vectors

ISSN: 1756-3305

Contact us