Background

Global climate change is widespread and affects zoonoses by increasing (1) the geographical range and/or pervasiveness of animal reservoirs and/or arthropod vectors; (2) introductions of competent vectors, and/or the occurrence, intensity; and (3) the duration of transmission cycles [1]. Canada continues to show signs of a changing climate, including a ~ 2 °C rise in annual surface temperatures since 1948 [2]. The Intergovernmental Panel on Climate Change (IPPC, 2018) has indicated that mosquito-borne diseases (MBDs) will be particularly impacted by climate change. Indeed, seasonal variations in the occurrence and abundance of mosquito populations are intricately tied to climatic factors, which in turn impact their vector potential [3]. In Canada, the burden of MBDs has increased by 10% in the past 20 years and is expected to continue to rise if the impacts of climate change are not mitigated [4].

Several studies have shown strong associations between mosquito vector abundances and weather factors [5,6,7,8]. Temperature, precipitation, and relative humidity are the three major weather variables influencing mosquito seasonal activity [9,10,11,12] and host–pathogen interactions [13]. Temperature can impact mosquito survival [14,15,16], development [4, 16, 17], geographical range [18, 19], vector competence [20,21,22], and host-seeking and other behaviors [23]. Precipitation can alter the occurrence of suitable larval habitats [22, 24, 25] and the viability of eggs and larvae [26]. Humidity can affect the mating, dispersal, longevity, bloodfeeding behaviors, and oviposition of mosquitoes [3, 27, 28]. However, there are numerous mitigating factors that can drastically alter mosquito population dynamics (e.g., forest cover), and the combined effects of multiple abiotic and biotic factors are often challenging to resolve.

Mosquitoes typically have species-specific ranges of weather conditions for optimal seasonal activity. Higher temperatures are generally favored by mosquitoes [29], but species can vary in their minimum thresholds. For instance, the minimum metabolic threshold for Aedes vexans is 12 °C, but is slightly lower (10 °C) for Culex tarsalis and Coquillettidia perturbans [30,31,32]. To this end, some studies associate mosquito abundance with degree days, which is a weather-based indicator that takes into account both ambient temperatures and minimum metabolic thresholds of a given species [30, 33]. Standing water from rainfall creates necessary breeding grounds for many species, but too much precipitation can wash away larval habitats [34,35,36]. Thus, species that utilize more permanent breeding grounds (e.g., lakes, marshes) are likely less susceptible to population fluctuations associated with rainfall. The overall relationship between mosquito abundance and precipitation is not straightforward, however, as some species abundances appear most influenced by rainfall occurring weeks to even months prior [5, 30, 37]. In contrast, high humidity conditions are typically preferred by mosquitoes, as sustained bouts of low moisture can impact their survival, behaviors, and development [38].

There are several species of mosquitoes found in the Canadian Prairies that can potentially harbor and transmit viruses of public health concern. Aedes vexans Meigen, the inland floodwater mosquito, is a cosmopolitan nuisance mosquito with broad vector potential. It is capable of transmitting West Nile virus (WNV), California serogroup viruses (CSGVs), Rift Valley fever virus, and Zika virus [39,40,41,42]. Ochlerotatus dorsalis Meigen, the summer saltmarsh mosquito, is found throughout North America and is a competent vector of Western equine encephalitis virus (WEEV), CSGVs, and WNV [43]. The cattail mosquito, Coquillettidia perturbans Walker, is found throughout the Prairies, breeding in permanent swamps containing cattails and aquatic plants [43]. This species is associated with the transmission of Eastern equine encephalitis virus (EEEV), WNV, and CSGVs [43, 44]. The geographical range of Culex tarsalis Coquillett extends from northern Mexico into Canada and from the west coast to the Mississippi River [45, 46]. The species is the primary vector of WNV in the Prairies and also capable of transmitting WEEV and CSGVs [43, 47]. Other mosquito vector species occurring in the Prairies include Aedes canadensis (CSGVs, WNV, EEEV), Ochlerotatus triseriatus (La Crosse virus, EEEV, WEEV), and Ochlerotatus trivittatus (CSGVs) [19, 43, 47,48,49]. Although these vectors presumably have varying optimal ranges for temperature, humidity, and precipitation, little information is presently available on the relationships between weather factors and mosquito seasonal activity in the Canadian Prairies.

Two Canadian Prairie provinces (Manitoba and Saskatchewan) carry out annual mosquito surveillance to detect the causal agents of MBDs at the provincial level. However, these programs focus their monitoring activities on Culex species capable of transmitting WNV, most notably Cx. tarsalis. To our knowledge no other mosquito vector species are identified or tested for human pathogens in these programs. Consequently, we carried out weekly surveillance during the active season over a 2-year period in Manitoba to characterize the population dynamics of nine commonly found mosquito species. We then used Generalized Linear Mixed Models to determine the relationships (if any) between mosquito trap counts and weather variables (temperature, precipitation, and relative humidity) for the four most abundant vector species. Since the life cycle (egg-to-adult) of most species is between 8 and 14 days, the combination of weather conditions over this period is likely to affect development and thus mosquito abundance and activity [32, 43]. Consequently, our analyses investigated how the 14 days preceding the trapping date impacted the number of mosquitoes captured.

Material and methods

Mosquito trapping and identification

Host-seeking mosquitoes were trapped using CDC Miniature Light Traps (Model 1012, John W. Hock, Gainesville, FL) with carbon dioxide (CO2) regulators set to 15 psi and the light disabled (to minimize non-mosquito collections). We placed traps on tree limbs ~ 1.5 m from the ground and activated them from dusk until dawn. Traps were operated twice weekly (Monday and Tuesday) in 2020 and 2021, from June to August (CDC weeks 23 to 36). A total of 24 traps were deployed in eight Western Manitoba communities in 2020, with one trap setup in each community in 2021 (Additional files 1 and 2). In 2020, collections from one-time satellite traps from nine additional locations in Central and Eastern Manitoba were provided to us by the City of Winnipeg Insect Control Branch with Culex species removed (Additional files 1 and 2). All mosquitoes were stored at − 80 °C in Petri dishes coded by date and collection site.

Five mosquito vector species were visually identified using dissecting microscopes in 2020: Ochlerotatus flavescens Muller, Oc. dorsalis, Ae. vexans, Cx. tarsalis, and Cq. perturbans. We expanded our identification efforts to include four less common and/or non-vector species in 2021: Aedes canadensis Thebald, Ochlerotatus trivittatus Coquillett, Ochlerotatus triseriatus Say, and Anopheles earlei Vargus. Specimens were identified to species using relevant mosquito identification keys [43, 50, 51]. For traps with high numbers of specimens (> 1000), we subsampled by counting a randomized ¼ sample of the trap and then extrapolated the numbers by a factor of four.

Weather factors associated with mosquito counts

For each trapping location over the 2-year surveillance period, we recorded three variables that may be connected to mosquito trap catch: temperature (°C), precipitation (mm), and relative humidity (%). These data were obtained from the Environment Canada weather station closest to each trapping location. Weather data was collected daily from each location between May and August in both years. The distances between trapping site and the closest servicing weather station ranged from < 1 to 63 km. The reason some of the stations are farther away than others is they service multiple towns that have had historically comparable weather indices (Environment Canada, personal communication).

We focused our analyses (see below) on the four most commonly found vector species: Oc. dorsalis, Ae. vexans, Cx. tarsalis, and Cq. perturbans. The specific variables explored were: (1) mean rainfall (mm) over the 14 days preceding the trapping date (ppm14); (2) mean relative humidity (%) 14 days prior to the trapping date (rhm14); and mean degree days 14 days preceding the trapping date (ddm14). The latter incorporated mean daily temperatures (Tmean; °C) and baseline metabolic temperature (Tb; °C) for each mosquito species [30,31,32], where ddm14 represents the number of degrees above Tbase over the 14 d period; thus if dd1: Tmean > Tbase, then dd1 = Tmean − Tbase, but if dd1: Tmean ≤ Tbase, then dd1 = 0 °C [30]. Since Tb is not published for Oc. dorsalis, we used the same value (12 °C) as Ae. vexans, which is also a floodwater species. Preliminary exploration suggested that trap count differences with local weather variation were species-specific. We therefore modelled weather variables separately for each species. Finally, as low temperatures can inhibit mosquito activity and therefore influence trap counts we included trapping day minimum temperature as a covariate to account for this effect.

Statistical analyses

Relationships among environmental variables and mosquito counts over 2 years (2020 and 2021) were assessed for each species (Oc. dorsalis, Ae. vexans, Cx. tarsalis, and Cq. perturbans) using R statistical software (v4.2.1; [52]). We explored two sets of models: (1) a single model exploring the effect of mosquito species on trap counts; and (2) four species-specific models exploring the effects of time (CDC week) and weather variables (ppm14, rhm14, ddm14) on trap counts. All models were Generalized Linear Mixed Models (GLMMs; glmmTMB package v1.1.2.3; [53]). To control for spatial, site-level effects we included trap location nested within site as a random intercept. To control for temporal, week-level effects we also included a categorical week-by-year variable as a random intercept. We used a negative binomial distribution because while we have count data, they were over-dispersed and did not match a Poisson distribution.

For the species model, trap counts were modelled using the complete data set with species, year, and their interaction as explanatory variables. Each trap count was represented by a row in the data with no pooling of counts within or between weeks/sites. Minimum temperature of the trap-day was included as a covariate. Because this full dataset showed significant temporal autocorrelation, we added an AR(1) covariance structure to week grouped by unique site (across years). We conducted Post-Hoc analyses to compare species differences among years (emmeans package v1.7.0; [54]) with the false discovery rate P-value adjustment.

For the weather models, curvilinear relationships in the weather variables and week were modelled as 2nd degree orthogonal polynomials. We also included interactions between year and each polynomial: trapcount ~ poly(week, degree = 2) * year + poly(ddm14, degree = 2) * year + poly(pt14, degree = 2) * year + poly(rha14, degree = 2) * year + ttmin + (1|week_year) + (1|site/sitespecific). Further, we included minimum temperature of the trap-day as a covariate. Where non-significant (alpha of 5%), individual polynomial terms and interactions were omitted from the models and linear terms and main effects retained alone. Type III ANOVA tables were computed with the car package (v3.0.12; [55]) and used to assess polynomial terms. Significant linear effects were additionally reported with summary table statistics (Estimate and Z-test) in order to capture the magnitude of the effect (i.e. the Estimate). To aid in the interpretation we converted the original estimates to incident rate ratios which can be interpreted as a multiplicative factor (i.e. an incident rate ratio of 2 indicates a 2 times increase).

All model fits and assumptions (including potential spatial and temporal autocorrelation) were assessed with the DHARMa package (v0.4.4; [56]); multicollinearity was assessed with the performance package (v0.8.0; [57]); and figures were created with the ggplot2 package (v3.3.5; [58]). Note that figure scales are log10 transformed after first adding 1 to better visualize patterns.

Results

Mosquito surveillance activities

More than 265,000 mosquitoes were collected throughout southern Manitoba over the 2-year surveillance period, with 57% captured in 2020. This included weekly collections in Western Manitoba and one-time satellite collections at various times between June and September in Eastern and Central Manitoba, though the latter represented a small proportion (11%) of the total mosquito catch. Trap counts tended to be highest between weeks 26 and 29, though this differed to some extent by species and year. Notable were fogging events in Brandon in both 2020 and 2021, which subsequently resulted in considerably reduced mosquito numbers in that community. Consequently, we omitted mosquito trap count data from this site post-fogging for all analyses (CDC weeks ≥ 30 in 2020, weeks ≥ 28 in 2021).

Aedes vexans was the most common mosquito species

There was considerable variation in the relative proportions of each mosquito species per trapping location (Fig. 1), as well as over time (Fig. 2). Of the mosquitoes caught, 40% (2020) and 80% (2021) represented the four primary vector species: Ae. vexans, Oc. dorsalis, Cx. tarsalis, and Cq. perturbans. This discrepancy in proportions between years is largely attributed to Cq. perturbans from a single location (Cypress River), where > 50,000 individuals were captured in 2021 and only 14,000 in 2020. As this site reflected drastically and systematically different trapping patterns (Figs. 1, 2) we removed the site from all subsequent analyses (outlier effect).

Fig. 1
figure 1

Relative trap counts for the eight most commonly found mosquito species in 2020 and 2021. Mosquitoes were captured on a weekly basis (May to September) from 17 sampling sites throughout Manitoba, Canada. Culex tarsalis counts are not included for all locations in the eastern part of the region (denoted with an asterisk*). Ae. canadensis, An. earlei, Oc. trivittatus, and Oc. triseriatus were not surveyed in 2020. We collected one Oc. triseriatus in 2021, which was not included on the figure

Full size image
Fig. 2
figure 2

Average weekly trap counts for each of the 17 sampling sites in 2020 a and 2021 b. Displayed are the four most commonly collected vector species: Ae. vexans, Cq. perturbans, Cx. tarsalis, and Oc. dorsalis. Culex tarsalis counts are not included for all locations in the eastern part of the region (denoted with an asterisk*)

Full size image

There was a significant interaction between species and year (χ23 = 33.57; P < 0.0001) and we therefore conducted Post-Hoc analyses separately for each year. These showed that in both years, we captured significantly more Ae. vexans than each of the three other primary vector species (P < 0.01 across all pairwise comparisons;). Ochlerotatus dorsalis trap counts were 2.4× higher than Cx. tarsalis in 2020 (P < 0.001) but 0.28× lower in 2021 (P < 0.001). However, both O. dorsalis and Cx. tarsalis had higher trap counts than Cq. perturbans in 2020 (9.3× and 3.4×; both P < 0.0001) as well as in 2021 (2.5× and 9.0×; both P < 0.05).

The influence of weather variables on trap counts was species-specific

Aedes vexans

There was a significant interaction between year and the 2nd order polynomial for trap week (χ22 = 6.34; P = 0.042). As such, trap counts increased and then decreased over the season (with a greater increase in 2021) (Fig. 3). Trap counts increased linearly with relative humidity (χ21 = 9.35; P = 0.002), with 1.12× (12%) more mosquitoes captured for every % increase in relative humidity (Est = 0.109; z = 3.057; Fig. 3a). In addition, there were significant quadratic effects of degree days (χ22 = 19.58; P < 0.0001) and precipitation (χ22 = 11.03; P = 0.004). Accordingly, trap counts were highest with intermediate values of degree days (Fig. 3b) and low precipitation (Fig. 3c). Trapping day minimum temperature had a significant effect (χ21 = 5.13; P = 0.024), where trap counts were 1.07× (7%) greater with every 1 °C increase in minimum temperature (Est = 0.072; z = 2.265). Notably the model fit (assessed by checking patterns in the residuals) for this species was marginal, suggesting other factors may be in play that were not been captured by this model.

Fig. 3
figure 3

GLMM model analyses showing the effects of time (CDC week) and weather variables on Aedes vexans trap counts in 2020 and 2021. Seasonal mosquito activity was significantly impacted by a relative humidity, b degree days, and c precipitation in the 2-week period preceding the trapping date. Week corresponds to the week of the year for 2020 and 2021 (e.g., week 24 is the 24th week of both 2020 and 2021). Points represent observed data

Full size image

Culex tarsalis

There was a significant interaction between year and the 2nd order polynomial for trap week (χ22 = 16.17; P = 0.0003), such that trap counts increased and then decreased over the season with a more distinct peak in 2020 (Fig. 4). There were no associations (quadratic or linear) between trap counts and relative humidity (χ21 = 2.52; P = 0.112; Fig. 4a). However, there was a linear relationship between trap counts and degree days (χ21 = 4.16; P = 0.041; Fig. 4b), such that the number of mosquitoes captured increased by 32% for every 1 °C increase in mean degree days (Est = 0.281; z = 2.040). There was a significant interaction between year and precipitation (χ21 = 4.73; P = 0.030). Accordingly, the relationship between trap counts and rainfall showed contrasting patterns, with a greater number of mosquitoes captured with high and low precipitation in 2020 and 2021, respectively (Fig. 4c). There was no effect of trapping day minimum temperature on trap counts (χ21 = 2.05; P = 0.152).

Fig. 4
figure 4

GLMM model analyses showing the effects of time (CDC week) and weather variables on Culex tarsalis trap counts in 2020 and 2021. Seasonal mosquito activity was significantly affected by b degree days and c precipitation in the 2-week period preceding the trapping date. Since there was no significant effect of a relative humidity, only one, black, line is shown (seasonal effect). Week corresponds to the week of the year for 2020 and 2021 (e.g., week 24 is the 24th week of both 2020 and 2021). Points represent observed data

Full size image

Coquillettidia perturbans

There was a significant interaction between year and the 2nd order polynomial for trap week (χ2x = 7.74; P = 0.0209), such that trap counts increased and then decreased over the season (with a relatively more consistent peak in 2021) (Fig. 5). We also identified a significant quadratic relationship between trap counts and relative humidity (χ22 = 9.10; P = 0.011), with elevated (but not extreme) humidity resulting in a greater number of mosquitoes captured (Fig. 5a). There were no effects (quadratic or linear) of degree days (χ21 = 0.81; P = 0.368; Fig. 5b) nor precipitation (χ21 = 0.24; P = 0.622; Fig. 5c). Trapping day minimum temperature had a significant effect (χ21 = 29.62; P < 0.0001), where trap counts were 1.43 times greater with every 1 °C increase in minimum temperature (Est = 0.36; z = 5.44).

Fig. 5
figure 5

GLMM model analyses showing the effects of time (CDC week) and weather variables on Coquillettidia perturbans trap counts in 2020 and 2021. Seasonal mosquito activity was significantly influenced by a relative humidity in the 2-week period preceding the trapping date. Since there was no significant effect of b degree days or c precipitation, only one, black, line is shown (seasonal effect). Week corresponds to the week of the year for 2020 and 2021 (e.g., week 24 is the 24th week of both 2020 and 2021). Points represent observed data

Full size image

Ochlerotatus dorsalis

There was a significant interaction between year and the 2nd order polynomial for trap week (χ22 = 15.94; P < 0.001). As such, trap counts in 2020 decreased almost linearly, but in 2021 showed a pattern of increased and decreased numbers over the season. The was no effect (linear or quadratic) of relative humidity (χ21 = 1.67; P = 0.196; Fig. 6a) nor degree days (χ21 = 0.47; P = 0.494; Fig. 6b) on trap counts. However, we identified a significant linear relationship between trap counts and precipitation (χ21 = 3.89; P = 0.049; Fig. 6c). Accordingly, trap counts increased by 1.01 times (1%) for each mm increase in precipitation (Est = 0.0069; z = 1.972). There was no effect of minimum trapping day temperature (χ21 = 0.21; P = 0.646).

Fig. 6
figure 6

GLMM model analyses showing the effects of time (CDC week) and weather variables on Ochlerotatus dorsalis trap counts in 2020 and 2021. Seasonal mosquito activity was significantly impacted by c precipitation in the two-week period preceding the trapping date. Since there was no significant effect of a relative humidity and b degree days, only one, black, line is shown (seasonal effect). Week corresponds to the week of the year for 2020 and 2021 (e.g., week 24 is the 24th week of both 2020 and 2021). Points represent observed data

Full size image

Discussion

The primary objective of our study was to explore the relationships between key weather variables and mosquito population dynamics in the Canadian Prairies. Our two consecutive years of weekly trapping throughout southern Manitoba yielded over 265,000 mosquitoes, of which the majority represented four noted vector species: Oc. dorsalis, Ae. vexans, Cx. tarsalis, and Cq. perturbans. It should be emphasized that the trap counts provide a good indication of mosquito activity (i.e., host-seeking) during the trapping period, but do not necessarily correlate with overall mosquito abundances at a given sampling site. Further, we discuss weather conditions favored by mosquitoes as it relates to higher trap counts rather than their true abundance/activity/biology. Aedes vexans was the most common mosquito in most sites/weeks, which is in agreement with historical records for our sampling region [59] and nearby regions [41, 47]. Both Oc. dorsalis and Cx. tarsalis are also well established in the Canadian Prairies [60, 61]. Interestingly, Cq. perturbans trap counts were relatively low with the exception of one site, Cypress River. This is likely due to habitat suitability, as the larvae of this species feed on cattails [43, 62, 63] and our traps at Cypress River were situated adjacent a marsh-like area with heavy aquatic vegetation that included abundant cattails. Consequently, Cq. perturbans activity at this site appears driven by breeding site conditions rather than weather variables.

In terms of seasonal activity, Ae. vexans, Cx. tarsalis, and Cq. perturbans all showed a similar (and expected) pattern, with trap counts gradually increasing to peak numbers before progressively declining. However, the peak in trap counts occurred later in the season for Cx. tarsalis (late-July to early August) in comparison to the other two species (early- to mid-July). This discrepancy is likely attributed to the overwintering behaviors of these species. While Cx. tarsalis overwinter as non-fed adults, Ae. vexans and Cq. perturbans overwinter in the egg stage and as larvae, respectively [31, 43, 64, 65]. Consequently, the former requires a bloodmeal prior to laying eggs thereby delaying the first generation in comparison to the other two species. Trap counts for all three species were higher in 2021 compared to 2020, presumably due to more favorable environmental conditions for adult survival, oviposition success, and/or host-seeking activities. Although Oc. dorsalis showed a seasonal trend similar to the other mosquito species in 2021, trap counts in the previous year were highest at or near the start of our surveillance activities. This suggests an early spring emergence of Oc. dorsalis in 2020, which is consistent with their known biology [66]. However, the steady decline in their numbers throughout the 2020 season was unexpected given Oc. dorsalis can have multiple generations per year [67]. This suggests some combination of environmental factors later in the season and comparatively low sample sizes may have impeded the success of subsequent generations.

Mosquito seasonal activity is largely driven by temperature, precipitation, and relative humidity [9,10,11,12]. Given most species complete development within 14 days, we focused on how temperature, precipitation, and relative humidity in the two weeks preceding the trapping date affected mosquito counts. Both Ae. vexans and Cq. perturbans favored (i.e., higher trap counts) high humidity (75–85%), which is consistent with studies from other geographic regions [23, 68]. High humidity has been associated with increased egg production, larval indices, adult survival and activity, including host-seeking at close range [9, 11, 26, 68, 69]. In contrast, bouts of low humidity can cause eggs to desiccate and reduce adult longevity and/or activity in favor of seeking shelter [27, 28, 70]. The lack of a relationship between Cx. tarsalis and Oc. dorsalis trap counts and humidity was a bit unexpected, though the biology of both species suggests that their activity is less impacted by humidity. Stuart (2020) found Culex mosquitoes to prefer breeding under hot and dry conditions [71] with adult Cx. tarsalis typically reared in laboratories under relatively low humidity [72]. Ochlerotatus dorsalis is found in a wide rage of habitats (e.g., coastal marshes, grasslands, forests, tidal areas, and semiarid deserts), which can vary considerably in relative humidity [73, 74].

Since mosquito development is temperature-dependent, their abundance typically increases with air temperature (i.e., faster lifecycle) and then declines once a threshold has been reached [32, 75]. This threshold varies depending on species, with reported ranges between 22 and 30 °C [32, 76,77,78,79,80]. Moreover, mosquitoes have minimum metabolic temperatures (e.g., [30,31,32]) for which their activity largely ceases below this level. Consequently, we focused only on the air temperatures fostering mosquito development over the 14 days preceding the trapping date. Culex tarsalis favored high degree days, which is consistent with their biology in the Canadian Prairies [61]. Aedes vexans favored intermediate degree days, presumably having increased mortality and/or reducing their activity by seeking refuge during bouts of extreme temperatures [81,82,83]. Temperature did not impact Cq. perturbans nor Oc. dorsalis trap counts, which may be attributed to the unique habitat requirements of the former and the diversity of suitable habitats for the latter. It is also possible that increasing the lag period to more than 14 days may better capture the effect of temperature on both species, as found for other species/regions [5, 30]. Finally, minimum trapping day temperature was associated with trap counts for Ae. vexans and Cq. perturbans but not the other three species. Temperatures between 15 and 24 °C are generally suitable for host-seeking activities in mosquitoes [23], but both species may have a narrower temperature range for optimal activity.

The influence of precipitation on mosquito life history traits is often complex and differs among species and studies [6, 30, 37, 84,85,86,87]. Even within a species the relationship is not always clear. For instance, some studies have found a positive effect of rainfall on Ae. alpbopictus abundance [88,89,90,91], whereas others have not [92, 93]. Although the larval stages of all mosquitoes are dependent on water availability, their breeding habitats, oviposition biology, and egg physiology can vary markedly. Indeed, rainfall over the 14 days prior to the trapping date influenced mosquito counts for Ae. vexans, Cx. tarsalis, and Oc. dorsalis, but the underlying biological reasoning was not always obvious. Larval breeding sites for Oc. dorsalis include temporary pools formed by precipitation [73], which is in line with the higher levels of rainfall that this species favors. However, Ae. vexans favoring dryer conditions was unexpected as this species oviposits on soil, relying on precipitation events to trigger egg hatching [94, 95]. Consequently, Ae. vexans should increase in abundance with higher levels of short-term rainfall. Precipitation influenced Cx. tarsalis activity, but the species favored higher and lower levels of rainfall in 2020 and 20,921, respectively. Given Cx. tarsalis and Cq. perturbans lay their eggs directly on the surface of water [96, 97], they are more likely to be influenced by longer-term precipitation. This would presumably dictate the number of suitable breeding sites available and in turn mosquito abundance several weeks later [37]. Indeed, abundance of some mosquito species has been positively associated with rainfall events occurring several weeks to even months later [5, 30, 37, 86].

There are several considerations related to experimental design that must be taken into account when forming conclusions from our study. The interactions between individual weather factors, additional factors not examined in this study (e.g., wind velocity, moonlight, and anthropogenic water sources), and their combined effects are nearly impossible to disentangle without controlled experiments. Some weather variables, particularly precipitation, may show improved modelling for some species with longer lag periods from the trapping date, though it would be challenging to infer the specific reason(s) for any significant relationships. The precision of our study could also be improved by setting up weather stations directly at each trapping site. Further, our traps predominately captured host-seeking females from dawn to dusk, with most traps set up on the interface between human dwellings and forest/agricultural land. Given the differences in host-seeking behaviors and ecologies among species, this design may not accurately reflect the true relative abundances of each species. Deploying multiple trap types (e.g., gravid, net, BG-Sentinel) to supplement our collections in conjunction with sampling sites with varying land cover (e.g., forest, urban and rural areas, agricultural areas) may better inform on the relative abundances of each species. Finally, extending our surveillance activities earlier in the spring may better determine the seasonal abundances of some mosquito species and the associated weather factors (e.g., Oc. dorsalis).

Conclusion

We carried out 2 years of seasonal surveillance in Manitoba, Canada to explore the seasonal population dynamics and associated weather factors. The environmental conditions varied markedly between years (e.g. 2020: abnormally high rainfall concentrated over a few days; 2021: very dry), allowing us to capture a wide range of weather variables. Previous work has investigated the associations between temperature/precipitation and Cx. tarsalis trap counts in the Canadian Prairies [61], but to our knowledge this is the first study to explore other commonly found vector species in this region. The pervasiveness, seasonal activity, and associations with weather variables differed among species, likely due to their unique ecologies and behaviors. From our experiences, future surveillance efforts in the Canadian Prairies may benefit from using multiple trap types and a breadth of sampling sites. Placing moisture and temperature probes adjacent to each trap would improve accuracy and investigating other meteorological elements could provide further insights. Future studies aimed at associating the population dynamics of these vector species with pathogen infection rates would provide valuable information for surveillance programs. Ultimately these discrete differences among mosquito species in optimal weather conditions will influence their vector potential on an annual basis.