Gravid Culex pipiens Exhibit A Reduced Susceptibility to Ultra–Low Volume Adult Control Treatments Under Field Conditions

Since its 1999 introduction, West Nile virus (WNV) has remained the most prevalent arboviral disease in North America (Curren et al. 2018). Frequently referred to as a hotspot of WNV activity, Chicago, IL, and the surrounding Cook County suburbs have an especially high annual human case incidence that from 2005 to 2013 exceeded 10 cases per 100,000 (Karki et al. 2018). WNV epizootic transmission and amplification is likely driven by Culex pipiens (L.) and Culex restuans (Theobald) mosquitoes with human transmission and occasional outbreaks initiated by the “bridge vector” Cx. pipiens (Hamer et al. 2008, 2009). Populations of Cx. pipiens in the Chicago area possess 2 features that may make them especially efficient vectors of WNV. First, the normally ornithophilic Cx. pipiens demonstrates extensive hybridization with the more anthropophilic Cx. pipiens f. molestus (Huang et al. 2009). A wider feeding preference that includes humans and other mammals has been demonstrated for Chicago-area populations of this species (Hamer et al. 2009). And second, Cx. pipiens f. molestus hybrids are a peridomestic mosquito capable of using a variety of household containers and storm water infrastructure for reproduction, placing them in close proximity to both avian hosts and humans (Barr 1967, Mutebi and Savage 2009). Risk estimations suggest that together Cx. pipiens and Cx. restuans could be responsible for up to 80% of human cases of WNV in this region (Kilpatrick et al. 2005).

Mosquito abatement districts in Cook County, IL, use a variety of strategies to mitigate the risks posed by vector Culex spp. mosquitoes. Public education programs, source reduction efforts, catch-basin larval control, and ground-based ultra–low volume (ULV) adult mosquito control applications are routinely employed in an attempt to reduce the occurrence of human illness from WNV. Ultra–low volume adult mosquito control is frequently used throughout the city of Chicago and surrounding suburbs during July, August, and September when mosquito infection rates and populations are the highest and the risk of human WNV infection is the greatest (Hayes et al. 2005, Karki et al. 2018). Typically, ground-based ULV treatments disperse a fine aerosol (12–20 μm diameter droplets) of pesticide, which impinges on flying mosquitoes within the treatment zone (Bonds 2012). In the Chicago area, pyrethroid-based control materials are commonly employed to rapidly reduce populations and the entomological risk posed by vector mosquitoes. Therefore, ensuring highly effective ULV treatments that can reduce infected adult mosquito populations and disease risk is an important component of any integrated mosquito management program (Hayes et al. 2005).

A tremendous amount of literature exists that details the effectiveness of ULV-based methodologies for killing adult mosquitoes (Mount et al. 1996, Mount 1998, Bonds 2012). In most of these prior works, the effectiveness of ULV treatments have been defined using caged adult trials (Mount et al. 1978, Chadeeq 1985). Caged trials, where insectary-reared or field-caught mosquitoes are caged and then subjected to an adult control treatment, predominate because of the ease of deployment and the elimination of migration and reinfestation of the treatment area as confounding factors (Mount 1998). Using multiple cages in a caged field trial can also provide important data on the spatial pattern of mosquito mortality, the distance of the effect, and the influence of wind or other local meteorological conditions. More recent studies detailing the effectiveness of adult control materials containing the active ingredients d-phenothrin and prallethrin have also relied on caged adult trials to characterize the treatment effect and determine effectiveness (Qualls and Xue 2010, Suman et al. 2012, Xue et al. 2013, Farajollahi and Williams 2013). While cage trials are very useful for standardizing treatment conditions, evaluating the magnitude of control exerted, and making clear comparisons between potential control materials, they do not provide any direct information about the effect of a treatment on existing natural mosquito populations. Therefore, the caged trial serves as a very useful, but imperfect proxy for understanding the effectiveness of ULV adulticide treatments.

Attempts to examine the effect of ULV treatments on natural mosquito populations (uncaged) are limited by a different set of factors. First, a large treatment area is required (≥1 mi²) to limit or reduce migration into the treatment area (Mitchell et al. 1970, Mount 1998). Second, trapping methodologies introduce substantial sampling bias because they are prone to capturing different subsets of the mosquito population (i.e., gravid vs. host-seeking mosquitoes) and may provide only a narrow perspective on the treatment effect (Acuff 1976, Burkett et al. 2001, Ball and Ritchie 2014, Brown et al. 2014). Third, a true experimental design would rely on sample randomization and the selection of a control location that mimics the landscape and population dynamics of a treated location—a very difficult feat to achieve (Andis et al. 1987, Ritchie and Devine 2016). Finally, trapping mosquitoes over a large area for an extended time with conventional trapping technologies is extremely labor intensive and expensive. For this reason, evaluations of the effectiveness of ULV adulticides on natural adult mosquito populations do exist but are limited (Mitchell et al. 1970, Chadeeq 1985, Andis et al. 1987, Fox and Specht 1988, Elnaiem et al. 2008, Sudsom et al. 2015, Stoddard 2018). Other strategies for assessing the effect of ULV treatments on mosquito populations have relied on egg counts or immature mosquito counts to detect changes in the population after a treatment with notably mixed results (Andis et al. 1987, Reiter et al. 1990, Chua et al. 2005, Reddy et al. 2006).

In this study, we assessed the effect of ULV adulticide treatments on Cx. pipiens and Cx. restuans mosquitoes in the Chicago suburb of Skokie, IL, using an experimental methodology that attempts to ameliorate many of the shortcomings of caged trials and trapping biases by continuously sampling the natural population of both gravid and host-seeking mosquitoes. We did this by greatly simplifying the trapping protocol while increasing the data resolution with the use of a BG-counter (Biogents GMBH, Germany), expanding the number of traps within a treatment zone, monitoring the local mosquito population for a longer duration, and using both Centers for Disease Control and Prevention (CDC) gravid traps and CO₂ baited traps to determine the effect of a treatment on a greater subset of the Culex spp. population. From this work it was determined that evaluating the effectiveness of a ULV treatment on a natural mosquito population reveals a dynamic population level response that is much more complicated than a simple reduction in mosquito numbers but instead depends on the physiological status of mosquitoes, the pattern of treatments, and possibly other undetermined factors.

This study was conducted within the boundaries of the North Shore Mosquito Abatement District (NSMAD). The NSMAD is a special-purpose taxing district comprised of ∼70 square miles of suburban and urban development in the northeastern part of Cook County, IL (Fig. 1A, 1B). The NSMAD serves 13 communities in the North Shore area (Fig. 1C) including the cities of Evanston and Skokie.

Two operational ULV adulticide treatment maps (hereafter referred to as study site “C-11” and study site “C-18”; Fig. 1D) were selected within the city of Skokie, IL, to conduct ULV effectiveness field trials. Each study site encompasses a ∼1 mi² portion of dense residential development. Houses in each study site are located along a tightly gridded road network that also contains a series of service alleys for utilities, garbage collection, storage buildings, and automobile parking (Fig. 1D). This region of the Chicago area is also characterized by generally poor drainage, frequent standing water, and an antiquated storm water/sewer system that holds water in catch basins and subterranean storm water connectors throughout the city. Therefore, larval habitat for Cx. pipiens and Cx. restuans is abundant and diverse throughout both study sites and also includes above-ground sites such as abandoned pools, clogged gutters, and household containers and refuse that are able to collect water.

This study was conducted from July 12, 2018, to August 31, 2018, to coincide with the seasonal peak of WNV activity and historically predictable nighttime meteorological conditions conducive to truck-based ULV adulticide treatments. Mean daily windspeeds and mean daily precipitation amounts were monitored and recorded during the study. Mean daily windspeed and mean daily precipitation data was obtained from the Nation Oceanic and Atmospheric Administration (NOAA) weather station at the Chicago Botanical Gardens.

Adult mosquito control operations via truck-based ULV (Clarke Cougar, Roselle, IL) were conducted using Duet^™ (5% d-phenothrin and 1% prallethrin as active ingredients synergized with 5% piperonyl butoxide). Applications were made starting between 8:50 p.m. and 9:30 p.m. on treatment nights and terminated when the study area was adequately covered. Ultra–low volume adulticide applications were made at an application rate of 1.25 oz/acre based on a 300 ft swath width. In study site C-18, treatments occurred on July 12, 2018; July 31, 2018; August 6, 2018; and August 14, 2018. In study site C-11, treatments occurred on July 31, 2018, and August 14, 2018. All study sites had adjoining spray maps treated except for the August 6, 2018, treatment in C-18, which occurred in isolation.

This study made extensive use of the BG-counter to continuously determine abundance of gravid and host-seeking mosquitoes simultaneously. The BG-counter is an automated mosquito population sampling device that can identify and count mosquitoes against a backdrop of larger or smaller insects as they come into a fan-powered trap. The BG-counter also contains sensors to monitor relative humidity and temperature. The BG-counter wirelessly transmits collected data with a cellular antenna back to an internet-based application (live.bg-counter.com) for viewing or downloading. The BG-counter was selected for this project because short, medium, and long-term trends in mosquito populations can be observed in response to control efforts, climate, or other factors with minimal trapping effort.

Gravid Cx. pipiens and Cx. restuans mosquitoes were continuously sampled using a standard CDC gravid trap modified to incorporate a BG-counter. The incorporation of a BG-counter into a standard CDC gravid trap enables an automatic and continuous count of gravid mosquitoes to be made (Willis 2017). Gravid mosquito populations are reported in 15-min increments (4 counts per hour) and can provide mosquito abundance data with high temporal resolution and in near real time. The gravid BG-counter trap oviposition pan was baited with an Alfalfa pellet infusion and treated with Vectolex FG (7.5% Lysinibacillus sphaericus Meyer and Neide) larvicide as needed to prevent mosquito development within the reservoir of the trap. The gravid BG-counter was allowed to run continuously (24 h) over the study period. Mosquito samples were removed from the trap catch bag 3 times/week and saved to determine species composition of trapped mosquitoes. Since Cx. pipiens and Cx. restuans are nearly morphologically indistinguishable when damaged by trapping, identifications of this species were combined (Kulasekera et al. 2001, Russell and Hunter 2005).

To sample mosquitoes that were host seeking, a BG-counter was affixed to a BG-Sentinel II (Biogents GMBH, Regensburg, Germany) and baited with pressurized CO₂. The traps and counters were operated continuously (24 h) during the study period. Host-seeking mosquito populations were reported via web application in 15-min increments (4 counts per hour). Trapped mosquitoes, which were collected 3 times/week to validate counted mosquitoes, were Cx. pipiens/restuans and not Aedes vexans (Meigen) or other nuisance species.

Traps were located in yards with vegetation and other habitat likely to harbor mosquitoes. Study site C-11 received 3 host-seeking BG-counters and 1 Gravid BG-counter. Study site C-18 received 4 host-seeking BG-counters and 2 Gravid BG-counters (Fig. 1D). Some individual trap locations were occupied by both gravid and host-seeking traps set 3–4 m apart. Traps were located away from the study site edges to minimize any migration effects from adjoining spray areas.

To calculate the mean number of mosquitoes per day from host-seeking and gravid BG-counter data, the total number of mosquitoes counted by the trap for each 15-min increment was summed for the 24-h period running from midnight to midnight each day (Table 1). Traps within the same study site and of the same type were averaged (mean) together for each 24-h period during the study, and the standard error of the mean (SEM) was calculated (Table 1). Daily mean relative humidity (RH) and daily mean temperature data recorded by the BG-counter was calculated in the same manner as mosquito counts (Data not shown).

The daily means for host-seeking and gravid mosquitoes in study site C-18 and C-11 were analyzed using Pearson's Correlation Coefficient (Pearson's r) in Graphpad statistical software (GraphPad Software, San Diego, CA) for any statistically significant correlational relationship between mean daily RH (BG-counter), mean daily temperature (BG-counter), mean daily windspeed (NOAA), mean daily precipitation (NOAA), and mosquito populations (BG-counter). Analyses were conducted on both the entire data set as well as a 96-h window following each treatment to identify meteorological explanations for the population pattern observed.

To analyze the differential response of gravid and host-seeking female mosquito populations to ULV adulticiding, a bivariate analysis was conducted in GraphPad software with daily mean host-seeking mosquitoes plotted as the independent variable and daily mean gravid mosquitoes plotted as the dependent variable. The relationship between host-seeking and gravid mosquitoes was further analyzed using a 2-tailed Pearson's r.

Nuisance (non-WNV vector) mosquito species richness and abundance across the NSMAD was monitored during the study period via New Jersey Light Traps (NJLT; Model 1112; John Hock Inc., Gainesville, FL). Twice weekly collections were made of trapped mosquitoes. Mosquitoes were identified and counted.

New Jersey Light Trap (NJLT) counts averaged only 6 mosquitoes/day due to the dry conditions during the study period. Heavy precipitation during August 25 and 29 (Fig. 2A) led to the development of large populations of nuisance Ae. vexans (>100 mosquitoes per day), thus obfuscating the pattern of Culex spp. populations via host-seeking BG-counter traps and terminating the study. During the entire study period, only 452 mosquitoes, 1.2% of the seasonal NJLT mosquito count, were captured by NJLTs across the entire district.

The mean daily temperature during the study period ranged from 18.8°C to 28.8°C (Fig. 2A). Precipitation during the 1st 4 wk of the study period was minimal, with daily mean rainfall amounts less than 1 cm. The last 1 wk of the study period brought rainfall greater than 1 cm and standing water throughout the district (Fig. 2A).

When host-seeking Cx. pipiens/restuans populations were compared 24-h before (pretreatment) an adulticide treatment versus the 24-h period after a treatment (posttreatment), a 65.3% decline was recorded (Table 2; n = 5, SEM 1.4%). The percent change in pretreatment versus posttreatment populations ranged from a −62.8% decrease to a −69.1% decrease and was reliably detected in all 5 ULV treatments at 2 locations (Table 2).

When gravid Cx. pipiens/restuans populations were compared 24-h pretreatment versus 24-h posttreatment, a mean decline of −29.2% was recorded (Table 2; n = 6, SEM 11.0%). The percent change in gravid mosquitoes was variable and ranged in pre vs. post populations from a −68.2% decrease to a 1.0% increase in 1 treatment (Table 2).

When host-seeking Culex spp. populations were compared 24-h posttreatment to 72-h posttreatment, sampled mosquitoes showed a strong rebound of 303.1% (Table 2; n = 5, SEM 94.4%). Populations of Cx. pipiens and Cx. restuans mosquitoes increased 72-h posttreatment to levels higher than the pretreatment population in every single adulticide treatment conducted.

The population of gravid mosquitoes responded differently than host-seeking mosquitoes. When populations 24-h posttreatment were compared to populations 72-h posttreatment a smaller increase of 5.7% was observed (Table 2; n = 6, SEM 15.9%). This increase was also highly variable, with changes ranging from −41.2% to +72.1% depending on the treatment (Table 2). Only 3 out of 6 treatments resulted in a lower gravid Cx. pipiens/restuans population 72-h posttreatment.

Mosquito adulticide treatments were followed by sharp but transient declines in the host-seeking mosquito population as measured by the BG-counter. After each treatment in study zone C-18, the mosquito population rebounded to levels at least as high as 24-h pretreatment (Fig. 2B). Between the dates August 18, 2018, and August 28, 2018, the host-seeking Culex spp. population oscillated dramatically with the same periodicity (6–8 days) and intensity as the 3 prior successive adulticide treatments despite receiving no treatment on August 19, 2018 (Fig. 2B).

In study zone C-11, ULV adulticide treatments were scheduled 2 wk apart. Each treatment was followed by a temporary decline of >60% and then a 1.5 to 2.5-fold rebound in the host-seeking population (Fig. 2C). The August 6, 2018, treatment in C-18 did not occur in study site C-11. Therefore, no rebound was observed in the host-seeking population during this time in C-11 (Fig. 2C).

A comparison of the mean number of host-seeking mosquitoes in study site C-18 to the mean number of gravid mosquitoes for each date during the study using a bivariate plot demonstrates that the relationship between gravid mosquitoes and host-seeking mosquitoes is not correlational or proportional (Table 3; Pearson's r = 0.116, n = 36, P = 0.5). Increases in the host-seeking population are not met with increases in the gravid population, and the gravid mosquito population appears to be have a limit (Fig. 3). A graphical comparison of the C-18 mosquito population 24-h before a treatment with the period 24-h after a treatment demonstrates that gravid mosquito populations did not decline proportionally to host-seeking mosquitoes following ULV adulticiding, and in 1 case (July 31, 2018) the gravid population increased slightly (Fig. 3).

Correlational comparisons of mean daily humidity, mean daily precipitation, and mean daily windspeed with gravid or host-seeking counts failed to yield any statistically significant correlations both across the July 28, 2018–September 1, 2018, timeframe (Table 3) and across the discrete 96-h window following a treatment (Data not shown). The mean daily count of gravid mosquitoes in C-18 was positively correlated with mean daily temperature (Table 3; Pearson's r, r = 0.396, n = 36, P = 0.0169).

Meteorological conditions during the study period were remarkably stable (Fig. 2A). Daily precipitation was generally insufficient to create a larval “flushing” event (thereby affecting adult mosquito populations) until the end of the study period (Fig. 2A; Koenraadt and Harrington 2008). A correlational relationship between mosquito populations in study site C-18 and C-11 and any meteorological factor (windspeed, temperature, humidity, and precipitation) could not be determined except for a statistically significant interaction between gravid mosquitoes and daily mean temperature in study site C-18 (Table 3). The lack of clear associations between mosquito populations and meteorological conditions suggests that the patterns seen in mosquito abundance in study site C-11 and C-18 were influenced primarily by ULV treatments and possibly other factors such as larval production dynamics (Fig. 2B, 2C).

Floodwater throughout the NSMAD was absent, which limited the production of floodwater species such as Ae. vexans to less than 2% of the catch total in the CO₂ baited BG-counters (data not shown). This period of low species richness was also observed in contemporaneous catches from NJLTs, which only collected 452 mosquitoes of species other than Cx. pipiens/restuans during the entire study period. Although BG-counters can count mosquitoes as they pass through the trap, the counters are unable to determine the species. The dearth of additional species attracted to traps (BG-counters and NJLTs) during the study period, along with periodic resampling of the BG-counter catches for species richness, indicates that the population trends seen in Fig. 2B, 2C are representative of the vector Culex spp. population and not Ae. vexans or any other species common to the area.

To our knowledge this is the 1st study that has attempted to describe the effect of multiple adulticide treatments over a long period of time on both gravid and host-seeking mosquitoes simultaneously. The lack of effective control observed with gravid mosquitoes when compared with host-seeking mosquitoes (Fig. 3) was unexpected but not without precedent. A variety of previous studies have also described unexpected results when adulticide treatments are paired with population assessments that rely on collecting gravid mosquitoes or egg rafts. In one notable example, Reddy et al. (2006) documented no effect on Cx. pipiens/restuans egg-raft production following a series of treatments with resmethrin despite treating a fully susceptible population and conducting several trials. Chadeeq (1985) and Chua et al. (2005) both documented a failure of adulticide fogging to influence egg counts of Aedes aegypti mosquitoes. Reiter et al. (1990) similarly documented an unexpected oscillation in Culex spp. egg-raft production following treatments with resmethrin. Gravid mosquitoes at the time of treatment were clearly surviving and ovipositing 3–4 days later (Reiter et al. 1990). Laboratory and model-based (in silico) studies demonstrated that an oscillation in egg-raft production can be caused by a transient physiological resistance to adult control materials during the 24–72-h period following a blood meal in Cx. pipiens, Ae. aegypti (L.), and Culiseta melanura (Coq.) (Eliason et al. 1990, Moore et al. 1990). Halliday and Feyreisen (1987) previously demonstrated the same transient physiological resistance to trans-permethrin, fenvalerate, and dichlorodiphenyltrichloroethane following a blood meal, and similar results have appeared in the literature since 1956 (Hadaway and Barlow 1956). In both the Eliason et al. (1990) and Halliday and Feyereisen (1987) examples, the amount of topical insecticide required to cause mortality in gravid females was approximately doubled, while in the present study the mortality of gravid mosquitoes was approximately halved when compared with host-seeking mosquitoes. This study was conducted near the maximum application rate permitted and, therefore, represents the likely maximum effectiveness obtainable by our truck-based ULV methodology.

Previous works demonstrating a transient tolerance to insecticides, taken together with the work presented here, suggest that insecticide materials may exhibit different effects on different physiological subsets of the active Culex spp. population and that any cohort of female mosquitoes is only fully susceptible: 1) after emergence and before the 1st blood meal and 2) between oviposition and the next blood feeding (Reiter et al. 1990, Eliason et al. 1990). In addition, blood-fed female mosquitoes frequently seek shelter, further limiting exposure to airborne adulticide materials. As Fig. 3 illustrates, host-seeking mosquitoes in study site C-18 were clearly reduced 24-h after a ULV treatment, while gravid mosquitoes showed a variable response, declining modestly after 2 treatments but increasing after another treatment. The implications of this result are troubling for ULV interventions aimed at disrupting viral transmission because, at any given moment, those mosquitoes most likely to be infected with an arbovirus (blood fed and/or gravid) are the least likely to be controlled with an ULV adulticide due to sheltering behavior and/or decreased susceptibility. In addition, these results suggest that the underlying infection rate of mosquitoes during periods of viral transmission is not likely to be reduced by a single ULV adulticide treatment (Newton and Reiter 1992). A cohort of infected and gravid mosquitoes will always be tolerant to the insecticide or sheltering during treatment and would return to host seeking and susceptibility at some point within the few days after a treatment, thus perpetuating the underlying infection rate and viral transmission.

In this study, the cohort of gravid, resistant, and possibly infected mosquitoes was not insignificant. The mean proportion of gravid mosquitoes was 32% of the total sampled population and did not increase proportionally to increases in host-seeking population (Fig. 3; Table 3). The lack of a clear positive correlational relationship between the number of host-seeking mosquitoes and the number of gravid mosquitoes was unexpected and suggests that the population of gravid mosquitoes has a definable asymptote. Although the reason for a limit to the gravid population is unknown, it may be an indication that the conversion of host-seeking to gravid mosquitoes are rate-limited by the availability of blood meals or some other factor.

Studies that demonstrate the relationship between ULV treatments and prevalence of arbovirus in vector mosquitoes remain rare (Ritchie and Devine 2016). Despite this, many authors have suggested that more frequent ULV adulticiding may be necessary to alter viral transmission dynamics (Andis et al. 1987, Eliason et al. 1990, Reiter et al. 1990, Moore et al. 1990, Pawelek et al. 2014). It is interesting to note that some of the only papers that have demonstrated a clear reduction in viral prevalence from ULV treatments in vector mosquitoes conducted aerial treatments over 3 sequential nights (Elnaiem et al. 2008, Macedo et al. 2010). The work presented here further supports the notion that multiple ULV treatments, timed to account for cohorts of mosquitoes transitioning into and out of susceptibility, may yield a greater reduction in viral transmission and vector populations than isolated ULV treatments (Andis et al. 1987, Eliason et al. 1990, Reiter et al. 1990, Moore et al. 1990). Although we did not assess mosquito infection rates in this study, the majority of gravid Culex spp. remained alive after ULV treatments (Tables 1 and 2).

Transient declines in mosquito population following an adulticide treatment have been described in a variety of previous works (Mitchell et al. 1970, Chadeeq 1985, Andis et al. 1987, Focks et al. 1987, Chua et al. 2005, Pawelek et al. 2014, Sudsom et al. 2015, Stoddard 2018). In some cases, a rebound in mosquito population has been attributed to immigration from surrounding treatment areas, and, in general, this seems possible (Mitchell et al. 1970, Reddy et al. 2006). However, the present study was conducted in conjunction with standard mosquito abatement operations in response to elevated mosquito WNV infection rates. Although only 2 ∼1 mi² treatment maps were sampled for mosquito abundance, the surrounding treatment maps were simultaneously treated via ULV as part of regular vector control operations. It is therefore unlikely that immigration into the treatment area alone can account for the dramatic rebound in populations. In 5 unique treatment events across 2 study areas, the application of ULV adulticides preceded a robust rebound effect (Fig. 2B, 2C). On 1 occasion (August 6, 2018) study site C-18 was treated while study site C-11 was not treated, thus serving as a nontreated control. By August 11, 2018 (5 days posttreatment) study site C-18 had experienced a 4-fold increase in the host-seeking population, while the untreated site (C-11) was nearly unchanged. The different patterns observed in C-18 and C-11 following the August 6, 2018 treatment could not be explained with meteorological data (Table 3). This result suggests that ULV adulticiding produces a transient decline followed by robust rebounds in the host-seeking mosquito population (Fig. 2B, 2C).

While the exact reasons for such a rebound effect are unknown, there are some possible explanations beyond only reinfestation and migration, which should be studied in more detail. One possibility is that the rebound in populations is caused in part by the conversion of gravid mosquitoes (and thus protected from the ULV treatment because of physiology or cryptic behavior) back into host-seeking mosquitoes in the days following a treatment. Certainly, some portion of the rebound is comprised of mosquitoes that survived the adulticide treatment, oviposited, and could then return to host seeking. Perhaps treatment survivors and newly emerged mosquitoes together can account for the dramatic rebound.

Another possibility is that sublethal exposure to pyrethroid adulticides may alter the vector mosquito's ability to locate a host or CO₂ baited trap. Under this scenario, the host-seeking traps would not only be recording a reduction in population but also a loss of olfactory sensitivity in surviving mosquitoes for host cues such as CO₂. Laboratory evidence for this possibility exists as well (Cohnstaedt and Allen 2011) and can potentially explain the difference seen between gravid traps and those meant to attract host-seeking mosquitoes after a treatment. Field trials using human landing rate counts may be especially sensitive to this confounder.

A final possibility suggests that some interaction between the ULV adulticide and larval development could be occurring. Under this scenario, trace quantities of adulticide may cause a surge in pupation and emergence or may induce density-dependent effects in the larval habitat by causing mortality of the earliest instars and a release from competition. Muturi et al. (2010) demonstrated that ULV-relevant trace quantities of malathion can enhance the survival of late mosquito instars by killing earlier instars and reducing competition. However, any interaction between ULV adulticide and larval mosquito development or mortality under field conditions has yet to be demonstrated and, in general, ULV adult control treatments are specifically designed to limit deposition.

At first glance, the rapid increase in mosquito populations following a ULV treatment may appear counter-productive for arboviral control. However, the age-structure of the mosquito population is an important aspect of viral transmission, since older mosquitoes are more likely to be infective, and strategies that shorten mosquito lifespan or cause mortality in older mosquitoes are also likely to impact viral transmission (Cook et al. 2008). Without knowing the age structure of the rebounding population, or the effect of a ULV treatment on the age structure of the population, conclusions about the “effectiveness” of ULV treatments based solely on percentage reductions of the population or pre/post comparisons should not be made. It is entirely possible that significant reductions in viral prevalence may occur despite rebounds in mosquito population.

Although not a fully randomized controlled field trial, the work presented here has identified 3 important points that should be considered in future field trials. First, the measurement tools of a field trial (i.e., trapping methodology) need to be carefully considered to observe the effect on the entirety of the mosquito population. Trials that rely on only 1 trap type may present highly biased data that represents only mosquitoes of a specific physiological state. Second, the “endpoint” measurement for effectiveness should be clearly defined. In every single treatment we observed a rapid reduction and then rebound of host-seeking populations. Deciding the “effectiveness” of our ULV treatments could be just a matter of selecting the correct timeframes to compare. Our ULV adulticide methodology was both very effective and simultaneously not effective at all depending on what was measured and how it was measured. Since the ultimate endpoint for public health ULV adulticiding is to control disease and not necessarily to control mosquito abundance (although the 2 can be related), more studies that examine the relationship between ULV interventions and viral prevalence should be conducted. Finally, the complexity of a dynamic mosquito population should be appreciated when conducting assessments of ULV materials.

This publication was supported by Cooperative Agreement #U01 CK000505, funded by the Centers for Disease Control and Prevention. Its contents are solely the responsibility of the authors and do not necessarily represent the official views of the Centers of Disease Control and Prevention or the Department of Health and Human Services.