Introduction
Battery-operated CDC-type light traps (Sudia & Chamberlain, Reference Sudia and Chamberlain1962) supplemented with CO2 are commonly used in mosquito surveillance programs (Moore et al., Reference Moore, McLean, Mitchell, Nasci, Tsai, Calisher, Marfin, Moore and Gubler1993). These devices capture a greater number and variety of adult mosquitoes than other trap types (e.g. resting boxes, malaise traps, ovitraps) (Williams & Gingrich, Reference Williams and Gingrich2007) and provide faunal composition and abundance data that are important for the implementation and evaluation of mosquito control activities (Amoo et al., Reference Amoo, Xue, Qualls, Quinn and Bernier2008). Similarly, risk assessment models for disease transmission and depictions of mosquito distribution produced by spatial analysis methods and mapping systems software rely on data provided by light traps (Diuk-Wasser et al., Reference Diuk-Wasser, Brown, Andreadis and Fish2006), including inputs that are used to estimate mosquito density; biting activity on humans; and age-structure, survivorship and pathogen infection rate(s) in the mosquito population (Garrett-Jones, Reference Garrett-Jones1964; Reisen & Pfuntner, Reference Reisen and Pfuntner1987; Gu et al., Reference Gu, Lampman and Novak2003; Kilpatrick et al., Reference Kilpatrick, Kramer, Campbell, Alleyne, Dobson and Daszak2005; Eisen & Eisen, Reference Eisen and Eisen2008).
Light traps have been studied extensively as a human host surrogate for the estimation of mosquito landing/biting rates. These evaluations have targeted primarily malaria vectors and used unbaited CDC-type traps. Odetoyinbo (Reference Odetoyinbo1969) in The Gambia, and later Gunasekaran et al. (Reference Gunasekaran, Jambulingam, Sadanandane, Sahu and Das1994) in India and Hii et al. (Reference Hii, Smith, Mai, Ibam and Alpers2000) in Papua were unable to demonstrate a relationship between the numbers of landing/biting mosquitoes on humans and those captured by light trap. However, Garret-Jones & Magayuka (Reference Garrett-Jones and Magayuka1975) combined mosquito responses to three indoor trapping techniques (the CDC portable trap and the Monks Wood light trap fitted with white light and mercury vapour tubes) and the use of bed nets by humans to estimate the ‘man-biting density per person per night’ for Anopheles spp. in Tanzania. The same workers showed that light traps did not increase the number of hungry mosquitoes entering human sleeping areas but rather intercepted those otherwise present and prevented from feeding by bed nets. Yet other studies in Thailand (Ismail et al., Reference Ismail, Pinichpongse, Chitprarop, Prasittisuk and Schepens1982), Tanzania (Lines at al., 1991; Davis et al., Reference Davis, Hall, Chee, Majala, Minjas and Shiff1995), Kenya (Mbogo et al., Reference Mbogo, Glass, Forster, Kabiru, Githure, Ouma and Beier1993), Venezuela (Rubio-Palis & Curtis, Reference Rubio-Palis and Curtis1992), Burkina-Faso (Costantini et al., Reference Costantini, Sagnon, Sanogo, Merzagora and Coluzzi1998) and Sierra Leone (Magbity et al., 2002) have demonstrated a correspondence between the numbers of Anopheles and Culex mosquitoes collected in light traps placed near bed net-protected human hosts and the numbers of mosquitoes biting human collectors.
The capture rates of mosquitoes in CDC-type light traps have been compared directly with those landing on human hosts in two additional studies. Vaidyanathan & Edman (Reference Vaidyanathan and Edman1997) explained 18% of landing Cx. salinarius Coquillett responses on a human host from the numbers of females trapped in CDC light traps in Massachusetts, USA. Strickman et al. (Reference Strickman, Miller, Kim and Lee2000), in The Republic of Korea, compared mosquito landing responses to human hosts with those to CO2-baited CDC light traps and developed a series of density thresholds based on the latter that were used to estimate attainment of a minimum significant potential biting rate by An. sinensis Wiedemann.
From a sampling perspective, light traps are useful for catching large numbers of certain mosquito species and for measuring relative changes in abundance of these species in time and space. They have limited value as an ecological tool, however, because they sample species and sub-populations within species unequally (Service, Reference Service1993; Southwood & Henderson, Reference Southwood and Henderson2000). In the latter case, group testing-based estimates of virus infection rate in pools of light-trapped mosquitoes (Chiang & Reeves, Reference Chiang and Reeves1962; Lanciotti et al., Reference Lanciotti, Kerst, Nasci, Godsey, Mitchell, Savage, Komar and Spielman2000; Gu et al., Reference Gu, Lampman and Novak2003), while effective for documenting virus transmission, can lead to an upwardly biased estimate of infection rate in the vector population (Katholi & Unnasch, Reference Katholi and Unnasch2006). This appears to be a critical issue, given an implicit assumption of equivalency between light trap-captured and natural mosquito population parameters (Reisen & Pfuntner, Reference Reisen and Pfuntner1987).
The study reported here was made under controlled conditions to compare, as precisely as possible, the capture rate of adult mosquitoes by light trap (LT) with a baseline capture rate determined using the human landing (HL) collection method. The objective of the study was to develop the procedures and techniques needed to observe, analyze and interpret mosquito responses to LT and HL and to propose hypotheses that can be tested using these methodologies with field populations of mosquitoes.
Materials and methods
Test arena
Observations were made in an aluminum-framed building (12.9 m L×4.3 m W×2.8 m [average] H) with a fiberglass window screen (approximately seven openings per linear cm) on four sides and a white-enameled aluminum sheet metal roof. Flooring comprised 20-cm-deep ‘pea gravel’ (0.50–0.75-cm dia) throughout. The building was partitioned (by a translucent vinyl-fiberglass screen panel) into two equal-sized rooms (6.4 m L×4.3 m W×2.8 m H) each with an external door. Window screen was fitted into channels in the exterior frame members of the building and held in place with rubber stripping. All joints and other openings in the building were sealed to prevent entry/escape of mosquitoes and invertebrate or vertebrate predators. The volume of each room was 77.7 m3.
Mosquitoes
The mosquito species selected for this study, while of laboratory origin, were intended to represent diverse taxa. The test populations were of finite density and were confined to screened enclosures to eliminate emigration/recruitment effects. Mosquitoes were reared outdoors so that we could observe their responses within the context of exposure to natural cycles of light and temperature.
Capture rate responses were studied for Aedes albopictus (Skuse) (Gainesville strain, 1992), Ae. triseriatus (Say) (Gainesville strain, 1996), Anopheles quadrimaculatus Say (Orlando strain, 1952), Culex nigripalpus Theobald (Vero Beach strain, 1999) and Cx. quinquefasciatus Say (Gainesville strain, 1995). Cohorts of eggs of each species were reared to the adult stage outdoors under ambient light and temperature conditions using the techniques described by Gerberg et al. (Reference Gerberg, Barnard and Ward1994). An approximately equal number of 4–9-day-old nulliparous female mosquitoes were available in each room at the beginning of a given test, although this number ranged among the tests from approximately 800 to 2200 females (density: 12.8–28.3 ♀ m−3), depending on the time of year in which tests were made (March through October) and the water temperature during mosquito development. Adult mosquitoes were released into each room from the holding cages 24 h in advance of the beginning of a test. During this time, they were provided 10% sucrose solution (in H2O) via cotton wick.
Collection methods
Each test lasted 24 h and comprised a matched pairs comparison of the numbers of mosquitoes captured by a continuously-operating miniature (CDC-type) light trap (LT) with the numbers captured using the human landing (HL) collection method in eight separate 15-min-long intervals (spaced throughout the 24-h period). The LT was operated in the center of one (randomly selected) room of the screened building. At the same time, in the center of the second room, mosquitoes that landed on the exposed forearm of a human subject were collected with a mechanical aspirator (Hausherr's Machine Works, Toms River, NJ). For LT collections, a programmable (model 4012, John W. Hock Company, Gainesville, FL) collection bottle rotator, fitted with a single CDC-type light trap (with light) (model 512, John W. Hock Company, Gainesville, FL) and eight collection bottles, was used to capture adult mosquitoes in each of eight consecutive collection intervals of variable duration (see below) in a 24-h period. Compressed CO2 gas was released continuously at the rate of 250 ml min−1 from the end of 0.5 mm O.D. Tygon® tubing attached 1.5 cm below the LT intake. HL collections were made using a mechanical aspirator to vacuum mosquitoes that alighted on the skin and commenced immediately to probe (i.e. touched the skin with their proboscis) or that remained on the skin for five seconds. None was allowed to bite. Approximately 415 cm2 of forearm skin was exposed for this purpose, comprising the area from a line of circumference 3 cm below the elbow to a line of circumference 3 cm above the wrist. To capture landing mosquitoes, the exposed forearm was extended forward in front of the body of the test subject and held approximately 45 cm above the ground throughout the collection period. The exposed skin surface was observed for landed mosquitoes during this time as the arm was rotated in a counterclockwise then clockwise fashion. A 6000 candle power VisorLIGHT™ (Model LT06, Donegan Optical Company, Lenexa, KS, USA), attached to the top of the aspirator and fitted with a red acetate lens cover, provided on-demand night-time illumination. All HL collections were made from the same human subject (DRB).
In each 24-h test, the schedule for LT operation and for HL collections was arranged according to the times of sunset and sunrise within the diel period and the length (min) of the corresponding photophase (sunrise to sunset) and scotophase (sunset to sunrise) (local times for each event were obtained from the Nautical Almanac, US Naval Observatory, for longitude W 82°20′ and latitude N 29°40′). To do this, the 24-h day was divided into eight ‘periods’ (periods 1 through 8). Periods 1 and 5 incorporated the two light-transition events in the diel period (day-sunset-night and night-sunrise-day, respectively). The day portion of these two periods comprised one-eighth of all minutes in the photophase, the night portion one-eighth of all minutes in the scotophase. In the same manner, periods 2, 3 and 4 (night) each comprised one-fourth of all minutes in the scotophase and periods 6, 7 and 8 (day) each comprised one-fourth of all minutes in the photophase. Thus, for example, in the case of tests made in June when the photoperiod was 14L:10D, the duration of periods 2, 3 and 4 was 150 min each, the duration of periods 6, 7 and 8 was 210 min each, and the duration of periods 1 and 5 was 180 min each. For period 1, the time intervals preceding and following sunset were 105 min and 75 min, respectively; and, for period 5, the time intervals preceding and following sunrise were 75 min and 105 min, respectively.
The LT was operated (and CO2 released) continuously for 24 h. The collection bottle rotator was programmed to move a new collection bottle in position beneath the light trap at the beginning of each period. Collections using the HL method were made for 15 min in each period. In periods 1 and 5, these were made at sunset and sunrise, respectively. In periods 2, 3 and 4 (night) and 6, 7 and 8 (day), HL collections were made midway through each period. Studies commenced July 2004 and ended October 2006. During this time, five matched-pair comparisons (replicates) of responses to the LT and HL collection methods were made for each mosquito species.
Catch per unit effort
The total number of female mosquitoes captured (nf) by each collection method in each period (i) in each replicate was transformed to log10 (nf i+1). Operating time (t) for the HL method in each period was 15 min. Operating time (t) for LT in a period ranged between 150 and 210 min. For analysis purposes, we calculated a catch-per-unit-effort response (R) (Southwood & Henderson, Reference Southwood and Henderson2000) for the nf i observed for each species according to collection method and replicate as the number of mosquitoes captured after 1 min of collection time:
Capture rate by period
Differences in mean R between periods were compared for each species according to collection method using the model: R=β0+β1(period). For each species, the pattern of diel activity indicated by LT and HL was compared by rank ordering all eight periods according to mean R (rank=1 for the period with highest R; rank=8 for the period with lowest R) and then testing the difference in ranks assigned LT and HL in the same period for departure from 0 using Student's t-test.
Capture efficiency index
A capture efficiency index (CEI=R LT/R HL) was used to compare the mosquito capture rate by LT with the capture rate using HL. A mean CEI≥1 indicated equivalent or greater efficiency for LT compared with HL for that mosquito species in that period. An index <1 indicates the LT collection method was less efficient than HL.
Daily capture rate
The daily capture rate is a commonly used operational index of mosquito density, but the comparability of LT and HL data for depicting seasonal trends in mosquito population density is unknown. Given the total LT operating time (T LT=1440 min) and the total HL collection time (T HL=120 min) used in each of our 24-h tests, we calculated the mean daily capture rate (R D) for each species according to collection method (R D(LT) or R D(HL)) as:
The model: R D(LT)=β0+β1(R D(HL)) (Neter et al., Reference Neter, Wasserman and Kutner1983) was used to determine if change in R D(LT) is related to change in R D(HL) and to evaluate the comparability of seasonal population data indicated by each collection method.
Relationship of capture rates by LT to HL
Initially, a linear regression model (R LT=β0+β1(R HL) (Neter et al., Reference Neter, Wasserman and Kutner1983) was used to evaluate the relationship of R LT to R HL for the following sampling regimens: (a) all eight periods, (b) period 1, (c) period 5, (d) periods 1 and 5, (e) periods 1 through 5, (f) periods 5 through 1, (g) periods 6 through 8, (h) periods 2 through 4, (i) periods 2 through 5, (j) periods 6 through 1, (k) periods 8 through 2 and (l) periods 4 through 6. We evaluated the linear and curvilinear response in each case by successive additions of quadratic (β2(R HL2)) and cubic effect coefficients (β3(R HL3)) to the linear model.
Data analysis
Statistical analysis (SAS Institute, 2003) utilized tabulation (PROC MEANS), analysis of variance (PROC ANOVA, PROC GLM) and regression (PROC REG, PROC GLM) procedures. Pre-planned comparisons of means were made using the Least Significant Difference (LSD) test at the 5% level of significance.
Results
Catch per unit effort
When R LT responses were compared with R HL responses, fitted models for Ae. albopictus, Ae. triseriatus and An. quadrimaculatus were not significant (fig. 1). High LT responses generally corresponded with high HL responses, although R LT at or near 0 were observed at R HL≥0.8 for Ae. albopictus, at R HL≥0.2 for An. quadrimaculatus and at R HL≥0.5 for Ae. triseriatus.
Fig. 1. Relationship of mosquito capture rate by LT (R LT) to mosquito capture rate by HL (R HL) for five mosquito species based on catch per one minute of collection effort.
Coefficients (±SE) for the fitted linear model for Cx. nigripalpus (s 2=0.0014) were: β0=0.0085 (0.0097) and β1=0.0508 (±0.0182) (fig. 1). For Cx. quinquefasciatus, the fitted regression line indicated a curved response when R HL<0.75 and a straight line response at higher values. Fitted model coefficients (s 2=0.0365) were: β0=0.1678 (±0.1030), β1=−0.4851 (±0.6144), β2=0.7948 (±0.9388) and β3=−0.2076 (±0.4032) (fig. 1).
Capture rate by period
There was no significant difference in the response to LT by Ae. albopictus from one period to the next. In contrast, average capture rates by HL were highest (F 7,32=5.74, P<0.001) in periods 6–7–8–1 and lowest in period 4 (table 1). HL responses for Ae. albopictus increased between early-morning and sunset (periods 6–7–8), but 25% of all landing females were collected at night (periods 2–3–4) compared with 31% at night by LT.
Table 1. Mean capture rate (±SE) of five mosquito species by LT and HL collection methods.
Tabulated means for LT and HL based on non-transformed data. Row means followed by the same letter are not significantly different (P=0.05, LSD test using log10 (nf i+1).
The effects of period were significant (F 7,32=5.11, P<0.001) for Ae. triseriatus responses to LT but not to HL (table 1). Most females were collected at sunrise and after sunset (periods 5, 2), whereas fewest were captured in daytime (periods 6–7–8, 1). In contrast, HL capture rates for Ae. triseriatus were highest in periods 7–8, 1 and 5 and lowest following sunrise (period 6).
The number of An. quadrimaculatus captured in LT increased before and at sunset (periods 8–1) and at sunrise (period 5). Mean responses differed only for periods 7–8 (LSD, P=0.05) (table 1). Period effects were significant for HL responses (F 7,32=6.93, P<0.001) with collections highest before sunset (periods 7–8) and higher between sunrise and sunset (periods 5–6–7–8–1) than at night (period 4). Sixty-five percent of An. quadrimaculatus were collected by HL during daylight (periods 5–6–7–8) compared with 50% of all females by LT during this time.
Culex nigripalpus responses to LT were not significantly influenced by period. Daily activity patterns indicated by this collection method were bimodal with highest capture rates at, and following, sunset (periods 1–2) and at sunrise (period 5) (table 1). The HL response pattern was unimodal with significant period effects (F 7,32=5.51, P<0.001). Landing rates were highest in periods 8–1 but decreased thereafter through sunrise (periods 2–3–4–5–6).
Response patterns of Cx. quinquefasciatus to LT and HL indicated a single peak of activity in each case but significant period effects for LT only (F 7,32=5.01, P<0.001) (table 1). In this case, responses were higher in periods 8–1–2 than in periods 4–5–6–7, whereas HL responses were higher at sunset (period 1) than before, during or after sunrise (periods 4–5–6) (LSD, P=0.05).
The daily modes of activity indicated by LT and HL compared poorly within species when ranked by period (table 2). There were eight instances of subjective correspondence (i.e. P≥0.59) among the 80 rankings and nine significant (P=0.05) departures from correspondence. Significant departures from correspondence indicate a disparate response (in terms of mean R) to each collection method in the same period. Aedine species accounted for a majority of these differences (64%). In the case of Ae. albopictus (table 2), for example, HL collections indicated period 8 as the time of peak activity (rank=1.2/8), whereas LT collections for period 8 were ranked 3.8/8. Similarly, for Ae. triseriatus, the period 8 rank for LT was 7.3/8 and 2.2/8 for HL. The overall comparability of ranks was greatest (based on P) for Cx. quinquefasciatus, particularly in periods 4–5–6 and 8.
Table 2. Mean rank of period based on R LT compared with mean rank of period based on R HL (i.e. R LT|R HL).
* Difference in mean ranks significantly different from 0 (P=0.05, Student's t-test).
Capture efficiency index
The CEI varied widely for each mosquito species (table 3). The range was greatest (3.5–108%) for Cx. quinquefasciatus and least (0.2–12.4%) for Ae. triseriatus. Period effects were significant only for the latter species (F 7,32=2.67, P=0.027). The CEI was highest on average in period 2, lowest in period 7, and higher between sunset and sunrise (periods 1–2–3–4–5) than during the day (6–7–8).
Table 3. Mean capture efficiency indices (CEI) for LT collection method for five mosquito species.
Tabulated means based on log10 (nf i+1).
Daily capture rate
The fitted daily capture rate model for An. quadrimaculatus was significant, but factors other than collection method influenced variability of R D for all other species (table 4). The results indicate that changes in the daily capture rate according to LT are not well correlated with changes indicated by HL.
Table 4. Mean daily capture rates for five mosquito species by LT (R D(LT)) and HL (R D(HL)) collection methods.
* month/year of observation. Tabulated means based on log10 (nf i+1).
Relationship of capture rates by LT to HL
None of the single or multi-period models evaluated for Ae. albopictus was significant at the 5% level (table 5). For An. quadrimaculatus, there was a significant curvilinear relationship between R LT and R HL in periods 1–5, but neither the model for periods 5–1 (>85% of R HL responses) nor for other sampling intervals for this species was significant. Four sampling interval models were fitted (P≤0.05) for Cx. nigripalpus (table 4), including a single period model for sunset (period 1). All multiple period models for this species included period 1 and period 2 responses, with the most robust model encompassing the time between sunset and sunrise. For Cx. quinquefasciatus, neither the period 1 nor the period 5 model was significant; conversely, fitted multiple period models included the daytime through sunset periods (table 5). For Ae. triseriatus, fitted multiple period models (table 5) included daytime through sunset (periods 6–7–8–1) and sunrise through sunset (periods 5–6–7–8–1).
Table 5. Fitted linear model coefficients (±SE) for R LT=R HL.
Discussion
Daily capture rate
The daily capture rate (R D) is a commonly used operational index of mosquito density. Our results suggest this index may lack meaning with respect to seasonal trends in mosquito landing rates when determined on the basis of responses to LT. The disparity can also potentially impact daily and seasonal estimates of the minimum infection rate, use of the maximum likelihood estimation procedure and the determination of mosquito population size in calculations of vectorial capacity (Garrett-Jones, Reference Garrett-Jones1964; Dye, Reference Dye1986; Gu et al., Reference Gu, Lampman and Novak2003) because each of these computational methods depends on estimates of mosquito density acquired using relative sampling methods, most often light traps. Sample representativeness is a concern in such cases, particularly when group-based pathogen assay methods are used to quantify infection rates in captured mosquito populations (Katholi & Unnasch, Reference Katholi and Unnasch2006).
Patterns of diel activity and the relationship of capture rates by R LT to R HL
Patterns of diel activity in mosquito populations are used to target the application of insecticides in time and space, measure repellency in field tests and to determine the risk of infection with mosquito-borne pathogens. The baseline HL responses observed here indicate a single peak of diel activity for all mosquito species except Ae. triseriatus. In contrast, responses to LT at sunrise by An. quadrimaculatus and Cx. nigripalpus, which compare poorly with HL responses at the same time, are likely the result of stimuli other than human host presence. Similarly, patterns of diel activity indicated by each collection method, when compared by the rank-order of periods, lacked congruency for all species except Cx. quinquefasciatus. For example, the highest ranked R HL-based periods indicate maximum activity before sunset for Ae. albopictus, whereas LT responses indicate peak activity after sunset. For An. quadrimaculatus and Cx. nigripalpus, discordance in the patterns of activity indicated by LT and HL was observed for the midday and sunrise periods. Taken in sum, these observations suggest that the patterns of diel activity indicated by LT collection do not accurately reflect temporal modes of mosquito landing on human hosts and that, in the field, such activity should be verified by observation of HL responses.
Trap efficiency
In a strict sense, trap efficiency indicates the number of mosquitoes available for capture that are actually captured. Under the proper conditions (i.e. mosquito availability to capture is constant; the rate of mosquito capture per unit of time is constant), trap efficiency can be used to estimate absolute population density (Southwood & Henderson, Reference Southwood and Henderson2000). Neither of the foregoing conditions was met in our study nor is either likely to be observed in nature. This fact notwithstanding, the crucial measure of mosquito availability for capture in a vector surveillance systems (using LT or any other device), and the response most relevant to disease agent transmission, is the number of female mosquitoes that land on a human/animal host per unit of time. LT efficiency determined on this basis was generally low in each period (e.g. 80% of CEI<0.17), regardless of mosquito species. An exception to this pattern was for Cx. quinquefasciatus (average CEI=0.43) in period 2, when capture rates using LT were 8% higher (more efficient) than those for HL.
Trap efficiency may also be considered in a relative sense as the ratio to one another of the numbers of each mosquito species captured using (in this case) LT compared with the same ratios as determined by HL. When we ranked the ratios of mean R D(LT) to mean R D(HL) observed for each species in this manner, the order of ranks was: Cx. quinquefasciatus (0.125:1)>An. quadrimaculatus (0.043:1)>Cx. nigripalpus (0.029:1)>Ae. albopictus (0.015:1)>Ae. triseriatus (0.011:1). Thus, in a hypothetical LT collection comprising these five species, Ae. triseriatus would be under-represented by 11% compared with Cx. quinquefasciatus, and Ae. albopictus would be under-represented by 3% compared with An. quadrimaculatus. And while these rankings clearly depend on the LT configuration and mosquito strains used in the present study, the relative efficiency of other trap designs has been compared in a similar manner in other studies (Kline et al., Reference Kline, Patnaude and Barnard2006, Reference Kline, Allan, Bernier and Welch2007). Our findings suggest that the merits of any trapping technology being considered for use in a vector surveillance system should be ascertained via the comparison of mosquito capture rates using that technology with the concomitant rate of mosquito landing on human/animal subjects before such traps are deployed in the field.
The ideal vector surveillance system would enable early detection of mosquito vectors and the timely/accurate prediction of disease agent transmission. The effectiveness of any such system will depend on the estimation of critical population parameters in an unbiased manner (Morris, Reference Morris1960; Bidlingmayer, Reference Bidlingmayer1985; Dye, Reference Dye1986). This requires extraction of a sample of the habitat and enumeration of the target organisms contained in it (Southwood & Henderson, Reference Southwood and Henderson2000) – an impractical approach for vector surveillance, given the mobility and constantly changing patterns of dispersion of adult mosquito populations. Nor are conventional mosquito traps, including recently developed mechanical and semiochemical-augmented trapping technologies (Kline, Reference Kline2007), designed to acquire unbiased estimates of mosquito density. In the case of light traps, for example, their range of attractiveness to mosquitoes is unknown, as is the volume of habitat they sample. The interpretation of data obtained using light traps and other relative sampling methods (Southwood & Henderson, Reference Southwood and Henderson2000) is thus limited by the confounding effects of trap location, by change in the density and behavior of mosquito populations in space and time, and by variations in trap efficiency caused by change in local weather conditions and/or other environmental factors (Bidlingmayer, Reference Bidlingmayer1985). A significant consequence of these sampling deficiencies is insensitivity of the vector surveillance system to arbovirus infection rates in the mosquito population (Reisen & Pfuntner, Reference Reisen and Pfuntner1987).
An objective of this study was to identify strategies for LT operation that would provide field-testable hypotheses relative to the accurate identification of mosquito landing rates on a human host. This was not possible for Ae. albopictus, although more recently devised trap configurations for other aedine species (Kröckel et al., Reference Kröckel, Rose, Eiras and Geier2006; Chambers et al., Reference Chambers, McClintock, Avery, King, Bradley, Schmaedick, Lammie and Burkot2009) may enable such comparisons in the future. For An. quadrimaculatus, a single plan of LT operation comprised trap operation from sunset through sunrise. Multiple schemes for LT operation were identified for Cx. nigripalpus, Cx. quinquefasciatus and Ae. triseriatus. For Cx. nigripalpus, the most precise index of the mosquito landing rate is from LT data collected between sunset and sunrise, even though LT data obtained for this species before, during and after sunset provide similar (albeit less precise) information. In this same context, optimal LT operation times for Cx. quinquefasciatus and Ae. triseriatus are sunrise through sunset. It is important to note that these strategies for LT operation are based on a correlation between mosquito responses to LT and HL for specific times of the diel period. These times may not be the same as those for maximum and/or minimum mosquito flight activity.
Evaluation of the results of this study under field conditions is an important next research step, particularly in cases where the objective of the monitoring/surveillance program is to understand, depict and/or accurately forecast the rate of human contact with pest and/or vector mosquito species. This may not be possible for Ae. albopictus, where LT responses do not accurately represent the timing or intensity of mosquito contact with humans. Actual measurement of the mosquito landing rate, in such cases, may be required. For other species, the results of our study suggest it is feasible to identify specific LT operating times and to interpret the resulting capture data in terms of the frequency of mosquito-human host contact. Under the conditions of this study, for example, LT operation between sunset and sunrise provided a reliable index of the mosquito landing rate on humans by Cx. nigripalpus and An. quadrimaculatus, whereas LT operation between sunrise and sunset or continuously for 24 h did not. Similarly, for Cx. quinquefasciatus and Ae. triseriatus, the times of LT operation for this purpose are best restricted to between sunrise and sunset.
Finally, in some operational venues, an index of mosquito activity such as provided by LT is considered as, or more, useful than obtained by other methods (including HL). This may be the case for pest mosquito species known to present little or no danger of disease agent transmission to humans or livestock. An important requirement, in such situations, is to obtain mosquito samples under the same conditions, keeping in mind that each trap location is unique and that microclimate, illumination levels and other local conditions profoundly influence mosquito flight (Bidlingmayer, Reference Bidlingmayer1985). Furthermore, in such cases, it may be prudent to develop a sampling plan that targets individual mosquito species, rather than the composite population of airborne mosquito species. This can be done using knowledge of the natural history of the target species and care in the selection of the habitat(s) in which traps are deployed and from which samples of the adult mosquito population are obtained.
