Animals with complex life cycles have traits related to oviposition and juvenile survival that can respond to environmental factors in similar or dissimilar ways. We examined the preference-performance hypothesis (PPH), which states that females lacking parental care select juvenile habitats that maximize fitness, for two ubiquitous mosquito species, Aedes albopictus and Culex quinquefasciatus. Specifically, we examined if environmental factors known to affect larval abundance patterns in the field played a role in the PPH for these species. We first identified important environmental factors from a field survey that predicted larvae across different spatial scales. We then performed two experiments, the first testing the independent responses of oviposition and larval survival to these environmental factors, followed by a combined experiment where initial oviposition decisions were allowed to affect larval life history measures. We used path analysis for this last experiment to determine important links among factors in explaining egg numbers, larval mass, development time, and survival. For separate trials, Aedes albopictus displayed congruence between oviposition and larval survival, however C. quinquefasciatus did not. For the combined experiment path analysis suggested neither species completely fit predictions of the PPH, with density dependent effects of initial egg number on juvenile performance in A. albopictus. For these species the consequences of female oviposition choices on larval performance do not appear to fit expectations of the PPH.  

Oviposition Experiments

The oviposition responses to environmental factors for each species were conducted at the University of Southern Mississippi Science Park (31º 21’12” N, 89º 21’35” W). Experiments occurred outside in large wooden screen boxes (hereafter, arrays). Each array was 2.4 m long x 1.2 m high and 1.2 m wide, constructed with pressure treated lumber, and covered on the sides with no-see-um mesh (Phifer Inc., Tuscaloosa, AL); the top was covered with clear plastic sheeting, had a concrete bottom, with a door at one end. Each array had a small segment of PVC attached to one side affixed with a stocking net sleeve to introduce adult mosquitoes. Inside each array (n = 4) we placed 8 automobile tires (rim size 16”, mean size from [38]) vertically along the perimeter. Each tire was provisioned with one or more factors depending on the experiment and washed with a 10% bleach solution and rinsed before experiments. Each array received four sets of artificial potted plants placed in cups of sand between pairs of tires, to provide additional adult resting sites. Each array also held 20 ml glass vials of 10% sucrose solution administered with a cotton wick as a food source for adult mosquitoes. Gravid females were introduced after all treatments, artificial plants, and sugar had been placed inside and the door sealed.

Pine detritus. Senescent pine needles were collected from the Lake Thoreau Environmental Center, located 13 km from the USM campus (31°19’37.63” N, 89°17’25.22” W) and dried at 50 °C for ≥ 48 hrs. Dry needles were weighed and placed into tires containing 2000 ml of water and 200 ml of strained inoculum water (collected and homogenized from several undisturbed tires and strained to removed invertebrates), and allowed to exist outside for 3 d prior to the start of experiments. The range of pine detritus reflected values collected from natural tires [38]: 0, 14, 26, 38, 50, 62, 74, and 86 g. Tires were lined with brown paper as an oviposition substrate. We released 20 gravid female A. albopictus into each array and after 48 hrs collected papers to count eggs. Four replicates were conducted, and detritus levels were haphazardly assigned to tires.           

Volume. This factor was important for A. albopictus and C. quinquefasciatus, and thus the design was the same for both species. We used eight volumes that spanned the range found in [38]: 0.1, 1.2, 2.4, 3.6, 4.8, 6.0, 7.2, and 8.4 L. The lowest volume corresponded to the lowest volume where larvae were collected [38]. Water added to tires consisted of a mixture of tap water and tire inoculum (prepared by taking a mixture of various detritus types and soaking in a 20 L bucket outside for several days). For each volume, we added 10 ml of inoculum for every 100 ml of volume (e.g., 1.2 L would receive 120 ml tire inoculum). Because Culex lay eggs in rafts on the water surface, experiments involving that species did not include eggs papers. All other elements of design were the same as for pine detritus.

Canopy cover. Culex quinquefasciatus trials involving variation in canopy cover used 2000 ml of water with 200 ml inoculum in each tire. Variation in cover was achieved using segments of black no-see-um mesh placed on the array above each tire to decrease light penetration. We first measured light intensity next to two arrays at several times during each of a cloudless and overcast day using a hand-held light meter (Extech Light Meter Model 401025, Nashua, NH). We used the mean of these two readings as the highest light level, then reduced the light by using more layers of black no-see-um mesh, with each layer reducing the level by ~50-60%. By using different layers of mesh, this allowed us to produce eight different levels of LUX (2400, 1200, 600, 300, 240, 180, 120, 60; higher values equate to more light). As above, all other elements of design were the same as for volume.

Statistical Analyses. As oviposition responses to treatment levels were not independent (i.e., egg placement could affect the placement of other eggs in a tire) we used a mixed model randomized block design for each species and treatment, separately using PROC MIXED in SAS. For the egg oviposition experiments we included array as a random factor in the analysis. To meet assumptions of normality and homoscedasticity, the number of eggs of A. albopictus were log_(x+1) transformed for volume. For C. quinquefasciatus, we used a 1/(x+1) transformation of egg rafts for volume, and log_(x+1) for canopy cover. Differences in egg or raft numbers among treatment levels were identified using a Tukey-Kramer HSD post-hoc test for multiple comparisons. All analyses were performed using SAS.

Larval Survival Experiments

To assess the effects of our field parameters on larval survival, we also conducted laboratory experiments for both species using the same factors as those for oviposition. In all cases, we used black rubber bowls made of recycled tires (Miller Manufacturing, Glencoe, MN) to rear larvae to mimic the environment of tires. Bowls were kept in environmental chambers set at 27° C on a 14:10 light/dark schedule. For each replicate within a trial 90 1^st instar Aedes albopictus were used whereas replicates for C. quinquefasciatus used 55 1^st instars; these numbers reflect the mean number of larvae per tires from field observations [38]. In all cases, we used pupation as a surrogate for survival. The experiment lasted until all larvae had pupated or died.

Volume. The treatments consisted of rubber bowls filled with different volumes of water as for oviposition experiments (0.1, 1.2, 2.4, 3.6, 4.8, 6.0, 7.2, and 8.4 L). Small bowls (2.5 L) were used for the two lowest volumes, whereas larger bowls (8.5 L) were used for the remaining volumes. Each bowl was filled with 95% by volume RO water and 5% by volume inoculum. After the water was placed in the bowls, 0.25g of freeze-dried animal (crickets, Gryllodes sigillatus, Fluker Farms, Port Allen, LA) detritus was placed in each bowl to provide food for developing larvae. Bowls were left in incubators for three days to encourage the growth of the microbial community before the addition of larvae.

Pine detritus. Senescent pine needles were obtained from Lake Thoreau and dried at 50 °C for 48 hrs before establishing treatment levels identical to those used in the oviposition trials (0, 14, 26, 38, 50, 62, 74, and 86 g). Detritus was placed in individual 2.5 L bowls with 2.0 L RO water and 0.2 L tire inoculum strained with a 150-micron sieve. Detritus sat for 48 hrs before larvae were added. Larvae A. albopictus densities were identical as in the volume levels.

Canopy Cover. Small bowls were filled with 2.0 L of water (1.95 L RO water, 0.05 L inoculate water). The highest light level within incubators registered at 1500 LUX and each subsequent layer of no-see-um mesh saw a decrease in light by ~50% (1500, 750, 375,180, 90, 45, 20, 10 LUX). Thus, this gradient of light levels was proportionally similar to but lower than those used in the oviposition trials. Detritus, soaking duration, larval C. quinquefasciatus, and other factors were identical as in the volume levels.

Statistical Analyses. We assessed differences among treatment levels for mosquito survival using 1-way ANOVA. To meet assumptions of normality and homoscedasticity, survival of A. albopictus was square root +1 transformed for volume. For C. quinquefasciatus, we used a log_(x+1) transformation for survival for volume and canopy cover. Differences in mosquito survival and oviposition among treatment levels were identified using the Tukey-Kramer HSD post-hoc test for multiple comparisons where applicable.

Linking oviposition to life history and survival

We performed a combined experiment to test the hypothesis that oviposition responses by females have consequences for life history and survival of progeny. Specifically, we repeated the oviposition experiment for each species but crossed each environmental factor within each array. For A. albopictus, we used three volumes (2.4, 4.8, 8.4 L) crossed with three levels of pine detritus (0, 14, 62 g). We excluded the 0 g pine detritus by 2.4 L volume, leaving us eight total combinations. For C. quinquefasciatus, we used three volumes (0.1, 2.4, 6.0 L) and canopy cover (10, 750, 1550 LUX). Here, we excluded the 0.1 L by 10 LUX combination. Values chosen for both species generally elicited the highest oviposition responses of females during the separate oviposition trials (see Results). Because we planned to quantify different life history traits, we used 40 females during these trials to generate as many eggs as possible. We used the same general methods as in the separate oviposition trials (e.g., artificial plants, oviposition water, sugar). After 48 hrs, we counted eggs and rafts as before, but unlike the oviposition trials, here we returned larvae to the arrays and followed their development. Specifically, we hatched eggs of both species and placed offspring in 8.5 L rubber bowls within each upright tire. This was done to facilitate the identification and removal of pupae but to maintain the same microenvironment as the tire. Aedes egg papers were kept out of water for one week to allow for embryonation, whereas C. quinquefasciatus eggs, which hatch within 24 hrs, were immediately placed into bowls after counting emerging larvae. In all cases, we used the number of hatched larvae added to bowls to reflect female reproductive output. For treatment combinations that lacked food for developing larvae (e.g., all C. quinquefasciatus treatments, no pine treatment levels for A. albopictus), we added 0.25 g of crickets to each bowl. We ran four replicates of each species with two replicates run for each species during each trial (two trials were run to achieve our four replicates per species). All bowls were monitored every other day for the presence of pupae, which when present were removed and isolated in 0.25-dram shell vials until they eclosed. Adults were then sexed, identified to species, and then dried at 50º C for > 48 hrs and weighed to the nearest 0.0001 mg using an XP2U ultra-microbalance (Mettler Toledo Inc., Columbus, Ohio). For each bowl, we determined the mean female weight (mg), female development time (from egg to adult), and percent survival (males were excluded from weight and development times given that they are not the main driver of population sizes in mosquitoes as they don’t lay eggs).

Oviposition patterns were linked to mosquito life history (female mass, development time) and survival using path analysis, which can test for the effects of multiple independent variables on multiple dependent variables. In path analysis, unlike multiple regression, a variable may be both dependent (i.e., affected by other variables) and independent (i.e., affecting other variables). Path coefficients, like standardized regression coefficients, quantify direct effects on a dependent variable caused by variation in an independent variable, independently of direct effects of other independent variables. The importance of a path was evaluated by testing the fit of reduced models in which paths are removed, relative to a full model. Reduced models were compared to the full model using goodness-of-fit χ² tests (PROC CALIS). We hypothesized three sets of paths based on the relationship between oviposition and larval survival in explaining life history parameters and survival:  a Combined Model (Fig. 3A), a Larval Model (Fig. 3B), and an Oviposition Model (Fig. 3C). The Combined model served as the full model, in which all variables were connected to all other variables. Moreover, if this model was supported in any way, it would suggest that both oviposition (i.e., number of hatched larvae) and larval survival and life history traits (female mass and development time) were both affected by habitat factors (i.e., ecological filters are consistent and important for both life stages). The Larval model contained direct effects of environmental factors on survival, female mass, and female development time. If this model was supported then it would suggest that habitat parameters largely worked to affect larval traits, but that egg laying was less important. The Oviposition model contained only the direct paths between environmental factors and the number of hatched larvae produced, and if supported, would suggest that filters only acted on females for each species. After analyzing the Combined model, paths were then removed in a systematic way, and if removing a link did not decrease the fit of the model, then it was assumed that the link was unimportant in explaining variation in the dependent variable. If this occurred, the path was eliminated and the next path was tested. Briefly, paths were removed between independent variables and eggs, then between independent variables and life history traits, and then between eggs and life history traits. We also ran the same model using λ’, a composite index that acts as a surrogate of population growth in place of survival, female mass, and development time. Unlike other studies that have considered only population growth we chose to focus on the individual life history variables for two reasons. First, the results were redundant in that we found similar results for λ’ to using each life history variable and survival, separately (not shown). And second, as traits could be affected independently by the environmental factors, using a surrogate that incorporated all three might mask important patterns. The final model for each species thus will represent all important paths included in the model to explain variation in oviposition and larval survival and could then be compared to our three hypothesized models to assess the importance of ecological filters in shaping patterns of these important insects.

One replicate was dropped for the Culex quniquefasciauts ovipostioin response to volume as no eggs were laid. No other data issues exist.