Effective arbovirus transmission requires that the principal vertebrate hosts and vectors have frequent contact. Vegetation and other landscape features used by roosting or nesting birds at night dictate their exposure to nocturnally active host-seeking Culex tarsalis Coquillett and therefore to western equine encephalomyelitis and St. Louis encephalitis viruses. Precipitin tests on 645 Cx. tarsalis that were collected resting and host-seeking near the Salton Sea in Coachella Valley, CA, indicated that passeriform birds (64%) and rabbits (25%) were the most frequent bloodmeal hosts and that the percentage of females feeding on birds varied temporally as an inverse function of mosquito abundance. Blood meals were not taken from communally roosting water birds. The spatial distribution of host-seeking females then was investigated by deploying dry ice baited traps within seven sites representative of habitats found along the Salton Sea. Mosquito catch was greatest at traps within elevated vegetation such as Tamarisk, mesquite, cattails, and orchards and lowest at traps positioned at snags over water, sand bars, open fields, and within housing in a small rural community. These data indicate that host-seeking Cx. tarsalis females congregated at specific landscape features that were not necessarily associated with large concentrations of potential bloodmeal hosts.
Vector and vertebrate host distribution and behavior determines when, where, and how frequently and effectively arbovirus transmission occurs. In California, western equine encephalomyelitis (WEE) and St. Louis encephalitis (SLE) viruses cycle enzootically during summer among the primary mosquito vector, Culex tarsalis Coquillett, and a variety of bird species (Reeves 1990). Our research continues to investigate ecological factors associated with the persistence, maintenance and amplification of these viruses at wetlands and associated habitats along the north shore of the Salton Sea, sites involved repeatedly in early season virus activity (Reisen et al. 1992).
The frequency of vector contact with vertebrate host species may be measured indirectly by the prevalence of infection with vector-transmitted parasites, by identifying the source of blood within the gut of engorged vectors, and by determining the spatial distribution of host-seeking vectors in time and space. The highly saline Salton Sea and its environs in Riverside County, CA, present a unique series of rapidly transitioning habitats that are exploited by a diverse avian fauna (Reisen et al. 2000). The nocturnal roosting, nesting, and in a few cases feeding sites where birds are exposed to host-seeking Cx. tarsalis may be divided into several general landscape categories (Table 1). Species such as grebes float on the sea surface, whereas others such as cormorants (Phalacrocoracidae) and herons (Ciconiiformes) roost and nest communally in snags over water. Sand beach and occasional spits are used intermittently for roosting by flocks of shore birds (Charadriiformes), pelicans (Pelecaniformes), and gulls (Laridae). Wave action has created a marginal berm that has been colonized by narrow stands of Tamarisk (Tamarix) and pickleweed (Salicornia). Immediately upland is a narrow belt (<200 m wide) of marsh created mostly by freshwater seepage and agricultural runoff. Open marsh is colonized by intermittent stands of cattails (Typhus), whereas the upland margin is bordered by a mixture of salt grass (Dystichilis), pickleweed, and stunted shoreline tamarisk. The shoreline tamarisk frequently was used by migratory warblers and sparrows, whereas the freshwater marsh and cattails are colonized by redwinged blackbirds, black-crowned night herons, song sparrows, and marsh wrens. The marsh edge gives way to alkaline desert brush dominated by pickleweed and arrow weed (Pulchchea) interspersed with clumps of mesquite (Prosopis). Gambel’s quail, Abert’s towhee and white crowned sparrows roost at night within this habitat. Cattail stands and inundated margins of the marsh provide the principal developmental sites for immature Cx. tarsalis. Further upland, desert brush is replaced by irrigated agriculture dominated by low row crops such as peppers and melons, and taller citrus orchards and grapes. Mourning doves, mocking birds, and house finches commonly roost and nest in orchards.
The number of bird species and total sera tested for antibodies against WEE and SLE virus (Reisen et al. 2000) have been summarized by habitat category in Table 1. Indicator bird species were included to characterize each habitat. WEE and SLE antibody rates were highest in species that roost and nest at night in upland vegetation and were detected rarely in species communally roosting over water, on sand spits, or within low marsh and alkaline brush vegetation. Although large flocks of cormorants, waterfowl, and shore birds were observed over water or on the shore, it was impossible to enumerate birds in dense reed, tamarisk thicket, or orchard habitats at night and occurrence here was inferred from the literature or by observations at dawn.
We hypothesized that the host-seeking behavior of Cx. tarsalis in relation to landscape features contributed to the results of our avian serological survey. Although well investigated in the Central Valley of California (Tempelis et al. 1965, Tempelis and Washino 1967), the bloodmeal hosts used by Cx. tarsalis in the desert biome of southeastern California are poorly understood. Blood fed females collected by light traps at farmhouses in Imperial Valley fed on both birds and mammals, with passeriform birds, chickens, horses, and cattle the most common hosts (Gunstream et al. 1971). However, data were not available for females collected at wetland habitats, and too few were collected to describe seasonal patterns. Therefore, an initial objective of our research was to describe the blood-feeding patterns of Cx. tarsalis at wetland habitats along the Salton Sea. Because the phenology of Cx. tarsalis in the Coachella Valley differs markedly from that in the Central Valley (Reisen and Reeves 1990), data were analyzed over time to describe how feeding patterns change seasonally in relation to mosquito and host abundance. Previous studies in the Central Valley showed that the proportion of blood meals taken from avian hosts decreased during summer concurrently with an increase in the density of host-seeking females (Reeves 1971), perhaps due to differences in defensive behavior in response to mosquito biting rates (Nelson et al. 1976).
Preliminary analysis of blood meal identifications (Lothrop et al. 1997) supported our bird seroprevalence data (Reisen et al. 2000) and indicated that abundant communally roosting birds such as cormorants, herons, shore birds and waterfowl were not fed upon by Cx. tarsalis. We hypothesized that the spatial distribution of host-seeking females may provide one explanation for these results and that this distribution could be explored in relation to landscape features using dry ice-baited traps (Meyer et al. 1991). Cx. tarsalis females collected in CDC style traps (Sudia and Chamberlain 1962) that were operated without lights and baited with dry ice mostly were unfed, inseminated and at Christophers’ follicular developmental stages I-II, and therefore presumed to be host-seeking (Reisen et al. 1995a). Catch of Cx. tarsalis at dry ice-baited traps typically peaked during the first few hours after sunset (Reisen et al. 1997), indicating a nocturnal host-seeking periodicity. The second objective of our research was to determine if the catch of Cx. tarsalis in dry ice-baited traps varied significantly among habitats along the shore of the Salton Sea. A significant clumping of host-seeking females at specific landscape or vegetation features would delineate the spatial risk of virus infection for vertebrates using those habitats at night when infective Cx. tarsalis were host-seeking.
Materials and Methods
Blood Feeding Patterns.
During 1995 and 1996, blood engorged female Cx. tarsalis were collected resting during early morning one to three times per week in five walk-in red boxes (Meyer 1985) positioned within 200 m of the margin of the Salton Sea. Boxes were positioned along stands of Tamarisk, the tallest vegetation at these areas. Previous attempts to aspirate females from natural resting refugia during early morning were unproductive, even during mild winter temperatures (Reisen et al. 1995b). Mosquitoes were enumerated by species, sex, and female abdominal appearance (i.e., empty, blood fed, or gravid). Blood fed females were frozen individually in gelatin capsules at -70°C, until hosts were identified using a microprecipitin test (Tempelis 1975). Host-seeking female abundance was measured biweekly along the Salton Sea by using 26–63 dry ice-baited CDC traps (CO2 traps) (Reisen and Lothrop 1999).
The number of females caught in CO2 traps measured the abundance of females host-seeking in the habitat in which the trap was placed. Abundance at habitats representative of landscape features in the southern Coachella Valley was compared during seven separate experiments conducted during 1997–1998. The location of the study areas in relation to the Salton Sea is shown in Fig. 1. At each site, four to six CO2 traps were positioned within different habitats along each of three parallel transects >50 m apart and perpendicular to the Salton Sea. Traps were operated on three consecutive nights during each experiment (n = 36–54 trap-nights per experiment). Dark green plastic garbage cans (121 liters) were placed under or within vegetation and oriented to face northward to sample resting mosquitoes. Resting mosquitoes were removed each morning concurrent with CO2 trap collection. Sites, month-year of sampling, and habitats were as follows.
Site 1 (April 1997 and April, June, August, October, and 1998). (1) Sandbar used by a variety of shorebirds, gulls and pelicans, (2) shoreline Tamarisk, (3) low brush dominated by salt grass and pickleweed, (4) mesquite clumps, (5) citrus orchard edge, and (6) citrus orchard interior at row 4. Traps were operated at the same six habitats at bimonthly intervals during 1998 to determine if the patterns observed during April 1997 remained consistent as temperature and mosquito abundance changed.
Site 2 (June 1997). (1) Offshore snags used for roosting by 50–100 double-crested cormorants, (2) sand beach, (3) cattail marsh edge, and (4) cattail marsh interior.
Site 3 (July 1997). (1) Narrow sandbar, (2) cattail stands surrounded by water, (3) shoreline Tamarisk, (4) low brush consisting mostly of pickle weed, (5) citrus edge, and (6) citrus interior (third row).
Site 4 (September 1997). (1) Offshore snags, (2) sandbar and adjacent berm, (3) shoreline Tamarisk, (4) open field (grass and annual plants), (5) margin of upland Tamarisk, and (6) interior of upland Tamarisk thicket (10–40 m from margin).
Site 5 (October 1997). (1) Mature cauliflower, (2) citrus row four (outside tree canopy), (3) citrus row four (under tree canopy), (4) citrus row 10 (outside canopy), (5) citrus row 10 (under canopy), and (6) vineyard.
Site 6 (April 1998). (1) Offshore snags (nesting sites for several species of egrets and herons), (2) mudflat, (3) berm with pickle weed, (4) shoreline Tamarisk, and (5) upland Tamarisk.
Site 7 (October 1998). (1) Brush dominated by arrow weed, (2) melon field, (3) citrus edge, (4) vineyard edge, (5) oleander hedge at town/melon field interface, (6) two blocks into town, and (7) four blocks into town.
Mosquitoes were enumerated by species, sex, and female abdominal status. Counts per trap were transformed to ln (y + 1) and then compared by a two-way analysis of variance (ANOVA), with habitats and transects as main effects and consecutive nights as replicates (Hintze 1998). Preliminary analyses of data from site 1 during 1998 indicated that variation among transects was not significant, and consequently catch size was analyzed using a two-way ANOVA with habitats and months as main effects. Means were compared by Fisher least significant difference (LSD) test (Hintze 1998). Data presented were back-transformed or geometric means.
Blood Feeding Patterns.
Overall, 645 blood meals produced a positive precipitin reaction (Lothrop et al. 1997). Passeriform bird (64%) was the most frequently positive host group, followed by rabbit (Lepus or Sylvilagus) (25%), canine (dog or coyote) (3%), chicken (1%), dove (pigeon or dove) (0.2%), and cat (0.2%); a blood meal from a single female reacted to both passeriform and dove antisera. In addition, 3% and 3% of the blood meals reacted positively with general mammal or bird antisera, respectively, but failed to react with group specific reagents. These meals were either too well digested for definitive testing or came from unidentified hosts. Of the 645 reacting blood meals, 101 were from females collected in CO2 traps and were host-seeking to complete a partial blood meal. Statistically, fewer females collected in CO2 traps fed on a passeriform bird (57%) than did females collected resting in walk-in red boxes (65%, χ2 = 11.02, P < 0.05). These data indicate that females feeding on perching birds were slightly less likely to be interrupted and continue host-seeking than females feeding on other hosts including rabbits.
The phenology of host-seeking Cx. tarsalis females was decidedly bimodal, with a large peak in late winter-early spring and a second smaller peak in late summer-early fall (Fig. 2 A). Conversely, the average number of females collected per walk-in red box per day during each month peaked in September and October (Fig. 2B); however, the number of blood fed females per collection remained <3 per box per day during all months. Consistent with previous observations in the Central Valley (Reisen and Reeves 1990), the percentage of females feeding on avian hosts was lowest during months when host-seeking abundance was greatest (Fig. 2C). The mean number of blood engorged females per box also was lowest during spring, even though the abundance of resident and migratory birds at wetland habitats was greatest at this time.
The abundance of Cx. tarsalis host-seeking females per trap-night was greatest at elevated vegetation (Fig. 3). Trap counts were significantly greatest at mesquite clumps and orchard habitats at site 1, at the interface and within cattail stands at site 2, within and at the edge of the citrus orchard at site 3, and at shoreline and upland Tamarisk at sites 4 and 6. A marked exception was the fairly homogenous pattern of abundance during October at site 5, which consisted of upland row crops, relatively low profile grape vineyard, and citrus habitat (Fig. 3). Cx. tarsalis abundance was relatively low during this experiment, and there were no significant differences among habitats. Counts at traps placed on sand beach or spits (sites 1–4) or suspended from snags over water (sites 1, 2, 4, and 6) were significantly less than at traps placed in the remaining habitats, even though these habitats often were used as communal roosting sites by large numbers of water and shore birds.
An interesting pattern was observed at habitats within and around the small town of Mecca (Fig. 4). Here, few females were collected at traps placed within the town, even though humans, domestic animals, and peridomestic birds were abundant. Catch was greatest at the edge of a citrus orchard and at moderate height brush dominated by arrow weed.
Because our results during the October 1997 experiment at site 5 (Fig. 3) appeared different, we investigated the influence of season on catch at traps in habitats at site 1 during four bimonthly experiments during 1998. April: mean daily temperature at Mecca (California Irrigation Management Information System station #141) during the experimental period = 25.1°C, period of maximum Cx. tarsalis abundance (Fig. 2). June: 28.6°C, moderate abundance. August: 29.9°C, low abundance. October: 22.7°C, moderate abundance. In the ANOVA, both month (F = 451.2; df = 3, 187; P < 0.001) and habitat (F = 140.0; df = 5, 187; P < 0.001) effects were highly significant (Table 2). The catch of host-seeking females at different habitats generally remained consistent over months (Fig. 5), although the interaction between month and habitat was significant (F = 5.51; df = 15, 187; P < 0.001). This interaction effect accounted for only 3% of the variability in the ANOVA and was not considered biologically meaningful. The pattern of host-seeking abundance among habitats generally followed that described at site 1 during April 1997 (Fig. 3), with few mosquitoes host-seeking at the sand spit along the shore of the Salton Sea. Relatively more females were collected hunting over low brush during August and at mesquite clumps during June and October than during April (Fig. 5). In agreement with the patterns observed for host-seeking females, the proportion of females (including blood engorged females) resting in garbage cans also was greatest at citrus habitat, indicating that females blood fed in greatest numbers where they were collected host-seeking (Fig. 6).
Vertebrates associated with elevated upland vegetation at night were the most frequent bloodmeal hosts of Cx. tarsalis in our survey in the southern Coachella Valley. Similar to the Central Valley (Reisen and Reeves 1990), precipitin tests identified perching birds in the order Passeriformes and rabbits, probably cottontails (Sylvilagus auduboni), as frequent bloodmeal hosts. The percentage of resting females that fed on birds was lowest during spring when the abundance of females host-seeking was greatest. Although this result agreed with previous findings in California (Tempelis et al. 1965) and elsewhere (Reeves 1971, Washino and Tempelis 1983), these data were somewhat unexpected in Coachella Valley, where wetlands along the Salton Sea were ecologically most productive, mosquitoes most abundant, and migratory and resident bird populations most abundant and diverse during spring. However, most new virus infections in birds were not detected by seroconversion or by infection of hatching year birds until after midsummer (Reisen et al. 2000), concurrent with the increase in the proportion of mosquitoes that fed on birds.
The catch of host-seeking females at CO2 traps positioned at different habitats agreed well with the results of our survey of avian populations for antibodies against WEE and SLE viruses (Reisen et al. 2000). In this survey, birds roosting/nesting in Tamarisk and orchards (house finches, house sparrows, ground doves, mourning doves), mesquite (Gambel’s quail, Abert’s towhees), and cattail stands (redwinged blackbirds, marsh wrens, least bitterns, black-crowned night-herons) were infected most frequently. Summer and permanent resident bird species not infected, but commonly seen in our study areas, included a variety of shore, wading, and water birds that spend the night on sand bars, roosting in snags over water, or floating on the water surface. By virtue of their nocturnal habitat selection, these birds seemed to avoid Cx. tarsalis bites and therefore virus infection. Collectively, these data question the idea that large numbers of communally roosting birds are a critical factor for the amplification of encephalitis viruses (Komar et al. 1999) and indicate that factors such as landscape may dictate which birds frequently are contacted by infected vectors. The failure to incriminate herons as an important reservoir host for SLE virus was unexpected, given their reputed importance in the epidemiology of closely related Japanese encephalitis virus (Buescher et al. 1959).
Host-seeking Cx. tarsalis female abundance at traps set in habitats around and in the town of Mecca exhibited epidemiologically revealing patterns. In this experiment, females were abundant at orchards and brush outside of town, but rarely were collected at traps operated 2 and 4 blocks inside the edge of the town. The small rural community of Mecca (3.4 km2) has a human population of ≈2,000 residents, with dogs and house sparrows visibly abundant and available to provide blood meals for host-seeking females. The low abundance of host-seeking females within the town agreed well with the low seroprevalence of antibodies to WEE and SLE viruses found in this and other human populations in the southern Coachella Valley (Reisen et al. 1996). These data indicate that humans would be at greatest risk of being bitten by Cx. tarsalis when entering or working near elevated vegetation at night in rural areas (e.g., orchards), but would have less mosquito contact in open areas including agricultural fields and preferred fishing sites such as sand spits and on boats over water.
Patterns of mosquito abundance in different habitats also indicate where CO2 traps should be placed to enhance mosquito surveillance. Our data indicate that to collect large numbers of Cx. tarsalis females, traps should be placed in ecotones near elevated vegetation such as Tamarisk, orchards, mesquite clumps, or cattail beds. Traps at these sites collected significantly more mosquitoes than traps placed in other habitats, and this placement should be considered when designing sampling programs for this species (Reisen and Lothrop 1999).
We especially thank G. Gorski (Coachella Valley Mosquito and Vector Control District) for assistance with mosquito collection and processing. C. H. Tempelis and J. Lui (University of California, Berkeley) identified the mosquito blood meals by precipitin test. This research was funded, in part, by grants from the University of California Mosquito Research Program, the Coachella Valley Mosquito and Vector Control District, and the National Institute of Allergy and Infectious Diseases, NIH (Grant No. 1-R01-[AI]32939).