Terrestrial vegetation and aquatic chemistry influence larval mosquito abundance in catch basins, Chicago, USA.

3.3 MB·English

by  Gardner, Allison M

Gardneretal.Parasites&Vectors2013,6:9
http://www.parasitesandvectors.com/content/6/1/9
RESEARCH Open Access
Terrestrial vegetation and aquatic chemistry
influence larval mosquito abundance in catch
basins, Chicago, USA
Allison M Gardner1*, Tavis K Anderson2, Gabriel L Hamer3, Dana E Johnson1, Kate E Varela1, Edward D Walker4
and Marilyn O Ruiz1
Abstract
Background: Animportantdeterminant ofmosquito-borne pathogentransmission is the spatial distribution of
vectors. The primary vectorsof WestNile virus (WNV) in Illinois are Culex pipiens Linnaeus (Diptera: Culicidae) and
Culex restuans Theobald. In urban environments,these mosquitoes commonly oviposit in roadside storm water
catch basins. However, use ofthis habitat is inconsistent, with abundance of larvae varying significantly across catch
basins ata fine spatial scale.
Methods: We testedthe hypothesis thatattributesofthebiotic and abiotic environment contribute to spatial and
temporal variation inproduction ofmosquitovectors, characterizing the relationship between terrestrial vegetation
and aquatic chemistry and Culex abundance inChicago,Illinois. Larvae were sampled from 60 catch basins from
June 14 to October 3, 2009. Density of shrubs and 14 tree genera surrounding thebasins were quantified, as well
as aquatic chemistry content of each basin.
Results: We demonstrate that thespatial pattern of Culex abundance in catch basins is strongly influenced by
environmental characteristics, resulting in significant variationacross theurban landscape. Using regression and
machine learningtechniques, wedescribed landscape features and microhabitat characteristics offour Chicago
neighborhoods and examined the implications ofthese measures for larval abundance inadjacent catch basins.
The important positive predictors of high larval abundance were aquatic ammonia,nitrates, and area ofshrubsof
height <1 msurrounding thecatch basins, whereas pH and area offlowering shrub were negatively correlated
withlarval abundance. Treedensity, particularly of arborvitae, maple, and pear, also positively influencedthe
distribution of Culex during thefruit-bearingperiods and early senescent periods inAugust and September.
Conclusions: This study identifiesenvironmental predictorsof mosquito production inurban environments.
Because anabundance of adult Culex is integral to efficient WNVtransmission and mosquitoes are found in
especially high densities near larval habitats, identifying aquatic sites for Culex and landscape features that promote
larval production are importantinpredicting the spatial pattern of cases of human and veterinary illness. Thus,
thesedata enable accurate assessment of regions atrisk for exposure to WNV and aid in thepreventionof
vector-borne disease transmission.
Keywords: Culex mosquitoes, Larval habitat, Landscape ecology, Vegetation, Aquatic chemistry, Geographic
information science, WestNile virus
*Correspondence:[email protected]
1DepartmentofPathobiology,UniversityofIllinoisatUrbana-Champaign,
2001SouthLincolnAvenue,Urbana,IL61802,USA
Fulllistofauthorinformationisavailableattheendofthearticle
©2013Gardneretal.;licenseeBioMedCentralLtd.ThisisanOpenAccessarticledistributedunderthetermsoftheCreative
CommonsAttributionLicense(http://creativecommons.org/licenses/by/2.0),whichpermitsunrestricteduse,distribution,and
reproductioninanymedium,providedtheoriginalworkisproperlycited.
Gardneretal.Parasites&Vectors2013,6:9 Page2of11
http://www.parasitesandvectors.com/content/6/1/9
Background adults as WNV vectors develop as the direct products of
Heterogeneous patterns of infection at various spatial environmental exposures in the larval stage [22]. Second,
scales are evident in many zoonotic disease systems, in- abatementeffortsarecenteredontheidentificationofim-
cluding mosquito-borne viruses, such as West Nile virus portant oviposition habitats and control of mosquitoes in
(WNV) [1,2], and those vectored by other arthropods the aquatic immaturestages. Therefore, it is of interest to
[3,4]. This variation in disease distribution is to be ex- publichealthdistrictsandothersresponsibleformosquito
pected, because biotic and abiotic habitat characteristics abatement to understand natural sources of variation in
differ between locations and these distinctions can either vector production, and apply this knowledge to disrupt
facilitate or inhibit the establishment and persistence of the ecological pathways that favor high vector abundance
pathogens [5]. Environmental features may influence viral andconcomitantelevatedpathogentransmission.
risk of human illness by altering the transmission compe- Prior research has revealed that weather is one of the
tenceandfitnessofvectors,theinfectionprevalenceinthe most important predictors of Culex larval abundance in
vertebrate reservoir host population, and the abundance the storm water catch basin system, explaining much of
and spatial distribution of all three organisms involved in the temporal variation in urban larval production. Large
the arboviral transmission cycle: hosts, parasites, and vec- (>3.50 cm) multi-hour rainfall events within four days
tors. In WNV, for example, high ambient temperature preceding collection have been shown to “flush” almost
increasesthemosquitodevelopmentrate[6]andpathogen all mosquito larvae from underground storm drains and
transmission capacity [7,8], decreases the viral incubation to limit adult production in catch basins for up to a
period [9], and alters amplification processes in the virus’ week in metropolitan Chicago [22]. High ambient and
avian reservoir hosts (in NorthAmerica, these include the aquatic temperatures have also been shown to accelerate
American robin,Turdus migratorius, and the house spar- larval production and development rates [6]. These
row,Passerdomesticus)[10].Humanmodificationofhabi- effects do not influence all catch basins equally: some
tat, such as introduction of pesticides, fertilizers, and catch basins are protected from heat and rainfall by
xenobiotics to the aquatic larval environment, has also overhanging trees, and catch basins surrounded by im-
been shown to influence the vector competence of adult pervious surfacesaremoresusceptible tostreetchemical
mosquitoes, affecting the development of mosquitoes and and fertilizer runoff due to high rainfall than those sur-
altering the outcome of inter- and intra-specific competi- rounded by grass. However, weather generally is a broad
tionamonglarvae[11,12]. scale exposure, and though it may contribute signifi-
An important driver of enzootic transmission in cantly to temporal variation in Culex production, it
mosquito-borne viruses is the abundance and distribu- likely is not a key determinant of the inconsistent use of
tion of the vector community. Culex pipiens Linnaeus catch basinhabitatswithin anarrow geographic range.
(Diptera:Culicidae)andCulexrestuansTheobald(here- Thepurposeofthisstudyistoexaminesourcesoffine-
after “Culex”) are the most common amplification vec- scale environmental heterogeneity in anaquatic container
tors for WNV in the east north central United States habitat that influence Culex larval production, and help
[13], and are also bridge vectors of WNV to humans, explain why two catch basins in close proximity may vary
equines, and other mammalian ancillary hosts [14]. dramatically in larval abundance. One landscape feature
These container breeders are generalists in aquatic that largely has been overlooked in relation to Culex
habitatuse,withlarvaeandpupaefoundinbothnatural abundance in field studies is terrestrial vegetation genus
and artificial habitats: gutters, metal and plastic con- and structure. Laboratory studies have demonstrated that
tainers, discarded tires, bird baths, rain barrels, vernal the substrates of different generas of tree leaves in the
ponds, and stagnant pools of water [15,16]. Roadside aquatic environment alter adult survival, development
storm water catch basins have also been identified as rate, and longevity, as well as outcomes of inter- and
important oviposition sites for Culex mosquitoes in intra-specific competition due to differential toxicity of
urbanlocales[17,18]. detritus genera to mosquito larvae and potential differ-
The abundance and infection prevalence of adult ences in the effects of plants on the aquatic chemistry of
Culexmosquitoeshavebeenlinkedtolandscapefeatures the larval habitat [23,24]. However, to the authors’ know-
including vegetation cover [19], demographic character- ledge no previous research has investigated the influence
istics andstructureofresidentialneighborhoods[20],and ofsurroundingvegetationonCulexproductionintheepi-
spatial and temporal weather patterns [21]. However, demiologically relevant man-made habitats provided by
fewerstudieshaveexploredtherelationshipbetweenland- stormwatercatchbasins.
scapecharacteristicsofurbanneighborhoodsanddistribu- Trees and shrubs may impact the spatial pattern and
tion of Culex larvae. These data are important for two abundance of Culex larvae at both broad and fine scales.
reasons. First, laboratory studies have found that many At the neighborhood level, dominant street tree genus
physiological traits that later drive the effectiveness of may have important implications for larval production;
Gardneretal.Parasites&Vectors2013,6:9 Page3of11
http://www.parasitesandvectors.com/content/6/1/9
for instance, areas with high oak densities may produce located on the edges of residential suburban streets. The
fewer mosquitoes because the tannins contained in their basins were sampled for larvae once per week from June
leaves are toxic to the aquatic stages [25]. On a local 14 to October 3, 2009 according to methods described
scale,trees and shrubs may determinethe distribution of in Hamer et al. (2011) [31]. Larvae and pupae were col-
mosquitoes due to their importance as sugar feeding lected using a 10.2 X 10.2 cm aquarium net attached to
sources for adult mosquitoes as well as resting sites for the end of a conduit pole, 3 m in length and 1.3 cm in
both adult mosquitoes and their avian blood meal hosts, diameter. The pole was inserted through the grate and
encouraging gravid females to oviposit in nearby catch passed over the water surface in a single figure eight.
basins and other proximate aquatic habitats. At the Thenet wastheninverted into acontainer and alllarvae
microhabitat level, organic detritus of vegetation may were collected, counted, and identified to species using
alter the aquatic chemistry in catch basins, potentially taxonomic keys[32].
influencing the attractiveness of these habitats to ovipo- Larval abundance of Culex pipiens and Culex restuans
siting females and adult emergence rates. Further, be- were aggregated for all statistical analysis. This is an ap-
cause larvae feed on microorganisms in the water propriate approach because larvae of both species often
column, the algal growth promoted by nitrogen con- share similar habitats [18], feed on similar hosts [33],
tained inleavesmaysupportCulexproduction[26,27]. and function similarly as enzootic disease vectors [34].
To test the hypothesis that attributes of the biotic and In our study region, Cx. restuans and Cx. pipiens occur
abiotic environment play a role in the variation observed at comparable abundances early in the summer and by
in the spatial distribution of important vector species, mid-summer,Cx.pipiens dominates (Figure2)[35].
we characterized the effect of vegetation and aquatic
chemistry on the abundance of WNV vectors in metro- Vegetation
politan Chicago, Illinois. We used regression and ma- To characterize the vegetation types and structures sur-
chine learning statistical techniques to identify vegetation rounding the catch basins, trees and shrubs in the four
featuresandaquaticchemistryparametersassociatedwith study neighborhoodsweremappedusingmethodsmodi-
high abundance of Culex larvae. We focused our analysis fied from the U.S. Forest Service Urban Forest Project
on density of 14 genera of common trees, area of shrubs [36]. On-site surveys were conducted with boundaries
and ornamental grasses surrounding catch basins, and surrounding catch basins by 25 m. Only trees greater
pH, nitrate, phosphate, and ammonia content of catch than 3 m in height were identified to genus. A small
basinwater. proportion of plants in this zone were unidentifiable to
the surveyor or were not visible from the street. They
Methods were classified as “unknown” or “obscured” respectively.
Larvalabundance Tree densities within 25 m of each basin were calculated
Sampling was conducted in four residential municipal- using kernel density estimation in ArcGIS 10 (ESRI Inc.,
ities in metropolitan Chicago, Illinois (Cook County; Redlands, CA) with a cell size of 10 m2, a search radius
87’44” W, 41’42” N). Evergreen Park is a suburban vil- of 100 m, and Silverman’s rule of thumb for bandwidth
lage with an area of 8.2 km2 and a population of 19,237 selection [37]. Shrub area was quantified within 25 m of
in 2009. Alsip has an area of 16.5 km2 and a population each basin. Shrubs were classified by foliage type (ever-
of 18,580. Oak Lawn (subdivided into north and south green, deciduous, flowering, or ornamental grass) and
regions) has an area of 22.3 km2 and a population of height (<1m,1–2m,or2–3m).
52,948. WNV became established in the region during
the summer of 2002 with 884 human cases state-wide Aquaticchemistry
[28], at the time the largest reported WN meningo- To quantify the aquatic chemistry of the catch basins,
encephalitis epidemic. We selected the study area for its water quality parameters were tested by calorimetric
notable concentration of virus-positive mosquitoes and analysis using Chemetrics CHEMets test kits (Chemtech
birds and cases of human illness over the past eight International Inc., Irvine, CA). We tested for ammonia
years [21,29], as well as its diverse composition of street (Kit K-1510, direct nesslerization method), phosphate
trees andshrubs inresidentialneighborhoods[30]. (K-8510, stannous chloride chemistry), and nitrate (K-
To estimate the mean abundance of mosquito larvae 6904, cadmium reduction method). Basin pH was mea-
per catch basin, we sampled 60 catch basins: 15 basins sured using the Chemetrics double junction pH meter
within each of four neighborhoods (named here as (Cat. No. I-1000). To quantify the broad seasonal trends
Evergreen Park, Alsip, Oak Lawn North, and Oak Lawn in water quality observed in other stream and stagnant
South) (Figure 1). The mean depth from street surface water ecosystems [38,39], these parameters were mea-
to the bottom of the basin was 68 ± 20.22 cm with a sured twice during the study period for all basins (weeks
diameter of 60 cm. All basins had open grates and were ofJune28andAugust16).Tocapturefine-scaletemporal
Gardneretal.Parasites&Vectors2013,6:9 Page4of11
http://www.parasitesandvectors.com/content/6/1/9
Figure1SixtylarvalsamplingcatchbasinsinfourmetropolitanChicagoneighborhoods(Alsip,EvergreenPark,OakLawnNorth,and
OakLawnSouth).Black-and-whitecrossesindicatelowlarvalabundance(average<11larvaepersample),greycrossesindicatemedium
abundance(average11–30larvaepersample),andblackcrossesindicatehighlarvalabundance(average>30larvaepersample).
variationinwaterquality,thesameparametersweremea- immediately. The sample was then stored in a cooler on
suredweeklyforeightbasinsspacedthroughoutthestudy icefornomorethan6hourspriortoprocessing.
area.Watersampleswereobtainedbyattachinganaquar-
iumnetframetotheendofaconduitpole,3minlength Statisticalanalysis
and1.3cmindiameter.Aplasticsealablebagwassecured The following environmental metrics were examined in
to the frame with foldback clips. The pole was inserted relation to Culex larval abundance: area of shrubs of
through the grate and submerged under water until the height <1 m, 1–2 m, and 2–3 m within 25 m of each
bag was at least half full. Aquatic pH was measured catch basin; area of shrubs of foliage types deciduous,
Gardneretal.Parasites&Vectors2013,6:9 Page5of11
http://www.parasitesandvectors.com/content/6/1/9
Figure2SeasonalabundanceofCulexpipiensandCx.restuansthroughoutthestudyperiodfromtheweekofJune14thtoSeptember
27th,2009.
evergreen, flowering, and ornamental grass within 25 m RTand RF belong to the Classification and Regression
of each basin; density of deciduous trees and evergreen Tree (CART) family of nonparametric decision tree mod-
trees within25m ofeachbasin;andpH,ammonia,phos- elsdescribedbyBreimanetal.(1984)[42].InRT,thevari-
phate, and nitrate content of each basin. To determine ation in the response variable (mean larval abundance per
which independent variables were correlated to larval catch basin) is recursively partitioned along binary nodes
abundance, we conducted forward stepwise regression in ofpredictivecovariates(densityofeachtreespecies),maxi-
SAS 9.2 (SAS Institute Inc., Cary, NC). The significance mizing the homogeneity within each partition. The result-
level α = 0.05 was required for entry into the model. The ing “tree” of nested covariates demonstrates the relative
dependent variable was average mosquito abundance for amountofvariationintheresponseexplainedbyeachpre-
each basin. To quantify the temporal structure of larval dictor.RFisabootstrappingmethodbasedoniterationsof
abundance, this analysis was conducted for early season RT, in which both predictors and responses are randomly
and late season time windows. The early season window permuted.Therobustnessoftheclassificationsdetermined
was based on larval abundance for June and July and the by RT is assessed based on the total decrease in node im-
aquatic chemistry measurements obtained on June 28th, purities from splitting the variable, measured by residual
2009 and the late season window was based on larval sumofsquares.Tree-basedmodelinghandlesmissingcov-
abundance for August and September and the aquatic ariates, may combine quantitative and qualitative covari-
chemistry measurements were obtained on August 16th. ates, and does not have the assumptions of generalized
Terrestrial vegetation was assumed to remain constant linear mixed models and neural networks, among other
throughoutbothtimegroups. alternatives for quantitative data [43]. These procedures
Toexaminewhichtreegeneraweremostcloselyrelated havebeenusedtoanalyzedatainmultiplemosquito-borne
to larval Culex abundance, we conducted regression tree diseasesystems[22,44].
(RT) and random forest (RF) nonlinear regression using
thepackagesrpart[40]andrandomForest[41]inR2.12.1 Results
(RFoundationforStatisticalComputing,Vienna,Austria). Larvalabundance
Themostcommonlyobservedlocaltreegenerawerecon- A total of 960 samples were collected from 60 catch
sidered in this procedure. Deciduous trees included ash basinsfromJune14toOctober3,2009.Themeannum-
(Fraxinus spp.), birch (Betula spp.), cottonwood (Populus ber of larvae per collection was 19 for the early season
spp.), crabapple (Malus spp.), elm (Ulmus spp.), locust and27forthelateseason,withthelargestlarvalsamples
(Gleditsia spp.), magnolia (Magnolia spp.), maple (Acer taken at site IDs OLN13, A5, A11, and A15 and the
spp.), oak (Quercus spp.), pear (Pyrus spp.), and plum smallest samples taken at site OLN5 (Figure 1). Culex
(Prunus spp.). Evergreen trees included arborvitae (Thuja restuans was the dominant species in the early season,
spp.), pine (Pinus spp.), and spruce (Picea spp.). All of comprising 65 percent of all larvae collected during that
these genera were observed more than 50 times within time window (n = 831). Culex pipiens was the only spe-
thestudyarea.Again,toexaminethevariableeffectoftree cies collected during the late season window (n = 10,770)
genera throughout the season, the analysis was repeated (Figure2).Therewerenonon-Culexmosquitoesrecorded
fortheearlyseasonandlateseasontimewindows. in catch basin samples throughout the period. Other
Gardneretal.Parasites&Vectors2013,6:9 Page6of11
http://www.parasitesandvectors.com/content/6/1/9
invertebratesincludingsomepotentialmosquitopredators with larval abundance (Figure 4A). These results were
such as copepods were collected in the basins, but these verified by the RF model, which suggested that spruce
occurredinlowabundanceandwerenottabulated. and arborvitae, in addition to maple, birch, and oak,
were most strongly associated with larval abundance
Vegetationandaquaticchemistry (Figure 4B). The late season RT model showed pear,
Forward stepwise linear regression fitted significant spruce, and elm tree density positively correlated to lar-
models describing variation in Culex larval abundance val abundance and ash and maple tree density negatively
among the 60 catch basins for the early (n = 57; R2 = correlated (Figure 4C). Again, these results were sup-
0.24; p < 0.001) and late season (n = 58; R2 = 0.31; p = ported by the RF model, which showed these genera and
0.003) time windows. The retained variable for the early pine and magnolia as the most important predictors of
season model was aquatic ammonia, indicating that am- larvalabundance(Figure 4D).
monia is positively correlated to larval abundance. The
retained variables for the late season model were aquatic Discussion
pH, ammonia, and nitrate, area of flowering shrubs and OurresultsdemonstratethatCulexlarvalabundance var-
shrubs<1mheight,anddensityofdeciduoustreeswithin iesspatiallyandtemporallyacrossurbanlandscapesandis
25 m of the basin. Of these, aquatic ammonia, aquatic strongly influenced by environmental characteristics, in-
nitrate, and short shrub area were positively correlated to cluding the surrounding vegetation structure and aquatic
larval abundance while aquatic pH and flowering shrub chemistry. Using regression and machine learning techni-
area were negatively correlated to larval abundance ques, we described the landscape features and fine-scale
(Table 1). Among the eight basins sampled weekly for microhabitat characteristics of four Chicago neighbor-
aquaticchemistry,therewasacomparabletemporalasso- hoods and examined the implications of these measures
ciation between larval abundance and pH, ammonia, and for larval abundance in adjacent catch basins. We deter-
nitratecontent(Figure3). mined that aquatic pH, ammonia, and nitrates, terrestrial
deciduous and flowering shrub area, and tree density are
Treegenera predictors of larval production in catch basins. However,
Regression tree (RT) and random forest (RF) nonlinear the relative importance of these effects varies temporally,
regression fitted models for larval abundance as a func- with aquatic chemistry influential in the early season and
tion of density of 14 tree genera within 25 m of each vegetation increasing in significance in the late season.
catch basin in the early season and late season time win- These data may be used to inform mosquito control
dows. The RT models outperformed the RF models for effortsbydemonstrating the consequences oflandscaping
both early season (R2=14.4)andlate season (R2=45.6). decisionsonlocalmosquitoproduction,andsubsequently
The RF model was a poor fit for early season data aid in the reduction of human risk of exposure to
(pseudo-R2 = −23.1), although the same model demon- mosquito-bornediseaseinresidentialneighborhoods.
strated substantially stronger predictive power for late Potential effects of vegetation on larval abundance gen-
seasondata (pseudo-R2=42.0). erally fall into two categories: First, trees and shrubs may
The early season RT model showed arborvitae, spruce, offer resting habitats and sugar sources to adult females
and elm tree density positively associated with Culex lar- and their avian blood meal hosts, encouraging the mos-
val abundance and ash tree density negatively associated quitoes to oviposit in nearby catch basins. Second, trees
Table1MultipleforwardstepwiseregressionmodelofCulexlarvaeinrelationtoterrestrialvegetationandaquatic
chemistrycharacteristicsforearlyseason(June-July)andlateseason(August-September)timewindows
Parameter Coefficient StandardError Std.Coefficient T P
Earlyseason(June-July)
Ammonia 5.51 1.33 0.49 4.16 <0.001*
Lateseason(August-September)
Ammonia 25.25 8.55 0.37 2.96 0.005*
pH −14.74 8.60 −0.21 −1.71 0.093†
Nitrate 16.43 8.61 0.26 1.91 0.062†
Floweringshrubs −1.40 0.70 −0.36 −2.01 0.050*
Shrubsheight<1m 0.67 0.46 0.29 1.45 0.152†
Deciduoustrees 1.69 1.41 0.14 1.19 0.238
*Indicatesastatisticallysignificantresult,α=0.05.
†Indicatesamarginallysignificantresult,α=0.20.
Gardneretal.Parasites&Vectors2013,6:9 Page7of11
http://www.parasitesandvectors.com/content/6/1/9
Figure3Larvalabundanceandaquaticchemistrycontent(pH,ammonia,andnitrate)foreightcatchbasinssampledweeklyfrom
June28thtoAugust16th,2009.
Gardneretal.Parasites&Vectors2013,6:9 Page8of11
http://www.parasitesandvectors.com/content/6/1/9
Figure4Regressiontreemodels(A,C)andvariableimportancescoresforrandomforestmodels(B,D)oftreegeneradensityas
predictorsoflarvalabundanceintheearlyseason(A,B)andlateseason(C,D)timewindows.
andshrubsmayprovidedirectdetritalinputstothe catch This relationship between vegetation proximate to
basins.Leaflitterandfallenfruitmayaffectmosquitopro- catch basins and larval production was observed in trees
ductionbyinfluencingaquaticchemistry,introducingbio- as well as shrubs. Although linear regression indicated
molecules to the larval habitat, and altering abundance that total tree density was not correlated to larval abun-
and composition of the microbial community within the dance, this finding did not reflect broadly posed unim-
catch basin. Collectively, these effects have the potential portance of trees in determining Culex production, but
to alter the attractiveness of the aquatic habitat to gravid ratherthesumofnon-uniformeffectsofdifferentgenera
females as well as the quality of the aquatic habitat for on larval abundance. Regression tree and random forest
developinglarvae. models showed that key tree generamay havesignificant
Because there is often a positive correlation between positive or negative effects on larval production. For ex-
abundance of adult mosquitoes and abundance of larvae ample, in the early season, arborvitae and spruce trees
in urban storm water systems [45,46], shrubs may have were positively associated with larval abundance, sug-
an important role in determining larval production in gesting that these shrubby, low-hanging trees serve a
catch basins by affecting the abundance of gravid females similar ecological function to shrubs in providing a rest-
nearby. We found a marginally significant positive associ- ing and feeding site for Culex females and their avian
ation between larval abundance and area of shrubs <1 m hosts. The increased importance of pear trees to larval
height within 25 m of each catch basin. In contrast, area abundance during the fruit-bearing period in the late
of flowering plants was negatively associated with larval season indicates that trees not only provide habitat to
abundance. A possible explanation for this pattern is that adult mosquitoes, but may also offer a possible sugar
male and female mosquitoes as well as potential avian source for females and males and an attractant to blood
blood meal hostsareattracted tonon-floweringshrubsas meal hosts [49]. Few other fruiting trees were observed
resting sites and sources of fruit, encouraging gravid in the studyarea in sufficiently high densities for consid-
femalestoovipositincatchbasinsandotheraquatichabi- eration in our models, and the effect of plums, mul-
tatsproximatetothesiteoftheirbloodmeals.Conversely, berries, and other fruit-bearing trees on mosquito diet
flowering plants containing pyrethroid compounds (e.g., andhost presence remains opentofuturestudy.
chrysanthemum) and natural volatile oils (e.g., geranium, Vegetation also provides a direct source of detritus to
peppermint, and rosemary) may repel adult mosquitoes, nearbycatchbasins,withseveralpotentialimplicationsfor
and thusreduce the numberof adults present and poten- larvalabundance.First,vegetationhasbeendemonstrated
tialforovipositioninnearbycatchbasins[47,48]. to alter the aquatic chemistry of larval environments in
Gardneretal.Parasites&Vectors2013,6:9 Page9of11
http://www.parasitesandvectors.com/content/6/1/9
naturally occurring and artificial container habitats, such microbial flora in catch basin mesocosms. It is under-
as tree holes and used tires [50], and our findings in the stoodthatmicrobialabundanceisrelatedtoleaf biomass
catchbasinecosystemwereconsistentwiththeseprevious in aquatic habitats [53,54], and that microbes contribute
results. We found that larval abundance in catch basins tothelarvaldietthrough nutrient cycling andbyprovid-
was positively correlated with broad, seasonal trends in ing a direct food source to developing larvae [55,56].
aquatic nitrate and ammonia variation. The organic de- However, microbial communities likely vary substantially
tritus of plants introduced throughout the year, including among substrates of different leaf genera and thereby
pollen, flowers, fruit, seeds, and leaves, likely is a primary may alter mosquito production and fitness. One mech-
source of nitrogen in catch basins, explaining an increase anism that may be responsible for associations between
in nitrates in catch basins during the late season after a tree genera surrounding catch basins and mosquito pro-
criticalmassoftreeshaddroppedfruitandenteredsenes- duction is differences in abundance and composition of
cence [26,27]. Elm trees, positively correlated to larval microbial communities that grow on each leaf genus.
abundance, may have contributed substantially to this ef- These relationships between leaf detritus type, bacterial
fectinourstudyareabecausetheirleaveslackathickwax flora, and mosquito abundance and distribution should
coatingandthereforedecomposequicklyinwater. beexamined infutureinvestigations.
The impact of natural inputs on aquatic chemistry in
the urban environment also may be compounded by Conclusions
artificial inputs and xenobiotics, such as lawn fertilizers, An important predictor of human illness in multiple
road salt, pesticides, and herbicides. Stemflow has been mosquito-borne disease systems [57,58] is local abun-
shown to dilute the chemical concentration of treehole dance of vectors. Studies of mosquito ecology in catch
microhabitats [51], and a similar effect of groundwater basin and underground storm drain systems [45,46]
could explain additional variance in our models. Further, among other larval habitats [59] have demonstrated that
it is noteworthy that catch basins are subject to constant adult mosquitoes are often found aggregated near stand-
fluctuationsinpH,ammonia,andnitrates,asweobserved ing water breeding environments. Because an abundance
in the eight catch basins tested weekly for water quality of adult Culex is integral to efficient WNV transmission
concurrent with larval sampling. From our qualitative in- and mosquitoes are found in especially high densities
terpretation of that sample size, weekly larval abundance near oviposition locations, identifying breeding sites for
appears to track with these parameters. Our early season Culex and assessing the landscape features that promote
and late season models based on fewer aquatic chemistry larval production are important in predicting the spatial
measurementsaresignificantbecausetheycapturethein- pattern ofcasesof human disease.
fluence of broad scale, seasonal patterns in water quality The current study contributes to our knowledge of the
[38,39] on larval production, knowledge that is important environmental conditions that favor larval production in
totheaccuratepredictionofannualandinter-annualvari- the urban ecosystem, thus enabling more accurate as-
ationinmosquitoabundance.Furtherresearchcouldclar- sessment of populations at risk for human illness. Fur-
ify the function of fine-scale temporal aquatic chemistry ther, our results may be used to guide mosquito district
fluctuationsindeterminingvectorproduction. protocols for larval control. Although catch basins may
Second, vegetation introduces biomolecules to catch be treated with larvicides to eliminate larvae or inhibit
basin aquatic communities. These compounds, including larval development and thus reduce adult emergence
tannins, cellulose, and glucose, may have significant effects rates [60], many public health departments lack the ma-
on Culex production. While the catch basin sample size terial and human resources to sample catch basins con-
and water quality parameters taken in our current study tinuously throughout the summer and treatments lose
wereinsufficientforcorrelationanalysisexaminingtherela- their effect over time. Focusing chemical treatment on
tionship between vegetation characteristics and biomole- neighborhoods with high shrub and deciduous tree
cules,priorstudieshavedemonstratedthatthecompounds densities may help target limited supplies to locations
contained in the leaves of certain tree species may create especiallylikelytoproduce vectors.
environments inhospitable to Culex [25]. For example, the
Competinginterests
tannins contained in oak leaves are toxic to the aquatic
Theauthorsdeclarethattheyhavenocompetinginterests.
stages of mosquitoes [52], so that catch basins near these
trees may have low larval abundance. Laboratory study Authors’contributions
AGconceptualizedandcarriedoutthestatisticalanalysisandwaslead
could reveal whether detritus of ash and maple trees, such
authorinwritingandrevisionofthemanuscript.TAandGHcontributedto
aswind-dispersedsamaras,leaves,andbranches,havecom- thebackgroundliterature,helpedwithinterpretationandconclusions,and
parablephysiologicaleffectsondevelopingmosquitoes. critiquedthestatisticalanalysis.DJandKVcollectedfielddata,including
mosquitoabundance,terrestrialvegetation,andaquaticchemistrydata.EW
Finally, vegetation detritus may alter aquatic habitat
conceptualizedfieldsamplingmethodsandcontributedtointerpretation
quality and mosquito production via its effect on the andconclusions.MRdefinedanddirectedtheresearchquestionsand
Gardneretal.Parasites&Vectors2013,6:9 Page10of11
http://www.parasitesandvectors.com/content/6/1/9
methods,contributedtothestatisticalanalysis,andhelpedtowritethe 15. CalhounLM,AveryM,JonesL,GunartoK,KingR,etal:Combinedsewage
manuscript.Allauthorsreadandapprovedthefinalversionofthe overflows(CSO)aremajorurbanbreedingsitesforCulex
manuscript. quinquefasciatusinAtlanta,Georgia.AmJTropMedHyg2007,77:478–484.
16. YeeDA:Tiresashabitatsformosquitoes:Areviewofstudieswithinthe
Acknowledgements easternUnitedStates.JMedEntomol2008,45:581–593.
WethankZachAllison,WilliamBrown,KellyDeBaene,JoannaGanning, 17. GeeryPR,HolubRE:SeasonalabundanceandcontrolofCulexspp.in
DianeGodhe,PatrickKelly,andTimothyThompsonforassistancein catchbasinsinIllinois.JAmMosquitoContr1989,5:537–540.
collectinglarvalmosquitoesandmicrohabitatdataandRonaldWeigelfor 18. CransWJ:Aclassificationsystemformosquitolifecycles:Lifecycletypes
adviceonthestatisticalanalysis.WealsothanktheVillageofAlsip,the formosquitoesofthenortheasternUnitedStates.JVectorEcol2004,
VillageofEvergreenPark,theVillageofOakLawn,theCityofChicago,and 29:1–10.
privatelandownersinthesemunicipalitiesforallowingustoconductthis 19. BrownH,Diuk-WasserM,AndreadisT,FishD:Remotelysensedvegetation
research,aswellastheVillageofOakLawnforprovidingfieldlaboratory indicesidentifymosquitoclustersofWestNilevirusvectorsinanurban
facilitiesandsupport.Thismaterialisbaseduponworksupportedbythe landscapeinthenortheasternUnitedStates.Vector-BorneZoonot2008,
NationalScienceFoundation/NationalInstitutesofHealthEcologyof 8:197–206.
InfectiousDiseaseprogramunderAwardNo.0840403andthroughfunding 20. RuizMO,TedescoC,McTigheTJ,AustinC,KitronU:Environmentaland
fromtheUniversityofIllinoisAdaptiveInfrastructureInformationSystems socialdeterminantsofhumanriskduringaWestNilevirusoutbreakin
Initiative. thegreaterChicagoarea,2002.IntJHealthGeogr2004,3:8.
21. RuizMO,ChavesLF,HamerGL,SunT,BrownWM,etal:Localimpactof
Authordetails temperatureandprecipitationonWestNilevirusinfectioninCulex
1DepartmentofPathobiology,UniversityofIllinoisatUrbana-Champaign, speciesmosquitoesinnortheastIllinois,USA.ParasitVectors2010,3:19.
2001SouthLincolnAvenue,Urbana,IL61802,USA.2VirusandPrionResearch 22. GardnerAM,HamerGL,HinesAM,NewmanCM,WalkerED,etal:Weather
Unit,NationalAnimalDiseaseCenter,USDA-ARS,Ames,IA50010,USA. variabilityaffectsabundanceoflarvalCulex(Diptera:Culicidae)instorm
3DepartmentofEntomology,TexasA&MUniversity,TAMU2475,College watercatchbasins.JMedEntomol2012,49:270–276.
Station,TX77843,USA.4DepartmentofMicrobiologyandMolecular 23. WalkerED,MerrittRW,KaufmanMG,AyresMP,RiedelMH:Effectsofvariation
Genetics,MichiganStateUniversity,2215BiomedicalPhysicalSciences,East inqualityofleafdetritusongrowthoftheeasterntree-holemosquito,
Lansing,MI48824,USA. Aedestriseriatus(Diptera:Culicidae).CanJZoolog1997,75:706–718.
24. DavidJP,ReyD,PautouMP,MeyranJC:Differentialtoxicityofleaflitterto
Received:16September2012Accepted:1January2013 Dipteranlarvaeofmosquitodevelopmentalsites.JInvertebrPathol2000,
Published:11January2013 75(1):9–18.
25. ReyD,DavidJP,CuanyA,AmichotM,MeyranJC:Differentialsensitivityto
vegetabletanninsinplanktoniccrustaceafromalpinemosquito
References
breedingsites.PesticBiochemPhysiol2000,67:103–113.
1. KomarN:WestNilevirus:EpidemiologyandecologyinNorthAmerica.
AdvVirusRes2003,61:185–234. 26. MerrittRW,DaddRH,WalkerED:Feedingbehavior,naturalfood,and
nutritionalrelationshipsoflarvalmosquitoes.AnnuRevEntomol1992,
2. Diuk-WasserMA,BrownHE,AndreadisTG,FishD:Modelingthespatial
37:349–374.
distributionofmosquitovectorsforWestNilevirusinConnecticut,USA.
Vector-BorneZoonot2006,6:283–295. 27. ChavesLF,KeoghCL,NguyenAM,DeckerGM,Vazquez-ProkopecGM,etal:
3. BrownsteinJS,HolfordTR,FishD:Aclimate-basedmodelpredictsthe Combinedsewageoverflowacceleratesimmaturedevelopmentand
spatialdistributionoftheLymediseasevectorIxodesscapularisinthe increasesbodysizeintheurbanmosquitoCulexquinquefasciatus.JAppl
UnitedStates.EnvironHealthPers2003,111:1152–1157. Entomol2011,135:611–620.
4. Cruz-ReyesA,Pickering-LopezJM:ChagasdiseaseinMexico:Ananalysis 28. HuhnGD,AustinC,LangkopC,KellyK,LuchtR,etal:Theemergenceof
ofgeographicaldistributionduringthepast76years-Areview.M.I. WestNilevirusduringalargeoutbreakinIllinoisin2002.AmJTropMed
OswaldoCruz2006,101:345–354. Hyg2005,72:768–776.
5. KalluriS,GilruthP,RogersD,SzczurM:Surveillanceofarthropodvector- 29. BertolottiL,KitronUD,WalkerED,RuizMO,BrawnJD,etal:Fine-scale
borneinfectiousdiseasesusingremotesensingtechniques:Areview. geneticvariationandevolutionofWestNilevirusinatransmission“hot
PlosPath2007,3:1361–1371. spot”insuburbanChicago,USA.Virology2008,374:381–389.
6. RuedaLM,PatelKJ,AxtellRC,StinnerRE:Temperature-dependent 30. McPhersonEG,NowakD,HeislerG,GrimmondS,SouchC,GrantR,
developmentandsurvivalratesofCulexquinquefasciatusandAedes RowntreeR:Quantifyingurbanforeststructure,function,andvalue:the
aegypti(Diptera:Culicidae).JMedEntomol1990,27:892–8. ChicagoUrbanForestClimateProject.UrbanEcosys1997,1:49–61.
7. DohmDJ,O’GuinnML,TurellMJ:Effectofenvironmentaltemperatureon 31. HamerGL,KellyPH,FocksDA,GoldbergTL,WalkerED:Evaluationofa
theabilityofCulexpipiens(Diptera:Culicidae)totransmitWestNile novelemergencetraptostudyCulexmosquitoesinurbancatchbasins.
virus.JMedEntomol2002,39:221–225. JAmMosquitoContr2011,27:142–7.
8. ReisenWK,FangY,MartinezVM:Effectsoftemperatureonthe 32. AndreadisTG,ThomasMC,ShepardJJ:Identificationguidetothemosquitoes
transmissionofWestNilevirusbyCulextarsalis(Diptera:Culicidae). ofConnecticut.NewHaven,CT:TheConnecticutAgriculturalExperiment
JMedEntomol2006,43:309–317. Station;2005.
9. KilpatrickAM,MeolaMA,MoudyRM,KramerLD:Temperature,viral 33. AppersonCS,HassanHK,HarrisonBA,SavageHM,AspenSE,etal:Host
genetics,andthetransmissionofWestNilevirusbyCulexpipiens feedingpatternsofestablishedandpotentialmosquitovectorsofWest
mosquitoes.PlosPath2008,4:e1000092. NilevirusintheeasternUnitedStates.Vector-BorneZoonot2004,4:71–82.
10. RappoleJH,DerricksonSR,HubalekZ:MigratorybirdsandspreadofWest 34. EbelGD,RochlinI,LongackerJ,KramerLD:Culexrestuans(Diptera:
Nilevirusinthewesternhemisphere.EmergInfectDis2000,6:319–328. Culicidae)relativeabundanceandvectorcompetenceforWestNile
11. MuturiEJ,KimC,AltoBW,BerenbaumMR,SchulerMA:Larval virus.JMedEntomol2005,42:838–843.
environmentalstressaltersAedesaegypticompetenceforSindbisvirus. 35. HamerGL,KitronUD,BrawnJD,LossSR,RuizMO,GoldbergTL,WalkerED:
TropMedIntlHealth2011,16:955–964. Culexpipiens(Diptera:Culicidae):AbridgevectorofWestNilevirusto
12. MuturiEJ,CostanzoK,KesayarajuB,LampmanR,AltoBW:Interactionofa humans.JMedEntomol2008,45:125–128.
pesticideandlarvalcompetitiononlifehistorytraitsofCulexpipiens. 36. McPhersonEG,NowakD,HeislerG,GrimmondS,SouchC,etal:
ActaTrop2010,116:141–146. Quantifyingurbanforeststructure,function,andvalue:TheChicago
13. AndreadisTG,AndersonJF,VossbrinckCR,MainAJ:EpidemiologyofWest urbanforestclimateproject.UrbanEcosys1997,1:49–61.
NilevirusinConnecticut:Afive-yearanalysisofmosquitodata,1999– 37. SilvermanBW:Densityestimationforstatisticsanddataanalysis.NewYork:
2003.Vector-BorneZoonot2004,4:360–378. ChapmanandHall;1986.
14. KilpatrickAM,KramerLD,CampbellSR,AlleyneEO,DobsonAP,etal:West 38. JohnsonLB,RichardsC,HostGE,ArthurJW:Landscapeinfluenceson
Nilevirusriskassessmentandthebridgevectorparadigm.EmergInfect waterchemistryinmidwesternstreamecosystems.FreshwaterBiol1997,
Dis2005,11:425–429. 37:193–208.