Regional and seasonal response of a West Nile virusvector to climate changeCory W. Morin1 and Andrew C. Comrie

School of Geography and Development, University of Arizona, Tucson, AZ 85721

Edited by Burton H. Singer, University of Florida, Gainesville, FL, and approved August 8, 2013 (received for review April 16, 2013)

Climate change will affect the abundance and seasonality of WestNile virus (WNV) vectors, altering the risk of virus transmission tohumans. Using downscaled general circulation model output, wecalculate a WNV vector’s response to climate change across thesouthern United States using process-based modeling. In the east-ern United States, Culex quinquefasciatus response to projectedclimate change displays a latitudinal and elevational gradient. Pro-jected summer population depressions as a result of increasedimmature mortality and habitat drying are most severe in thesouth and almost absent further north; extended spring and fallsurvival is ubiquitous. Much of California also exhibits a bimodalpattern. Projected onset of mosquito season is delayed in thesouthwestern United States because of extremely dry and hotspring and summers; however, increased temperature and latesummer and fall rains extend the mosquito season. These resultsare unique in being a broad-scale calculation of the projectedimpacts of climate change on a WNV vector. The results show that,despite projected widespread future warming, the future seasonalresponse of C. quinquefasciatus populations across the southernUnited States will not be homogeneous, and will depend on spe-cific combinations of local and regional conditions.

disease | insect | ecology

Projections of disease-related climate change impacts arecurrently limited and constitute a key research priority (1).

During its expansion across North America, West Nile virus(WNV) led to major epidemics during the summers of 2002–2004 (2) and is now endemic in most areas. Although host-birdspecies behavior and viral strain temperature tolerances arecritical components of WNV dynamics (2), vector ecology is alsoa key element of the virus ecology. Although climate and me-teorological factors can moderate mosquito vector populationdynamics (3, 4) and infection rates (5), the projected impactsof climate change on WNV vectors are not yet known. From ahuman health perspective, a better understanding of this com-plex system will facilitate the implementation of more effec-tive control measures and reduce virus transmission to humanpopulations (6).Studies relating climate to the incidence of other mosquito-

borne diseases have been performed with varying success formalaria (7–9) and dengue fever (10, 11). The most effective wayto study the complex feedbacks and impacts of climate change onvector populations is through dynamic modeling because itresolves some of the limitations associated with empiricallybased statistical techniques. These limitations include the lack oflong-term and continuous mosquito surveillance data and theinappropriate extrapolation of projections from temperature andprecipitation combinations outside the bounds of past measure-ments. Some of these issues, however, still manifest themselves indynamic modeling because mosquito response to climate variablesmay change in time and between places as a result of evolutionarypressures. Limited test data from mosquito trapping and inac-curacies in climate change projections are limitations in bothanalysis techniques.Few studies have quantified the connection between climate

and WNV vectoring mosquito populations through process-

based modeling (12). Hopp and Foley (11) explored the world-wide distribution of dengue fever through the container-inhab-iting mosquito simulation model (CiMSIM) at a 1° grid cellresolution. CiMSIM is a dynamic life-table model driven withclimate input to predict Aedes aegypti populations (13). Anotherprocess-based model, created by Ahumada et al. (14), simulatesCulex quinquefasciatusmosquito populations in Hawaii or similarenvironments. Other region-based climate-driven models havebeen created to simulate vector dynamics by Gong et al. (15) forWNV-transmitting mosquito species in the Northeast UnitedStates and by Schaeffer et al. (16) for yellow fever in theIvory Coast.Recently, Morin and Comrie (17) created the Dynamic Mos-

quito Simulation Model (DyMSiM) to examine C. quinquefasciatuspopulation dynamics across a wide range of environments and atdifferent spatial scales. The model captured much of the sea-sonal and subseasonal vector dynamics in two contrasting envi-ronments: the humid climate of southern Florida and the aridclimate of southeastern California. Although the model did notalways replicate daily dynamics well, our focus in this study islarger scale (weekly–monthly) where the model performed better.Because it feeds on avian and human hosts and it inhabits

urban environments, this species is suspected to be a primaryvector of WNV in many southern US states (18, 19). In thisstudy, we use DyMSiM to discern the effects of climate changeon C. quinquefasciatus populations across the southern UnitedStates, using a general circulation model (GCM) output run onguidelines established by the Intergovernmental Panel on Cli-mate Change (IPCC) Fourth Assessment Report (AR4).Because of a scientific consensus that climate change is occur-

ring with associated human health consequences, public healthresearch has focused on identifying and implementing effective

Significance

The potential impacts of climate change on human health arepossibly large and not yet well understood, especially forvector-borne diseases. This study provides projections of howclimate change may affect the population of a West Nile virusmosquito vector across the southern United States. Usinga climate-driven mosquito populationmodel, we simulate vectorabundance under base and future climate. Under future climate,many locations exhibit a lengthening of the mosquito seasonwith a decrease in summer populations. These impacts are notuniform geographically and vary with local temperature andprecipitation conditions. The results imply that disease-trans-mission studies and vector-control programs must be targetedlocally to maximize their effectiveness.

Author contributions: C.W.M. and A.C.C. designed research; C.W.M. performed research;C.W.M. and A.C.C. contributed new reagents/analytic tools; C.W.M. analyzed data; andC.W.M. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1To whom correspondence should be addressed. E-mail: cmorin@email.arizona.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1307135110/-/DCSupplemental.

15620–15625 | PNAS | September 24, 2013 | vol. 110 | no. 39 www.pnas.org/cgi/doi/10.1073/pnas.1307135110

Dow

nloa

ded

by g

uest

on

Oct

ober

21,

202

1

mitigation and adaptive strategies (20–22). Challenges for publichealth responses to climate change include the need for location-specific assessment of risk (20, 21) and the lack of quantitativeresearch (22). This study addresses both challenges by quantifyinghow projected climate change will affect the spatial and temporaldynamics of a WNV vector across a large spatial domain ata spatial resolution necessary for effective public health action.

Results and DiscussionOur results show that projected changes in temperature, pre-cipitation, and evaporation modify modeled water/habitat avail-ability, life-cycle development, and the number of days withmosquitoes, summarized as mosquito days per month (MDM)(Fig. 1). During the cooler seasons there are widespread increasesin MDM that follow latitudinal and elevational gradients. Thispattern occurs despite drying conditions in many locations, indi-cating that temperature is the controlling climate variable. An-thropogenic responses to drying conditions like lawn watering,however, could replenish habitats. During late spring and sum-mer, widespread decreases in MDM throughout much of thecountry, particularly Texas and the lower Midwest, suggest thatprecipitation limits mosquitoes. This finding contradicts theoften-held assumption that projected warmer conditions alwaysfavor mosquitoes and clarifies some of the uncertainty in com-plex feedbacks involving climate and climate change influenceson vectors and virus transmission, enabling more targeted publichealth action by using location-specific knowledge of vectorresponses to climate (23–25). Most notably, these results revealthat the relative importance of individual climate variables formosquito population development and maintenance are stronglydependent on time and location.The analysis illustrates the sensitivity of mosquito populations

to the complex effects of projected climate change at broad andlocal scales. For example, drying in high rainfall areas may havelittle or no effect on mosquito populations if precipitation is stillsufficient to create and maintain immature mosquito habitat. Sim-ilarly, increased precipitation in an arid region may have no effecton mosquito populations if it is insufficient to support container-breeding mosquito populations, as exemplified in southeastern

California during the summer. Several studies have even sug-gested that drought conditions can lead to increased vectorpopulations (26, 27) and vector infection rates (5) caused byincreased contact between vectors and avian hosts. Becauseclimate change is temporally and spatially heterogeneous, thesensitivity of mosquito populations to climate change is dependenton local climatic context. The seasonal timing of precipitation (28,29) and the intensity of events (30) are also important factorsmitigating WNV vector population development and infection.Consequently, the impacts on mosquito populations can varyconsiderably between locations within a region.To generalize beyond local variation, temporal patterns of

change across the study locations are regionalized for summarypurposes into five geographic zones based on similar changes intheir mosquito seasons (Fig. 2). Time series of climate variablesand mosquito populations averaged by region reveal the factorsresponsible for population change (Fig. 2). Whereas most loca-tions are clustered in their respective geographic zones, some aredistant. Although two locations have different climates, similarchanges in temperature and precipitation can produce a similarresponse in mosquito population dynamics. Additionally, Cal-ifornia’s complex topography produces spatially heterogeneousclimates, some of which may be similar (in a mosquito pop-ulation-related sense) to climate regimes further east. It is alsopossible for areas experiencing different changes in climate tohave similar changes in population dynamics. For example, in-creased spring rains in an arid environment and warmer springtemperatures in a moist environment could both create similarincreases in spring mosquito populations. All locations experiencean extended season, with mosquito populations rising earlier inthe spring and lasting later into the fall as a result of elevatedtemperatures under projected climate change. Although risingtemperatures could influence WNV ecology by extending theseason during which mosquitoes are active, it could also in-fluence vector competence. Warmer temperatures increase viralreplication, which can shorten the extrinsic incubation period,increase virus infection and dissemination within the mosquito,and result in increased transmission rates (2, 31–33). Thus,warmer fall conditions may benefit the ecology of WNV by

Fig. 1. Projected climate change (1970–1999 to2021–2050) from the AR4 GCM ensemble and result-ing changes in mosquito days per month (see text),averaged to 2-mo periods (JF, January–February;MA, March–April; MJ, May–June; JA, July–August;SO, September–October; ND, November–December).There is widespread warming throughout the studyregion, with the most intense warming occurringduring the summer in the Southwest and lowerMidwest. Precipitation changes are more variable inspace and in time. The greatest drying occurs duringspring and summer in the central United States.The drying extends to the Southwest during spring;however, precipitation is enhanced during the sum-mer and fall. Precipitation changes in the easternUnited States are relatively limited, with some dryingin Florida and the southeast and small increasesthroughout the rest of the region. Higher springand fall temperatures increase the number of dayswith mosquitoes across much of the United States,except in the Southwest during spring where severedrying inhibits population development. Decreasedprecipitation during summer and early fall decreasesthe number of days with mosquitoes across thecentral United States.

Morin and Comrie PNAS | September 24, 2013 | vol. 110 | no. 39 | 15621

ENVIRONMEN

TAL

SCIENCE

S

Dow

nloa

ded

by g

uest

on

Oct

ober

21,

202

1

allowing the virus to circulate within the avian and mosquitopopulations for a greater period.Changes in the start date (Fig. 3A) and end date (Fig. 3B) of

the mosquito season between baseline (1970–1999) and future(2021–2050) climate conditions also suggest a lengthening ofvector activity. Warmer temperatures during the cool season arelikely responsible for the earlier and larger spring and earlysummer populations. Studies have shown that climate during theprevious fall, winter, and spring are good predictors of current-year vector populations (3, 34). During an exceptionally warmCalifornia winter, vector population and reproductive activityincreased compared with previous years (35). As stated pre-viously, increased mosquito activity during cooler parts of the yearextends the time of virus circulation in the avian and mosquitopopulation. Warmer winter temperatures, for example, have beenassociated with WNV-related declines in crow populations (36).Increased temperatures and extended seasonality of mosquito

activity could also signify an expansion in the range of WNVvectors. Elevated summer temperatures in British Columbia,Canada, for example, have been associated with the expansion ofWNV in this area (37). Under projected climate change, mod-eled Culex pipiens habitat suitability has been shown to increasefurther north in Canada (38). Climate change could also facili-tate the introduction or reintroduction of new vectors and dis-eases, especially in conjunction with globalization as exemplifiedwith dengue fever (39). Although we discuss the relevance of

changing vector dynamics for WNV transmission, the complicatedecology of WNV prevents projections of changes in infection riskbased on vector populations alone.During the summer, mosquito populations under projected

future climate conditions often decline because of decreasedprecipitation and increased evaporation and immature mortalityat high temperatures (Fig. 4). This decline may not decreaseWNV transmission risk, however, as it has been suggested thatbirds and mosquitoes can be forced into greater contact bycongregating around scarce water sources serving as mosquitohabitat, thereby increasing transmission (28, 29). In addition,vector species that rely on more permanent water sources may beless affected by drying conditions and thereby gain a larger roleas disease vectors. Even within a local area microclimate, sea-sonality and surrounding vegetation can influence mosquitospecies composition (40).Although these results provide an important new understanding

of the potential effects of climate change on WNV vector ecology,they have some important limitations. DyMSiM has been eval-uated successfully across a range of environments, but not forall of the study locations because suitable validation data arenot available. Improved integration between model output andmosquito observations is required to advance validation and de-velop location-specific predictions of mosquito population dy-namics. The model also does not account for species interactionsor for the effects of human interventions, such as pesticide

Fig. 2. Regionally averaged time series of pro-jected climate change and mosquito populations(relative to the base maximum population) underbase and future climate scenarios (A–E), with a mapof locations (F). Increasing temperatures extend theduration of mosquito season in all regions. Extremeheat combined with projected decreases in pre-cipitation result in fewer mosquitoes during sum-mer. This trend is most severe in southern locationsand almost absent in the northern and easternlocations where more mild temperatures prevailand precipitation remains steady or increases. In thesouthwestern United States change in populationstrongly follows that of precipitation because thisregion is water-stressed and mosquitoes often mustrely on seasonal rains for habitat.

15622 | www.pnas.org/cgi/doi/10.1073/pnas.1307135110 Morin and Comrie

Dow

nloa

ded

by g

uest

on

Oct

ober

21,

202

1

use, water storage, or large-scale irrigation. Additionally, GCMprecipitation projections, especially for summer convective pre-cipitation, have much greater uncertainty than those for tem-perature, although climate models are improving in accuracy(41, 42). This study also focuses predominantly on large-scaleclimate circulation patterns simulated by GCMs, and thereforethese results represent projected changes over large regions andare not predictions of local mosquito populations, which are af-fected by a variety of local environmental variables. Differencesin land use and microclimates, for example, can cause differencesin dynamics even across small scales.In this study, precipitation and temperature are found to work

interdependently to influence future mosquito population dynamics.In the southcentral and southwestern United States, water scarcityis often the dominant limiting variable, causing mosquito pop-ulations to be sensitive to changes in precipitation and vulnerableto larval/pupal habitat drying at higher temperatures. The south-eastern United States generally receives sufficient precipitation tosupport mosquito populations throughout the year, making tem-perature the dominant variable influencing population dynamics.Overall, the seasonal response of C. quinquefasciatus population

dynamics to projected climate change across the southern UnitedStates varies geographically; however, some general trends emerge.Projected mosquito-season length increases by several weeks at thestart and end of the current season in many locations. Most areasare also projected to experience a decrease in mosquitoes duringthe summer. Changes are sometimes subtle and in contrast to in-tuitive assumptions. This finding is especially true in locations thatare water-stressed or experience strongly seasonal precipitation.These results suggest that climate change will modify seasonal

mosquito population levels across the United States, with possible

consequences for vector ecology and public health policy. Thesefindings also explain statistical associations between climatevariables, such as warmer temperatures and heavy precipitation,and WNV infection rates in human populations (26, 43). Loca-tion-explicit knowledge of the ecology of WNV is also critical forpredictive and preventative purposes (44). Public health inter-ventions must account for site-specific changes and character-istics to implement the most effective control strategies toprevent WNV transmission to humans (20, 21).These results also highlight the need for future research on

WNV ecology under climate change in the United States. Mostof the locations in this study also harbor other important WNVvectors whose response to climate change may differ from that ofC. quinquefasciatus. Furthermore, additional work is needed onthe avian response to climate change and how it may impact theecology of WNV from a human health perspective. Finally, theimplications of changing land cover and land use need to beexamined as they can make an area more or less resistant todrying and alter the magnitude of the mosquito population byallowing more water to collect. Urban infrastructure (27), agri-culture and water sources (45), and wetlands (46) have all beenreported to influence vector populations or WNV transmission.With a more complete understanding of WNV ecology and moreknowledge about the response of the individual components toclimate change, more effective measures can be used to limitviral transmission to humans.

Materials and MethodsObserved climate data for each location were obtained through the USHistorical Climatology Network (USHCN) (http://cdiac.ornl.gov/epubs/ndp/ushcn/ushcn.html) at the National Climatic Data Center. Site-selectionrequirements included locations within the habitat range of C. quinque-fasciatus (47), sufficient historic climate data at each site to train theweather generator, and spatial representation across the southern UnitedStates. We consider only C. quinquefasciatus within our study range, al-though genetic introgression with C. pipiens has been reported along thenorthern part of our study area (48, 49) and the possibility exists that thesehybrids may have different temperature tolerances and diapause behaviorthat could affect their population dynamics. Climate records for the years1970–1999 were selected if at least 98% complete with no more than 1 moof data completely missing. Eighty-four locations met these requirements(Table S1). Daily maximum temperature, minimum temperature, and totalprecipitation were obtained for the years 1970 through 1999 at each loca-tion. These data were used to train the statistical weather generator.

Fig. 3. Changes in start date (A) and end date (B) of the mosquito season(days earlier/later than base case). Most locations experience a 1- to 2-wk earlierstart and 1- to 2-wk later end to the mosquito season because of increases intemperature, with a few notable exceptions. Some southern-tier western loca-tions experience a later seasonal start date because of the dry conditions duringspring and early summer, which are exacerbated under climate change. South-ern Florida experiences little change because this area is capable of sustainingmosquito populations throughout the year. Drier summer and fall conditionstrigger a premature end to the mosquito season in northwestern Oklahoma.Extremely dry and hot summer conditions in inland California cause an earlierend to the mosquito season, whereas the coastal locations experience a laterdecline because of a second population spike during the fall, when the rainsreturn and temperatures are sufficient for mosquito development.

Fig. 4. Number of summer days (June 1 to September 30) with future mos-quito population less than base mosquito population. In most locationsa summer decrease in mosquito population is projected because of increasedimmature mortality at high temperatures and decreased larval and pupalhabitat because of enhanced evaporation. The central and Gulf states expe-rience the longest summer population dip because of the greater length andintensity of projected summer warming along with considerable drying.Locations further north and at higher elevations (especially along the Appa-lachians) experience a shallow dip because damaging high temperatures areless frequent and eastern locations receive greater precipitation under futureclimate. In the western United States, extremely dry summers limit sustainedmosquito populations under both base and future climate conditions.

Morin and Comrie PNAS | September 24, 2013 | vol. 110 | no. 39 | 15623

ENVIRONMEN

TAL

SCIENCE

S

Dow

nloa

ded

by g

uest

on

Oct

ober

21,

202

1

GCMdatawere acquired through the IPCC Data Distribution Centre (www.mad.zmaw.de/IPCC_DDC/html/ddc_gcmdata.html). We obtained monthlyaveraged temperature and total monthly precipitation for the years 1970–1999 and 2021–2050 under the A2 climate change scenario in the IPCC AR4.We collected these data for 16 of the models used in the AR4 (Table S2).

Weather generators provide a useful and practical method for down-scaling monthly gridded GCM output to daily point data. The Long AshtonResearch Station–Weather Generator 4 (LARS-WG) is a stochastic weathergenerator developed by M. A. Semenov (50, 51) at Rothamsted Research inthe United Kingdom (www.rothamsted.bbsrc.ac.uk/mas-models/larswg.php).LARS-WG can generate a daily time series of weather data of specifiedlength for a chosen site. A random seed allows the user to generate anumber of different time series that have the same statistical properties asthe original observed data but that represent different weather scenarioscharacteristic of the location. The weather generator requires input of his-toric climate information from the selected location to calculate site analysisstatistics. In this study we used the USHCN climate data (1970–1999) to trainthe model independently at each location.

To produce time series that incorporate the results of climate changescenarios, we calculated the GCM projected change in temperature andprecipitation at the monthly level at each location. For each month, changesin temperature (absolute) and precipitation (proportion) for each locationwere calculated as the average difference between the modeled base values(1970–1999) and the projected future values (2021–2050). These calculationswere performed for each GCM separately and then averaged to generatea GCM ensemble mean. By incorporating the GCM ensemble changes intothe site analysis statistics in LARS-WG, local weather time series with ad-justed statistics consistent with projected site-specific future climate con-ditions were generated. For each location we produced 50 base and 50future yearly weather scenarios representing the range of atmospheric vari-ability reported in the station data. This approach has been used widely forcalculating agricultural climate change impacts using crop models (52, 53).

We chose to use the GCM-derived 1970–1999 conditions as our base cli-mate scenario instead of the raw station data to keep the changes betweenpresent and future GCM runs comparable. Using the differences betweenthe raw station data for the present and GCM data for the future is notappropriate because they would include not only differences because ofmodeled climate change but also validation differences, a known issue withGCM precipitation data. Furthermore, the station data represent informationfor a particular site and the GCM data represent a grid box.

DyMSiM is a discrete, deterministic model (17) that uses Euler’s method ofintegration. The model requires the input of daily average temperature (C°),precipitation (cm), amount of impermeable land cover per area (cm2),amount of permanent water per area (cm2), and latitude. The modelincludes options for container depth, irrigation, and infiltration rates ofpermeable surface cover, but these options were not used in this study. Wefocus only on climate factors, which include temperature and precipitation,and thus land cover remains constant in time and space. Because containersand shallow surface puddles are the main breeding sources, impermeableland cover is set to 2,500 cm2 and permanent water is excluded.

Because some new metrics of measurements were used in this study thatwere not evaluated previously, we performed a validation analysis on modeland trap data for Coachella Valley, CA and Pasco County, FL (Tables S3 and S4).The metrics had to be altered slightly, however, to account for the smallerpopulation sizes and because the trap and model data behave differently. Thevalidation data should be interpreted with caution, as they are limited ina number of ways. For example, trapping ceases or greatly decreases duringthe shoulder seasons when we were attempting to calculate season start/endand the trap data are very location-specific and do not fully represent anentire area. Still, these data are the best source of information to evaluate themodel. In future work, improved validation analysis can be achieved by con-ducting mosquito trapping to enable location-specific predictions and furtherenhance model performance. Further improvement can be made by includingmore location-specific habitat sources, such as storm drains and pools, insteadof focusing predominantly on containers and puddles.

We ran DyMSiM with climate data generated from LARS-WG for the 84locations across the southern United States under each climate scenario. Inthe model, containers and puddles fill with water from precipitation and losewater through evaporation. Themodel output consists of two 50-y time series(base and future) of daily mosquito populations for each location. The first-year output was not included in the analysis because the model requires theinitial year as a spin-up period.

To decipher geographic trends in the data, we mapped the change inMDMbetween the two climate scenarios. This process was done to determineif the effects of future climate would be unique at each location or if there

were general trends associated with areas sharing a broad geographicalregion. Similar maps of temperature and precipitation change across the timeperiods were produced to assess their similarity with the MDM maps anddetermine the relative importance of each variable geographically. MDMwasdefined as the total number of days during each month that the dailymodeled mosquito population was above 10. The results for each monthrepresent the average number of mosquito days over the 49-y run at eachlocation. During the validation analysis, MDM was calculated as the numberof days during themonthwhen the average population across the countywasgreater than 0.5. This number was adjusted because the trap numbers weremuch lower because of their small effective range. There is general agree-ment between model and trap MDMwith the exception of January/Februaryand March/April in Florida during 1997 (Table S3). It is difficult to tell if this isthe result of an inaccurate model parameter or if there was some form ofhuman intervention, such as pesticide spraying or elimination of breedingsources. The accuracy of the model during the other tests suggests the latter.

Because of noise in model output at the daily level, we used meanweekly mosquito populations for our time-series analysis. An individualweek’s mosquito population at a location and time period is calculatedas the mean mosquito population during that week, averaged across thatweek for all 49 y, in each of the base and future model runs. Time series ofmean weekly temperature and total weekly precipitation were also pro-duced. We calculated the change in the weekly mosquito population be-tween the base and future climate conditions at each location as follows:

CMij = FMij −BMij , [1]

where CM is the change in week i’s mosquito population at site j, FM is themosquito population during week i at location j under the future climatescenario, and BM is the mosquito population during week i at location junder the base climate scenario.

To identify overall patterns of change in mosquito population dynamicswe performed an S-mode principal components (PCs) analysis on the pop-ulation change data series across all sites. A similar method of regionalizationhas been used on climate data (54). We used the correlation matrix in ourcalculation to enable comparison of locations with high and low mosquitopopulations. The first five PCs were retained (based on variance explainedand eigenvalues greater than 1) and each location was initially assigned toa PC based on its maximum loading. Once mapped, some locations werereassigned to the PC with the second- or third-highest loading to createcontiguous groupings on the map. There was generally little differencebetween the strengths of the loadings when this was done. Time series oftemperature, precipitation, and mosquito population were created by av-eraging all sites within each PC group.

The change in the season start date and season end date were calculatedandmapped to identify geographic trends. To calculate the change in season,we first averaged the daily mosquito population across the 49 y of the baseand future model runs to create two time series of data. Averaged dailypopulations were calculated rather than performing the analysis on eachindividual year, because it limited noise in the data and provided a clearersignal. For each dataset (base and future) we identified the dates at which thepopulation first reached a percentage of the base maximum population. Thisprocess was repeated using a range of values from 7% to 25% of the basemaximum population at 1% intervals for each time series. From this pro-cedure we used the median date as the start date, thereby eliminatingoutliers or biases that could result from choosing an arbitrary percentage.Still, these dates are subject to our chosen parameters and represent qual-itative estimates of season shift rather than hard predictions. The season enddate was calculated in a similar way by identifying the last day at which thepopulation was above each percentage of the base population maximum. Amore detailed description of this procedure and an example is provided in theSI Materials and Methods and Tables S5–S7. For each location we also cal-culated and mapped the number of summer days when the future mosquitopopulation was less than the base mosquito population (by at least twomosquitoes), with summer defined as June 1 to September 30. In the vali-dation analysis the start/end dates are defined as the day when the mos-quito population has increased for 2 consecutive days. It was not possible toperform this analysis with the same algorithm because of the small amountof trap data and because trapping was limited or ceased during the coolertimes of the year when the start/end would normally occur. Once again thetest on the 1997 Florida trap data shows considerable disagreement with themodel; however, model and trap start/end dates for most other years andlocations are within a week or a few days of each other (Table S4).

15624 | www.pnas.org/cgi/doi/10.1073/pnas.1307135110 Morin and Comrie

Dow

nloa

ded

by g

uest

on

Oct

ober

21,

202

1

ACKNOWLEDGMENTS. This research was supported in part by the NationalOceanic and Atmospheric Administration Regional Integrated Sciences

and Assessments program via the Climate Assessment for the Southwestprogram at the University of Arizona.

1. Confalonieri U, et al. (2007) in Human Health in Climate Change 2007: Impacts, Ad-aptation and Vulnerability, eds Parry ML, Canziani OF, Palutikof JP, van der Linden PJ,Hanson CE (Cambridge Univ Press, Cambridge), pp 391–431.

2. ReisenWK, Fang Y, Martinez VM (2006) Effects of temperature on the transmission ofWest Nile virus by Culex tarsalis (Diptera: Culicidae). J Med Entomol 43(2):309–317.

3. Reisen WK, et al. (2008) Impact of climate variation on mosquito abundance in Cal-ifornia. J Vector Ecol 33(1):89–98.

4. Trawinski PR, Mackay DS (2008) Meteorologically conditioned time-series predictionsof West Nile virus vector mosquitoes. Vector Borne Zoonotic Dis 8(4):505–521.

5. Ruiz MO, et al. (2010) Local impact of temperature and precipitation on West Nilevirus infection in Culex species mosquitoes in northeast Illinois, USA. Parasit Vectors3(19).

6. Ladeau SL, Marra PP, Kilpatrick AM, Calder CA (2008) West Nile virus revisited: Con-sequences for North American ecology. Bioscience 58(10):937–946.

7. Rogers DJ, Randolph SE (2000) The global spread of malaria in a future, warmerworld. Science 289(5485):1763–1766.

8. Small J, Goetz SJ, Hay SI (2003) Climatic suitability for malaria transmission in Africa,1911–1995. Proc Natl Acad Sci USA 100(26):15341–15345.

9. Ebi KL, et al. (2005) Climate suitability for stable malaria transmission in Zimbabweunder different climate change scenarios. Clim Change 73(3):375–393.

10. Hales S, de Wet N, Maindonald J, Woodward A (2002) Potential effect of populationand climate changes on global distribution of dengue fever: An empirical model.Lancet 360(9336):830–834.

11. Hopp JH, Foley JA (2003) Worldwide fluctuations in dengue fever cases related toclimate variability. Clim Res 25(1):85–94.

12. Hartley DM, et al. (2012) Effects of temperature on emergence and seasonality ofWest Nile virus in California. Am J Trop Med Hyg 86(5):884–894.

13. Focks DA, Haile DG, Daniels E, Mount GA (1993) Dynamic life table model for Aedesaegypti (Diptera: Culicidae): Analysis of the literature and model development. J MedEntomol 30(6):1003–1017.

14. Ahumada JA, Lapointe D, Samuel MD (2004) Modeling the population dynamics ofCulex quinquefasciatus (Diptera: Culicidae), along an elevational gradient in Hawaii.J Med Entomol 41(6):1157–1170.

15. Gong H, DeGaetano AT, Harrington LC (2011) Climate-based models for West NileCulex mosquito vectors in the Northeastern US. Int J Biometeorol 55(3):435–446.

16. Schaeffer B, Mondet B, Touzeau S (2008) Using a climate-dependent model to predictmosquito abundance: Application to Aedes (Stegomyia) africanus and Aedes (Dicer-omyia) furcifer (Diptera: Culicidae). Infect Genet Evol 8(4):422–432.

17. Morin CW, Comrie AC (2010) Modeled response of the West Nile virus vector Culexquinquefasciatus to changing climate using the dynamic mosquito simulation model.Int J Biometeorol 54(5):517–529.

18. Hayes EB, et al. (2005) Epidemiology and transmission dynamics of West Nile virusdisease. Emerg Infect Dis 11(8):1167–1173.

19. Zinser M, Ramberg F, Willott E (2004) Culex quinquefasciatus (Diptera: Culicidae) asa potential West Nile virus vector in Tucson, Arizona: Blood meal analysis indicatesfeeding on both humans and birds. J Insect Sci 4(20).

20. Ebi KL (2009) Public health responses to the risks of climate variability and change inthe United States. J Occup Environ Med 51(1):4–12.

21. Frumkin H, Hess J, Luber G, Malilay J, McGeehin M (2008) Climate change: The publichealth response. Am J Public Health 98(3):435–445.

22. Hosking J, Campbell-Lendrum D (2012) How well does climate change and humanhealth research match the demands of policymakers? A scoping review. EnvironHealth Perspect 120(8):1076–1082.

23. Khasnis AA, Nettleman MD (2005) Global warming and infectious disease. Arch MedRes 36(6):689–696.

24. Weaver SC, Reisen WK (2010) Present and future arboviral threats. Antiviral Res 85(2):328–345.

25. McMichael AJ, Lindgren E (2011) Climate change: Present and future risks to health,and necessary responses. J Intern Med 270(5):401–413.

26. Landesman WJ, Allan BF, Langerhans RB, Knight TM, Chase JM (2007) Inter-annualassociations between precipitation and human incidence of West Nile virus in theUnited States. Vector Borne Zoonotic Dis 7(3):337–343.

27. Deichmeister JM, Telang A (2011) Abundance of West Nile virus mosquito vectors inrelation to climate and landscape variables. J Vector Ecol 36(1):75–85.

28. Shaman J, Day JF, Stieglitz M (2005) Drought-induced amplification and epidemictransmission of West Nile virus in southern Florida. J Med Entomol 42(2):134–141.

29. Shaman J, Harding K, Campbell SR (2011) Meteorological and hydrological influenceson the spatial and temporal prevalence of West Nile virus in Culex mosquitoes, Suf-folk County, New York. J Med Entomol 48(4):867–875.

30. Koenraadt CJM, Harrington LC (2008) Flushing effect of rain on container-inhabitingmosquitoes Aedes aegypti and Culex pipiens (Diptera: Culicidae). J Med Entomol45(1):28–35.

31. Dohm DJ, O’Guinn ML, Turell MJ (2002) Effect of environmental temperature on theability of Culex pipiens (Diptera: Culicidae) to transmit West Nile virus. J Med Entomol39(1):221–225.

32. Richards SL, Mores CN, Lord CC, Tabachnick WJ (2007) Impact of extrinsic incubationtemperature and virus exposure on vector competence of Culex pipiens quinque-fasciatus Say (Diptera: Culicidae) for West Nile virus. Vector Borne Zoonotic Dis 7(4):629–636.

33. Kilpatrick AM, Meola MA, Moudy RM, Kramer LD (2008) Temperature, viral genetics,and the transmission of West Nile virus by Culex pipiens mosquitoes. PLoS Pathog4(6):e1000092.

34. Walsh AS, Glass GE, Lesser CR, Curriero FC (2008) Predicting seasonal abundanceof mosquitoes based on off-season meteorological conditions. Environ Ecol Stat 15(3):279–291.

35. Reisen WK, et al. (2010) Effects of warm winter temperature on the abundance andgonotrophic activity of Culex (Diptera: Culicidae) in California. J Med Entomol 47(2):230–237.

36. LaDeau SL, Calder CA, Doran PJ, Marra PP (2011) West Nile virus impacts in Americancrow populations are associated with human land use and climate. Ecol Res 26(5):909–916.

37. Roth D, et al.; Members of the British Columbia West Nile Virus Surveillance Team(2010) West Nile virus range expansion into British Columbia. Emerg Infect Dis 16(8):1251–1258.

38. Hongoh V, Berang-Ford L, Scott ME, Lindsay LR (2012) Expanding geographicaldistribution of the mosquito, Culex pipiens, in Canada under climate change. ApplGeogr 33(1):53–62.

39. Brunkard JM, Cifuentes E, Rothenberg SJ (2008) Assessing the roles of temperature,precipitation, and ENSO in dengue re-emergence on the Texas-Mexico border region.Salud Publica Mex 50(3):227–234.

40. Bolling BG, Kennedy JH, Zimmerman EG (2005) Seasonal dynamics of four potentialWest Nile vector species in north-central Texas. J Vector Ecol 30(2):186–194.

41. Gates WL, et al. (1999) An overview of the results of the Atmospheric Model In-tercomparison Project (AMIP I). Bull Am Meteorol Soc 80(1):29–55.

42. Reichler T, Kim J (2008) How well do coupled models simulate today’s climate? BullAm Meteorol Soc 89(3):303–311.

43. Soverow JE, Wellenius GA, Fisman DN, Mittleman MA (2009) Infectious disease ina warming world: Howweather influencedWest Nile virus in the United States (2001–2005). Environ Health Perspect 117(7):1049–1052.

44. Tabachnick WJ (2010) Challenges in predicting climate and environmental effects onvector-borne disease episystems in a changing world. J Exp Biol 213(6):946–954.

45. Liu H, Weng Q, Gaines D (2008) Spatio-temporal analysis of the relationship betweenWNV dissemination and environmental variables in Indianapolis, USA. Int J HealthGeogr 7(66).

46. Johnson BJ, et al. (2012) The roles of mosquito and bird communities on the prev-alence of West Nile virus in urban wetland and residential habitats. Urban Ecosyst15(3):513–531.

47. Barr AR (1957) The distribution of Culex p. pipiens and C.P. quinquefasciatus in NorthAmerica. Am J Trop Med Hyg 6(1):153–165.

48. Cornel AJ, et al. (2003) Differences in extent of genetic introgression between sym-patric Culex pipiens and Culex quinquefasciatus (Diptera: Culicidae) in California andSouth Africa. J Med Entomol 40(1):36–51.

49. Huang S, Molaei G, Andreadis TG (2011) Reexamination of Culex pipiens hybridiza-tion zone in the Eastern United States by ribosomal DNA-based single nucleotidepolymorphism markers. Am J Trop Med Hyg 85(3):434–441.

50. Semenov MA, Barrow EM (1997) Use of a stochastic weather generator in the de-velopment of climate change scenarios. Clim Change 35(4):397–414.

51. Qian BD, Hayhoe H, Gameda S (2005) Evaluation of the stochastic weather generatorsLars-WG and AAFC-WG for climate change impact studies. Clim Res 29(1):3–21.

52. Barrow BM, Semenov MA (1995) Climate-change scenarios with high spatial andtemporal resolution for agricultural applications. Forestry 68(4):349–360.

53. Trnka M, Dubrovsky M, Semeradova D, Zalud Z (2004) Projections of uncertainties inclimate change scenarios into expected winter wheat yields. Theor Appl Climatol77(3–4):229–249.

54. Comrie AC, Glenn EC (1998) Principal components-based regionalization of pre-cipitation regimes across the southwest United States and northern Mexico, with anapplication to monsoon precipitation variability. Clim Res 10(3):201–215.

Morin and Comrie PNAS | September 24, 2013 | vol. 110 | no. 39 | 15625

ENVIRONMEN

TAL

SCIENCE

S

Dow

nloa

ded

by g

uest

on

Oct

ober

21,

202

1