The use of early summer mosquito surveillance to predict late summer West Nile virus activity



Similar documents
Guide for public health units: Considerations for adult mosquito control

West Nile Virus Risk Assessment and the Bridge Vector Paradigm

Comparison of light traps, gravid traps, and resting boxes for West Nile virus surveillance

West Nile Virus Weekly Surveillance Report

Research Paper. West Nile Virus Antibody Prevalence in Red-Winged Blackbirds (Agelaius phoeniceus) from North Dakota, USA

VECTOR SURVEILLANCE IN NEW JERSEY EEE, WNV, SLE and LAC CDC WEEK 36: September 5 to September 11, 2010 Data Downloaded 9:36 am 14 Sep 2010

Algorithm for detecting Zika virus (ZIKV) 1

9.0 PUBLIC HEALTH (MOSQUITO ABATEMENT)

Saint Louis Encephalitis (SLE)

excerpted from Reducing Pandemic Risk, Promoting Global Health For the full report go to

Saint Louis Encephalitis (SLE)

Interactions between rodent borne diseases and climate, and the risks for public and animal health

Host Feeding Patterns of Culex Mosquitoes and West Nile Virus Transmission, Northeastern United States

[Table 1] [Figure 2] [Figure 3] Table 7] [Table 1]

West Nile Virus in the United States: Guidelines for Surveillance, Prevention, and Control

Zika Virus. History of Zika virus

The Use of Ovitraps Baited with Hay Infusion as a Surveillance Tool for Aedes aegypti Mosquitoes in Cambodia

Next Generation Sequencing in Public Health Laboratories Survey Results

If emergenc y trap, please check. date

AN ENSEMBLE SEASONAL FORECAST OF HUMAN CASES OF ST. LOUIS ENCEPHALITIS IN FLORIDA BASED ON SEASONAL HYDROLOGIC FORECASTS

West Nile virus in the WHO european region

Disease surveillance and outbreak prevention and control

Oviposition Preferences for Infusion-Baited Traps and Seasonal Abundance of Culex Mosquitoes in Southwestern Virginia

Assessment of Vulnerability to the Health Impacts of Climate Change in Middlesex-London

West Nile Encephalitis Professional Fact Sheet

Mosquito Surveillance Report Vector Surveillance Program

Principles of Disease and Epidemiology. Copyright 2010 Pearson Education, Inc.

Sentinel Chicken Screening Here are the mosquito larvae! Ground Application Aerial Larvicide Applications

Excess mortality in Europe in the winter season 2014/15, in particular amongst the elderly.

West Nile Virus Encephalitis Fact Sheet

Climatic and landscape correlates for potential West Nile virus mosquito vectors in the Seattle region

Effects of Temperature on the Transmission of West Nile Virus by Culex tarsalis (Diptera: Culicidae)

Fact Sheet for Health Care Providers: Interpreting Results from the Aptima Zika Virus Assay. June 17, 2016

Victims Compensation Claim Status of All Pending Claims and Claims Decided Within the Last Three Years

Florida Arbovirus Surveillance Week 25: June 19-25, 2016

WEST NILE VIRUS DEPARTMENT OF HEALTH AND HUMAN SERVICES

Certified in Public Health (CPH) Exam CONTENT OUTLINE

JENNIFER S. ARMISTEAD. Molecular Microbiology and Immunology

West Nile virus cluster analysis and vertical transmission in Culex pipiens complex mosquitoes in Sacramento and Yolo Counties, California, 2011

Eastern Equine Encephalitis Virus. History and Enhanced Surveillance in Ontario

Date of Commencement: January, 2004 Duration: One Year Status: Ongoing. Objectives

Isolation of West Nile Virus from Mosquitoes (Diptera: Culicidae) in the Florida Keys, Monroe County, Florida

Obesity in America: A Growing Trend

Pandemic Risk Assessment

PREPARING FOR A PANDEMIC. Lessons from the Past Plans for the Present and Future

California Department of Public Health. Operational Plan for Emergency Response to Mosquito-Borne Disease Outbreaks

Arboviral Surveillance Results

GIS and Public Health (GEOG 5190/6190) Course Syllabus. Spring 2015 University of Utah Department of Geography

4

Yale New Haven Health System Center for Healthcare Solutions

Alberta Health Services Speakers Series

4A. Types of Laboratory Tests Available and Specimens Required. Three main types of laboratory tests are used for diagnosing CHIK: virus

WEST NILE VIRUS QUESTIONS ABOUT SPRAYING AND MOSQUITO CONTROL

Risk Factors for Alcoholism among Taiwanese Aborigines

GAO WEST NILE VIRUS OUTBREAK. Lessons for Public Health Preparedness. Report to Congressional Requesters. United States General Accounting Office

Town of Concord Board of Health. Minutes of the Meeting Tuesday, August 21, 2012

EPIDEMIOLOGY OF HEPATITIS B IN IRELAND

COMPREHENSIVE MOSQUITO SURVEILLANCE AND CONTROL PLAN

AARHUS UNIVERSITY JUNE 15, 2010 BED BUGS OLE KILPINEN DANISH PEST INFESTATION LABORATORY INSTITUTE OF INTEGRATED PEST MANAGEMENT DENMARK

Use of Electronic Health Records in Residential Care Communities

ORANGE COUNTY MOSQUITO AND VECTOR CONTROL DISTRICT INTEGRATED VECTOR MANAGEMENT AND RESPONSE PLAN

Arboviral Encephalitis

Biology 3998 Seminar II. How To Give a TERRIBLE PowerPoint Presentation

Short Report: Failure of Burkholderia pseudomallei to Grow in an Automated Blood Culture System

Patient reported symptoms of psoriasis: results from the Psoriasis SELECT Patient Study

Exploratory Factor Analysis of Demographic Characteristics of Antenatal Clinic Attendees and their Association with HIV Risk

Resource development to maximise effective control of an outbreak of African Horse Sickness in Great Britain

Curriculum Vitae Updated April Stephanie Lynn Richards, PhD

TRACKS INFECTIOUS DISEASE EPIDEMIOLOGY

WHO Regional Office for Europe update on avian influenza A (H7N9) virus

EPIDEMIOLOGY OF TICK-BORNE ENCEPHALITIS AND LYME DISEASES IN ESTONIA

South Africa. General Climate. UNDP Climate Change Country Profiles. A. Karmalkar 1, C. McSweeney 1, M. New 1,2 and G. Lizcano 1

Guidelines for Animal Disease Control

The PCA Peanut Butter Outbreak - Minnesota s Involvement and Perspective

The 2015 African Horse Sickness season: Report

SAS and R calculations for cause specific hazard ratios in a competing risks analysis with time dependent covariates

Discovery, Distribution, and Abundance of the Newly Introduced Mosquito Ochlerotatus japonicus (Diptera: Culicidae) in Connecticut, USA

Transatlantic Masters Degree Program in Forest Resources (ATLANTIS)

Mosquito Abatement District Annual Report for 2013 (As required by 70 ILCS 1005 / et. seq.)

WHO-TDR PRIORITY DISEASES. EcoEpidemiology. Durland Fish, Ph.D. Yale University School of Public Health School of Forestry and Environmental Studies

A Method of Population Estimation: Mark & Recapture

How To Determine The Effects Of Hurricane Ivon On Seagrass Meadows In Alabama

Global Influenza Surveillance Network (GISN) Activities in the Eastern Mediterranean Region

The interface between wild boar and extensive pig production:

West Nile Virus Infection Surveillance in Québec 2013 SEASON

April King-Todd 2014 National TB Conference Atlanta, Georgia June 10, 2014

Illinois Influenza Surveillance Report

West Nile Virus and Mosquito Control

Epidemiology and Transmission Dynamics of West Nile Virus Disease

National FMD Response Planning

Gail Bennett, RN, MSN, CIC

Updated ECDC Public Health Microbiology Strategy and Work Plan

VECTOR-SURVEILLANCE IN NORTHERN ITALY

The Impact of Climate Change on Vector-Borne Infectious Diseases. Namasha Schelling. Graduate Research Paper

Canadian Public Health Laboratory Network. Core Functions of Canadian Public Health Laboratories

Core Functions and Capabilities. Laboratory Services

Ebola: Teaching Points for Nurse Educators

Transatlantic Masters Degree Program in Forest Resources (ATLANTIS)

Chapter 20: Analysis of Surveillance Data

statement Climate change and infectious diseases in Europe

Transcription:

Vol. 35, no. 1 Journal of Vector Ecology 35 The use of early summer mosquito surveillance to predict late summer West Nile virus activity Howard S. Ginsberg 1, Ilia Rochlin 2, and Scott R. Campbell 3 1 USGS Patuxent Wildlife Research Center, Coastal Field Station, Woodward Hall-PLS, University of Rhode Island, Kingston, RI 02881, U.S.A. 2 Division of Vector Control, Suffolk County Department of Public Works, Yaphank, NY 11980, U.S.A. 3 Arthropod-Borne Disease Laboratory, Suffolk County Department of Health Services, Yaphank, NY 11980, U.S.A. Received 12 June 2009; Accepted 28 September 2009 ABSTRACT: Utility of early-season mosquito surveillance to predict West Nile virus activity in late summer was assessed in Suffolk County, NY. Dry ice-baited CDC miniature light traps paired with gravid traps were set weekly. Maximumlikelihood estimates of WNV positivity, minimum infection rates, and % positive pools were generally well correlated. However, positivity in gravid traps was not correlated with positivity in CDC light traps. The best early-season predictors of WNV activity in late summer (estimated using maximum-likelihood estimates of Culex positivity in August and September) were early date of first positive pool, low numbers of mosquitoes in July, and low numbers of mosquito species in July. These results suggest that early-season entomological samples can be used to predict WNV activity later in the summer, when most human cases are acquired. Additional research is needed to establish which surveillance variables are most predictive and to characterize the reliability of the predictions. Journal of Vector Ecology 35 (1): 35-42. 2010. Keyword Index: Aedes, Culex, surveillance, West Nile virus. INTRODUCTION Surveillance programs that give accurate predictions of disease outbreaks are critical to effectively manage the risk of arboviral infections. In the case of West Nile virus (WNV), signs of current epizootic activity, such as increasing numbers of bird deaths or multiple positive mosquito pools, can indicate imminent risk of human disease (Kulasekera et al. 2001). However, recent research indicates that WNV amplification can be very focal and rapid (Ruiz et al. 2007, Bertolotti et al. 2008, Eisen and Eisen 2008, Hamer et al. 2008) so that timely identification of epizootic activity with the potential for human infection can be difficult. Gu et al. (2008) suggested that surveillance early in the season can help focus subsequent surveillance activities on highrisk areas that are likely to be transmission foci later in the season. This can provide more efficient surveillance programs than broadly-dispersed surveillance schemes and is more likely to identify areas of disease risk in time for preventive interventions. Several investigators have used remotely-sensed geospatial data to identify sites with higher than average abundance of vector mosquitoes, presumably with increased risk of viral activity (Brown et al. 2008, Rochlin et al. 2008). This approach can help focus surveillance, but epizootic activity can occur at different sites in different years because of changes in ecological factors, such as the effects of weather on local mosquito populations and on the abundance of appropriate bird species. Several studies have shown that sites with bird deaths attributable to WNV early in the season have a high risk of human disease later in the season (Eidson et al. 2001, Guptill et al. 2003, Johnson et al. 2005). However, most sites with WNV positive birds don t have human cases later that year, and the utility of bird surveillance changes with local changes in bird populations and immunity to WNV (Gu et al. 2008). Brownstein et al. (2004) found that on a national scale, mosquito surveillance data better predicted human disease than did bird data. However, the specific early-season mosquito surveillance results predictive of WNV activity, and the reliability of these surveillance data, have not been well characterized. In this study, we utilize data from sites with paired mosquito traps to assess the utility of entomological surveillance to predict WNV activity in August and September, when most humans acquire WNV infection. We employed traps typically used in WNV surveillance programs (CDC miniature light traps baited with dry ice paired with gravid traps baited with rabbit chow infusion) and compared various estimators of WNV infection levels in mosquitoes. We then examined the utility of early summer entomological data to predict WNV activity later in the season and thus to focus late summer surveillance and management efforts. MATERIALS AND METHODS Mosquito sampling Mosquito traps were set at three sample areas in Suffolk County, NY. The areas were selected based on an analysis of Suffolk County surveillance data on human and equine WNV cases using spatial cluster detection software (SaTScan TM, Kulldorff 1997) in an ArcGIS 9.1 (ESRI Inc, Redlands, CA) environment (Rochlin et al. 2009). Three pairs of traps (a CDC light trap paired with a gravid trap)

36 Journal of Vector Ecology June 2010 were set in each sample area, for a total of nine pairs of traps. One pair of traps was set at a residential site, one pair at a commercial site, and one pair at a natural site in each sample area. At each trapping site a dry ice-baited CDC trap was placed about 20 m from a rabbit chow-baited gravid trap, and the traps were set overnight for one night each week. The most distant of the nine pairs of traps were about 18 km apart (see Figure 1 in Rochlin et al. 2009). Traps were set weekly from 1 June to 30 September in 2005 and 2006. Mosquito specimens were anesthetized on dry ice and determined to species. We wanted to simulate surveillance data, where large numbers of frequently-taken trap collections are rapidly processed and sorted, so we used genus-level categories in our analyses. Aedes mosquitoes are generally determined to species in surveillance data, but individual species are often present only intermittently, in individual traps or at restricted times of year. Therefore, we combined Aedes species potentially involved in WNV transmission (Turrell et al. 2005) into one category, Aedes spp. This category included Ae.canadensis, Ae. cantator, Ae. japonicus, Ae. sollicitans, Ae. triseriatus, and Ae. vexans at our sites. Three species of Culex at our study sites (Cx. pipiens, Cx. restuans, and Cx. salinarius) are potentially important in WNV transmission (details of the distributions of these species at our sites are given by Rochlin et al. 2009). Fieldcollected specimens are difficult to distinguish without molecular techniques (Crabtree et al. 1995) so surveillance programs generally report pools of these species as Culex pipiens/restuans or Culex pipiens/restuans/salinarius (Slaff and Crans 1982,White et al. 2001, Lucacik et al. 2006). Therefore, we compiled collections into two categories (Aedes spp. and Culex spp.) for our analysis. Analysis Mosquito pools that were tested for WNV varied in size, with a maximum size of 50 specimens. We assessed levels of infection in pools of Culex spp. using three measures; 1) the proportion of pools positive for WNV, 2) the minimum infection rate (MIR), which is the number of positive pools per thousand mosquitoes tested, and 3) a maximum-likelihood estimate (MLE) of the actual number of mosquitoes infected per thousand (using PooledInfRate estimation software as an Add-In to Microsoft Excel; Biggerstaff 2004). Correlations of various prevalence estimators at each of the nine trap pairs were analyzed using BIOMstat (Exeter Software, Setauket, NY). Regressions to assess predictions of late summer WNV activity using mosquito trap surveillance data (at each of the nine trap pairs) were analyzed using SAS 9.1 (SAS Institute Inc., Cary, NC). The surveillance variables utilized in the regressions were: July-CxG - mean number of Culex spp. in gravid trap samples in July July-CxL - mean number of Culex spp. in CDC trap samples in July July-AeL - mean number of Aedes spp. in CDC trap samples in July Aug-CxG - mean number of Culex spp. in gravid trap samples in August Aug-CxL - mean number of Culex spp. in CDC trap samples in August Aug-AeL - mean number of Aedes spp. in CDC trap samples in August July-Nspp - mean number of mosquito species (= species richness) in CDC trap samples in July Aug-Nspp - mean number of mosquito species (= species richness) in CDC trap samples in August Date 1 st + - day number of first date of positive mosquito pool in any trap Late summer WNV activity was estimated using MLE values for the number of adult female Culex mosquitoes infected per 1,000 in combined trap catches from August and September. RESULTS Overall WNV positivity in our samples from August and September are shown in Table 1. Correlations between various estimators of infection rates in mosquitoes trapped during August and September are shown in Table 2, and the numbers of mosquitoes in the pools tested are shown in Figure 1. Mosquito pools varied substantially in the numbers of specimens per pool. Nevertheless, MLE, MIR, and percent of pools positive were generally highly correlated in both light and gravid trap samples. In contrast, WNV infection rates among mosquitoes in gravid trap samples were not correlated with infection rates in paired light trap samples, Table 1. Overall WNV positivity in mosquito pools collected in August and September from study site in Suffolk County, NY. Year Trap type # pools tested # pools positive Average MIR Average MLE 2005 2006 Gravid 62 22 15.1 20.7 Light 64 14 13.3 16.4 Gravid 69 11 7.3 9.4 Light 60 7 9.1 11.6 MIR = Minimum Infection Rate (number of positive pools per 1,000 mosquitoes tested). MLE = Maximum-Likelihood Estimate of number of mosquitoes positive per 1,000 tested.

Vol. 35, no. 1 Journal of Vector Ecology 37 Table 2. Comparison of various estimates of WNV infection rates in mosquitoes trapped in August and September. 2005 2006 Estimates compared Trap type r P r P % positive vs MLE Gravid 0.793 0.011 0.901 0.0009 CDC Light 0.969 0.000016 0.871 0.0022 MLE vs MIR Gravid 0.898 0.001 0.992 0.0000002 CDC Light 0.992 0.00000014 0.985 0.0000014 MIR vs % positive Gravid 0.538 0.135 0.908 0.0007 CDC Light 0.955 0.000062 0.920 0.0004 % positive = percent of pools that were positive for WNV. MLE = Maximum-Likelihood Estimate of number of mosquitoes positive per 1,000. MIR = Minimum Infection Rate (number of positive pools per 1,000 mosquitoes tested). N = 9 for all comparisons. NUMBER OF POOLS.. 25 20 15 10 5 0 2005 Light traps Gravid traps 1 to 10 11 to 20 21 to 30 31 to 40 41 to 50 2006 NUMBER OF POOLS.. 40 30 20 10 0 1 to 10 11 to 20 21 to 30 31 to 40 41 to 50 NUMBER OF MOSQUITOES IN POOL Figure 1. Numbers of mosquitoes in all pools tested for WNV infection from trap collections in August and September.

38 Journal of Vector Ecology June 2010 Table 3. Comparison of estimates of infection rates in gravid trap vs. light trap data from August and September. 2005 2006 Estimation method r P r P Maximum-Likelihood Estimate 0.397 0.290-0.0304 0.938 % positive pools 0.408 0.275-0.203 0.601 Minimum Infection Rate 0.301 0.432-0.169 0.665 N=9 for all comparisons regardless of the estimator used (Table 3). To assess the individual surveillance variables as predictors of WNV activity in late summer, we regressed MLE values for WNV positivity in August and September on surveillance data from light traps and gravid traps (Table 4). The best predictive variables in this preliminary analysis included date of first positive pool, number of Aedes in light traps in July, and mosquito species richness. In all cases, the regression coefficients were negative. Therefore, WNV activity in late summer was best predicted by early collection of positive pools, low numbers of Aedes in July, and low mosquito species richness, especially in July. Numerous tests using single variables can lead to spurious relationships, so we then calculated multiple regression models for these surveillance data. The best two-variable regression models (highest R 2, lowest AIC in all cases) are shown in Table 5. The variables that contributed significantly to these models included early isolations of WNV in mosquito pools, low mosquito populations and species richness in July, and high mosquito populations in August. DISCUSSION Minimum Infection Rate is a commonly-used estimator of WNV infectivity in mosquito surveillance samples. However, MIR indicates only the minimum possible level of infection, assuming only one mosquito per pool is infected, so maximum-likelihood approaches to estimating actual infection rates provide estimates that are generally closer to the true infection rates in mosquitoes (Gu et al. 2004). The high correlations of these estimates in our samples (Table 2), even with highly variable pool sizes (Figure 1), suggest that either estimator can be used to assess potential disease risk. The general agreement of MIR and MLE in our samples results, in part, from the likelihood that only one or a few mosquitoes in each positive pool were infected. Since the infection rates at our sample sites were in the 1 2 % range (Table 1), most positive pools (which contained less than 50 mosquitoes each) probably contained only one infected specimen. Nevertheless, as infection rates increase both estimators would increase, though at different schedules depending on how infected mosquitoes were distributed among pools. MIR has the advantage of simple calculation, but software programs (such as Excel Add-Ins) are now available to rapidly calculate MLEs of infection rates. We conclude that either estimator can be useful to evaluate WNV infection levels in vector mosquitoes, and if used consistently and interpreted carefully, to assess risk of WNV transmission. The lack of a correlation between levels of infection in gravid trap samples and those in dry ice-baited CDC light trap samples is of considerable interest. In general, gravid trap samples are considered more sensitive indicators of the presence of WNV because the mosquitoes trapped tend to be gravid Culex females that have already had blood meals, and thus have an increased probability of having been exposed to the virus (Moore et al. 1993). CDC light traps, in contrast, tend to capture a variety of blood-meal seeking mosquitoes, many of which may be newly emerged with no chance of viral exposure. Therefore, light trap samples are expected to have lower WNV prevalence than gravid trap samples but might be better predictors of human risk because they reflect infection levels in potentially humanbiting mosquitoes. Our results confirm that positivity in gravid traps does not match positivity in light traps (Table 3), but gravid trap samples did not consistently have higher MLE values than light trap samples (Table 1), and we cannot determine from our results which is a better predictor of human disease risk. Indeed, only one human case of WNVrelated disease was reported from our study area during 2005 and none in 2006 (of a total of nine cases in 2005 and two in 2006 reported from all of Suffolk County, NY). Interestingly, our models generally gave better predictions in 2005, when human risk was greater, than in 2006 (Tables 3 and 4). Several studies have shown relationships between estimators of WNV activity in mosquitoes and human cases of disease (Kulasekera et al. 2001, Brownstein et al. 2004, Winters et al. 2008), but they have not separated light trap from gravid trap samples. This point is worth further study, in order to develop more accurate predictors of human WNV risk from surveillance data. Our results suggest that early-season mosquito surveillance data can be used to predict WNV activity levels in late summer when most humans acquire WNV infection. Reliable predictive models could be used to analyze results from a diffuse early-season monitoring program so as to design a spatially targeted surveillance program for August and September that could provide more accurate and more timely predictions of human disease risk than are currently available (Gu et al. 2008). Our results come from three proximate areas over two years, so more replicates are needed to develop reliable predictors of late summer WNV activity in our study area and to determine their utility at other sites. Recent authors have utilized weather and hydrological data to develop predictive models of WNV activity (Landesman et al. 2007, Day and Shaman 2008). Combining entomological

Vol. 35, no. 1 Journal of Vector Ecology 39 Table 4. Regression of individual surveillance variables on WNV positivity in Culex mosquitoes (estimated by MLE) in August and September. Trap type Variable Year Std. coeff. P R 2 Gravid July-CxG 2005 0.149 0.701 0.022 2006 0.125 0.748 0.016 July-CxL 2005-0.264 0.492 0.070 2006 0.259 0.502 0.067 July-AeL 2005-0.947 0.0001 0.898 2006-0.127 0.744 0.016 Aug-CxG 2005 0.204 0.598 0.042 2006 0.084 0.829 0.007 Aug-CxL 2005 0.312 0.414 0.097 2006 0.305 0.425 0.093 Aug-AeL 2005-0.620 0.075 0.384 2006 0.379 0.315 0.144 July-Nspp 2005-0.764 0.017 0.584 2006-0.038 0.924 0.001 Aug-Nspp 2005-0.766 0.016 0.587 2006 0.258 0.503 0.066 Date 1 st + 2005-0.406 0.279 0.165 2006-0.773 0.025 0.597 Light July-CxG 2005 0.573 0.107 0.329 2006-0.386 0.304 0.149 July-CxL 2005 0.644 0.061 0.415 2006 0.333 0.382 0.111 July-AeL 2005-0.284 0.460 0.080 2006-0.150 0.699 0.023 Aug-CxG 2005 0.389 0.301 0.151 2006-0.117 0.764 0.014 Aug-CxL 2005 0.186 0.631 0.035 2006 0.220 0.570 0.048 Aug-AeL 2005 0.362 0.338 0.131 2006-0.300 0.433 0.090 July-Nspp 2005-0.745 0.021 0.555 2006-0.449 0.225 0.202 Aug-Nspp 2005-0.238 0.538 0.057 2006-0.192 0.621 0.037 Date 1 st + 2005-0.612 0.080 0.375 2006-0.336 0.416 0.113 Significant relationships in boldface. N = 9 for all regressions.

40 Journal of Vector Ecology June 2010 Table 5. Best 2-variable models to predict mosquito positivity a for WNV in August and September. Model first variable second variable Trap year R 2 P variable coeff. P variable coeff. P Gravid 2005 0.969 <0.0001 July-AeL -2.645 <0.0001 date 1 st + b -0.163 0.01 2006 0.680 0.058 date 1 st + b -0.459 0.032 Aug-AeL 0.606 0.31 Light 2005 0.923 0.0005 Nspp-Jl -11.017 0.0002 Aug-AeL 2.991 0.002 2006 0.682 0.084 July-CxG -0.350 0.023 July-AeL 0.676 0.096 a Maximum-Likelihood Estimates of number positive per 1,000. b Date of first positive pool in either gravid or light traps. Significant relationships in boldface. N = 9 for all regressions. data with weather and hydrological data can presumably be used to generate considerably more reliable predictions of WNV activity than are currently possible. The surveillance variables that most consistently predicted high late summer WNV activity were early date of first positive pool, low population sizes, and low species richness of mosquitoes in July (Tables 3 and 4). Early collection of positive pools indicates elevated levels of viral circulation early on, which would be expected to correlate (with some degree of variability resulting from weather and other factors during the summer) with increased levels of viral activity later in the summer. However, the other indicators, low mosquito numbers and species richness, are intriguing because they are counterintuitive. One possible explanation for these results comes from recent work on tick-borne diseases, which suggests that high species diversity tends to lower pathogen transmission by diffusing the importance of major reservoirs (LoGiudice et al. 2003). Similar patterns have been reported for WNV and avian species diversity (Ezenwa et al. 2006, Swaddle and Calos 2008). However, this hypothesis is based on diversity of reservoirs and not vectors as in our results. Vector diversity likely has a different relationship with transmission patterns than reservoir diversity (Mather and Ginsberg 1994). Furthermore, this possible explanation seems unlikely to apply to our results because it deals only with diversity and cannot explain the predictive value of low mosquito numbers. Low vector numbers predictive of high pathogen prevalence would contradict the diffusion hypothesis. Perhaps a more plausible explanation comes from recent studies suggesting that dry conditions tend to foster WNV transmission, perhaps by concentrating mosquitoes and bird reservoirs at the few remaining wetspots (Shaman et al. 2005, Day and Shaman 2008). Taken spatially, areas that are drier would have fewer mosquitoes and fewer species of mosquitoes in surveillance traps, but these sites might provide relatively few isolated wetspots that would tend to concentrate mosquitoes and birds, resulting in efficient WNV transmission. A third possible explanation is that WNV prevalence was so low at our study sites that the results are actually spurious relationships. The consistency of our results, with the signs of the regression coefficients being the same in nearly all cases (in all statistically significant relationships in Tables 4 and 5, the regression coefficients for date of first positive pool, mosquito abundance in July, or species richness in July were negative), argues against this possibility. Although we did not attempt to distinguish in this study whether diversity or host/vector concentration is responsible for these patterns, our results suggest strongly that additional research to develop more effective surveillance programs for WNV transmission using early season entomological data is likely to yield useful results. We have not considered bird surveillance or weather patterns in this study, but adding variables from these types of data would presumably result in stronger models with more reliable predictive capacities. To summarize, estimates of levels of WNV positivity in mosquitoes calculated using MIR, MLE, or percent positive pools were generally well correlated. However, positivity in gravid trap samples was not correlated with positivity in samples from CO 2 -baited light traps. Entomological surveillance data from early summer show considerable potential to identify sites that are likely to have high viral activity later in the summer. Combining results from entomological samples with weather and hydrological data can potentially give powerful predictions of WNV activity. More research is needed to characterize the utility of various types of early-season surveillance data. Acknowledgments The authors thank Melissa Zanini, Kerri Harding, and Ralph Narain for assistance with technical aspects of the project, including field and lab work, and data management. We thank N. Petti, R. Chayes, C. Provenzano, and A. Culkin for help processing mosquito specimens. We appreciate arboviral analysis of mosquito pools by Laura Kramer and staff at the NY State Department of Health Arboviruses Laboratories. C. Barry Knisley and Michael Higgins kindly provided comments on early drafts of the manuscript. The research was partially funded by Specific Cooperative Agreement # 58-5410-4-338 from the United States Department of Agriculture, and by the U.S. Geological Survey.

Vol. 35, no. 1 Journal of Vector Ecology 41 REFERENCES CITED Bertolotti L., U.D. Kitron, E.D. Walker, M.O. Ruiz, J.D. Brawn, S.R. Loss, G.L. Hamer, and T.L. Goldberg. 2008. Fine-scale genetic variation and evolution of West Nile Virus in a transmission hot spot in suburban Chicago, USA. Virology 374: 381-389. Biggerstaff B. 2004. PooledInfRate, version 2.0. Excel Add- In. Centers for Disease Control and Prevention, Fort Collins, CO. Brown H., M. Duik-Wasser, T. Andreadis, and D. Fish. 2008. Remotely-sensed vegetation indices identify mosquito clusters of West Nile Virus vectors in an urban landscape in the northeastern United States. Vector-Borne Zoonotic Dis. 8: 197-206. Brownstein J.S., T.R. Holford, and D. Fish. 2004. Enhancing West Nile Virus surveillance, United States. Emerg. Infect. Dis. 10: 1129-1133. Crabtree M.B., H.M. Savage, and B.R. Miller. 1995. Development of a species-diagnostic polymerase chain reaction assay for the identification of Culex vectors of St. Louis encephalitis virus based on interspecies variation in ribosomal DNA spacers. Am. J. Trop. Med. Hyg. 53: 105-109. Day J.F. and J. Shaman. 2008. Using hydrologic conditions to forecast the risk of focal and epidemic arboviral transmission in peninsular Florida. J. Med. Entomol. 45: 458-465. Eidson M., L. Kramer, W. Stone, Y. Hagiwara, K. Schmit, and N.Y. State WNV avian surveillance team 2001. Dead bird surveillance as an early warning system for West Nile Virus. Emerg. Infect. Dis. 7: 631-635. Eisen R.J. and L. Eisen. 2008. Spatial modeling of human risk of exposure to vector-borne pathogens based on epidemiological versus arthropod vector data. J. Med. Entomol. 45: 181-192. Ezenwa V.O., M.S. Godsey, R.J. King, and S.C. Guptill. 2006. Avian diversity and West Nile virus: testing associations between biodiversity and infectious disease risk. Proc. Biol. Sci. 273: 109-117. Gu W., R. Lampman, and R.J. Novak. 2004. Assessment of arbovirus vector infection rates using variable size pooling. Med. Vet. Entomol. 18: 200-204. Gu W., T.R. Unnasch, C.R. Katholi, R. Lampman, and R.J. Novak. 2008. Fundamental issues in mosquito surveillance for arboviral transmission. Trans. R. Soc. Trop. Med. Hyg. 102: 817-822. Guptill S.C., K.G. Julian, G.L. Campbell, S.D. Price, and A.A. Marfin. 2003. Early-season avian deaths from West Nile Virus as warnings of human infection. Emerg. Infect. Dis. 9: 483-484. Hamer G.L., E.D. Walker, J.D. Brawn, S.R. Loss, M.O. Ruiz, T.L. Goldberg TL, A.M. Schotthoefer, W.M. Brown, E. Wheeler, and U.D. Kitron. 2008. Rapid amplification of West Nile virus: the role of hatch-year birds. Vector- Borne Zoonotic Dis. 8: 57-67. Johnson G.D., M. Eidson, K. Schmit, A. Ellis, and M. Kulldorff. 2005. Geographic prediction of human onset of West Nile Virus using dead crow clusters: an evaluation of year 2002 data in New York State. Am. J. Epidemiol. 163: 171-180. Kulasekera V.L., L. Kramer, R.S. Nasci, F. Mostashari, B. Cherry B, S.C. Trock, C. Glaser, and J.R. Miller. 2001. West Nile virus infection in mosquitoes, birds, horses, and humans, Staten Island, New York, 2000. Emerg. Infect. Dis. 7: 722-725. Kulldorff, M.1997. A spatial scan statistic. Comm. Stat. Theor. Meth. 26: 1481-1496. Landesman, W.J., B.F. Allen, R.B. Langerhans, T.M. Knight and J.M. Chase. 2007. Inter-annual associations between precipitation and human incidence of West Nile Virus in the United States. Vector-Borne and Zoonot. Dis. 7: 337-343. LoGiudice K., R.S. Ostfeld, K.A. Schmidt, and F. Keesing. 2003. The ecology of infectious disease: effects of host diversity and community composition on Lyme disease risk. Proc. Natl. Acad. Sci. 100: 567-571. Lucacik G., M. Anand, E.J. Shusas, J.J. Howard, J. Oliver, and H. Chen, 2006. West Nile Virus surveillance in mosquitoes in New York State, 2000-2004. J. Am. Mosq. Contr. Assoc. 22: 264-271. Mather T.N. and H.S. Ginsberg. 1994. Vector-host-pathogen relationships: transmission dynamics of tick-borne infections. In: D.E. Sonenshine and T.N. Mather (eds.) Ecological Dynamics of Tick-borne Zoonoses. pp. 68-90. Oxford University Press, NewYork. Moore C.G., R.G. McLean, C.J. Mitchell, R.S. Nasci, T.F. Tsai, C.H. Calisher CH, A.A. martin, P.S. Moore, and D.J. Gubler. 1993. Guidelines for arbovirus surveillance programs in the United States. Centers for Disease Control and Prevention, Fort Collins, CO, 83 pp. Rochlin, I., H.S. Ginsberg, and S.R. Campbell. 2009. Distribution and abundance of host-seeking Culex species at three proximate locations with different levels of West Nile virus (WNV) activity. Am. J. Trop. Med. Hyg. 80: 661-668. Rochlin I., K. Harding, H.S. Ginsberg, and S.R. Campbell. 2008. Comparative analysis of distribution and abundance of West Nile and Eastern Equine Encephelitis virus vectors in Suffolk County, New York using human population density and land use/cover data. J. Med. Entomol. 45: 563-571. Ruiz M.O., E.D. Walker, E.S. Foster, L.D. Haramis, and U.D.Kitron. 2007. Association of West Nile virus illness and urban landscapes in Chicago and Detroit. Int. J. Hlth. Geogr. 6:10. Shaman J., J.F. Day, and M. Stieglitz. 2005. Drought-induced amplification and epidemic transmission of West Nile Virus in southern Florida. J. Med. Entomol. 42: 134-141. Slaff M. and W.J. Crans. 1982. Impounded water as a major producer of Culex salinarius (Diptera: Culicidae) in coastal areas of New Jersey, USA. J. Med. Entomol. 19: 185-190. Swaddle J.P. and S.E. Calos. 2008. Increased avian diversity is associated with lower incidence of human West Nile

42 Journal of Vector Ecology June 2010 infection: observation of the dilution effect. PLoS ONE. 3:e2488. Turell M.J., D.J. Dohm, M.R. Sardelis, M.L. Oguinn, T.G. Andreadis, and J.A. Blow. 2005. An update on the potential of North American mosquitoes (Diptera: Culicidae) to transmit West Nile Virus. J. Med. Entomol. 42: 57-62. White D.J., L.D. Kramer, P.B. Backenson, G. Lukacik, G. Johnson, J. Oliver, J.J. Howard, R.G. Means, M. Eidson, I. Gotham, V. Kulasekera, S. Campbell, and the West Nile Virus Response Team. 2001. Mosquito surveillance and polymerase chain reaction detection of West Nile Virus, New York State. Emerg. Infect. Dis. 7: 643-649. Winters A.M., B.G. Bolling, B.J. Beaty, C.D. Blair, R.J. Eisen, A.M. Meyer, W.J. Pape, C.G. Moore, and L. Eisen. 2008. Combining mosquito vector and human disease data for improved assessment of spatial West Nile virus disease risk. Am. J. Trop. Med. Hyg. 78: 654-665.