The study area consisted of the City of Fort Collins, CO. The city has a population of roughly 152,000 and covers ~84 km² (United States Census Bureau 2013b). Fort Collins is located near the foothills of the Rocky Mountains in the high plains ecological zone (Brown et al. 2011). The climate is semiarid, with cold winters and hot and dry summers, low humidity, and variable precipitation (Winters et al. 2008). Fort Collins borders on extensively irrigated agricultural lands to the north, east, and south, whereas the western edge lies along the uncultivated foothills. For analytical purposes, the city was divided into four zones using major thoroughfares: College Avenue was the East–West boundary, and Drake Road was the North–South boundary (Fig. 1).

Colorado Mosquito Control, Inc. conducted weekly trapping and identification of adult mosquitoes for the entirety of the study. Trap locations within Fort Collins were consistent over the 8-yr study period. Trapping was performed weekly using 42 CO₂ (dry ice)-baited Centers for Disease Control and Prevention (CDC) miniature light traps (BioQuip Products, CA) distributed in a grid-like pattern across the city, ~1.3km apart (Fig. 1). In addition to the light traps, up to 10 gravid traps were operated in any given week to attract oviparous female mosquitoes. Trapping began in Morbidity and Mortality Weekly Report epidemiological week 23 (early June) and was continued through week 35 (early September) in all years except 2011 when trapping was concluded in week 32 due to low values for entomological risk indices. Each trap was run one night per week from late afternoon until the following morning. Mosquitoes were collected, and then sorted by site, date, species, and sex. Female Culex mosquitoes were pooled, typically in pools of no more than 50 specimens, and submitted for WNV screening. Pools were submitted as Cx. tarsalis, Cx. pipiens, or, for some specimens lacking certain body parts required for species identification, as Culex species (spp.). Based on the results from molecular identification assays showing that the vast majority of Culex spp. pool specimens were Cx. pipiens (CDC, unpublished data), Culex spp. mosquitoes were included as Cx. pipiens in calculations for mosquito abundance, infection rate, and VI.

From 2006–2008, mosquito pools were processed by CDC Division of Vector-Borne Diseases personnel for the presence of WNV RNA as described previously (Lanciotti et al. 2000, Nasci et al. 2001b). From 2009–2013, mosquito pools were processed by the Colorado State University Arthropod-Borne and Infectious Diseases Laboratory using the following methodology. Mosquito pools were homogenized in 1 ml of mosquito diluent (80% PBS, 20% FBS, supplemented with penicillin, streptomycin, gentamicin, and amphotericin B) with a single steel ball bearing using a Retsch Mixer Mill 400 (Retsch GmbH, Haan, Germany) at 24Hz for 45 s. Homogenates were then centrifuged at 20,000 × g for 5 min, and 50 µl of cleared supernatant was used for RNA extraction. RNA was extracted using either QIAamp Viral RNA Mini Kit (Qiagen, CA), or Mag-Bind Viral DNA/RNA kit (Omega, GA) with the KingFisher Flex Magnetic Particle Processor (Thermo Fisher Scientific, MA) according to manufacturer’s protocol. The extracted RNA was amplified via reverse transcriptase polymerase chain reaction (RT-PCR) using the following primers: forward 212 5′ TTGTGTTGGCTCT CTTGGCGT 3′, reverse 619c 5′ CAGCCGACAGCACTGGACATT 3′ (Giladi et al. 2001). RT-PCR products were run on a 1% agarose gel stained with ethidium bromide in order to visualize the 408 base pair target sequence.

Weekly Culex abundance and infection rate were calculated separately for females of 1) Cx. tarsalis, 2) Cx. pipiens, and 3) all Culex. Abundance was based solely on collections from light traps, whereas collections from both light traps and gravid traps were included in the WNV infection rate calculation, following the protocol established by Nasci et al. (2005). Infection rates were calculated per 1,000 females using the bias-corrected maximum likelihood estimate (MLE) in the Excel add-in, PooledInfRate (Biggerstaff 2009). MLE was used in favor of the minimum infection rate (MIR) because it is more accurate at high infection rates and mosquito abundances (Gu et al. 2003). The VI is calculated by multiplying the abundance per trap night of a given mosquito species with the estimated proportion of infected females for that species (CDC 2013); VI=∑N¯iP^iwhere i refers to the specific mosquito species, N is the abundance per trap night of the ith species, and P is the estimated infection rate per one female of the ith species. The VI is calculated separately for each vector species in a given area, in this case Cx. tarsalis and Cx. pipiens, and the subsequent addition of the VIs for all vector species provides an overall VI value. Abundance per trap night, which is the total Culex abundance divided by the number of traps operated in that zone, infection rates, and VI were calculated on a weekly basis for each year in the study for each zone.

Weather data from 2006–2013 were obtained from the Colorado Agricultural Meteorological Network. The average daily temperature and precipitation was recorded at the Fort Collins Agricultural Engineering Research Center on the CSU Foothills campus. The average weekly temperature for all years was calculated by taking the average of the daily temperatures for the week. Total precipitation was calculated by taking the sum of all precipitation for the week.

Anonymized human case data from 2006–2013 were provided by the Larimer County Department of Health and Environment (LCDHE). Cases were assigned to zones by LCDHE using Google Fusion Tables (an open source geocoding service) based on the home address. Annual zone populations were estimated by determining the correspondence of 2000 and 2010 census blocks with a physical LCDHE map of Fort Collins (United States Census Bureau 2013a, 2014). The resulting census block lists were verified by matching them to U.S. Census Bureau TIGER/Line 2000 and 2010 census block shapefiles using ArcGIS (Esri 2015, United States Census Bureau 2015). The 2000 and 2010 census population data for each zone by census block were then obtained using American Fact Finder (United States Census Bureau 2012a, b). Annual zone populations were estimated using a compound annual growth rate calculated based on the change between the 2000 and 2010 census population falling within each zone by census block (Parker 2002). These estimates were used with mapped case data to calculate annual and cumulative zone incidence rates. Relative risks among zones were calculated using cumulative zone incidence rates. The NW zone is used as the reference group for the analysis, as it has the lowest cumulative incidence.

Because the entomological data did not follow assumptions for normality or equal variance, a Friedman’s test was conducted followed by Dunn’s test of multiple comparisons to determine differences between zones for Culex abundance per trap night, infection rate, and VI. The Spearman correlation was used to determine the relationship between the VI and human cases. Significance of relative risk for human cases was determined through calculation of 95% confidence intervals; a 95% confidence interval that does not include 1 (reference zone risk) is considered statistically significant. Statistics were calculated with Prism (GraphPad, CA).

Over the 8-yr study period a total of 602,420 female mosquitoes were trapped city wide, of which 131,777 were Culex species (21.9%). Of the Culex, 115,882 (87.9%) were identified as Cx. tarsalis, and 15,895 (12.1%) were identified either as Cx. pipiens or Culex spp. and assigned to Cx. pipiens. Using data from the whole city for all years, weekly 8-yr averages for abundance, infection rate, and VI were calculated (Fig. 2). Cx. tarsalis weekly 8-yr average abundance peaked in week 29, whereas Cx. pipiens abundance peaked in week 32 (Fig. 2A). There was no discernible difference between the two species for the seasonal peak in WNV infection rate (Fig. 2B). Infection rates did not exceed 1 per 1,000 females until week 29, and gradually increased through the end of the trapping season. Infection rates tended to peak at the end of the trapping season in week 35. The weekly 8-yr average VI rose sharply and peaked in week 29 for Cx. tarsalis, and rose gradually and peaked in week 32 for Cx. pipiens (Fig. 2C). The 8-yr average VI was higher for Cx. tarsalis.

The general trends for Culex abundance, infection rate, and VI remained constant year to year; however, the values varied substantially between years (Fig. 3). Culex abundance peaked prior to infection rate and VI in all years. The infection rate continued to rise throughout the trapping season, while the VI decreased with Culex abundance late in the season. 2007 had the highest Culex abundance with a total of 34,608 females captured, which accounted for over 25% of the total abundance for the 8-yr study. The majority of Culex captured in 2007 were Cx. tarsalis (32,314; 93.3%) as opposed to Cx. pipiens (2,294; 6.7%). 2012 had the highest average weekly infection rate with 10.3 per 1,000 females infected. 2007 and 2013 had the highest average weekly VIs of 0.40 and 0.43, respectively. Average weekly temperature fluctuated mildly within years, typically being lowest at the beginning of the season (May), rising throughout the season, and decreasing at the end (September). The temperatures between years varied more substantially. Weekly precipitation was sporadic and varied between years. In several years, there appeared to be a positive relationship between peaks in precipitation and subsequent peaks in vector abundance (Fig. 3—2006, 2009, 2010), but in other years that pattern was missing or greatly reduced (Fig. 3—2007, 2008, 2012).

Fort Collins was divided into four zones in order to test the hypothesis that the city is heterogeneous for mosquito-based risk measures (Fig. 1). We found significant differences for each zone comparison for Cx. tarsalis and Cx. pipiens abundance per trap night (CO₂-baited light traps), save the NE versus SE zones for Cx. tarsalis (Table 1). The only significant difference in infection rate was between the NW and SE zones for Cx. tarsalis, with the SE zone having a higher infection rate. The SE zone had a significantly higher VI when compared with the NW and SW zones for Cx. tarsalis and all Culex VI. The differences in VI for each zone, week, and year are visually represented in Fig. 4. In general, the eastern zones had higher Culex abundance and VI when compared with the western zones. Total human cases for each zone in each year are highly correlated with the sum of the zone VI in the same year (Fig. 5). Relative risk follows the same pattern as the VI. Inhabitants of the SE and NE zones have a significantly higher risk (P-value < 0.05) of contracting WNV as opposed to those living in the western zones (Table 2).