Comparisons of mosquito populations before and after construction of a wetland for water quality improvement in Pitt County, North Carolina, and data-reliant vectorborne disease management.

Introduction

With human cases of West Nile virus (WNV) and Eastern Equine encephalitis (EEE) again appearing in many states, (U.S. Geological Survey [USGS], 2005), citizens in all states and locations must be kept informed and continually educated about mosquitoborne diseases. Many citizens, however, still harbor misconceptions about mosquito life cycles, habits, and habitats, which continue to confuse, frustrate, and anger them each summer. For example, new residents from Northern states may not be familiar with the species of mosquitoes that occur in Mid-Atlantic tidal areas, or with the periodicity of freshwater floodwater mosquitoes. Other examples are confusion about which mosquito species are vectors for disease and what types of habitat are required for any of the 60 species that occur in North Carolina. In a local small town (Simpson, North Carolina), concern was publicly voiced over the possible negative consequences of constructing a wetland near a residential area because of confusion over the species of mosquito that carry West Nile virus, the habitat that the species use, and the periodicity of the mosquito population. Water quality improvements and other wetland services often come into conflict with citizen comfort and approval in wetland construction plans (City of Phoenix, 1999; Reid, 2003). Even in well-funded mosquito control districts (as in Florida and New Jersey), the negative mosquito effects of wetland construction near residential areas is a public concern. The study reported here was a coordinated effort of the North Carolina Clean Water Management Trust Fund (NCCWMTF), the Greenville North Carolina Resource Conservation and Development Office (NCRCD), the North Carolina Public Health Pest Management Section (NCPHPM), and East Carolina University to provide data that would address the question of whether mosquito populations would significantly rise and the question of what species would be present in a community after a new wetland was introduced.

[ILLUSTRATION OMITTED]

The town of Simpson, North Carolina, was selected as the study site (Figure 1) because of an NCRCD-planned small-wetland construction project. Simpson is a small but growing town east of Greenville with a population of about 500 citizens. Historically, the town has had severe problems with drainage during times of significant rainfall (Pope, Tasker, & Robbins, 2001). Following Hurricane Floyd in 1999, the Simpson area had a relentless mosquito problem, similar to that experienced by other small towns in Eastern North Carolina (Anderson et al., 2000). In an effort to correct and improve drainage problems and to contribute to water quality, NCRCD's Greenville office initiated a project to construct a wetland along the Mill Branch Creek in Simpson (Figure 1). There was concern within the NCRCD organization that local citizens would object to a constructed wetland because of the perceived potential for additional mosquito breeding. Therefore, North Carolina Public Health Pest Management was contracted to do a research project to aid in addressing this concern. Recent data support the claim that constructed wetlands add to mosquito problems in a local area (Evans, 1996; Shafer, Lundstrum, Pfeiffer, Lundkvist, & Landin, 2003), but many factors besides wetland construction must be considered before cause is ascribed. Since some mosquitoes, such as Aedes vexans, travel long distances from breeding sites, a jurisdiction-wide survey of mosquito populations in Simpson was needed to determine the current mosquito-breeding levels in the town. This survey would provide a baseline for determination of whether mosquito populations in and near the wetland construction site changed after wetland construction. Measures of rainfall, time of year, and temperature were included in the analysis to account for natural yearly variability. Local citizens living in the areas under surveillance were involved in the process of surveying for mosquitoes through door-to-door questioning, explanations of surveillance, and distribution of literature, including information about mosquito breeding. At town meetings, the options for management of mosquito population levels, such as larvaciding and adulticiding with Permethrin, were presented.

[FIGURE 1 OMITTED]

One important aspect of mosquito control that is new to most small-town local citizens and necessary for conducting efficient and effective mosquito control is the need for long-term data on mosquito species and breeding locations in a local area. There are about 200 different species of mosquitoes in the United States and nearly 60 in North Carolina (North Carolina Department of Environment and Natural Resources [NCDER], 2006). Of these, Culex and Aedes species appear to be the most common mosquito vectors (Godsey et al., 2005). Culex species are the predominant vectors of WNV. Of the 60 species found in North Carolina, the Asian tiger mosquito (Aedes [Stegomyia] albopictus) is the mosquito that generates the most complaints in urban areas. It is a potential vector of dengue fever, a disease currently not found in North Carolina. This mosquito breeds readily in any container that holds water for at least a week. Populations can rise dramatically after rainy weather (North Carolina Department of Health and Human Services, 2006). In light of the current sweep of West Nile virus across the United States, one of the main public health reasons for surveying and identifying mosquito species is to determine when and where there is elevated risk from WNV- or EEE-transmitting mosquitoes, so that people can be alerted and control measures can be increased before a human case is detected (Kilpatrick et al., 2005; Shaman & Day, 2005). The Culex species is a WNV vector, and EEE involves several species of mosquito vectors, including Culiseta melanura, a bird-biting mosquito.

The town of Simpson resembles a small suburban habitat. In suburban Minnesota, Aedes vexans and Coquillettidia perturbans, both WNV-transmitting species, are the dominant species found in surveys, according to Jim Stark from the Metropolitan Mosquito Control District in Minneapolis. In the Midwest, Aedes vexans, a floodwater variety, accounts for 80 percent of all infestations (Kronenwetter-Koepel, Meece, & Miller, 2005). Culex pipiens, found in many other suburban locations, is known as a stagnant-water variety and in many areas is one of the most dangerous species because it is a vector for many diseases, including WNV (Molaei, Andreadis, Armstrong, Anderson, & Vossbrinck, 2006). Other problematic Culex species that are present in North Carolina include Cx. restuans, Cx. salinarius, and Cx. quinquefasciatus. If these species or other vectors are found in a jurisdictional area survey, it is important to inform citizens of the potential disease risk. The only way to be aware of the species present in a local area is to do biological surveillance throughout each mosquito season. Unfortunately, disease risk values are not easily correlated to specific mosquito population numbers (Shaman & Day, 2005); therefore, most mosquito control agencies manage mosquitoes to a given nuisance tolerance level. When WNV, EEE, or other mosquitoborne diseases have been detected in a town or county through surveys of dead birds, animal infections, or testing of collected mosquitoes, mosquito control is typically increased in the entire region. Work is progressing in the modeling of habitat and mosquito populations with the object of more closely tying control procedures to disease risk (Shaman & Day, 2005).

Methods

The town of Simpson is located in Pitt County, North Carolina, and was one of the many small towns in Eastern North Carolina that were flooded by Hurricane Floyd in 1999. Since weather patterns change the habitat landscape each year and construction of new homes continually changes the physical landscape, continual surveillance was needed--but was not affordable. A base map, with current breeding sites, was a necessary starting place for the investigation. New flood zone maps were made available in Pitt County following flooding from Hurricane Fran in 1999. Flood zone maps do not show where wetlands or flooded areas are, but they do map potential flood zones by soil and vegetation types. They can be helpful in locating breeding habitat, but they do not cover every site, nor are all mapped flood zones breeding sites.

The initial survey plan for Year 1 of the study could also be used as a model for any local jurisdiction needing to determine mosquito-breeding levels and possible elevated risk for vectorborne disease. The plan consisted of deploying a trained team to investigate every neighborhood and the surrounding fields and woodlands for possible mosquito habitat. Since the first year of the study was 2000, after the flooding of Hurricane Fran in Eastern North Carolina, federal money was available for training and employing students from East Carolina University in mosquito surveys for the Hurricane Floyd local recovery effort. The standard method for larval surveys is to use a 1-pint mosquito dipper (a plastic cup on a 5-inch dowel) to sample water from every water-holding location found. Five to 10 dips were taken at water-holding sites. Sites that were positive for mosquito larvae over one entire year were marked on local county maps. Larval species were not identified, and sites at which first-instar larvae were found were not recorded as positive sites since it was not known at that stage whether the larvae would survive to third instar and adulthood.

[FIGURE 2 OMITTED]

A second step in the areawide surveillance was to trap adult mosquitoes. Standard Centers for Disease Control and Prevention (CDC) traps with dry ice providing C[O.sub.2] attractant were set at six locations, with a focus on protected vegetated areas near known breeding sites. These traps were set weekly during the second half of the first season (2000). Two locations were near the planned wetland. Mosquitoes from the traps were sorted and identified by species, frozen, and stored for species confirmation and photography. Total numbers of mosquitoes per collection date and total numbers of each species of mosquito per date were recorded.

Ideally, weekly surveillance should be conducted in a wide area each year (Harne, 2004). In this study, however, limitations in resources necessitated that the focus be on only the proposed wetland site and one adult-trapping method. During the second, third, and fourth seasons, mosquitoes were collected from the wetland site and from other sites when possible. Construction of the wetland occurred in early 2002. For part of the construction year and for one year following construction, standard adult CDC mosquito traps with C[O.sub.2] bait were set at the north and south ends of the new wetland (Figure 1). The two primary wetland mosquito collection sites were at the furthest downstream end of the wetland (north) and at the furthest upstream section of the wetland (south). Collections were made at these two sites each week, adult mosquitoes were sorted out, and the mosquito species were identified. General Pitt County rainfall data were also collected for the four years of the study.

Wetland construction consisted of excavating and expanding the creek bed to cover a wider flood zone in a small portion of the Mill Creek Branch leading to the Tar River. Vegetation was installed by a volunteer group from the town and the university as part of community involvement. A floodgate was installed at the culvert under Jethro Mills Road. Water levels in the new wetland could be adjusted with risers according to water flow in the creek. The department of transportation assisted in the construction by replacing the culvert and adding riprap along the ditches near the culverts.

Statistical analysis of mosquito count data for the wetland trap sites was carried out by a statistician in Allied Health Sciences at East Carolina University. The Statistical Package for the Social Sciences (SPSS) Version 12.0 was used for data management and analysis. Descriptive measures and graphs (box plots, histograms, and scatter plots) were used to examine the distributions of all variables and to identify outliers, or anomalous or influential values. The analysis proceeded from a crude analysis of unadjusted preconstruction and postconstruction mosquito count means to several linear regressions adjusting the counts for month of the year (a surrogate for temperature and day length) and rainfall measured in inches. The residuals from these linear regressions were examined to determine the adequacy of the modeling strategies. Results from the two periods, preconstruction and postconstruction, were calculated as the mean, or mean difference in mosquito counts [+ or -] standard deviation, with 95 percent confidence intervals.

[FIGURE 3 OMITTED]

Results

Mosquito Population Characteristics and Rainfall

Figure 2 shows species of mosquitoes collected in the upstream and downstream traps near the constructed-wetland site for the years 2000 to 2003 with rainfall amounts included. Construction of the wetland was completed in early spring 2002.

Rainfall and temperature are typically the most influential environmental factors in mosquito population fluctuations (Anderson et al., 2000; Brower, 2001). In Figure 2, the rainfall columns following each monthly mosquito collection column indicate relative rainfall amounts. Two examples of rainfall influence on mosquito populations are the relatively high rainfall amounts and high mosquito counts in the last two monthly samples in 2002, when there was hurricane activity in coastal North Carolina. Numbers like these should trigger control measures regardless of species. Recent thinking in mosquito control is that nuisance = disease threat, because of the large number of species capable of transmitting diseases such as WNV and because of the threat of new or re-emerging diseases. In North Carolina, hurricane season starts in July, and typically mosquito populations build up during the latter part of the summer to peaks in September, when many storm-driven rain events occur. By the end of October, cool weather slows the activity of mosquitoes, and mosquito reproduction slows dramatically.

In raw numbers, 5,426 mosquitoes were collected in 2002 and 2003, compared with 8,631 in 2000 and 2001, in the two wetland sites combined.

Statistical Analysis of Time and Rainfall on Mosquito Counts

Essentially there is no difference between the preconstruction and postconstruction periods when the preconstruction period is defined as 2000-2001 and the postconstruction period as 2002-2003. These results hold for the crude analysis based on the two-sample t-test and the Mann-Whitney-Wilcoxon test (with and without outliers, or counts over 3,000). The adjusted analysis using a linear regression including rainfall, month of the year, and period also shows no significant difference between preconstruction period and the postconstruction period.

The mosquito count is not normally distributed, as is easily noted through graphs or formal tests of normality. The natural log of the counts, however, appears to be normal, and graphs and tests do not indicate that the log count is nonnormal. In addition, no outliers are indicated in the graphs for the log-transformed counts.

An analysis based on the natural log of the counts essentially shows the same results as for the counts--no significant differences between periods.

Crude Analysis

When mean counts of mosquitoes between periods are compared with no adjustment for rainfall or month of the year, t-tests provide the same results: The t-test value was 1.099, based on 22 degrees of freedom (p = .284).

The 95 percent confidence interval for the change in mosquito count was [-388.3, 1,263.6]. This interval can be interpreted as showing that the data support anywhere from 388 fewer mosquitoes to upward of 1,264 more on average after construction. There was no significant difference in variability between the two periods (Levene's test for equality of variances p-value = .588). Removal of the outliers does affect these results slightly; the t-test comparing counts of mosquitoes was significant, but the nonparametric tests were not. The t-test value for a student's t based on equal variances was t = 2.113, with 20 degrees of freedom, p = .047. The 95 percent confidence interval for the change in mosquito count was [4.9, 756.9]. This interval supports anywhere from an additional 5 mosquitoes to 757 more postconstruction. The results of Levene's test for the equality of variances for the two periods was also significant, however (F = 6.79, p = .017). If the t-test for unequal variances is employed, then t = 2.002, with 13.08 degrees of freedom and p-value = .066, not quite statistically significant at the .05 level. Given the considerations of normality required for the t-test, the nonparametric tests would be considered more valid. Additional analysis based on the natural log of mosquito counts showed that this transformation led to a variable that met the assumptions for normality. The t-test for the transformed variable was t = 1.522, with 22 degrees of freedom, p = .142. Levene's test was also not significant, p-value = .570. Nonparametric test results would be the same as those above for the transformed variable. Box plots for the log-transformed variable also did not indicate any outliers, as compared to the nontransformed variable (Figure 3 and Figure 4). The transformed plots show that the two distributions are quite similar for the two periods, with slightly greater counts in the post-reclamation period, but not statistically significantly so, as discussed above.

Adjusted Analysis

In a model summary of data with outliers of over 3,000 included, the [R.sup.2] in regression analysis was .102. None of the variables was noted to be significant. Removing the two outliers above 3,000 resulted in the model summaries A and B (untransformed and transformed) (Table 1, Table 2).

The influence of the two outliers can be noted in the results shown in Table 1 and Table 2. Month of the year is significant, and period, adjusted for rainfall and month of the year, is nearly so (p = .070). Even with the outliers removed, the untransformed mosquito counts are not normally distributed. Since the overall sample size is small (n = 23 with outliers and n = 21 without), the inference drawn from the above regressions may be questioned. Running the regression on the natural log-transformed counts (which, from a statistical perspective, cannot be said to be nonnormal) gives results indicating that there is not a statistically significant increase in mosquito counts postconstruction after rainfall and month of the year are adjusted for.

In summary, it appears that there was some increase in the average count of mosquitoes, as well as in variability, from preconstruction to postconstruction. The increase was not, however, statistically significant.

Discussion

Although there were no statistically significant differences in total population numbers after wetland construction, the data on mosquito species collected did yield significant findings, including information on the presence of potential mosquito vectors of WNV in the town and surrounding area.

[FIGURE 4 OMITTED]

Over the past five years of WNV transmission in the United States, many species of mosquitoes have been found capable of transmitting the West Nile virus. The earliest species found positive for WNV were Aedes vexans, Culex pipiens, Culex restuans, and Culex salinarius (Godsey et al., 2005). Recent research shows that these species are also the ones most likely to be transmitting WNV in future years, because of their ubiquity. Local management groups can use similar research information to direct intervention and control.

In the study reported here, for example, seven genera and 24 species of mosquitoes were collected in 2000 and 2001. In 2002 and 2003, six genera and 28 species of mosquitoes were collected. All of the currently listed common WNV vectors (Ae. vexans, Cx. pipiens, Cx. restuans, and Cx. salinarius) were commonly found in these collections (CDC, 2005). In addition, of the 20 or more species of mosquitoes in North Carolina capable of WNV transmission, six were collected in Simpson.

Mosquito census data show five species groups as percentages of the total monthly collections in Simpson in 2003 (Figure 5). Aedes vexans and Culex restuans were the most numerous species collected in 2003. Both of these species are vectors of WNV. Ae. vexans was present in higher numbers in the spring and the fall; Cx. restuans peaked in July and August. Since Ae. vexans is a floodwater species, it can be assumed that this species responded to spring and fall rainfall and consequent flooding activities in the Mill Creek tributary that flows through Simpson.

In 2001 and 2002, when six sites were sampled with CDC light traps, Ae. vexans was also found in the non-wetland locations and in high percentages in all but two sites. It is therefore not only the wetland habitat in which this species is breeding.

In 2001, fewer mosquitoes were collected overall, but Ae. vexans or Culex species were found in each of the five collection sites.

Local survey work can inform mosquito management agencies about the amounts and timing of vector mosquito species in their locales, thus providing information for potential disease risk reduction strategies. Survey information can aid in decision making to direct mosquito control to critical times and places for prevention of disease transmission, and can save funds from unnecessary weekly spraying. A data-based response to vector-borne disease health problems is as valid and important in mosquito management as it is in all other health areas.

Although rainfall patterns in North Carolina vary from year to year, flooding rains are most often seen in the fall during hurricane activity and often occur in the spring following storms (Pope et al., 2001). Further investigation into the resulting flooding patterns of the tributary creek in the Millcreek subdivision would also be useful in helping to abate the mosquito population in Simpson.

Since each of the other light trap collection sites yielded a variety of mosquitoes similar to those found in the wetland area, it can also be concluded that mosquito breeding is widespread in this town. Many breeding sites were small and could easily be altered, especially those on private property. Drainage ditches in many places needed minor cleaning and grading to prevent ponding of water. Farmland contiguous with housing developments had small ponded sections that bred mosquitoes. Filling or providing drainage for these depressions would improve crop yield and cut down on mosquito production.

No manmade alteration in the environment, such as a constructed wetland, can escape having both beneficial and harmful consequences, so community managers and planners need to be aware of all the costs (including health costs) and benefits of each project that is proposed. When decisions about wetland-construction projects are based on the analysis of data, such as the mosquito surveillance data reported in this paper, they can give more rational and predictable outcomes to community changes. In addition, a study of mosquito numbers for two years after a wetland construction project may not be adequate to assess changes in insect populations. Each wetland construction situation has different ecological parameters that can affect results (Sarneckis, 2002).

Conclusion

The construction of a wetland in the town of Simpson, North Carolina, did not significantly increase the population of mosquitoes collected (Figure 2). In four years of comparative study, more total mosquitoes were collected after a wetland construction project than before. Through regression analysis, however, it was shown that this difference was not significant. Month of the year (a surrogate for temperature and day length) was a significant variable, explaining the variability of counts. The wetland was small, and mosquito breeding was widespread throughout the town and surrounding area, according to the authors' initial survey. For this reason, the authors concluded that under this ecological regime, increased mosquito populations are not a significant side effect of a small wetland construction project such as the one discussed in this paper.

As seen in the data on mosquito species and abundance, the composition and numbers of mosquitoes changed over the years of the study. Total numbers of mosquitoes were quite variable during the study, but were not significantly higher in the years after wetland construction. A thorough, longer-term study of actual mosquito breeding and production in a small wetland, with data obtained through mosquito emergence traps, would provide needed further information on mosquito breeding in suburban constructed wetlands in North Carolina.

Further work on the succession of mosquito species and their predators inhabiting constructed wetlands would also aid in the development of more specific guidelines for building new wetlands. In an unpublished survey done in 1979 in Arkansas rice fields, one of the authors found that mosquitoes and aquatic insect-predator species do not follow a set sequential succession, but, as in unstable communities in many disturbed habitats, reach an assemblage of species depending on surrounding "feeder" habitats. Then succession is again disrupted by drought or reflooding. The concept of "balance" of predators and prey involves averaging population fluctuations over many years, and balance is not readily measurable or achievable in the creation of wetlands. The concept is not useful for public defense of wetlands, especially since, in an unpredictable weather situation, disease can be transmitted in a single week's brood of mosquitoes. The timing of the natural or human introduction of predatory fish and predatory insects is a critical factor not yet included in a total systems approach to designing and managing newly constructed wetlands, although predator presence is often claimed as a natural means of mosquito prevention. The complete story on the succession of mosquitoes and mosquito predator life in these wetlands has still not been told, and research in this area would help in the creation of more useful, sustainable, healthy wetlands for water quality improvement and for disease prevention.

[FIGURE 5 OMITTED]

Acknowledgements: This work was funded in part by the North Carolina Resource Conservation and Development Organization, through a grant from the North Carolina Clean Water Management Trust Fund.

Corresponding Author: Alice L. Anderson, Assistant Professor, East Carolina University, Environmental Health Sciences & Safety, Belk AH Building, Room 310, Greenville, NC 27858. E-mail: [e-mail].

REFERENCES

Anderson, A.L., Engber, B., Harrison, B., Brown, J., Andrews, J., Rueda, L.M., Whitt, P., & Newton, N. (2000, Summer). Emergency aerial spraying in North Carolina after Hurricane Floyd, 1999. Wing Beats, 11(2), 4-5.

Brower, V. (2001). Vector-borne diseases and global warming: Are both on an upward swing? EMBO Reports, 2, 755-757.

Centers for Disease Control and Prevention. (2005). Mosquito species producing WNV positives by year [Table]. Retrieved September 6, 2005, from http://www.cdc.gov/ncidod/dvbid/westnile/mosquitoSpecies.htm.

City of Phoenix. (1999). Tres Rios Research. Retrieved March 29, 2006, from http://phoenix.gov/TRESRIOS/research.html.

Evans, D.L. (1996). Aquatic macroinvertebrate communities of constructed and natural freshwater marshes in central Florida. Unpublished doctoral dissertation, University of Florida, Gainsville.

Godsey, M.S., Jr., Nasci, R., Savage, H.M., Aspen, S., King, R., Powers, A.M., Burkhalter, K., Colton, L., Charnetzky, D., Lasater, S., Taylor, V, & Palmisano, C.T. (2005, September). West Nile virus-infected mosquitoes, Louisiana, 2002. Emerging Infectious Diseases, 11(9). Retrieved September 30, 2005, from http://www.cdc.gov/ncidod/EID/volllno09/04-0443.htm.

Harne, E. (2004). 43rd annual report of the Charleston County mosquito control program, 2004. Charleston, SC: City of Charleston.

Kilpatrick, A. M., Kramer, L.D., Campbell, S.R., Alleyne, E.O., Dobson, A.P., & Daszak, P. (2005). West Nile virus risk assessment and the bridge vector paradigm. Emerging Infectious Diseases, 11(3), 425-429.

Kronenwetter-Koepel, T., Meece, J.K., & Miller, C.A. (2005). Surveillance of above- and below-ground mosquito breeding habitats in a rural Midwestern community: Baseline data for larvicidal control measures against West Nile virus vectors. Clinical Medicine and Research, 3(1), 3-12.

Molaei, G., Andreadis, T.G., Armstrong, P.M., Anderson, J.F., & Vossbrinck, C.R. (2006). Host feeding patterns of Culex mosquitoes and West Nile transmission, northeastern United States. Emerging Infectious Diseases, 12(3). Retrieved April 7, 2006, from http://www.cdc.gov/ncidod/EID/voll2no03/pdfs/05-1004.pdf.

North Carolina Department of Environment and Natural Resources. (2006). Public health pest management [Web site]. Retrieved September 30, 2005, from http://www.deh.enr.state.nc.us/phpm/in-dex.html.

North Carolina Department of Health and Human Services. (2006). State to spray Bladen, Edgecombe, Nash, Pender, & Pitt counties for mosquitoes [press release]. Retrieved September 30, 2005, from http://www.dhhs.state.nc.us/pressrel/10-6-99a.htm.

Pope, B.E, Tasker, G.D., & Robbins, J.C. (2001). Estimating the magnitude and frequency of floods in rural basins of North Carolina--Revised (USGS Water-Resources Investigations Report No. 01-4207). Retrieved November 28, 2005, from http://nc.water.usgs.gov/reports/abstracts/wri014207.html.

Reid, M. (2003). Population dynamics of mosquitoes at the Clark County Wetland Park Nature Preserve: A constructed wetland in Las Vegas, Nevada, 2003. Unpublished master's thesis, University of Nevada, Las Vegas.

Sarneckis, K. (2002). Mosquitoes in constructed wetlands. Adelaide, South Australia: Government of South Australia, Environmental Protection Agency. Retrieved April 7, 2006, from http://www.environment.sa.gov.au/epa/pdfs/mosquitoes.pdf.

Shafer, M.L., Lundstrum, J.O., Pfeiffer, M., Lundkvist, E., & Lan-din, J. (2003). Biological diversity versus risk for mosquito nuisance and disease transmission in constructed wetlands [Abstract]. Proceedings of the 2003 Zoonotic Ecology and Epidemiology (CRZEE) 2003 Meeting. Accessed March 29, 2006, from http://www.zooeco.org/zooeco/soczee/meetings/CRZEE/abstracts.asp?auth=schafer_martina.

Shaman, J., & Day, J.F. (2005, September). Achieving operational hydrologic monitoring of mosquitoborne disease. Emerging Infectious Diseases, 11(9). Retrieved October 5, 2005, from http://www.cdc.gov/ncidod/EID/volllno09/05-0340.htm.

U.S. Geological Survey. (2005). Disease maps. Retrieved November 24, 2006, from http://diseasemaps.usgs.gov/index.html.

Alice L. Anderson, Ph.D.

Kevin O'Brien, Ph.D.

Megan Hartwell, M.S.E.H.

TABLE 1 Adjusted Regression Analysis for Untransformed and Transformed
Data

                                                             Standard
                                                             Error of
                                                             the
Model               R         [R.sup.2]  Adjusted [R.sup.2]  Estimate

A. Untransformed    .697 (a)  .485       .399                352.133
B. Transformed (b)  .519 (a)  .269       .160                  1.632

(a) Predictors: constant Period 2, month, rainfall.
(b) Dependent variable: log mosquito count.

TABLE 2 Coefficients for Adjusted Regression Analysis for Untransformed
and Transformed Data

                       Unstandardized
Model Coefficients     Coefficients        Standardized Coefficients

                              Standard
A. Untransformed    Beta      Error     Beta   T       Significance

  1 (constant)      -506.745  238.911          -2.120   .048
  Rainfall             2.223    1.551   0.263   1.433   .169
  Month               77.731   31.635   0.430   2.457   .024
  Period 2           305.551  158.374   0.343   1.929  0.070

B. Transformed

  1 (constant)         2.572    1.062           2.422   .025
  Rainfall             0.005    0.007   0.136   0.641   .529
  Month                0.226    0.145   0.365   1.830   .082
  Period 2             0.831    0.714   0.238   1.630   .258