• View in gallery

    a, Scatter plot of water table depth (WTD) one-half month prior versus WTD four months prior at all 42 sites in southcentral Florida, 1990–1998. Blue circles denote county-months with no human cases of St. Louis encephalitis (SLE) and red circles denote county-months with one or more cases of SLE. More than one site may exist in a county. b, Same scatter plot as in Figure 1a. Red circles now randomly selected from among the meteorologic stations.

  • View in gallery

    Maps of modeled water table depth (WTD) for May 1990 and June 16–30, 1990 at all 42 stations in the study area. Location of the July 1990 human cases of St. Louis encephalitis (SLE) is also shown.

  • View in gallery

    Map of modeled water table depth (WTD) for July 16–31, 1990 at all 42 station sites in the study area. Location of the August 1990 human cases of St. Louis encephalitis (SLE) is also shown.

  • View in gallery

    Map of modeled water table depth (WTD) for August 16–31, 1990 at all 42 station sites in the study area. Location of the September 1990 human cases of St. Louis encephalitis (SLE) is also shown.

  • View in gallery

    Maps of modeled water table depth (WTD) for May 1993 and August 16–31, 1993 at all 42 station sites in the study area. Location of the September 1993 human cases of St. Louis encephalitis (SLE) is also shown.

  • View in gallery

    Maps of modeled water table depth (WTD) for May 1997 and August 16–31, 1997 at all 42 station sites in the study area. Location of the September 1997 human cases of St. Louis encephalitis (SLE) is also shown.

  • 1

    Day JF, Curtis GA, 1999. Blood feeding and oviposition by Culex nigripalpus (Diptera: Culicidae) before, during, and after a widespread St. Louis encephalitis virus epidemic in Florida. J Med Entomol 36 :176–181.

    • Search Google Scholar
    • Export Citation
  • 2

    Shaman J, Day J, Stieglitz M, 2002. Drought-induced amplification of St. Louis encephilitis virus, Florida. Emer Infect Dis 8 :575–580.

  • 3

    Stieglitz M, Rind D, Famiglietti J, Rosenzweig C, 1997. An efficient approach to modeling the topographic control of surface hydrology for regional and global climate modeling. J Climate 10 :118–137.

    • Search Google Scholar
    • Export Citation
  • 4

    Day JF, Stark LM, 2000. Frequency of Saint Louis encephalitis virus in humans from Florida, USA: 1990–1999. J Med Entomol 37 :626–633.

    • Search Google Scholar
    • Export Citation
  • 5

    Chamberlain RW, Sudia WD, Coleman PH, Beadle LD, 1964. Vector studies in the St. Louis encephalitis epidemic, Tampa Bay area, Florida, 1962. Am J Trop Med Hyg 13 :456–461.

    • Search Google Scholar
    • Export Citation
  • 6

    Dow RP, Coleman PH, Meadows KE, Work TH, 1964. Isolation of St. Louis encephalitis viruses from mosquitoes in the Tampa Bay area of Florida during the 1962 epidemic. Am J Trop Med Hyg 13 :462–474.

    • Search Google Scholar
    • Export Citation
  • 7

    Shroyer DA, 1991. The 1990 Florida epidemic of St. Louis encephalitis: virus infection rates in Culex nigripalpus. J Fla Mosq Control Assoc 62 :69–71.

    • Search Google Scholar
    • Export Citation
  • 8

    Day JF, Curtis GA, 1993. Annual emergence patterns of Culex nigripalpus females before, during and after a widespread St. Louis encephalitis epidemic in south Florida. J Am Mosq Cont Assoc 9 :249–253.

    • Search Google Scholar
    • Export Citation
  • 9

    Shaman J, Day J, Stieglitz M, 2003. St. Louis encephalitis virus in wild birds during the 1990 south Florida epidemic: the importance of drought, wetting conditions, and the emergence of Culex nigripalpus to arboviral amplification and transmission. J Med Entomol 40 :547–554.

    • Search Google Scholar
    • Export Citation
  • 10

    Beven KJ, Kirkby MJ, 1979. A physically based, variable contributing area model of basin hydrology. Hydrol Sci 24 :43–69.

  • 11

    Ducharne A, Koster RD, Suarez MJ, Stieglitz M, Kumar P, 2000. A catchment-based approach to modeling land surface processes. Part II: Parameter estimation and model demonstration. J Geophys Res 105 :24823–24838.

    • Search Google Scholar
    • Export Citation
  • 12

    Shaman J, Stieglitz M, Engel V, Koster R, Stark C, 2002. Representation of storm flow and a more responsive water table in a TOPMODEL-based hydrology model. Water Resources Res 38 : Art No. 1156.

    • Search Google Scholar
    • Export Citation
  • 13

    Shaman J, Stieglitz M, Zebiak S, Cane M, 2003. A local forecast of land surface wetness conditions derived from seasonal climate predictions. J Hydrometeorol 4 :611–626.

    • Search Google Scholar
    • Export Citation
  • 14

    DeGaetano A, Eggleston K, Knapp W, 1995. Daily solar radiation estimates for the northeastern United States using the northeast regional climate center and National Renewable Energy Laboratory models. Solar Energy 55 :185–194.

    • Search Google Scholar
    • Export Citation
  • 15

    Dow RP, Gerrish GM, 1970. Day-to-day change in relative humidity and the activity of Culex nigripalpus (Diptera: Culicidae). Ann Entomol Soc Am 63 :995–999.

    • Search Google Scholar
    • Export Citation
  • 16

    Day JF, Curtis GA, 1994. When it rains they soar–and that makes Culex nigripalpus a dangerous mosquito. Am Entomologist 40 :162–167.

 

 

 

 

 

THE SPATIAL-TEMPORAL DISTRIBUTION OF DROUGHT, WETTING, AND HUMAN CASES OF ST. LOUIS ENCEPHALITIS IN SOUTHCENTRAL FLORIDA

View More View Less
  • 1 Department of Earth and Planetary Sciences, Harvard University, Cambridge, Massachusetts; Florida Medical Entomology Laboratory, Institute of Food and Agricultural Sciences, University of Florida, Vero Beach, Florida; Lamont-Doherty Earth Observatory, Columbia University, Palisades, New York

Using a dynamic hydrology model, we simulated land surface wetness conditions at 42 sites in 28 counties in southcentral Florida from 1990 to 1998 and compared these simulations with the incidence of human cases of St. Louis encephalitis (SLE) within these counties. Within counties, drought four months prior and wetting one-half month prior were significantly associated with human cases of SLE. Simulated land surface wetness conditions resolved transmission loci in both space and time, and May drought was significantly associated with the subsequent occurrence of human SLE cases. These findings are consistent with previous results associating simulated land surface wetness conditions with the transmission of SLE virus as measured in sentinel chickens, and support our working hypothesis that springtime drought facilitates SLE virus amplification in mosquito and wild bird populations.

INTRODUCTION

Within Florida, four general patterns of transmission of St. Louis encephalitis (SLE) virus to humans have been observed since 1952.1 During the vast majority of years, no human cases have been reported. During some years, such as 1993, small, focal outbreaks, usually involving fewer than 10 human cases, were reported in various locations throughout the state. The 1993 SLE outbreak involved eight human cases from a small area on the border of Lee and Collier Counties. Another transmission pattern involves sporadic human cases reported over a wider geographic area. This type of transmission was most recently observed in 1997 when nine human SLE cases were reported from Brevard, Charlotte, Hillsborough, Lee, Palm Beach, and Polk Counties. Sporadic transmission of West Nile virus to humans was also reported throughout Florida in 2001 and 2002. The final transmission pattern is of greatest concern: a full-blown epidemic in which hundreds or thousands of cases, some appearing in well-defined clusters, are reported over large parts of south Florida. The most recent Florida SLE epidemic occurred in 1990 when 226 clinical cases and as many as 30,000 infections were reported throughout southcentral Florida. Indian River County was the epicenter of this outbreak.1

Recently, we found an association between antecedent drought, coincident wetting, and transmission of SLE virus in Indian River County Florida.2 We used a dynamic hydrology model3 to hindcast mean area water table depth (WTD) in Indian River County, and compared this model simulation to sentinel chicken seroconversion data. Seroconversion of sentinel chickens, as measured by serum assay for hemagglutination inhibition (HI) antibodies to SLE virus, has been strongly correlated with the clinical disease in humans.4 Using logistic regression, we found the probability of sentinel chicken seroconversion, i.e., transmission of SLE virus, to be strongly associated with low WTDs 11–17 weeks prior and higher WTDs 0–2 weeks prior.2

A mechanism for this empirical relationship was suggested by mosquito collection data obtained in Indian River County from 1986 to 1991. Culex nigripalpus Theobald is the demonstrated enzootic and epidemic vector of SLE virus in south Florida.5–7 Collections of Cx. nigripalpus were made in densely vegetated “hammock” (an island of dense native vegetation) habitats used by this species for daytime resting.8 During the driest conditions (low modeled WTD) preceding heavy SLE virus transmission, Cx. nigripalpus collections dramatically increased.2 Rather than indicate an increase of mosquito abundance, these data suggest that drought restricts Cx. nigripalpus activity to the more humid hammock habitats. Extreme drought periods in south Florida tend to occur during the spring, a time when nesting wild birds also make use of the hammocks. Thus, drought drives the mosquitoes and birds into contact with one another. This forced interaction of vector mosquitoes and susceptible avian amplification hosts provides an ideal environment for the rapid epizootic amplification of SLE virus. In addition, confinement of blood fed and gravid Cx. nigripalpus females to the hammock habitats for extended periods forces more infected females to complete the extrinsic incubation of acquired arboviruses during a single gonotrophic cycle.1,8 Subsequently, when the drought ends and water resources increase, infected mosquitoes and birds disperse and carry the virus from the hammocks. Gravid female mosquitoes oviposit, re-feed, and if infective, they initiate the early transmission phase of the Florida SLE cycle.

Additional supporting evidence was found from wild bird serologic data obtained during the 1990 SLE epidemic. We showed that the presence of HI antibodies to SLE virus in wild birds was significantly associated with modeled antecedent drought, coincident wetting, and the emergence of the Florida SLE virus vector, Cx. nigripalpus, five weeks prior.9 Our findings suggest that three factors conspired to create the 1990 epidemic: 1) a large population of susceptible wild birds; 2) severe springtime drought, which facilitated amplification of the SLE virus among the Cx. nigripalpus and a portion of the wild bird population; 3) continued rainfall and wetting of the land surface in the summer and early fall, which sustained a large, active Cx. nigripalpus population. The continued biting and reproductive activity of Cx. nigripalpus maintained levels of epizootic transmission throughout the summer and early fall in Indian River County. These constant high levels of SLE virus amplification facilitated spillover transmission to humans throughout much of southcentral Florida.

In this study, we shift our analysis to human cases of SLE reported throughout southcentral Florida. We hypothesize that the spatial-temporal organization of human SLE cases in Florida is tied to hydrologic variability. Specifically, we propose that the epidemic transmission with Indian River County as the epicenter in 1990, the focal transmission with Lee and Collier Counties as the focus in 1993, and the sporadic transmission in 1997, correspond to the spatial-temporal patterns of drought and wetting during these years. We expect that differences in the number and location of human cases of SLE among years reflect differences in land surface wetness conditions across southcentral Florida. We further hypothesize that foci of drought and wetting should be demonstrable within or near Indian River County during 1990, at the Lee/Collier County border in 1993, and scattered among several counties during 1997. To support our model of SLE virus amplification and transmission, these hydrologic epicenters should be manifest as spring droughts followed by wetting concurrent or prior to that of surrounding areas.

Finally, we hypothesize that antecedent drought is a necessary, although not sufficient, condition for SLE virus amplification and transmission in Florida. Thus, all counties recording human SLE cases should include or be near localities that experienced drought prior to the onset of human infections. To test these hypotheses we acquired 1988–1998 meteorologic data from available meteorologic stations in all Florida counties lying between 26.5°N and 29°N. These data were used to force a dynamic hydrology model and retrospectively simulate surface wetness conditions in the area of each station.

METHODS AND MATERIALS

The topographically based hydrology model.

Hydrologic modeling follows the methods set forth by Shaman and others.2 We use a dynamic hydrology model,3 here referred to as the topographically based hydrology (TBH) model, to simulate variations in WTD at each station site. Mean area WTD provides an integrated measure of near surface soil wetness. It is the rise and fall of the water table that determines where and when pools of water form at the land surface, thus creating potential mosquito breeding habitats. To model WTD, a suite of meteorologic variables, including precipitation and temperature, area soil and vegetation type, and antecedent conditions, must be accounted for so that evapotranspiration, water movement within the soil column, and river runoff can be quantified. Topography must also be constrained if the flow of water across the land surface, runoff rates, and the local convergence of water in lowlands (surface pooling) are to be modeled accurately. By using the TBH model we are able to track these variables and simulate variations in WTD.

The TBH model combines a soil column, which simulates the vertical movement of water and heat within the soil and between the soil surface, vegetation and the atmosphere, with the TOPMODEL approach,10 which incorporates the statistics of topography to track the horizontal movement of shallow groundwater from the uplands to the lowlands. The TOPMODEL formulations permit calculation of both the saturated fraction within the watershed (partial contributing area), and the groundwater flow that supports this area, from knowledge of the mean WTD and a probability density function for soil moisture deficit derived from topographic statistics. This approach to modeling the land surface has been validated at several catchments, ranging in scale from the Red Arkansas Basin (570,000 km2)11 to the Black Rock Forest catchment (1.34 km2).12

Hourly meteorologic data.

Seven hourly meteorologic variables, precipitation, temperature, surface pressure, wind speed, mixing ratio, incident solar radiation, and downwelling longwave radiation, are needed to force the TBH model. Because hourly data is of limited availability we use daily precipitation and temperature data and a resampling procedure to generate the hourly forcing data sets. This resampling procedure is performed to accommodate the TBH model interface; in so doing, daily variability is maintained in its entirety. See Shaman and others13 for a more complete explanation of this procedure.

The hourly meteorologic data used for the resampling procedure were assembled from National Climate Data Center (NCDC) archives for Vero Beach, Florida. Gaps in the hourly Vero Beach record were filled with hourly data from NCDC archives for Melbourne and West Palm Beach, Florida. In addition, hourly meteorologic data were assembled from the NCDC Solar and Meteorological Surface Observation Network (SAMSON) dataset for Daytona Beach, Miami, Tampa, and West Palm Beach. For the Vero Beach hourly record, solar radiation data were provided by the Northeast Regional Climate Center (NRCC) from analysis of the NCDC data using the NRCC solar energy model.14 Solar radiation data are included in the SAMSON dataset.

Daily meteorological data.

Daily precipitation and temperature data were assembled from NCDC archives for all stations with near continuous records spanning 1988–1998 for the 28 Florida counties lying partly or wholly between 26.5°N and 29°N. Small gaps in the daily records were filled with data from adjacent stations. A total of 42 records of 1988–1998 daily data were assembled. Table 1 provides a listing of the daily record station sites. Using the closest hourly meteorologic record (Daytona Beach, Miami, Tampa, Vero Beach, or West Palm Beach) and the resampling procedure,13 we then created hourly forcing data sets for each of the 42 station sites.

The TBH model initialization and validation data.

The TBH model was calibrated and validated at the Vero Beach 4W site.2 Using a routing algorithm and subsequent analysis, we generated topographic statistics for the Vero Beach area from a 30-meter cell United States Geological Survey (USGS) National Elevation Dataset Digital Elevation Model of southcentral Florida. Soil and vegetation types were derived from U.S. Department of Agriculture sources and personal inspection of the Vero Beach landscape.

At the Vero Beach 4W site, the TBH model was run from 1949 through 2001, and model output provided a daily time series of mean WTD and river runoff for this site area. Model validation was performed using groundwater well measurements and surface (canal) water levels, provided by the St. John’s Water Management District and USGS sources. Partitioning of runoff and evapotranspiration matched bulk estimates derived from USGS sources.

The TBH model simulations.

Model simulations at all 42 station sites were performed using the calibrations established at the Vero Beach 4W site. Most of southcentral Florida has similar flat topography, similar tropical vegetation, and is subject to channelization and water control. These commonalities permit this regional validation and application of the TBH model to these other areas, many of which lack local validation data.

Mean area WTD, as simulated by the TBH model, is not representative of a measurable quantity (it is the mean for the area) but rather serves as an index of relative wetness for the area through time. Use of the single, regional model calibration permits intercomparison of mean area WTD among the site simulations.

Human case data.

Summaries of mosquito-borne arboviral human case data are compiled, analyzed, and reported weekly, monthly, and annually by the Florida Department of Health (Tallahassee, FL). Table 2 shows the monthly human cases of SLE as reported by county for 1990–1998.

Statistical analysis.

We define human SLE incidence as a dichotomous variable: one, if one or more human SLE cases were recorded within a given county for a given month; and zero, if no cases were recorded. We also defined human SLE prevalence as a polytomous variable with the following groupings: zero cases, one case, 2 cases, 3–4 cases, 5–8 cases, and 9–16 cases. (These boundaries are arbitrary and merely meant to test whether hydrologic variability can distinguish the probability of numbers of human SLE cases.) The WTD was aggregated in both half-monthly and monthly averages. Lag comparisons were made in which the half-monthly average WTD for May 16–31 was considered lagged one-half month with June human SLE incidence or prevalence. Similarly, the monthly average WTD for May 16–June 15 was considered lagged one-half month with June human SLE incidence or intensity.

Bivariate logistic regression analysis was used to associate the probability of countywide human SLE incidence or prevalence with lag combinations of half-monthly modeled WTD at each station site within the county. All counties were analyzed in aggregate. Whole model goodness-of-fit was measured by log-likelihood ratio and the pseudo r2 (uncertainty) coefficient. Individual parameter estimates were made using a maximum likelihood procedure. Wald chi-square tests were used to determine whether these estimates were significantly different from zero.

Additional analyses were performed to determine whether drought or drought and wetting were necessarily associated with human SLE cases. Between 1990 and 1998 and for the counties represented in this study, there were 71 instances in which a county reported one or more monthly cases of human SLE, i.e., human SLE incidence. We hypothesized that antecedent drought should have occurred within or near all of these counties. We therefore calculated the number of these 71 occurrences for which antecedent drought fell below a given mean monthly WTD (e.g., below −1.35 meters, representative of deep drought) somewhere within the county. The antecedence lag length used for this analysis was the best fit of the logistic regression analysis. Many counties have more than one station record, and for this analysis we only required that at least one station meet the antecedent drought criterion. A similar calculation was also performed for both antecedent drought and near-coincident wetting (e.g., a WTD below −1.4 meters, and a near coincident wetting above −1.2 meters).

Bootstrap confidence intervals were then estimated using a Monte Carlo procedure. A total of 10,000 combinations of the 71 county-months were randomly sampled among the years 1990–1998 (1988 and 1989 were excluded to allow for TBH model spin-up), and a distribution of the totals meeting each defined criterion of drought (or drought and wetness) was constructed. For example, because Indian River County recorded human SLE cases in July 1990, a random selection of antecedent WTDs among the years 1990–1998 (only for Indian River County and July) would be made, and whether the drought criterion was met within the county for that randomly selected year would be determined. This random selection would then be repeated for the remaining 70 county-months of human SLE incidence, the total recorded, and the process repeated 10,000 times. The significance of the actual number of antecedent drought instances (or antecedent drought and near-coincident wetting) occurring with human SLE incidence was then assessed based on this distribution of 10,000. The null hypothesis is that the number of counties meeting the drought (or drought and wetness) criterion is no greater than that due to chance.

RESULTS

Regression findings.

A range of time lags (3–6 months for drought, 0–1 month for wetting) of half-monthly mean WTD was significantly associated with human SLE incidence. The best-fit logistic regression model (P < 0.0001) uses drought four months prior and near-coincident wetting one half-month prior and yields the following equation:

P(SLE)=(1+exp(10.40+6.05×WTD40.67×WTD0.5))1

where P(SLE) is the probability of human SLE incidence, WTD4 is the half-monthly mean WTD four months prior, and WTD0.5 is the half-monthly mean WTD one-half month prior. This pattern of drought and wetting associated with human SLE incidence is similar to the pattern of drought and wetting found to be associated with SLE virus transmission to sentinel chickens in Indian River County.2 Using the best-fit model time lags found for the above analysis, we repeated our logistic regression analysis using human SLE prevalence. Drought four months prior and wetting one-half month prior were significantly associated with all levels of human SLE prevalence.

Figure 1a shows a scatter plot of mean WTD four months prior versus mean WTD one-half month prior for all station sites for 1990–1998. The human SLE incidence associated with each plot value is indicated in color: blue if no human SLE cases occurred within the county, and red if one or more human SLE cases occurred within the county. Stations for which there was human SLE incidence within the county are clustered at the driest conditions. As a comparison, Figure 1b shows the same plot but with an equal number of station sites chosen randomly and colored red.

Drought as a necessary prerequisite of human SLE incidence.

We have hypothesized that drought is necessary for SLE virus amplification and the realization of mosquito infection rates sufficient to cause SLE virus spillover into human populations. To test this, we determined the number of instances in which antecedent drought was simulated at one or more meteorologic stations within each county reporting human SLE cases. If drought is indeed necessary, antecedent drought should always occur four months prior to the appearance of human SLE cases. Table 3 shows the results of this analysis for two magnitudes of drought. Also shown in Table 3 is the same calculation, but for both antecedent drought and near-coincident wetting (one-half month prior) at the same stations within each county.

These results again show that both antecedent drought and antecedent drought and near-coincident wetting are significantly associated human SLE incidence. A drought of defined magnitude < −1.35 meters WTD occurred four months prior in 52 of 71 instances (73%) of human SLE incidence. However, if drought is indeed a necessary prerequisite of intense SLE virus amplification and transmission, all human SLE incidence should be preceded by drought in the immediate area of the human SLE cases.

Our previous analyses2 compared human SLE incidence with drought using fixed time lags. However, our working hypothesis, based on the biology of the system, is that drought during spring brings nesting wild birds and Cx. nigripalpus into increased contact facilitating amplification.2 We therefore repeated our analysis, but instead of using drought four months prior to the onset of human SLE cases, we restricted the analysis of drought to the preceding May (Table 4). Our findings are significant and indicate that the onset of human SLE cases was predicated on drought during the previous May. Specifically, a drought of defined magnitude < −1.35 meters WTD occurred in May prior to the onset of human SLE cases in 59 of 71 possible instances (83%).

Maps of land surface wetness.

We have demonstrated that May drought occurred in 83% of Florida counties prior to the report of at least one human SLE case. However, if drought is a necessary prerequisite for SLE virus amplification and the production of mosquito infection rates sufficient to result in human disease, then localized drought should always be evident prior to the onset of human SLE cases. We propose that several factors may account for this incomplete association. 1) The movement of infected mosquitoes, wild birds, and humans might be confounding the spatial link between drought and human SLE cases. Culex nigripalpus populations may acquire infection in one county during a local drought, and then move and transmit SLE virus in another location. Humans are also mobile and may be infected outside their resident county, which serves as the location of infection for reporting purposes. 2) The network of 42 meteorologic stations (in 28 counties) used in our study might be too sparse to provide a comprehensive representation of land surface wetness conditions. Many counties contain only one station for which there exists an adequate meteorologic record with which to model local hydrology. Much of the rainfall in Florida is convective and varies over short spatial scales (< 10 km), and thus, land surface wetness can also vary over such short spatial scales. Consequently, a denser network of meteorologic stations may be needed to capture the full spatial variability of land surface wetness and adequately detect all pockets of drought.

To illustrate the possible effects of both these factors, we present plots of land surface wetness for the three study years in which human SLE cases were reported in south Florida: the 1990 epidemic; the 1993 focal event; and the 1997 sporadic event. Figure 2 shows the May 1990 mean modeled WTD for the 42 station sites, the June 16–30, 1990 mean modeled WTD for the 42 station sites, and the July occurrence of human SLE cases in Florida by county. Indian River County was the only county reporting human SLE cases. Pervasive drought is evident in May throughout southcentral Florida; most of the sites show a mean WTD of −1.3 meters or lower. The Vero Beach station in eastern Indian River County shows a May drought, but with no visible wetting in late June. Western Indian River County has no meteorologic station, yet just over the border at Fort Drum in Okeechobee County, a strong May drought and late June wetting is apparent. This wetting (denoted by the arrow), by virtue of its location, may have facilitated mosquito dispersal into Indian River County. It also suggests that the network of meteorologic stations may be too sparse for comprehensive modeling of land surface wetness (i.e., wetting may have occurred, but was undetected, in western Indian River County).

Figure 3 shows the July 16–31, 1990 mean modeled WTD for the 42 stations, and the August occurrence of human SLE cases by county. Human cases were reported in August by 16 of the 28 counties within the study area. By the end of July many of the stations have a modeled WTD above −1.2 meters. Figure 4 shows the August 16–31, 1990 mean modeled WTD for the 42 stations, and the September human SLE cases by county. Human SLE cases were reported in September by 21 of the 28 counties within the study area. By the end of August conditions are wetter than a month earlier; most of the station sites have a modeled WTD above −1.2 meters. This continued wetting of the landscape was the result of continual rainfall, which also resulted in continual mosquito biting activity.8 The wet land surface persisted broadly throughout southcentral Florida through October of 1990. Reported human SLE cases were high in October, then decreased through November and December, reflecting a decrease in the number of SLE virus-infected mosquitoes.

The hydrologic evolution of the landscape during the focal transmission event of 1993 presents a different picture. Figure 5 shows the May 1993 mean modeled WTD for the 42 meteorologic station sites, the August 16–31, 1993 mean modeled WTD for the 42 stations, and the October occurrence of human SLE cases in southwest Florida. Human SLE cases first occurred in Lee County in August and continued there through November. In Collier County, the first confirmed human SLE cases were in October. All of the cases were focused on the border of Lee and Collier Counties. In contrast with May 1990, there was little drought evident in south-central Florida during May 1993; however, in the southwest corner of the study area in the Lee, Hendry, and northern Collier County area (denoted by the oval) a cluster of sites simulate May drought. In late September, there was some evidence of wetting in the area, particularly at the Collier County station (denoted by the arrow); however, this site was not among the cluster showing May drought.

Thus in 1993, there was May drought in Lee County but no wetting, and wetting in Collier County but no May drought. The cluster of sites presenting May drought in Lee and Hendry Counties suggests that northcentral Collier County may have also been in drought; however, there is no meteorologic record in this region with which to model land surface wetness. Alternatively, stations immediately adjacent one another can have very different modeled wetness states (Figure 5). Given this spatial variability, it is difficult to determine the extent to which the 1993 drought in Lee and Hendry Counties delineated a countywide event. The spatial variability, however, indicates that the network of 42 stations in the study area is too sparse to detect all regions of drought or wetting.

We also mapped the sporadic SLE transmission event in 1997. Figure 6 shows the May 1997 mean modeled WTD for the 42 stations, the August 16–31, 1997 mean modeled WTD for the stations, and the September occurrence of human SLE cases by county. Human cases were reported in September by 4 of the 28 counties within the study area. May drought was pervasive in the northern half of the study area, and in patches along the southern extreme. By the second half of August, most stations reported a modeled WTD above −1.2 meters. May drought and wetting one-half month prior to the onset of SLE cases occurred in all four counties where human SLE cases were reported in 1997.

DISCUSSION

In this report we have presented three hypotheses. 1) The spatial-temporal variability of human SLE cases is associated with the spatial-temporal variability of modeled land surface wetness. 2) The epicenters of human SLE outbreaks can be identified as foci of drought and wetting. 3) Drought is a necessary precursor for human SLE incidence. We now evaluate our findings in the context of these hypotheses.

We have shown that human SLE cases in southcentral Florida from 1990 to 1998 were significantly associated with drought 3–6 months prior and wetting 0–1 months prior to onset. This range of lags reflects the slow variability of modeled WTD, particularly the slow drop of the WTD as the model soil column dries during drought conditions. The best-fit model of human SLE incidence was shown to be drought four months prior and wetting one-half month prior to case onset, and these lags were also significant predictors of the prevalence of human SLE cases. These logistic regression models illustrate that the spatial-temporal distribution of human SLE cases is indeed associated with spatial-temporal patterns of drought and wetting in south-central Florida. These empirical models using TBH model simulations may also enable development of skillful forecasts of human SLE epidemics in southcentral Florida.

Using maps of modeled WTD we were able to identify the spatial-temporal evolution of land surface wetness in south-central Florida from 1990 through 1998. In 1990, there was pervasive drought throughout the region. Within the network of meteorologic stations used for this study, we were able to detect initial surface wetting in late June just west of Indian River County. The first human SLE case in 1990 was reported in eastern Indian River County. Wetting may also have occurred in the western region of this county where there was no meteorologic record. In 1993, a focus of drought was identified in Lee and Hendry Counties. This drought may have also affected northern Collier County where again, there was no meteorologic station record. In 1997, May drought occurred in many areas of southcentral Florida, and sporadic human SLE cases were reported in counties affected by drought and subsequent wetting. These maps (Figures 2–6) demonstrate the resolution with which modeled land surface wetness can be used to detect loci of SLE virus amplification and transmission. A denser network of meteorologic stations would provide a higher resolution picture of evolving land surface wetness conditions.

Differences among the years suggest that the spatial extent of May drought delimits where amplification occurs and the type of transmission that will follow (i.e., epidemic, focal, or sporadic). Thus, the spatial structure of the May 1993 drought appears to provide an explanation of why human SLE cases occurred focally in Lee and Collier Counties. Similarly, the May 1990 drought and subsequent wetting throughout south-central Florida separates this year from 1993 and indicates that the 1990 SLE epidemic was broadly hydrologically preconditioned and driven.

In our study, we also show that human SLE cases are less strongly associated with near-coincident wetting than with antecedent (May) drought. This finding is consistent with previous results, in which near-coincident wetting was more weakly associated than antecedent drought with the transmission of SLE virus to sentinel chickens.2 Within our working hypothesis of the Florida epidemic SLE transmission cycle, drought induces amplification of SLE virus in select habitats. In addition, these drought conditions allow more Cx. nigripalpus females to become infective during a single gonotrophic cycle.1 Once land surface wetting occurs these infectious Cx. nigripalpus can range more broadly and thus are more likely to encounter humans. However, rainfall may also permit dispersal of infected Cx. nigripalpus.1,8 While a single pulse of rain may not be sufficient to increase the availability of oviposition sites, it can, and does, raise humidity levels allowing host-seeking females to disperse farther. The daily flight behavior of Cx. nigripalpus is significantly correlated with increased relative humidity one hour after sunset.15 Local rainfall increases humidity and allows Cx. nigripalpus females, infected and uninfected alike, to increase their daily flight time and flight range.16 Thus, the behavioral response of Cx. nigripalpus to local rainfall events might explain the weaker association found between near-coincident land surface wetting and SLE virus transmission.

Our final hypothesis is that drought should always precede the onset of human SLE cases. We found that drought occurred four months prior to the onset of human cases in 73% of the counties reporting SLE cases. By focusing exclusively upon May, when wild birds are nesting and when we hypothesize the bulk of SLE amplification occurs, we showed that drought occurred in 83% of the counties reporting human SLE cases. However, if drought-induced amplification of SLE virus is a necessary precondition for human SLE cases, pre-transmission drought should always be present in counties reporting human SLE cases.

This shortcoming may be a consequence of a mismatch between the spatial variability of subtropical precipitation and the incomplete geographic coverage presently provided by the existing network of meteorologic stations. To first order, hydrologic conditions in subtropical Florida should be localized at the spatial scales at which storms organize. With the exception of irrigation, precipitation is the sole source of water to near surface soils, and it is these wet soils that support the surface pooling that forms the preferred breeding sites of Cx. nigripalpus. Because much of the rainfall in Florida is convective, land surface wetness can vary over short spatial scales. Examination of TBH model-simulated land surface wetness maps for 1990, 1993, and 1997 in southcentral Florida (Figures 2–6) suggests that not all of the spatial variability of wetness is being captured by the present network of meteorologic stations. Where station sites are close, modeled land surface wetness can vary considerably over short distances, on the order tens of kilometers. However, over much of the study area, the network of meteorologic stations is too sparse to comprehensively detect such smaller scale variability. These gaps in our simulation of the land surface wetness seem to explain why May drought is not found within all counties subsequently reporting human SLE cases. Mosquito, wild bird, and human movement in and out of a county may also have introduced noise to our analysis.

While it has not been clearly established to be a necessary precondition for human SLE incidence, overall, the analysis does demonstrate that May drought is significantly associated with subsequent human SLE incidence. This finding supports our working hypothesis that during drought mosquitoes and nesting wild birds are brought into close contact, facilitating epizootic amplification and generating the mosquito infection rates necessary to support high levels of SLE virus transmission. Further testing of this working hypothesis requires more evidence from the field including sampling of mosquito and wild bird abundance and infection rates within southcentral Florida hammock habitats. Using these data we will be able to document arboviral amplification during periods of drought, and more precisely determine how mosquitoes respond to rainfall and land surface wetting. We will also be able to explore the ability of the TBH model to predict the reproductive response of mosquitoes to short periods of wetting.

Future predictions of human SLE incidence may also need to consider virologic factors, such as increased herd immunity to the SLE virus from heavy exposure in preceding years. The effects of a variety of behaviors on human SLE incidence might also be explored; these could include occupational and avocational habits, such as whether SLE is more prevalent in agricultural workers and outdoorsmen.

Our findings support three hypotheses: 1) the spatial-temporal variability of antecedent drought and near-coincident wetting is associated the spatial-temporal variability of human SLE cases; 2) epicenters of human SLE are identifiable as foci of drought and wetting; and 3) drought is a measurable precondition associated with subsequent human SLE incidence.

Large-scale epidemics are of primary concern to public health workers. The findings of this study suggest that pervasive May drought and summer wetting throughout southcentral Florida enables such an epidemic event. Such large-scale hydrologic events are easily monitored using records from the current network of meteorologic stations to model land surface wetness. A denser network of stations may be necessary to monitor and predict smaller-scale focal and sporadic transmission events in southcentral Florida.

Table 1

List of the 42 meteorologic recording stations in Florida used to force the topographically based hydrology model

Station siteCountyLatitudeLongitude
MelbourneBrevard28° 06′ N80° 39′W
TitusvilleBrevard28° 38′N80° 50′W
Punta GordaCharlotte26° 55′N82° 00′W
InvernessCitrus28° 48′N82° 19′W
NaplesCollier26° 10′N81° 52′W
ArcadiaDesoto28° 06′N80° 39′W
Moore HavenGlades26° 50′N81° 05′W
WauchulaHardee27° 33′N81° 48′W
ClewistonHendry26° 45′N80° 56′W
Devils GardenHendry26° 36′N81° 08′W
La BelleHendry26° 45′N81° 26′W
BrooksvilleHernando28° 37′N82° 22′W
Weeki WacheeHernando28° 31′N82° 35′W
ArchboldHighlands27° 11′N81° 21′W
Avon ParkHighlands27° 36′N81° 32′W
TampaHillsborough27° 58′N82° 32′W
Vero Beach (4W)Indian River27° 41′N80° 26′W
ClermontLake28° 27′N81° 45′W
LisbonLake28° 52′N81° 47′W
Fort MyersLee28° 35′N81° 52′W
BradentonManatee27° 27′N82° 30′W
ParrishManatee27° 37′N82° 21′W
StuartMartin27° 11′N80° 14′W
Fort DrumOkeechobee27° 35′N80° 51′W
OkeechobeeOkeechobee27° 12′N80° 50′W
OrlandoOrange28° 26′N81° 20′W
KissimmeeOsceola28° 17′N81° 25′W
Belle GladePalm Beach26° 39′N80° 38′W
Canal PointPalm Beach26° 52′N80° 38′W
West Palm BeachPalm Beach26° 41′N80° 06′W
St. LeoPasco28° 20′N82° 16′W
St. PetersburgPinellas27° 46′N82° 38′W
Tarpon SpringsPinellas28° 09′N82° 45′W
Mountain LakePolk27° 56′N81° 36′W
Winter HavenPolk28° 01′N81° 44′W
Fort PierceSt. Lucie27° 28′N80° 21′W
Myakka RiverSarasota27° 15′N82° 19′W
VeniceSarasota27° 06′N82° 26′W
SanfordSeminole28° 48′N81° 16′W
BushnellSumter28° 40′N82° 05′W
Daytona BeachVolusia29° 11′N81° 03′W
DelandVolusia29° 01′W81° 19′W
Table 2

Reported human cases of St. Louis encephalitis (SLE) by month of onset in southcentral Florida for the 28 counties in this study, 1990–1998

YearCounty of residenceNumber of human SLE casesMonth of onset
1990Indian River2July
1990Brevard2August
1990Hardee1August
1990Hendry1August
1990Highlands2August
1990Hillsborough1August
1990Indian River11August
1990Lake2August
1990Manatee1August
1990Martin1August
1990Orange3August
1990Palm Beach3August
1990St. Lucie4August
1990Brevard6September
1990Charlotte1September
1990Collier2September
1990DeSoto1September
1990Highlands1September
1990Hillsborough3September
1990Indian River4September
1990Lee4September
1990Manatee1September
1990Martin4September
1990Orange11September
1990Osceola2September
1990Palm Beach15September
1990Polk4September
1990St. Lucie8September
1990Seminole2September
1990Brevard3October
1990Charlotte5October
1990Collier2October
1990DeSoto2October
1990Hendry2October
1990Highlands1October
1990Hillsborough4October
1990Indian River2October
1990Lake3October
1990Lee10October
1990Manatee4October
1990Martin2October
1990Orange14October
1990Osceola2October
1990Palm Beach8October
1990Pasco1October
1990Polk9October
1990St. Lucie7October
1990Sarasota6October
1990Seminole2October
1990Sumter1October
1990Brevard1November
1990Hillsborough1November
1990Lee1November
1990Polk1November
1990Sarasota1November
1990Polk4December
1993Lee1August
1993Lee1September
1993Lee2October
1993Lee1November
1993Collier2October
1993Collier1November
1997Brevard1July
1997Hillsborough1September
1997Lee1September
1997Palm Beach1September
1997Polk2September
1997Charlotte1October
1997Lee1October
1997Polk1October
1998Palm Beach1October
Table 3

Total number of counties in Florida reporting one or more human cases of St. Louis encephalitis compared with station sites simulating drought (of a specified magnitude) four months prior and wetting (of a specified magnitude) one-half month prior to disease onset*

Drought (modeled WTD)Wetting (modeled WTD)NumberPossibleP
* Significance was determined from bootstrap confidence intervals estimated using a Monte Carlo procedure. WTD = water table depth. m = meters.
< −1.4 m4371< 0.0001
< −1.4 m> −1.2 m2871< 0.02
< −1.4 m> −1.1 m2671< 0.02
< −1.35 m5271< 0.0001
< −1.35 m> −1.2 m3371< 0.02
< −1.35 m> −1.1 m2971< 0.02
Table 4

Total number of counties in Florida reporting one or more human cases of St. Louis encephalitis with station sites simulating drought of a specified magnitude during the preceding May and wetting of a specified magnitude one-half month prior to case onset dates*

Drought (modeled WTD)Wetting (modeled WTD)NumberPossibleP
* Significance was determined from bootstrap confidence intervals estimated using a Monte Carlo procedure. WTD = water table depth. m = meters.
< −1.4 m5471< 0.0005
< −1.4 m> −1.2 m3671< 0.02
< −1.4 m> −1.1 m3271< 0.05
< −1.35 m5971< 0.0005
< −1.35 m> −1.2 m4171< 0.05
< −1.35 m> −1.1 m3871< 0.05
Figure 1.
Figure 1.

a, Scatter plot of water table depth (WTD) one-half month prior versus WTD four months prior at all 42 sites in southcentral Florida, 1990–1998. Blue circles denote county-months with no human cases of St. Louis encephalitis (SLE) and red circles denote county-months with one or more cases of SLE. More than one site may exist in a county. b, Same scatter plot as in Figure 1a. Red circles now randomly selected from among the meteorologic stations.

Citation: The American Journal of Tropical Medicine and Hygiene Am J Trop Med Hyg 71, 3; 10.4269/ajtmh.2004.71.251

Figure 2.
Figure 2.

Maps of modeled water table depth (WTD) for May 1990 and June 16–30, 1990 at all 42 stations in the study area. Location of the July 1990 human cases of St. Louis encephalitis (SLE) is also shown.

Citation: The American Journal of Tropical Medicine and Hygiene Am J Trop Med Hyg 71, 3; 10.4269/ajtmh.2004.71.251

Figure 3.
Figure 3.

Map of modeled water table depth (WTD) for July 16–31, 1990 at all 42 station sites in the study area. Location of the August 1990 human cases of St. Louis encephalitis (SLE) is also shown.

Citation: The American Journal of Tropical Medicine and Hygiene Am J Trop Med Hyg 71, 3; 10.4269/ajtmh.2004.71.251

Figure 4.
Figure 4.

Map of modeled water table depth (WTD) for August 16–31, 1990 at all 42 station sites in the study area. Location of the September 1990 human cases of St. Louis encephalitis (SLE) is also shown.

Citation: The American Journal of Tropical Medicine and Hygiene Am J Trop Med Hyg 71, 3; 10.4269/ajtmh.2004.71.251

Figure 5.
Figure 5.

Maps of modeled water table depth (WTD) for May 1993 and August 16–31, 1993 at all 42 station sites in the study area. Location of the September 1993 human cases of St. Louis encephalitis (SLE) is also shown.

Citation: The American Journal of Tropical Medicine and Hygiene Am J Trop Med Hyg 71, 3; 10.4269/ajtmh.2004.71.251

Figure 6.
Figure 6.

Maps of modeled water table depth (WTD) for May 1997 and August 16–31, 1997 at all 42 station sites in the study area. Location of the September 1997 human cases of St. Louis encephalitis (SLE) is also shown.

Citation: The American Journal of Tropical Medicine and Hygiene Am J Trop Med Hyg 71, 3; 10.4269/ajtmh.2004.71.251

Authors’ addresses: Jeffrey Shaman, Department of Earth and Planetary Sciences, Harvard University, Cambridge, MA 02138, E-mail: jshaman@fas.harvard.edu. Jonathan F. Day, Florida Medical Entomology Laboratory, Institute of Food and Agricultural Sciences, University of Florida, Vero Beach, FL 32962, E-mail: JFDA@mail.ifas. ufl.edu. Marc Stieglitz, Lamont-Doherty Earth Observatory, Columbia University, Palisades, NY 10964, E-mail: marc@ldeo.columbia.edu.

Acknowledgment: We thank Carina Blackmore, Caroline Collins and their staff at the Florida Department of Health in Tallahassee for their work tracking and reporting human arbovirus cases in Florida. This report is Florida Agricultural Experiment Station Journal Series R-101091.

Financial support: This work was supported by National Aeronautics and Space Administration Earth System Science Fellowship NGT5-50323 and the National Oceanic and Atmospheric Administration Postdoctoral Program in Climate and Global Change, administered by the University Corporation for Atmospheric Research.

REFERENCES

  • 1

    Day JF, Curtis GA, 1999. Blood feeding and oviposition by Culex nigripalpus (Diptera: Culicidae) before, during, and after a widespread St. Louis encephalitis virus epidemic in Florida. J Med Entomol 36 :176–181.

    • Search Google Scholar
    • Export Citation
  • 2

    Shaman J, Day J, Stieglitz M, 2002. Drought-induced amplification of St. Louis encephilitis virus, Florida. Emer Infect Dis 8 :575–580.

  • 3

    Stieglitz M, Rind D, Famiglietti J, Rosenzweig C, 1997. An efficient approach to modeling the topographic control of surface hydrology for regional and global climate modeling. J Climate 10 :118–137.

    • Search Google Scholar
    • Export Citation
  • 4

    Day JF, Stark LM, 2000. Frequency of Saint Louis encephalitis virus in humans from Florida, USA: 1990–1999. J Med Entomol 37 :626–633.

    • Search Google Scholar
    • Export Citation
  • 5

    Chamberlain RW, Sudia WD, Coleman PH, Beadle LD, 1964. Vector studies in the St. Louis encephalitis epidemic, Tampa Bay area, Florida, 1962. Am J Trop Med Hyg 13 :456–461.

    • Search Google Scholar
    • Export Citation
  • 6

    Dow RP, Coleman PH, Meadows KE, Work TH, 1964. Isolation of St. Louis encephalitis viruses from mosquitoes in the Tampa Bay area of Florida during the 1962 epidemic. Am J Trop Med Hyg 13 :462–474.

    • Search Google Scholar
    • Export Citation
  • 7

    Shroyer DA, 1991. The 1990 Florida epidemic of St. Louis encephalitis: virus infection rates in Culex nigripalpus. J Fla Mosq Control Assoc 62 :69–71.

    • Search Google Scholar
    • Export Citation
  • 8

    Day JF, Curtis GA, 1993. Annual emergence patterns of Culex nigripalpus females before, during and after a widespread St. Louis encephalitis epidemic in south Florida. J Am Mosq Cont Assoc 9 :249–253.

    • Search Google Scholar
    • Export Citation
  • 9

    Shaman J, Day J, Stieglitz M, 2003. St. Louis encephalitis virus in wild birds during the 1990 south Florida epidemic: the importance of drought, wetting conditions, and the emergence of Culex nigripalpus to arboviral amplification and transmission. J Med Entomol 40 :547–554.

    • Search Google Scholar
    • Export Citation
  • 10

    Beven KJ, Kirkby MJ, 1979. A physically based, variable contributing area model of basin hydrology. Hydrol Sci 24 :43–69.

  • 11

    Ducharne A, Koster RD, Suarez MJ, Stieglitz M, Kumar P, 2000. A catchment-based approach to modeling land surface processes. Part II: Parameter estimation and model demonstration. J Geophys Res 105 :24823–24838.

    • Search Google Scholar
    • Export Citation
  • 12

    Shaman J, Stieglitz M, Engel V, Koster R, Stark C, 2002. Representation of storm flow and a more responsive water table in a TOPMODEL-based hydrology model. Water Resources Res 38 : Art No. 1156.

    • Search Google Scholar
    • Export Citation
  • 13

    Shaman J, Stieglitz M, Zebiak S, Cane M, 2003. A local forecast of land surface wetness conditions derived from seasonal climate predictions. J Hydrometeorol 4 :611–626.

    • Search Google Scholar
    • Export Citation
  • 14

    DeGaetano A, Eggleston K, Knapp W, 1995. Daily solar radiation estimates for the northeastern United States using the northeast regional climate center and National Renewable Energy Laboratory models. Solar Energy 55 :185–194.

    • Search Google Scholar
    • Export Citation
  • 15

    Dow RP, Gerrish GM, 1970. Day-to-day change in relative humidity and the activity of Culex nigripalpus (Diptera: Culicidae). Ann Entomol Soc Am 63 :995–999.

    • Search Google Scholar
    • Export Citation
  • 16

    Day JF, Curtis GA, 1994. When it rains they soar–and that makes Culex nigripalpus a dangerous mosquito. Am Entomologist 40 :162–167.

Save