GIS Analysis of Arbovirus Activity in Florida

Arbovirus is a condensation of arthropod-borne virus, a large group of viruses that are spread by blood-sucking insects. In the United States arboviruses are most commonly spread by mosquitoes. Many vertebrate hosts can transmit the virus, but birds are the most common. St Louis Encephalitis (SLE) is the most common human disease caused by mosquitoes in the United States, with an average of 130 cases per year. Historically, severe SLE epidemics have occurred with as many as 3000 cases in some years, as in the 1933 epidemic that occurred in St. Louis, Missouri, and from which the disease got its name. The virus is found throughout the country with periodic outbreaks occurring in the Mississippi valley, the Gulf coast and Florida. SLE poses a major public health problem.

Florida experienced significant SLE outbreaks in 1959, 1961 and 1962. Following a widespread 1977 SLE outbreak in central Florida, the Florida State Department of Health initiated a Sentinel Chicken Arbovirus Surveillance Program to monitor arboviral transmission and provide an early warning system of future epidemics (Day & Stark, 1994). More recently, SLE outbreaks have occurred in Florida in 1990 and 1998. Prior to these outbreaks, the sentinel chicken flocks did exhibit significantly increased levels of arbovirus transmission in the affected areas. Climatological conditions, especially excessive rainfall, have been identified as an influential factor in the location and timing of arbovirus activity in general as well as specific SLE epidemics (Hayes & Hess, 1968; Day and Curtis, 1989). Analysis of SLE activity in the context of a Geographic Information System (GIS) assists health officials with understanding the geographic dimensions of potential outbreaks. The GIS analysis may also contribute to a better understanding of the rainfall patterns that contribute to increased arbovirus activity and SLE outbreaks.

Participating counties locate sentinel chicken flocks at various sites. Blood samples are taken from the chickens weekly and submitted to the Department of Health State Laboratory-Virology Section for testing. The Laboratory test results are recorded in a weekly report. Once an individual chicken has tested positive for two consecutive weeks, it is considered a confirmed positive, and noted as such in the weekly report. (Stark, 1999). The weekly reports are stored in a digital format in Microsoft Excel files.

The majority of participating counties recorded site locations using Global Positioning Satellite (GPS) technology. The Florida Department of Health compiled a listing of participating counties' sentinel chicken flock surveillance sites (Figure 1). The sites in the listing that included location coordinates were used to create an ArcView shape file of chicken flock locations. Each site location was assigned a unique seven-digit reference number based on the five-digit FIPS code (state code and county code) and a two-digit site ID number. For example, Site #1 in Bay County, Florida was encoded as 1200501, where 12 is the FIPS code for Florida, 005 is the FIPS code for Bay county, and 01 is the assigned code for site one.

The Department of Health Arbovirus Surveillance Summary weekly report (in digital format) is provided to the Florida State University (FSU) College of Social Sciences (COSS) Geographic Information Systems (GIS) Laboratory. Visual Basic Application (VBA) scripts for Microsoft Excel 2000 are developed that extract the flock site locations and sum the number of confirmed sentinel chicken seroconversions to St. Louis Encephalitis virus from the report for an individual week. The extracted information is written to a new dbase file that is opened in ArcView and joined to the Attribute table of the chicken flock locations shape file. Avenue scripts are developed that automate the process of creating a map layout from the weekly report in dbase format. The resulting map layout is exported to JPEG format, and submitted digitally to the Department of Health for publication on the Department's web site. A series of maps are created for the active 1999 arbovirus season reports from July 30 to Nov 24 that display the sites and the number of confirmed chicken seroconverions to SLE virus.

Monthly rainfall data are compiled for meteorological stations in Florida for the period from June 1999 to November 1999. The monthly rainfall data is classified as follows:

Rainfall Amount Value Description

Greater than 6.00 inches: 8 Very Wet

4.00 to 5.99 inches 6 Wet

2.00 to 3.99 inches 4 Moist

0.01 to 1.99 inches 2 Dry

0.00 0 No Data

The data are stored in a dbase file that is opened in ArcView and used with the Spatial Analyst extension to create a grid representing the monthly rainfall regime "surface" for the months of September and October. The total confirmed SLE seroconversions by county from October 1 to October 24 are summed and assigned to a manually derived county centroid. This process is repeated for the period from October 31 to November 24, 1999. There is a time delay of several weeks between the onset of favorable meteorological conditions and the actual observation of seroconversions in sentinel chickens. Therefore, the October 1 to October 24 conversions are expected to reflect the September rainfall regime's influence, while the October 31 to November 24 period is expected to reflect the October rainfall regime's influence. The Spatial Analyst is used to create grids reflecting the October 1 - October 24 and Oct 31 - November 24 SLE seroconversion "surfaces". The September and October rainfall surfaces are subtracted from the respective seroconversion surfaces to determine what type of relationship, if any, can be established between pattern of rainfall and the pattern of seroconversions of SLE.

Beginning with the August 20, 1999 Summary report, and continuing into early September, a few sentinel chicken flocks had confirmed seroconversions for SLE, primarily single bird confirmations. Spatially, the confirmed seroconversions were isolated and scattered across sites in central Florida. The pattern began to change with the September 17 Summary report. While the confirmed seroconversions were still only single bird situations, the sites were clustered in Lee and Manatee counties, in the south central Gulf coast region near Ft. Myers. By October 8, this region centered near Lee County was reporting confirmed seroconversions of multiple chickens at multiple sites.

The Florida State Department of Health issued an advisory subsequent to these results. The map produced from the October 15 Summary report indicated the focus of confirmed seroconversions was in Lee County, with arbovirus activity extending along the south central Gulf coast (Figure 2). The arbovirus activity slowly subsided during late October and into November. The November 12 Summary report was the last one of the season that included confirmed seroconversions of multiple birds, and those multiple birds were located at sites in the same focus area of the south central Gulf coast.

During the 1999 season, rainfall conditions tended to be drier than usual during June and July. In August, more typically wet conditions began to prevail, and by September significantly wet conditions were experienced at many of the reporting weather stations, especially along the Gulf and Atlantic Coasts. In October, very wet conditions were primarily confined to the southeastern coastal counties of the peninsula, and the central Gulf coast area was drier. The rainfall data for September and October are classified as previously described and utilized to create the respective monthly rainfall regime "surfaces". These surfaces are subtracted from the October 1 to October 24 and October 31 to November 24 arbovirus activity "surfaces" to create the differential surfaces of October arbovirus/September rainfall and November arbovirus/October rainfall differential surfaces.

In general, the differential between the October 1 to October 24 arbovirus surface and the September rainfall surface corresponded reasonably well (Figure 3). As expected, the focal area of observed arbovirus activity (the south central Gulf coast) had minimal difference between the October SLE seroconversions and the September rainfall regime surfaces. However, the fairly large area of the central portion of the peninsula had a positive difference between the surfaces, indicating much greater arbovirus activity than expected based on the prevailing rainfall regime. Several locations had a significantly negative difference between the surfaces, most notably the Tampa Bay area, as well as smaller isolated pockets along the Atlantic coast from Jacksonville southward to Dade County. These areas displayed less arbovirus activity than suggested by the rainfall regime.

The differential between the October 31-November 24 arbovirus surface and the October rainfall surface does not correspond very well. An extensive region from the center of the Florida peninsula to the Atlantic coast has a significantly negative difference between the surfaces. Also notable, the focal area of the seroconversions for SLE (Ft Myers region) has a significantly positive difference between the surfaces (Figure 4).

The 1999 maps created from the Department of Health Arbovirus Summary weekly reports are a very useful tool for state health officials, and their publication on the web provides a valuable feedback to the county officials. Inconsistent labeling of flock sites proved to be problematic for developing the automated process for generating these maps. In fact, seven counties submitted seroconversion test results that could not be mapped because the coordinates for their sites had never been submitted to the State Health Department. Several others submitted data, but it could not be mapped because the county labeled their sites with names or numbers that did not correspond to those used in the state database. For the 2000 season, state officials are coordinating and compiling site coordinates for ALL participating counties. Some rural counties that lack GPS equipment will submit hardcopy maps showing the location of their sentinel chicken flocks, and the sites will be manually digitized into the site location database.

In addition, the importance of using consistent site labels has been communicated to the county officials, and the Department of Health's Virology Lab will assign the proper labels to the digital Summary reports- even if the hardcopy reports submitted by the county(s) are incorrect. Implementing these procedures should ensure that the all of the 2000 arbovirus data collected would be able to be mapped.

The 1999 maps clearly illustrate the geographic dimensions of confirmed SLE seroconversions and were used in the decision-making process of the state's health officials when they opted to issue arbovirus warnings. The use of the customized GIS facilitates the production of maps from the Summary reports in near real time. The publication of the maps on the web is very useful to the local officials, some of who have expressed an interest in being able to do interactive mapping over the web. The Department of Health does not have currently have ArcView IMS, but obviously it would be extremely useful if the interactive mapping option were to be pursued.

The differentials between the arbovirus activity surfaces and the rainfall regime surfaces had mixed results. While the September rainfall regime indicated a relationship to the observed October arbovirus activity, the October rainfall regime's relationship to the November arbovirus was not clear at all. This analytical technique may be useful in the future, but issues of scale, both spatially and temporally, must be carefully considered when compiling the inputs for the GIS. The differential surfaces examples used in this paper are extremely preliminary. Additional research should be done with a much more systematic approach. For example, weekly rainfall data should be used that corresponds to the time period of the weekly Summary reports. Actual flock sites should be used, rather than a county centroid. In fact, the level of detail required for this type of research suggests that a county or regional level of analysis may be more appropriate than at the state level.

The appearance of the West Nile (WN) Virus, and the associated epidemic in New York during the summer of 1999 has become a key concern for health officials in Florida, as well as the rest of the eastern United States. The West Nile Virus is an arbovirus that had not previously been encountered in the Western Hemisphere. The consensus among health officials is not if the WN virus will spread to Florida, but WHEN will it get there. The well-established and long time span of the Florida Sentinel Chicken Arbovirus Surveillance Program provides a valuable baseline of arbovirus activity data, and should prove useful in detecting the WN virus if and when it arrives in the state. The GIS tools developed to map results of the weekly Summary reports may also be effective in detecting the WN virus's arrival and providing advance warning to Floridians.

We acknowledge the commitment of the Florida Department of Health Officials who assisted us with this work, namely Dr. Lisa Conti, Don Ward, and Robin Oliveri in the Tallahassee bureau and Dr. Lillian Stark and Dr. Deno Kazanis at the Virology laboratory in Tampa.

Day, J. and Curtis, A. 1989: Influence of Rainfall on Culex nigripalpus (Diptera:

Day, J. and Stark, L.. 1996: Transmission Patterns of St. Louis Encephalitis and Eastern

Hayes, R. and Hess, A. 1964 Climatological conditions associated with outbreaks of

Stark, L. 1999: (personal communication) – August 18, 1999. Tampa, Florida.

Michael J. Daly

GIS Lab Director

College of Social Sciences

Florida State University

360 Bellamy

Tallahassee, FL 32306

Phone: (850) 644-8373

Fax: (850) 644-5913

mdaly@garnet.acns.fsu.edu