Meredith G. Morrow Literature Review: Use of Remote Sensing/GIS to predict tick habitat and Lyme disease risk

The nymphal stages of the ticks Ixodes pacificus (western United States) and I. scapularis (eastern United States) are considered the primary vectors of Borrelia burgdorferi, the causative agent of Lyme disease in North America. Current studies concerning the epidemiology of this disease have focused on mapping potential vector habitat and spatial distribution of the tick vectors and Lyme disease cases through

the

use

of

Geographic

Information

Systems

and

Remote

Sensing

technologies (Vine et al, 1997).

In the past, research has been published on spatial distribution patterns of adult I. pacficus and I. scapularis showing where the ticks are established and/or reported. Preliminary studies are now showing that these studies may be inadequate at predicting Lyme disease risk as exposure to nymphal stages is minimal in adult habitats (Eisen, 2006a, Eisen, 2006b). For instance, host-seeking nymphs are readily collected by drag sampling or during normal human activities in woodlands with ground cover dominated by leaves or fir needles, whereas adults rarely are found questing in such areas (Eisen, 2006a; Eisen, 2006b). At present, many studies are more focused on creating predictive models for nymphal density patterns as risk of Lyme disease since the disease is often transmitted during the nymphal stage (Eisen, 2006a; Eisen, 2006b; Kitron, 1998).

Eisen et al (2006a) researched Ixodes pacificus nymphal density patterns in California through the use of a supervised classification model to identify the risk of human exposure. They determined that the best predictor of habitat type was to calculate the normalized difference vegetation index (NDVI), brightness, greenness,

and wetness (BGW). Brightness is related to soil reflectance, greenness is strongly related to amount of vegetation, and wetness is related to canopy and soil moisture. Differences between images in BGW and NDVI were calculated using the raster calculator in ArcMap.

They found that this GIS/RS model was 22% more accurate in predicting nymphal density at 16 sites compared to a field-derived model. They concluded the study with a risk map that showed that ~12% of the county was classified as habitat posing at least moderate risk of human exposure to nymphs (>6.4 nymphs per 100m2). They also discovered that high-risk areas (>10.5 nymphs per 100 m2) tended to occur in more populated areas in the interior of the county, but were rare in proximity to coastal populated areas.

Another study that utilized GIS/RS was carried out by Foley et al (2005) where coyotes were tested for two tick-borne diseases: Lyme disease and Human Granulocytic Anaplasmosis (human GA, formerly ehrlichiosis). Counties were chosen where a previous study of dogs indicated high rates of anaplasmosis and included Ventura, El Dorado, Mendocino, Napa, Placer, Sonoma, and Yuba counties. Two hundred and fifteen coyotes were trapped in these counties by USDA Wildlife Services as part of a predator control program, and coyotes were tested for both diseases with

a

95% confidence interval. Associations with

vegetation

and

precipitation were evaluated by comparing the locations of seropositive and seronegative coyotes within vegetation regions provided by the spatial vegetation GIS coverage CALVEG 2000 (California Forestry and Fire Protection, Sacramento, CA) and precipitation isohyetal polygons in the layer ‘Precipitation’ (Teale GIS Solutions, Sacramento, CA). Spatial joining of latitude and longitude for each coyote location to precipitation and vegetation polygons was performed in ArcMap.

The overall prevalence of IgG in 215 coyotes to human GA was 39.5%, and the overall Lyme disease seroprevalence was 19.0%. Human GA-seropositive coyotes appeared most commonly in areas which included the following types of vegetation: redwood, montane hardwood, montane hardwood-conifer, oak and pine. Lyme disease-seropositive coyotes were not found near redwood as often as Human GA-seropositive coyotes, although they did occur in montane hardwood, montane hardwood-conifer, oak, and pine forest. There was a significantly increased risk for Lyme disease present only in blue oak woodland. By the end of this study, they confirmed that there was greater risk of Human GA and Lyme disease in vegetation regions with higher rainfall. They were also able to create a risk map with areas that were at risk for either Human GA and Lyme disease, both, or neither. Areas that were protective or not at risk for Human GA and Lyme disease included sagebrush, chaparral, agricultural areas, and habitats in the Central Valley, all of which have either inhospitable microenvironments for ticks or are at too low an altitude for I. pacificus.

A similar study using a canine surveillance system was utilized by Guerra et al (2001). A seroprevalence survey for B. burgdorferi was conducted among the healthy canine (pet) population in selected counties of Wisconsin and Illinois to determine the distribution of Lyme disease and associated risk factors. For each dog, information was gathered including: place of residence, Lyme disease vaccination status, history of travel and tick exposure, and medical history. To determine environmental risk factors, GIS was used to integrate environmental information along with the exact location of each canines’ residence. Seropositivity was associated with living in forested and urban areas, and on sandy, fertile soils. Since Lyme disease incidence in canines was significantly correlated with human incidence

of Lyme disease and with abundance of the tick vector, the authors concluded that a canine surveillance system is a useful method for assessing the geographic distribution of Lyme disease.

In another study paid for by the Centers of Disease Control and Prevention, Guerra et al (2002) used GIS to predict habitat suitability for I. scapularis in the Northern Central U.S. The purpose of this study was to (1) determine the distribution of Ixodes scapularis in the Northern Central U.S., (2) determine how habitat plays a role in tick establishment and survival in order to (3) predict where tick-borne diseases might occur. They determined the association between tick presence and habitat and found that tick presence was positively associated with dry, deciduous forests, organic soils, and sedimentary rock; while tick absence was noted in grasslands, conifer forests, wet forests, acidic soils, clay soils, and Precambrian bedrock. GIS was used to create a risk map (noting suitable habitats and where I. scapularis is already established). The study concluded that habitat is the most important factor in determining where ticks occur and if they can successfully become established in an area. However, I would be interested to see data on numbers of white-tailed deer and white-footed mice in the areas where I. scapularis is predicted to thrive after being introduced since reservoir-host density was not examined in this study. Determining the environmental factors that limit survival, and predicting areas where ticks are likely to survive once introduced to an area, may help public health officials facilitate health program planning in order to control tick populations, thereby reducing tick-borne disease transmission rates.

Dister et al (1997) carried out another study which evaluated landscape composition for Lyme disease exposure risk on residential properties in two communities (Armonk and Chappaqua) in New York. Each of the 337 properties were

categorized as no, low or high risk based on densities of I. scapularis nymphs (which were determined via drag sampling). They determined that an increase in moisture and vegetation abundance were the two most important predictive characteristics of Lyme disease risk. Chappaqua was significantly greener and wetter than Armonk and also had more higher-risk properties. They also found that high-risk properties had a greater proportion of broadleaf trees, while lower-risk properties had a greater proportion of non-vegetative cover and/or open lawn. This study concluded that a remote sensing/GIS based approach was adequate and efficient at identifying risk of Lyme disese over larger geographic areas.

Another study carried out by Kitron and Kazmierczak (1997) created a risk map for the state of Wisconsin which took into consideration all three variables: human cases of Lyme disease, tick distribution, and vegetation coverage. They took satellite data to calculate a NDVI for each county, and then created a GIS that merged all three variables onto one map. They found that human case distribution by county was associated with tick distribution and high NDVI values. In addition, this was the only study that called to attention the imprecision of diagnosis and false positive/false negative serologic results when it comes to testing for Lyme disease. The goal of their study was to assist in the development of an accurate description of the geographic distribution and transmission risk of Lyme disease for the Centers for Disease Control and Prevention. Since the CDC only maps human surveillance cases, the authors hoped their research could add additional data to the CDC map in order to display which cases are more or less statistically probable based on whether they were diagnosed near a high density tick habitats (mainly in western Wisconsin). While this was an interesting approach to take, it calls to attention the potential limitations of GIS/remote sensing. Satellite imagery does not prove correlations, it simply shows random associations that should be researched further by other

methods. For instance, utilizing a GIS risk map from this study may show that it is more likely for a person to become infected with Lyme disease in western Wisconsin, however, that should not indicate that a diagnosis in northern Wisconsin should be viewed as a definitive false positive. Again, while studies like this are helpful in understanding the epidemiology of the disease, it should not be mistaken as a singular approach.

References Cited: Dister et al. (1997) Landscape characterization of peridomestic risk for Lyme disease

using satellite imagery. The American Journal of Tropical Medicine and Hygiene. 57(6): 687-692. Eisen et al. (2006a) Predicting density of Ixodes pacificus nymphs in dense woodlands in Mendocino County, California, based on geographic information systems and remote sensing versus field-derived data. The American Journal of Tropical Medicine and Hygiene. 74(4): 632-640. Eisen et al. (2006b) Geographical distribution patterns and habitat suitability models for presence of host-seeking ixodid ticks in dense woodlands of Mendocino County, California. Journal of Medical Entomology. 43(2): 415-427. Foley et al. (2005) GIS-facilitated spatial epidemiology of tick-borne diseases in coyotes (Canis latrans) in northern and coastal California. Comparative Immunology, Microbiology and Infectious Diseases. 28(3): 197-212. Guerra et al. (2001) Canine surveillance system for Lyme borreliosis in Wisconsin and northern Illinois: geographic distribution and risk factor analysis. The American Journal of Tropical Medicine and Hygiene. 65(5): 546-552. Guerra et al. (2002) Predicting the risk of Lyme disease: habitat suitability for Ixodes scapularis in the north central United States. Emerging Infectious Diseases. 8(3): 289-297. Kitron U and Kazmierczak J. (1997) Spatial analysis of the distribution of Lyme disease in Wisconsin. American Journal of Epidemiology. 145(6): 558-566. Kitron U. (1998) Landscape ecology and epidemiology of vector-borne diseases: tools for spatial analysis. Journal of Medical Entomology. 35(4): 435-445. Vine et al. (1997) Geographic information systems: their use in environmental epidemiologic research. Environmental Health Perspectives. 105(6): 598-605.