CHAART Logo

Remote Sensing Characterization and Prediction of Hyperendemic Foci for Lyme Disease: First Year Report



Introduction | YPH Main Page | CHAART Projects | CHAART Home

Targeting Lyme Disease Prevention with Remote Sensing and Geographic Information Systems

Current methods for preventing peridomestic exposure to Lyme disease in hyperendemic communities include area application of insecticides, deer fencing, and personal protection against tick bites. However, these methods seem to have little impact upon human case reports, which tend to track natural fluctuation in the vector tick population. Obstacles to acceptance of these methods are environmental and safety concerns (insecticides), cost (deer fencing), and inconvenience (personal protection). Some of these obstacles might be overcome by targeting educational efforts to select residences with high exposure to peridomestic tick bites. We used remote sensing and geographic information system technologies to identify individual high-risk residences in Westchester Co., NY, a county where Lyme disease has been hyperendemic since 1982.

A model was developed to predict high-risk residential properties using Landsat Thematic Mapper (TM) data and entomological survey data from 377 properties in two Westchester Co. communities. Spectral indices of wetness and greenness, which characterize vegetation moisture and structure, and vegetation abundance, respectively, were found to discriminate between properties with high tick density and those with low or no tick density. High-risk properties were characterized by landscapes comprised of a greater proportion of broad-leaved canopy trees than lower-risk properties which had more non-vegetation cover and/or open lawn. Regression equations were developed to predict risk from TM data covering the average land area for each residential property in the county.

The model was tested by comparing predicted risk to data obtained from questionnaires sent to 1000 randomly selected county residences and 100 additional residences from each of three communities stratified according to latitude. An overall response rate of 32% provided data on tick-bite incidence and history of tick-borne illness for 405 properties. These properties were located by address-matching in a geographic information system and a property location layer was constructed to obtain coordinates for the TM data. Remote sensing indices for each location categorized properties as either high-risk, or low/no-risk. Questionnaire data were categorized as high-risk when reporting a peridomestically acquired tick bite within the last year or a history of tick-borne disease within the household. Comparison of predicted and observed data revealed 71% accuracy of the remote sensing data in predicting risk. The model was found to have high specificity, but low sensitivity which underestimated risk for many properties. These parameters did not vary latitudinally along an urban-rural landscape gradient characteristic of the county.

Validation of the remote sensing model enables rapid assessment of Lyme disease risk for individual residential properties over large geographic areas. Prevention efforts targeted to specific households with high risk should increase the effectiveness of public health educational programs and increase compliance by providing residents with individual property-based data that specifies risk.


Back to Lyme Project Introduction

Last updated: Mar 2000