TY - JOUR
T1 - Modeling the spatial distribution of mosquito vectors West Nile virus in Connecticut, USA
AU - Diuk-Wasser, Maria A.
AU - Brown, Heidi D.
AU - Andreadis, Theodore G.
AU - Fish, Durland
PY - 2006/9
Y1 - 2006/9
N2 - The risk of transmission of West Nile virus (WNV) to humans is associated with the density of infected vector mosquitoes in a given area. Current technology for estimating vector distribution and abundance is primarily based on Centers for Disease Control and Prevention (CDC) light trap collections, which provide only point data. In order to estimate mosquito abundance in areas not sampled by traps, we developed logistic regression models for five mosquito species implicated as the most likely vectors of WNV in Connecticut. Using data from 32 traps in Fairfield County from 2001 to 2003, the models were developed to predict high and low abundance for every 30 × 30 m pixel in the County. They were then tested with an independent dataset from 16 traps in adjacent New Haven County. Environmental predictors of abundance were extracted from remotely sensed data. The best predictive models included non-forested areas for Culex pipiens, surface water and distance to estuaries for Cx. salinarius, surface water and grasslands/ agriculture for Aedes vexans and seasonal difference in the normalized difference vegetation index and distance to palustrine habitats for Culiseta melanura. No significant predictors were found for Cx. restuans. The sensitivity of the models ranged from 75% to 87.5% and the specificity from 75% to 93.8%. In New Haven County, the models correctly classified 81.3% of the traps for Cx. pipiens, 75.0% for Cx. salinarius, 62.5% for Ae. vexans, and 75.0% for Cs. melanura. Continuous surface maps of habitat suitability were generated for each species for both counties, which could contribute to future surveillance and intervention activities.
AB - The risk of transmission of West Nile virus (WNV) to humans is associated with the density of infected vector mosquitoes in a given area. Current technology for estimating vector distribution and abundance is primarily based on Centers for Disease Control and Prevention (CDC) light trap collections, which provide only point data. In order to estimate mosquito abundance in areas not sampled by traps, we developed logistic regression models for five mosquito species implicated as the most likely vectors of WNV in Connecticut. Using data from 32 traps in Fairfield County from 2001 to 2003, the models were developed to predict high and low abundance for every 30 × 30 m pixel in the County. They were then tested with an independent dataset from 16 traps in adjacent New Haven County. Environmental predictors of abundance were extracted from remotely sensed data. The best predictive models included non-forested areas for Culex pipiens, surface water and distance to estuaries for Cx. salinarius, surface water and grasslands/ agriculture for Aedes vexans and seasonal difference in the normalized difference vegetation index and distance to palustrine habitats for Culiseta melanura. No significant predictors were found for Cx. restuans. The sensitivity of the models ranged from 75% to 87.5% and the specificity from 75% to 93.8%. In New Haven County, the models correctly classified 81.3% of the traps for Cx. pipiens, 75.0% for Cx. salinarius, 62.5% for Ae. vexans, and 75.0% for Cs. melanura. Continuous surface maps of habitat suitability were generated for each species for both counties, which could contribute to future surveillance and intervention activities.
KW - Culex
KW - Light traps
KW - Spatial modeling
KW - West Nile virus
UR - http://www.scopus.com/inward/record.url?scp=33751243608&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=33751243608&partnerID=8YFLogxK
U2 - 10.1089/vbz.2006.6.283
DO - 10.1089/vbz.2006.6.283
M3 - Article
C2 - 16989568
AN - SCOPUS:33751243608
SN - 1530-3667
VL - 6
SP - 283
EP - 295
JO - Vector-Borne and Zoonotic Diseases
JF - Vector-Borne and Zoonotic Diseases
IS - 3
ER -