ESTIMATING OLIGORYZOMYS LONGICAUDATUS (CRICETIDAE: SIGMODONTINAE) POTENTIAL DISTRIBUTION AND ITS RELATIONSHIP WITH REPORTED HANTAVIRUS PULMONARY SYNDROME (HPS) CASES IN ARGENTINA

VERÓNICA ANDREO; GREGORY GLASS; TIMOTHY SHIELDS; CECILIA PROVENSAL; FRANCISCO POLOP; SILVANA LEVIS; NOEMI PINI; JAIME POLOP

Atenas

Conferencia; VIII International Conference on HFRS, HPS and Hantaviruses; 2010

The goal of this study was to establish the association between presence of Oligoryzomys longicaudatus (host of Andes virus) and environmental factors to build a spatial model that predicted its potential distribution in Argentina. We used an extensive database of species occurrence records obtained from published studies. It consisted of 188 presence and 340 absence records. We excluded 32 presence and 62 absence records as validation test points. We compared two methods to model the distribution in terms of probability of presence of O. longicaudatus; logistic regression models and the MaxEnt algorithm. The environmental variables used were altitude, tree cover, grass cover, bare soil cover, 19 bioclimatic variables from WorldClim database and 5 different classification schemes from MODIS imagery. We identified the ‘best’ logistic model according to the AIC value and used a Moran test to assess whether the unexplained variation was randomly distributed. We used the MaxEnt algorithm with the same variables used in the logistic model. We evaluated and compared the performance of the models using the receiver-operator curve (ROC) and the area under the curve (AUC). We overlapped the modeled distribution of O. longicaudatus with sites of confirmed HPS cases (Government records). The best models included a classification scheme based on net primary productivity, altitude, tree and grass cover and 8 bioclimatic variables. Both models predicted the highest probabilities of host occurrence in the southwest Andean region and performed very similar in terms of AUC (0.95 and 0.98 for logistic and MaxEnt models, respectively). The HPS cases coincided with occurrence probabilities higher than 85% for the logistic model and higher than 55% for MaxEnt prediction. These results may provide potentially useful information to establish risk of human disease and direct prevention programs.