Modelling the potential spatial distribution of mosquito species using three different techniques

Area of Science:

• Ecology
• Epidemiology
• Biomathematics

Background:

• Spatial distribution models are crucial for assessing vector-borne disease risk.
• Understanding mosquito habitat suitability aids in predicting disease spread.
• This study evaluates different modeling techniques for mosquito distribution.

Purpose of the Study:

• To define environmental conditions suitable for three indigenous mosquito species using species distribution modeling.
• To compare the performance of non-linear discriminant analysis, random forest, and generalized linear models.
• To identify key environmental variables influencing mosquito occurrence.

Main Methods:

• Employed three distinct modeling techniques: non-linear discriminant analysis (NLDA), random forest (RF), and generalized linear model (GLM).
• Investigated environmental suitability for Culiseta annulata, Anopheles claviger, and Ochlerotatus punctor in the Netherlands.
• Compared models based on suitability maps, associated environmental variables, and evaluation metrics.

Main Results:

• Precipitation, temperature, and population density were linked to Cs. annulata and An. claviger.
• Land surface temperature and vegetation indices correlated with Oc. punctor presence.
• Random forest models exhibited the highest classification accuracy (specificity 0.89–0.91, sensitivity 0.64–0.95).

Conclusions:

• Consistent spatial patterns were observed across models, but prediction levels varied.
• Random forest provided the most accurate predictions, outperforming NLDA and GLM.
• Model outputs can inform epidemiological models for vector-borne disease management.