Abstract:
Pakistan is recently experiencing a wave of Dengue epidemics that has claimed numerous lives.
Lahore the provincial metropolis of Punjab and Swat a major city in Khyber Pakhtunkhwa are
among the hard hit areas. The present study targeted the two cities for determining spatial
distribution of Dengue mosquitoes; a task essential for an effective control strategy and
predicting near future outbreaks of the dengue fever. We used general-purpose ( "Maxent")
maximum entropy model; undertaking an assembly of mosquitos' occurrence records, associated
environmental covariates also taking into account the issue of multi-collinearity.
Our results suggest: 1. Spatial patterns of species occupancy has a widespread distribution in
Lahore. This is perhaps due to the ecological homogeneity of the study area 2. Aedes aegypti has
a fragmented population structure in Swat where the urban patches provide refuge to the species
in an otherwise hostile heterogenous environment. Perhaps the ecological overlap of vector and
human population has led to the invasion of this species in Swat. Thus the distribution and
spread of Aedes aegpti in this area can be traced back to anthropogenic activities. 3. Modeling
results reveal that Maxent tends to perform better in dynamic regimes like Swat (AUC=0.98) that
are characterized by vast landscape diversity and biogeographic gradients as compared to
homogenous areas like Lahore (AUC=0.85). 3. Species distribution models are sensitive to
multi-collinearity, but sometimes changes in collinearity structure of Predictors can lead to
dramatic loss of prediction accuracy (both the training AUC (AUC=0.923 (for all variables)
>AUC= 0.882(for few variables)) and Test AUC (AUC= 0.881 (for all variables)
>AUC=0.850 (for Jew variables)) decreased as we remove so-called correlated variables, and
use rest of the plain variables both for Lahore.
