West Nile Virus (WNV) fever is a vector-borne disease maintained in nature in a mosquito-bird-mosquito transmission cycle (Campbell et al., 2002; Hayes, 1989). Data of WNV in the US clearly reveals decline of its maximal prevalence observed in the states where the disease first appeared in 1999. This fact reveals that WNV is evolving towards an endemic situation in temperate North America, with annual amplification events and regular epizootics and epidemics (Cruz-Pacheco et al. 2009). From a public health perspective, it is very important to understand how new large epidemic peaks could be generated from this endemic situation. Here we develop a mathematical model to understand the transmission dynamics of WNV in context of two sources of heterogeneity: preferential feeding of available host species and spatial clustering of vectors and hosts in order to predict infection risk. WNV transmission is model at post-breeding communal roosting sites because those are sites where and when West Nile transmission is most intense and therefore where infection risk to humans is highest. Preliminary results of our 2 vectors, 2 hosts model imply post-breeding communal roosting behavior to present certain conditions that create a favorable environment for infection to spillover but this largely depends on the competency of the host reservoir and requires more exploration. The modeling method developed here is broadly applicable to other vector-born zoonoses and the results facilitate public health agency to indentify sites and target specific avian population to implement cost-effective control measures to prevent WNV transmission.