West Nile virus is the most significant arbovirus in the United States in terms of both morbidity and mortality. West Nile exists in a complex transmission cycle between avian hosts and the arthropod vector, Culex spp. mosquitoes. Human spillover events occur when humans are in close proximity to vector populations with high rates of infection. Predicting these rates of infection and therefore the risk to humans is not straightforward. In this study, we evaluate the hydrological and meteorological drivers associated with mosquito biology and viral development to determine if these associations can be used to forecast seasonal West Nile risk in the Coachella Valley of California. To test this, we developed and tested a spatially-resolved ensemble forecast model of West Nile virus transmission in the Coachella Valley using 17 years of mosquito surveillance data and NLDAS-2 environmental data. Our multi-model inference system indicated that the combination of a cooler and dryer winter followed by a wetter and warmer spring and a cooler than normal summer was most predictive of West Nile positive mosquitoes in the Coachella Valley. The ability to make accurate early season predictions of West Nile risk could allow local abatement districts and public health entities to implement early season interventions such as targeted adulticiding and public messaging before human transmission occurs. Such early and targeted interventions could better mitigate the risk of West Nile virus to humans in the Coachella Valley.