Attempts to understand the distribution of disease and disease risk in populations has assimilated data and model-based analyses since before Snow first mapped cholera cases around a London water pump in 1854. Models that accurately forecast the behavior of infectious diseases represent one of the most appreciated contributions of ecology to human welfare world wide. The assimilation of data with disease models has followed two somewhat disparate traditions.  The epidemiological tradition has emphasized the statistical analysis of relationships in data, but has not explicitly represented disease processes. The theoretical tradition has emphasized the mathematical analysis of models representing disease processes, notably transmission, recovery and the epidemic dynamics that result from them. 

Results/Conclusions We review recent developments in these approaches with an emphasis on methods for identifying and estimating uncertainties needed to support reliable forecasting. We point to examples that incorporate uncertainty in model structure into estimates of interest, particularly R₀, the net reproductive rate of the disease.  We also focus on work that has explicitly partitioned process variance and observation error.  Using examples from modeling brucellosis in bison in Yellowstone National Park, we advocate combining traditional methods in population modeling, (e.g., age and stage structured models and mark-capture recapture methods) with traditional methods for representing disease transmission, as in the SIR family of models. Advances in data acquisition that deploy consistent methodology across diverse spatial scales (e.g., NEON) to examine distributions of vectors hosts and related pathogen burden could fundamentally advance understanding of zoonotic disease and the ability to forecast disease intensity. We specifically explore potential for improved understanding of zoonotic disease using prominent case studies including West Nile virus, Lyme disease, Hanta Pulmonary Syndrome.