The utility of agglomerative clustering methods for understanding dynamic systems that do not have a well-defined periodic structure has not yet been explored. We propose using this approach to examine the association between disease and weather parameters, to compliment the traditional harmonic regression models, and to determine specific meteorological conditions favoring high disease incidence. We utilized daily records on reported salmonellosis and non-specific enteritis, and four meteorological parameters (ambient temperature, dew point, humidity, and barometric pressure) in Barnaul, Russia in 2004-2011, maintained by the CliWaDIn database. The data structure was examined using the t-distributed stochastic neighbor embedding (t-SNE) method. The optimal number of clusters was selected based on Ward distance using the silhouette metric. The selected clusters were assessed with respect to their density and homogeneity. We detected that a well-defined cluster with high counts of salmonellosis occurred during warm summer days and unseasonably warm days in spring. We also detected a cluster with high counts of non-specific enteritis that occurred during unusually "very warm" winter days. The main advantage offered by the proposed technique is its ability to create a composite of meteorological conditions-a rule of thumb-to detect days favoring infectious outbreaks for a given location. These findings have major implications for understanding potential health impacts of climate change.