We propose a Bayesian hierarchical model for high-dimensional “multi-type" spatio-temporal data. By “multi-type" data we mean a collection of correlated datasets that have different distributional assumptions (e.g., continuous and count-valued observations). The literature on parametric models for multi-type spatio-temporal data is severely limited, and development is needed for these kind of datasets. For example, the Centers for Disease Control and Prevention (CDC) database provides counts of moralities related to lung, cardiovascular, and stroke in 2011 and Daily Fine Particulate Matter (PM2.5) over US counties. To jointly model high-dimensional multi-type spatio-temporal data, we capitalize on the shared conjugate structure between the multivariate log-gamma (to model PM2.5) and Poisson (to model mortality) distributions. This general hierarchical modeling framework can be applied to many different datasets besides CDC, and hence, is of independent interest. To illustrate the performance of our model, we use a simulation study and a demonstration using a large dataset from CDC.