Under anthropogenically-altered conditions, cyanobacteria may form harmful algal blooms (cHABs) that can be toxic and disrupt ecosystem function. Developing tools to predict cHABs is an increasingly important task, but such tools have been difficult to develop. In our study, we contribute to the development of a predictive model for cHAB formation in the waters of southern New Jersey by statistically screening water quality data from five polymictic reservoirs that were sampled weekly from June through September 2019. The correlation structure of water quality variables differed between the reservoirs in a way that suggests that the mean correlation coefficient is elevated for reservoirs experiencing a cHAB. Some water quality variables are unlikely to be useful for predictive modeling, but among those that do have utility, those measurements were obtained under natural field conditions, semi-controlled conditions in the field, and controlled conditions in the lab. The number of principal components (PC axes) required to describe variation in the water quality data differed between reservoirs in a way that suggests reservoirs experiencing cHABs have less complex covariance structures. Collectively, these results indicate that predictive modeling of cHAB formation should be possible.