Harmful algal blooms (HABs), in particular those consisting of the cyanobacteria \textit{Microcystis}, are becoming increasingly more common across the globe. Despite the growing body of evidence that suggests vertical heterogeneity of \textit{Microcystis} can be a precursor to HAB formation, the abiotic drivers of vertical distribution of \textit{Microcystis} are poorly understood in the field environment. The prediction of subsurface cyanobacteria is also pertinent because subsurface concentrations are not easily recognizable to the public or lake system managers, creating an unnoticed safety hazard. High-frequency temporal and vertical data were collected from an Eulerian research station anchored in a stratified and eutrophic lake for five months. Data show that the magnitude of the subsurface \textit{Microcystis} concentration peak and the center of gravity of the deep cyanobacteria layer are statistically significantly mediated by the thermal structure of the lake. The peak subsurface cyanobacteria biovolume scales linearly with the thermocline depth and temperature, whereas the center of gravity of the subsurface cyanobacteria biovolume scales linearly with the mixed layer depth and temperature. Furthermore, our data suggest there is a seasonal evolution of the subsurface cyanobacteria center of gravity that could potentially indicate timing of HAB onset. Based on easily measured parameters associated with the vertical lake temperature profile and meteorological conditions, we provide evidence of predictable trends in subsurface cyanobacteria variables.

1. Harmful algal blooms are increasing in both severity and frequency across the globe. Many bloom-forming species are capable of vertical motility and colony formation. The cyanobacterium Microcystis aeruginosa is a common example of such a species, yet current models poorly predict vertical distributions of M. aeruginosa. 2. To couple the hydrodynamics, buoyancy, and the colony dynamics of Microcystis, we present a system of one-dimensional advection-diffusion-aggregation equations with Smoluchowski aggregation terms. 3. Results indicate Smoluchowski aggregation accurately describes the colony dynamics of M. aeruginosa. Further, transport dynamics are strongly dependent on colony size, and aggregation processes are highly sensitive to algal concentration and wind-induced mixing. Both of these findings have direct consequences to harmful algal bloom formation. 4. While the theoretical framework outlined in this manuscript was derived for M. aeruginosa, both motility and colony formation are common among bloom-forming algae. As such, this coupling of vertical transport and colony dynamics is a useful step for improving forecasts of surface harmful algal blooms.