The seawater samples were stained with primuline (Direct Yellow 59, Sigma-Aldrich Co), fixed with a solution containing 3.6% (v/v) glutaraldehyde and 10% (v/v) glycerol (final concentrations), and gently filtered onto black-stained 25 mm diameter 0.4-µm pore-size Nucleopore filters. Filters were mounted on slides and analyzed by epifluorescence microscopy. The method is a modification from Grebecki (1962), Hobbie et al. (1977) and Caron (1983) with the glycerol added to reduce the damage of particularly small delicate protists during filtration as described in Sazhin et al. (2007). FlowCAM analysis of mesocosm seawater The FlowCAM was run in autoimage-mode, using 4x magnification to analyze particles ranging between 15 and 1000 µm. The water samples were kept in dim light at 4°C until analyzed, within 4 h after sampling. Each sample was run for ca. 30 min, corresponding to 5.7 ml of analyzed volume. The relevant context capture property chosen to do the analysis was distance to nearest neighbor = 50 µm. All the image collages were manually post-analyzed in order to identify the particles in question. Particle sizes were estimated from area-based diameter determined by the FlowCAM VISP 2.2 software following Jakobsen & Carstensen (2011). Colonies of P. pouchetii are composite clusters of cells spaced in colonies of various sizes. When the colonies are forming dense blooms, the FlowCAM cannot distinguish whether a cluster is a part of colony or if a cluster is a part of a neighbouring colony. The FlowCAM therefore captures clusters of colonies. S. marinoi and P. pouchetii are colonial structures of multiple cells, and we developed scaling relationships between colony area and cell number by manually counting the number of cells in either a diatom chain or a P. pouchetii colony (n > 100) and related the cell number to particle area. We further developed scaling relationships between cluster areas by manually counting the number of cells in P. pouchetii clusters (n > 100). The same procedure was used for estimating S. marinoi cells, where we manually counted the number of cells per chain. The cell numbers were related to either cluster area (P. pouchetii clusters) or chain length (S. marinoi), and the relationships were then used to estimate the total number of P. pouchetii or S. marinoi cells in colonies or chains, respectively, by sample. Cell concentrations were estimated by correcting for the analyzed volume. The resulting regression equation used for estimation the number of P. pouchetii cells in colonies and colony fragments was Cells = ABD 1.834 * 0.028 (R 2 = 0.83), where ABD (area base diameter) is an output parameter from