Appendix C: from S. R. Hall et al., “Inedible Producers in Food Webs: Controls on Stoichiometric Food Quality and Composition of Grazers”

We used a fully factorial experimental design to examine the simultaneous effects of light availability, absolute nutrient (nitrogen and phosphorus) availability, relative N : P availability, and trophic structure on zooplankton community composition and the abundance and stoichiometry of algal producers. During May–June 2000, we created gradients of light and nutrient supply and predation risk in 300-L cattle tanks. To each tank, we added silica sand substrate, well water, and inorganic nitrogen (N) and phosphorus (P; NaNO3 and NaH2PO4, respectively) to raise nutrient concentrations to target N : P ratios (nutrient ratio treatments: 5 : 1, 14 : 1, and 50 : 1) and supply (nutrient supply level treatments: high was 10 times the low value). Also, we reduced light availability in half of the tanks (light treatment) by 90% using neutral shade cloth. In other studies, we report results from other treatments of this experiment (one with predators and one without grazers; Hall et al. 2004, 2005), but these are not relevant here. We inoculated each mesocosm with diverse assemblages of algae and zooplankton at fortnightly intervals. These inoculae were collected along wide natural light–nutrient gradients from local ponds (proximate to Kellogg Biological Station and within Barry and Middleville State Game Areas, Kalamazoo and Barry Counties, MI, respectively). Each cattle tank also received 30 Physa spp. snails and 30 Rana catesbeina tadpoles to graze and remineralize nutrients bound in benthic algae. Finally, we added nutrients weekly to approximately maintain target levels (TN and TP levels of 700 and 14, 370.4 and 26.5, and 221.4 and 44.3 mg nutrient L 1 for 50 : 1, 14 : 1, and 5 : 1 nutrient ratio treatments, respectively; high-nutrient-supply-level treatments received 10 times these levels) by assuming a 5% day 1 loss rate from the water column (M. A. Leibold and V. H. Smith, unpublished data). These target levels of N and P correlated perfectly with each other on log scales ( ) and with N : P ratio ( for P, for N). The factorial experimental design R p 1 R p 1 R p 1 was