Experimental designThe experiment was conducted from 16 to 26 June 2008. In the first treatment, oxygen was added to the hypolimnion. In the second, nutrients were added to the epilimnion. The third treatment simulated a mixing event (overturn). There also was a control enclosure with no treatment and sampling of the ambient lake water. Throughout this manuscript, we refer to these as Oxygen, Nutrient, Mix, Control and Ambient.Twelve limnocorrals were constructed as enclosures for the experiment. Each limnocorral was cylindrical and extended vertically from the surface of the lake to the sediment (approximately 4 m). The total volume was approximately 5050 l. Details of limnocorral construction are provided in online Supporting Information.The limnocorrals were deployed on 15 June 2008 to allow the sediment and water column to stabilize before treatment on 16 June 2008. The limnocorrals were deployed in a random spatial arrangement throughout the lake, at a maximum depth of 3.25 to 3.5 m. Replicates from each treatment were instrumented with a chain of HOBO temperature sensors (Onset), and one replicate from each had a self-logging DO sonde (Yellow Springs Incorporated) in the hypolimnion (3 m depth). More thermistors were deployed in the Oxygen and Mix treatments because thermal stratification was important for evaluating success of these treatments.For the Mix treatment, a 60 cm flat disk was raised and lowered between 3.5 m depth and the lake surface. Holes were drilled through the disk surface to increase turbulence (Sanford, 1997; Regel et al., 2004). A brick was tied underneath the disk to maintain stability. We manually oscillated the disk every 10 min for 1 h and then, after a 1 h break, continued for an additional hour. Temperature and DO profiles were monitored within the limnocorral with a hand-held probe to track mixing progress.The goal of the Oxygen treatment was to aerate the hypolimnion water without allowing it to mix with the epilimnion. This treatment was achieved by pumping hypolimnion water from the bottom of the limnocorral into an external cooler where the water was aerated with bubble diffusers, and then returned to the bottom of the limnocorral (Fig. S1). Valves on a compressed air cylinder were used to control the delivery of air to the coolers. One cooler was maintained for each replicate limnocorral. Thermally insulated tubing was used to transport water. A thermistor was deployed in each cooler to ensure ambient hypolimnion temperature was maintained. The water was removed and returned using two linear diffusers that were 0.6 m in length, spanning a depth range of approximately 2.5–3.1 m within the hypolimnion. The diffusers faced inward with 0.5 m fixed distance between them, retained by a plastic divider. This treatment was applied continuously over 3 h, until DO concentrations increased.The Nutrient treatment was achieved by adding ammonium chloride (NH4Cl) and potassium phosphate monobasic (KH2PO4) as N and P sources. These compounds were chosen because they are commonly bioavailable sources of nutrients. P was added to the epilimnion to achieve a final concentration of 3 micro g P l−1, which was approximately the average concentration expected in the mixed water column. This value was based on nutrient analyses from integrated water collected on 9 June 2008 in North Sparkling Bog, a week prior to experiment start. Similarly, N was added to achieve a final concentration of 70 µg N l−1. The limnocorral s epilimnion volume (0–2 m integrated depth) was calculated to be 2520 l, and we used the molar mass to determine the amount of each nutrient added to the epilimnion to achieve the expected mixed concentration. Dry chemicals were dissolved into to 500 ml of surface water from each limnocorral, and then added into each separately. Rationale for directly manipulating only one layer in the Oxygen and Nutrient treatments is given in the online Supporting Information.The Control limnocorrals were left undisturbed. To prevent mixing during equipment removal, all tubing was left inside the limnocorrals until the experiment ended.