Methods

Data were collected at 5 minute intervals using in-situ automated sensors (sondes). All measurements and samples were collected from a stationary raft over the deepest part of the lake.
Sondes were suspended from floats with probes at a depth of 0.75m below the surface. Sonde sensors were cleaned daily in the field and calibrated monthly following manufacturer guidelines. Peter and Paul lakes were each monitored with two YSI multiparameter sondes (model 6600 V2-4) fitted with optical DO (model 6150), pH (model 6561), optical Chl-a (model 6025), and conductivity-temperature (model 6560) probes. Sensor measurements were made at 0.75 m every 5 min and were calibrated weekly. PAR was measured and the UNDERC meteorology station maintained by the University of Notre Dame or by the North Temperate Lakes Weather Station at Woodruff Airport.
Outliers were replaced by NA. Occasional gaps in the record due to instrument cleaning are NA.

Methods

Light (PAR) extinction coefficient is calculated by linearly regressing ln (FRLIGHT (z)) on depth z where the intercept is not constrained. FRLIGHT(z) = LIGHT(z) or DECK(z) where LIGHT(z) is light measured at depth z and DECK(z) is light measured on deck (above water) at the same time. For open water light profiles, the surface light measurement (depth z = 0) is excluded from the regression. For winter light profiles taken beneath the ice, the first light data are taken at the bottom of the ice cover and are included in the regression. The depth of uppermost light value is equal to the depth of the ice adjusted by the water level in the sample hole, i.e., the depth below the surface of the water. The water level can be at, above or below the surface of the ice. If the water level was not recorded, it is assumed to be 0.0 and the calculated light extinction coefficient is flagged. If ice thickness was not recorded, a light extinction coefficient is not calculated. For light data collected prior to March, 2007, light values less than 3.0 (micromolesPerMeterSquaredPerSec) are excluded. For light data collected starting in March 2007, light values less than 1.0 (micromolesPerMeterSquaredPerSec) are excluded. Except for bog lakes before August 1989, a light extinction coefficient is not calculated if there are less than three FRLIGHT values to be regressed. For bog lakes before August 1989, a light extinction coefficient is calculated if there are least two FRLIGHT values to be regressed. In these cases, the light extinction coefficient is flagged as non-standard. FRLIGHT values should be monotonically decreasing with depth. For light profiles where this is not true, a light extinction coefficient is not calculated. For samples for which light data at depth are present, but the corresponding deck light are missing, a light extinction coefficient is calculated by regressing ln (LIGHT (z)) on depth z. Note that if actual deck light had remained constant during the recording of the light profile, the resulting light extinction coefficient is the same as from regressing ln(FRLIGHT(z)). In these cases, the light extinction coefficient is flagged as non-standard. Oxygen and Temperature: We sample at the deepest part of the lake, taking a temperature and oxygen profile at meter intervals from the surface to within 1 meter of the bottom using a YSI Pro-ODO temporDO meter. We sample biweekly during open water and approximately every five weeks during ice cover. Protocol Log: Prior to 2011, we used a YSI Model 58 temporDO meter.