Methods

Study area
The experimental (McDermott Lake; 46.00299280, -90.16081610) and reference (Sandy Beach Lake (46.10614350, -89.97131020) systems are located in Iron County in Northern, Wisconsin. McDermott Lake has a surface area of 33.1 ha and a mean depth of 3.0 m, while Sandy Beach Lake has a surface area of 44.5 ha and a mean depth of 2.1 m. Maximum depth in McDermott Lake is 5.7 m and in Sandy Beach Lake is 4.0 m. Both lakes include a variety of substrates (e.g., rock, gravel, and sand) and areas of submerged and emergent vegetation. At the start of the study, McDermott and Sandy Beach Lake fish communities were similar with high Centrarchidae spp. (i.e., centrarchid) abundances, few adult walleye, and a history of natural walleye recruitment. Other species present include yellow perch, northern pike, muskellunge, black bullhead, and white sucker (Table 1).
Fish sampling
Standardized surveys
From 2017-2020, we conducted fish population sampling on the experimental (McDermott) and reference (Sandy Beach) lakes. Every year of the study, we have conducted standardized monitoring surveys employing numerous sampling techniques to detect changes in the fish community relative to 2017-baseline information (Fig. 1). Sampling began immediately following ice-out (~mid-April) with the deployment of five fyke nets for one week. The fyky surveys served two purposes: 1) to serve as the capture method for marking walleye as part of the mark-recapture survey to attain a population estimate, and 2) to estimate other focal species (i.e., black crappie, yellow perch, muskellunge, northern pike) relative abundances. During these surveys, all collected walleye were measured (total length (TL); mm), sexed, checked for a Passive Integrated Responder (PIT) tag, and if one was not present, marked with a unique PIT tag. We also removed a dorsal spine sample for aging. Adult (mature) walleyes were defined as all fish 381 mm and all fish for which sex could be determined (regardless of length). Walleye of unknown sex <381 mm were classified as juvenile (immature). McDermott and Sandy Beach Lakes have both had walleye population estimates previously conducted by the Wisconsin Department of Natural Resources (WDNR) therefore the goal was to mark 10% of the anticipated spawning population (based off of previous population estimates). Marking continued until the target number was reached or spent females began appearing in the fyke nets. Walleye were recaptured using an AC boat electrofishing survey within one week (typically 1-4 days) after netting and marking were completed. In each lake, the entire shoreline was sampled. All captured walleyes were measured and examined for marks. Based on electrofishing mark-recapture data, population estimates were calculated using the Chapman (1951) modification of the Petersen Estimator as:
N=((M+1)(C+1))/((R+1))
where N was the population estimate, M was the number of fish marked and released, C was the total number of fish captured and examined for marks in the recapture sample, and R was the total number of marked fish observed in C. The Chapman Modification method was used because it provides more accurate population estimates in cases when R is relatively small (Ricker 1975).

From early-May to mid-June, we sampled larval fishes using a 1,000m mesh conical ichthyoplankton net towed for five minutes immediately below the lake surface. Weekly samples were taken at night at five sampling locations in each lake. Each lake was divided into five quadrats and sites were established at a randomly selected nearshore (<100 m of shore) location in each quadrat on each sampling date. Once selected, locations remained fixed throughout the study. Volume of water filtered during each tow was estimated using a General Oceanics© model 2030R flowmeter mounted in the center of the net frame. Samples were transferred to containers and stored in 90% ethanol. Collected fishes were identified to species according to Auer (1982) and enumerated.
In addition to the adult walleye population, we were interested in estimating the size of the adult largemouth bass population. We performed early summer (late-May) mark-recapture surveys using AC boat electrofishing (Wisconsin‐style; AC; 2.0–3.0 amps, 200–350 V, 25% duty cycle with two netters) to sample largemouth bass. Collected largemouth bass were measured, checked for a top caudal fin clip, and if not present, marked with a top caudal fin clip and released. Adult (mature) largemouth bass were defined as all fish 203 mm. We aimed to recapture 10% of the marked population. Largemouth bass are in very low abundance in Sandy Beach Lake therefore a population estimate was not possible. In McDermott Lake, due to the small population size of largemouth bass we completed multiple marking surveys from late-May to early June to achieve this recapture rate. From electrofishing mark-recapture information, population estimates were calculated using the Schnabel (1938) modification of the Lincoln-Petersen method:
where N was the population estimate, M was the number of fish marked and released in sample t, C was the number of fish captured in sample t, and R was the number of fish already marked when caught in sample t.
To obtain centrarchid population demographic data, current standardized WDNR surveys of inland lakes consist of early summer (water temperature range = 13.0–21.0°C) AC boat electrofishing surveys or mid‐summer (18.3–26.7°C) mini‐fyke net (Simonson et al. 2008). To encompass this range of water temperatures, we performed a combination of surveys starting at the end of May with an AC boat electrofishing survey. Then, fish were sampled once monthly when lake surface water temperatures were ≥13.0°C in both lakes (June–September). Both lakes were sampled during 1‐week each month using three gears (AC boat electrofishing, mini-fyke nets, cloverleaf traps). Lakes were sampled on consecutive nights in each 1‐week period but only one gear type was employed per night.
All gears sampled shallow shorelines (0–5 m from bank, depth ≤2 m) and were deployed in fixed locations following standard approaches (Bonar et al. 2009). Sampling locations were evenly distributed along the shoreline of the lake, and all gears were deployed in similar habitat types. Five 10‐min nighttime boat electrofishing (Wisconsin‐style; AC; 2.0–3.0 amps, 200–350 V, 25% duty cycle) transects were conducted using two dipnetters. Five mini‐fyke nets (0.9‐m × 0.61‐m frames, 3.2‐mm mesh [bar measure], 7.6‐m‐long lead, and a double throat) were deployed in areas where the net frames would be in 1.0–1.5 m of water, and leads were fixed onshore. Five cloverleaf traps (three lobed, height = 41 cm, 50 cm diameter, 6.0‐mm bar wire mesh with 12.7‐mm‐wide openings between lobes, and an attractant [liver]) were deployed in littoral habitats. Both mini‐fyke nets and cloverleaf traps were set in early afternoon, fished overnight, and retrieved the following afternoon (~24‐h soak time). All catches were standardized according to gear-specific effort. For boat electrofishing, catch per unit effort (CPUE) is presented as fish/hr. For mini-fyke nets and cloverleaf traps, CPUE is calculated as fish/net night or fish/trap night.
To quantify walleye recruitment in each lake, we employed multiple gears throughout the sampling season including micromesh gillnets, beach seines, and boat electrofishing. In late July/early August, we deployed four 46-m x 1.2-m gillnets with 0.95-cm bar mesh. Sampling locations were evenly distributed along the shoreline and locations were fixed each year. Gillnets were set at night and at depths ranging from 0-5 m. Set duration ranged 1-2 hours to minimize bycatch, thus catches were standardized to age-0 walleyes collected per 10 hours of soak time. In late August, we pulled .24-m long beach seines with 0.64-cm mesh at five sites in each lake. Sites were chosen to represent a variety of habitat types and based on ability to effectively use the seine. Seining sites remained fixed for the duration of the study. Seines were used during daylight hours on each lake. Catch per unit effort was calculated as the number of individuals per seine haul. When water temperatures fell below 21°C (early September), we sampled age-0 walleyes using nighttime boat electrofishing (Wisconsin‐style; AC; 2.0–3.0 amps, 200–350 V, 25% duty cycle, two netters). The entire shoreline of each lake was sampled. Surveys were conducted prior to walleye fingerling stocking. All collected walleye were measured (TL, mm). Catch per unit effort was calculated as the number of age-0 walleyes per meter shoreline.
Removal efforts
In addition to standardized surveys in our experimental lake, in 2018 we began centrarchid removal efforts using a variety of techniques including fyke nets, boat electrofishing, mini-fyke nets, and cloverleaf traps (Fig. 1). Following annual spring fyke net surveys, fyke nets remained in the experimental lake to remove centrarchids. In 2018, we sampled 10 fyke nets from May 14 to June 7 when centrarchid catches started to decline. Due to personnel limitations, in 2019 and 2020 only five fyke nets were used from late spring (May 9, April 30) until late June (June 27, June 25). Additionally, we sampled five mini-fyke nets and 21 cloverleaf traps from late May through mid-August. All gears were emptied every 1-2 days and sites were rotated to maximize centrarchid catches. Collected fish were identified to species and measured (TL, mm). Centrarchid species were retained while other species were returned to McDermott Lake.

Methods

Ice and snow sampling was conducted weekly from 14 January to 30 March, 2010 on Lake Mendota when the ice was safe to walk on. A Kovacs Mark III core drill, manufactured by Ice Coring and Drilling Service (ICDS), Space Science and Engineering Center (SSEC) UW Madison, was used to collect ice cores. Snow depth was also measured at the locations where ice cores were sampled. All measurements were made in centimeters. Blue ice can be defined as the portion of the ice core that is strictly frozen lake water. White ice can be defined as “snow ice,” which occurs when water rushes through cracks in the ice and soaks the overlying snow, resulting in a mixture of ice and snow that subsequently freezes. Total ice is blue ice + snow ice. Finally, snow depth was calculated as the average of 10 snow depth samples at each sampling location.

Methods

Raw data (in English units) were assembled by Douglas Clark - Wisconsin State Climatologist. Data were converted to metric units and adjusted for temporal biases by Dale M. Robertson. For adjustments applied to various parameters see Robertson, 1989 Ph.D. Thesis UW-Madison. Adjusted data represent the BEST estimated daily data and may be raw data. Data collected at Washburn observatory, 8-1-1883 to 9-30-1904. Data collected at North Hall, 10-1-1904 to 12-31-1947 Data collected at Truax Field (Admin BLDG), 1-1-1948 to 12-31-1959. Data collected at Truax Field (Center of Field), 1-1-1960 to Present. Much of the data after 1990 were obtained in digital form from Ed Hopkins, UW-Meteorology. Data starting in 2002-05 were obtained from Sullivan at <a href="http://www.weather.gov/climate/index.php?wfo=mkx%20">http://www.weather.gov/climate/index.php?wfo=mkx</a> ,then go to CF6 and download monthly data to Madison_sullivan_conversion. Since Robertson (1989) adjusted all historical data to that collected from 1884-1989; no adjustments were applied to the recent data except for (1) wind and (2) estimated vapor pressure:(1) Wind after January 1997, and only wind from the southwest after November 2007, was extended by Dale M. Robertson and Yvonne Hsieh.In 1996, a discontinuity in the wind record was caused by change in observational techniques and sensor locations (Mckee et al. 2000). To address the non-climatic changes in wind speed, data from MSN were carefully compared with those collected from the tower of the Atmospheric and Oceanic Science Building at the University of Wisconsin-Madison, see http://ginsea.aos.wisc.edu/labs/mendota/index.htm. Hourly data from both sites (UMSN,hourly and UAOS,hourly) during 2003&ndash;2010 were used to form a 4&times;12 (four components of wind direction &times; 12 months) matrix (K4,12) of wind correction factors, yielding UAOS,daily= Ki,j&times;UMSN,daily. The comparison results indicated that the MSN weather station reported a higher magnitude in winds out of the east by 5% and lower magnitude in winds out of the west and south by 30% and 10%. The adjusted wind data (=Ki,j&times;UMSN,daily) were therefore employed and used in the model simulation. After adjustments, there was a decrease in wind velocities starting shortly before 1996. Overall the adjusted wind data had a decline in wind velocities of 16% from 1988&ndash;93 to 1994&ndash;2009) compared to a 7% decline at a nearby weather station with no known observational changes (St. Charles, Illinois; 150 km southeast of Lake Mendota). (2) Estimated vapor pressure was updated (after April 2002) by using the equation from DYRESM for estimation of vapor pressure (a function of both air temperature and dew point temperature); where a=7.5, b=237.3, and c=.7858.