Detailed descriptions of study sites, data collection, and metabolism modeling
Sequoia Lakes
The Sequoia lakes are a cluster of 5 subalpine and alpine lakes and ponds in Sequoia National Park, including Emerald Lake, a long-term research site 1,2. These lakes and ponds are shallow (1.9 - 10 m maximum depth), occupy granitic basins with little soil development, and are ice and snow-covered for up to 8 months each year3. Emerald Lake is dimictic, whereas Topaz Lake and all the ponds are polymictic during the ice-free season. Each lake or pond was instrumented with dissolved oxygen (DO; PME miniDOT) and temperature (Onset HOBO) sensors in the deepest part of the water column during the open-water seasons in 2020 and 2021. DO sensors in Emerald Lake and Topaz Lake were equipped with wiper units that cleaned sensors every 12 hours. The DO sensors in ponds (EML Pond 1, TOK 11 Pond, Topaz Pond) were manually cleaned every 2 weeks in 2020. Wiper units were installed on pond DO sensors in 2021. Meteorological data were available from two weather stations, one near the shore of Emerald Lake (elevation 2810 m.a.s.l.) and one near the shore of Topaz Lake (3230 m.a.s.l.). Water samples for chemical analysis (TDN, TDP, and chlorophyll-a) were collected 1- 6 times per season via surface grabs in the ponds, or from 1 m depth using a peristaltic pump or Van Dorn sampler in Emerald Lake and Topaz Lake. Water samples were processed and analyzed following methods in Sadro et al. 4. A single light profile in August was collected from Emerald Lake and Topaz Lake to estimate light attenuation. Light profiles were collected at the deepest point in each lake using a PME miniPAR lowered at 0.5 m intervals. We did not collect light profiles in ponds due to their shallow depth, so we used the kd value from Topaz Lake. PM2.5 data from 2021 were downloaded from a PurpleAir sensor in Three Rivers, CA. No PM2.5 data were available in 2020.
Lake Tahoe and Dulzura Lake
The two study sites in the northern Sierra Nevada mountains include Lake Tahoe and Dulzura Lake (Figure 2, Table 1). Both lakes occupy forested, subalpine watersheds. Lake Tahoe is a large, deep, monomictic and ice-free lake. Dulzura Lake is a small and shallow lake 25 km northwest of Lake Tahoe. DO and temperature were measured with RBR³T.ODO sensors at two nearshore stations on the western shore of Lake Tahoe in 2020 and 2021: one at Tahoe City (TC) and one at Tahoe Vista (TV)5. Sensors were installed at ~ 1 m depth and were equipped with wipers. Hourly SW radiation, wind speed, and air temperature data were obtained from a weather station in Tahoe City (USCG)5. Water samples for TDN, TDP and chlorophyll-a were collected from the lake surface 5 - 6 times between June 1 and November 1 at a mid-lake station (MLTP) and were processed and analyzed according to methods in Liston et al. 6. PM2.5 data from 2020 and 2021 were downloaded from a PurpleAir sensor in Tahoe City, CA.
In Dulzura Lake, DO and water temperature were measured with PME miniDOT loggers at 3-m depth in a pelagic and nearshore location in 2021. DO sensors were manually cleaned once a month. Wind speed and shortwave radiation were measured hourly at a weather station (TADC1; https://mesowest.utah.edu/cgi-bin/droman/mesomap.cgi) located 10 km northeast of the lake. Light (PAR; 400–700 nm) and temperature profiles were collected every month at 0.5 m intervals using a Biospherical Instruments 2104P radiometer at the deepest part of the lake. Surface water samples for TDN, TDP, and chlorophyll-a were collected 4 times at the deepest part of the lake. PM2.5 data were downloaded from a PurpleAir sensor in Truckee, CA.
Castle Lake
Castle Lake is a dimictic, seasonally ice-covered lake situated in a sub-alpine watershed in the Klamath mountains (Figure 2, Table 1)7. DO and temperature measurements were made using PME MiniDOT loggers at 3-m depth in a pelagic and littoral location during 2018. DO sensors were manually cleaned every two weeks. Wind speed and SW radiation were measured hourly at a weather station (Weatherhawk) on the shore of Castle Lake. Light (PAR; 400–700 nm) and temperature profiles (every 0.5 m) were obtained weekly using a Biospherical Instruments 2104P radiometer at the deepest part of the lake. Water samples were collected at 0 m depth 6 times between June 1- November 1 (see Scordo et al.7 for collection and analytical methods). PM2.5 data from 2018 were downloaded from a PurpleAir sensor in Yreka, CA.
Clear Lake
Clear Lake is a large, shallow, hyper-eutrophic lake located in the northern Coast Range mountains (Figure 2, Table 1). Clear Lake has three different sub-basins: the Upper Arm, Lower Arm, and Oaks Arm. The Upper Arm (UA) is the largest and shallowest basin, and directly receives >90% of the watershed runoff. A passage at the east end of the UA connects it with two smaller and deeper basins, the Lower Arm (LA), which connects to the only outlet at the southeast end of the lake, and the Oaks Arm (OA), which is the smallest basin. The lake is polymictic but can stratify for weeks during summer. Temperature (RBR soloT thermistors) and DO (PME miniDOTs with wipers) measurements were recorded at the deepest location within the LA and OA in 2020 and 2021. Meteorological data were obtained from a weather station installed on the shoreline at Buckingham Point (BKP)8. Water chemistry samples were collected 3 - 4 times during the study periods, at the same location as the DO sensors, from 0.5 m depth, and analyzed according to methods in Sharp et al.9. Measurements of photosynthetically active radiation (PAR) were made throughout the water column at each site, 2 - 4 times between June 1 – Nov 1, using a LiCOR L1400 light meter equipped with an upwelling quantum sensor. PM2.5 data from 2020 and 2021 were downloaded from a PurpleAir sensor near the east shore of the lake.
Delta (Sacramento Deep Water Ship Channel)
The Sacramento deep water ship channel, located in the northern Sacramento-San Joaquin River Delta (Figure 2), is a highly turbid freshwater channel connecting the West Sacramento Port to the Sacramento River. The 69 km long channel has a width of ~150 m, a depth of ~10 m, and is tidally forced and functionally resembles a dead-end slough10. DO and temperature measurements (PME miniDOTs with wiper units) were made in the upper portion of the channel (landward of channel marker 74) during summer and autumn in 2020. The landward portions of the channel have a long residence time (weeks-months) and very little net flow, resulting in a lake-like environment, including occasional vertical stratification11,12. Water samples were not collected from the Delta site in 2020 due to COVID19-related public health restrictions, so we used the most recent chemistry data and light attenuation measurements available (June - October of 2019; Smits et al. 2023). Hourly air temperature, wind speed, and SW radiation were downloaded from https://cimis.water.ca.gov/ (Station #121 in Dixon, CA). PM2.5 data were downloaded from an EPA sensor in West Sacramento, CA.
Description of data processing and metabolism models
We modeled daily rates of gross primary production (GPP; mg DO L-1 day-1), ecosystem respiration (R), and net ecosystem production (NEP = GPP - R) in the surface mixed layer of our study sites using hourly DO (mg L-1), water temperature (℃), SW radiation (W m-2), and wind speed (m s-1) data. Prior to use in metabolism models, all hourly data (DO, water temperature, SW radiation, wind speed) were visually inspected, and extreme outliers (> 3 standard deviations from the daily mean value) were removed and replaced using linear interpolation, following methods in Phillips13.
Hourly DO time series were modeled using the following equation: DOₜ₊₁ = DOₜ + GPP - R + F + ε; F is the flux of oxygen between the lake and atmosphere, and ε is the process error associated with vertical or horizontal mixing. GPP and R were estimated using the ‘metab’ function and bayesian model in the Lake Metabolizer R package14. The models in Lake Metabolizer have been used to estimate metabolic rates across diverse lake types15,16 and are described in detail in Winslow et al.14. The bayesian model estimates daily parameters for GPP and R and associated uncertainty in each estimate (expressed as a standard deviation; reported in Supplementary Table 1) within a Bayesian statistical framework. PAR (μmol m⁻² s⁻¹) and water temperature are covariates used to model rates of GPP and R, respectively. In addition to hourly DO, temperature, SW radiation, and wind speed the following model inputs were used: the depth of the surface mixed layer at each time step (zmix; m), the attenuation coefficient for PAR (kd; m-1), and lake surface area (m²).
We used several approaches to estimate the depth of the surface mixed layer (zmix), depending on data availability and habitat types (pelagic versus littoral). We calculated zmix using depth-distributed water temperature measurements from fixed depth sensors or vertical profiles in the following pelagic sites: Emerald Lake, Topaz Lake, Clear Lake (OA and LA), Castle Lake, and Dulzura Lake. Zmix was calculated for each time step from temperature data using the ‘ts.thermo.depth’ function in the LakeAnalyzer R package17. For littoral sites within larger stratified lakes (Castle, Dulzura, Tahoe), zmix was set to the depth of the DO sensor. In shallow water bodies that did not stratify (TOK 11, EML Pond 1, Topaz Pond), zmix was set to the lake depth at the location of the DO sensor, which was measured approximately monthly and linearly interpolated between measurements. In the tidally-influenced Delta site, zmix was set to the mean depth of the channel within the range of the tidal excursion (7.5 m).
We estimated mean PAR within the surface mixed layer by converting hourly SW radiation measurements to surface PAR using the ‘sw.to.par’ function in LakeMetabolizer, and then using the attenuation coefficient for PAR (kd; m -1) and zmix to estimate mean water column PAR using equation 5 from Staehr et al.18. In Castle Lake, Dulzura Lake, Lake Tahoe, and Clear Lake, light profiles were collected multiple times per season, and kd was estimated from each profile and linearly interpolated between profiling dates (Supplementary Figure 4). For three of the lakes (Emerald Lake, Topaz Lake, Delta), light profiles were collected once per summer, and thus kd did not vary through time. In ponds that were too shallow to perform light profiles (TOK 11, EML Pond 1, Topaz Pond), we set kd to the value of the nearest lake (Topaz Lake).
We calculated oxygen fluxes across the air-water interface at each time step using the following equation: F = k₆₀₀ (DO – DOsat)/zmix, where DO is the measured oxygen concentration in water, DOsat is the concentration of DO at equilibrium with the atmosphere at the measured water temperature and pressure, and k₆₀₀ is the transport coefficient for O₂. k₆₀₀ was calculated from a wind-based gas exchange model that accounted for lake surface area (vachon model in LakeMetabolizer)19. We set gas exchange to zero during periods when the DO sensor was below the diel or seasonal thermocline.
Days with unrealistic metabolism estimates (negative GPP, positive R) were excluded from results. Unrealistic estimates occurred most often on days when water column mixing resulted in irregular diel DO curves and were most common in polymictic, hyper-eutrophic Clear Lake (35 - 55 % of days between July 1 - Oct 1) and Emerald Lake during periods of thermocline deepening in autumn (24 - 33%; Supplementary Table 1).
