River Monitoring (Streamflow and Water Quality)

All streamflow data used in this study were collected by the USGS (U.S. Geological Survey, 2022). This data collection primarily consisted of flow measured at Appleton, Wis. (USGS station 04084445; latitude 44°14ʹ53ʺ, longitude −88°25ʹ23ʺ), and the mouth of the Fox River (at the oil depot in Green Bay, Wis.; USGS station 040851385; latitude 44°31ʹ43ʺ, longitude −88°00ʹ36ʺ). Flows at both locations were monitored using acoustic velocity meters because of backwater effects resulting in little relation between water elevation and flow. Instantaneous flow data were collected at 5–15-minute intervals and averaged to obtain mean daily flows. Occasionally, the mean daily flow at the mouth of the Fox River was negative. To account for short-term (5-minute to daily) negative flows, rather than estimating positive and negative loading, all negative flows were set to 2.83 cubic meters per second (m³/s) (100 cubic feet per second [ft³/s]). Then, to conserve mass, the total amount of added flow was subtracted from the preceding and succeeding days (Robertson and Saad, 2019). The daily flows measured at Appleton were used to estimate the daily flows at the Lake Winnebago outlet, and the daily flows measured at the mouth of the Fox River were used to estimate daily flows at De Pere. To complete these estimates, the ratios of the mean annual flows at Appleton and the mouth of the Fox River were first compared with the ratio in their total drainage areas. Annual flows at the mouth of the Fox River were on average 1.063 times that measured at Appleton, which is similar to the ratio in their drainage areas (1.064). Therefore, simple drainage ratios were used to extrapolate flows measured at Appleton to those at the Lake Winnebago outlet (outlet flow = Appleton flow × 0.988), and flows measured at the mouth of the Fox River were used to estimate the flows at De Pere (De Pere flow = mouth of the Fox River flow × 0.965).

Water-quality data at the Lake Winnebago outlet between Neenah and Menasha, Wis. (often referred to as “Neenah Menasha”), were only collected by the WDNR (WDNR station 713002; latitude 44°12ʹ17ʺ, longitude −88°28ʹ13ʺ). The sampling site was at the end of the pier near the public boat launch at Fritse Park in the Village of Fox Crossing. Water-quality samples were typically collected monthly, except during WYs 1989–90, when they were collected about two times per month.

Water-quality data were collected at De Pere by the WDNR (WDNR station 953210) and NEW Water (NEW Water station GB05) (latitude 44°12ʹ17ʺ, longitude −88°28ʹ13ʺ). The sampling site used by the WDNR was just upstream from the De Pere dam off the pier near the public boat launch, and the sampling site used by NEW Water was in the center of the river between the boat launch and St. Norbert College. The WDNR typically collected water-quality samples monthly, except during 2010–12, when they were collected about two to three times per month. NEW Water typically collected water-quality samples weekly from late May through early October.

Water-quality data were collected at the mouth of the Fox River by the USGS, WDNR, and NEW Water (USGS station 040851385, Fox River at the oil depot at Green Bay, Wis., west bank or a transect across the river at that location; WDNR station 053222, Fox River at Green Bay Yacht Club; and NEW Water station GB16 near the center of the river). The USGS collected detailed TSS samples during 1994–5 and then approximately monthly samplings during 2005–21, but few samples were collected during 2007–8 and more frequent samples were collected during high flows in 2011–13. During 2006–10, the WDNR typically collected one to two samples per month during October, November, March, and April; during 2011–18, they collected one sample per month during December, January, and February and two to three samples per month during March, April, October, and November. The WDNR only collected samples in October and March between October 2019 and December 2020 because of field sampling restrictions due to the COVID-19 pandemic. NEW Water typically collected samples weekly from late May through early October.

The WDNR collected samples in quality-assured polyethylene bottles from free flowing, well-mixed areas of the river (Wisconsin Department of Natural Resources, 2015). The sample bottles were dipped about 15 centimeters below the surface, facing upstream, and allowed to fill. Samples for TP analyses were immediately acidified with sulfuric acid. Samples for DP and TSS analysis were collected in separate sample bottles. All samples were placed in coolers with ice and returned to the WDNR laboratory for shipping. Water-quality samples were sent overnight on ice to the Wisconsin State Laboratory of Hygiene in Madison, Wis., where they were analyzed following standard methods and (or) approved U.S. Environmental Protection Agency (EPA) methods (Wisconsin State Laboratory of Hygiene, 2020; American Public Health Association and others, 2022). Samples for DP analyses were filtered through 0.45-micrometer membranes and measured as orthophosphorus or SRP, and TSS was measured as total suspended solids. All water-quality data are available in the WDNR Surface Water Integrated Monitoring System (more commonly known as SWIMS) database (Wisconsin Department of Natural Resources, 2023).

NEW Water collected discrete water-quality samples 1 m below the surface (top sample) and 1 m above the bottom of the water column (bottom sample). Samples were collected at De Pere using a Niskin sampler and at the mouth of the Fox River using a peristaltic pump with submersible tubing. All samples were collected in a 1-liter sterile polyethylene container and then 60 milliliters of the sample were filtered through a 0.45-µm membrane into a plastic container for orthophosphorus or SRP analysis. Samples were transported on ice to the NEW Water Laboratory for further analysis. The NEW Water Laboratory is in Green Bay, Wis., and it is certified by the WDNR for TP and TSS analyses. DP was measured as orthophosphorus or SRP, and TSS was measured as total suspended solids. TP and DP were analyzed using the ascorbic acid method (EPA method 365.4 for TP and EPA method 365.1 for DP; American Public Health Association and others, 2022). TSS was analyzed using standard methods (EPA method 2540D) described by the American Public Health Association and others (2022).

The USGS collected water-quality samples near the mouth of the Fox River from 1993 to 1995, from 2005 to 2006, and from 2008 to 2021 following standard procedures (U.S. Geological Survey, 2006; Wagner and others, 2006). Samples were collected at two locations (nearshore near the oil depot and transects across the river). During frozen months of these periods, samples were collected nearshore or from the drain of a flow-through tank at the station, which was on the west bank of the channel. During 2011–15, samples were collected with an automated, refrigerated sampler with tubing extending down into a fixed position in the river. During 2011–15, additional samples were collected during high-flow events with the autosamplers. During 2016–21, samples were collected from three locations in transects across the river. At each location, samples were collected from two depths (0.2 and 0.8 of the location depth) using a peristaltic pump with submersible tubing. During 2016–21, samples were collected monthly at fixed intervals, regardless of flow conditions. All water samples were placed into a churn splitter to subdivide the sample. Samples for DP analyses were filtered through 0.45-micrometer membranes. All nutrient samples were then chilled at around 4 degrees Celsius; samples for TP analyses were acidified with sulfuric acid, and samples were shipped to the USGS National Water Quality Laboratory in Denver, Colorado, for analysis. DP was measured as orthophosphorus or SRP, and suspended sediment was measured as total suspended sediment. Data collection and analytical procedures are described in detail by Robertson and others (2018a). All streamflow and water-quality data are available in the USGS National Water Information System (more commonly known as NWIS) database (U.S. Geological Survey, 2022).

Combining Agency Data

All water-quality data for each site were merged into one dataset. In general, the WDNR and NEW Water measured total suspended solids (TSS, EPA parameter code 00530) and the USGS measured total suspended sediment (EPA, parameter code 80154). All the TSS and total suspended sediment were used together with no adjustment factor and are referred to as simply “TSS.” For concentrations measured by NEW Water, all concentrations measured independently in top and bottom samples were averaged before use.

Even after combining the data from the three agencies, few samples were available for the mouth of the Fox River during November–April from 1989 to 2011 (table 1), but data collected at De Pere by the WDNR were available for many of these months with missing data. Gaps in data, especially at the beginning and end of the record, have been determined to cause poorly estimated concentrations and loads when using the WRTDS modeling approach (Lee and others, 2016). Therefore, the coinciding monthly water-quality data collected by the WDNR from 2012 to 2021 at De Pere and the mouth of the Fox River were used to determine monthly relations (slopes or multiplicative factors) between these two sites (table 2). Then, the data measured at De Pere during 1989–2011 were used to estimate concentrations at the mouth of the Fox River during these months for this early period. By estimating concentrations at the mouth of the Fox River from those measured at De Pere, fewer gaps in the data record existed (table 1). All final water-quality data for these three sites are available in an associated USGS data release (Diebel and Robertson, 2023).

Computation of Annual Mean and May–October Median Concentrations

Because of differences in the number of samples per month in each year, annual mean concentrations were computed by first computing mean monthly concentrations; any month with missing data was estimated to be the mean of the concentrations of the previous and preceding year with data. Similarly, median May–October concentrations were computed from the six monthly median concentrations for each year.

Computation of Daily and Annual Loads

Daily and annual loads for each constituent were computed using the flow and all concentration data from October 1988 to September 2021 (WYs 1989–2021) with the WRTDS modeling approach (Hirsch and De Cicco, 2015) that accounts for the autocorrelation structure of the model residuals using a first-order autoregressive model. This modified approach is called weighted regressions on time, discharge, and season with a first-order autoregressive Kalman filter (WRTDS–K; Zhang and Hirsch, 2019). WRTDS computes nonlinear, time-varying relations between the logarithms of the constituent concentrations and explanatory variables consisting of decimal time, the logarithm of daily discharge, and sine and cosine transformations of decimal time (Hirsch and others, 2010; Lee and others, 2016; eq. 1).

In other more traditional regression techniques used to compute loads, such as LOADest (Runkel and others, 2004), the fitted coefficients (β_n values) are computed using the entire dataset. In WRTDS, one regression model is estimated for each day of the record to estimate concentration and flux using only data that are “close” to the time t. With the known time, discharge, and season of the estimation day, WRTDS prescreens the concentration record and selects samples that are sufficiently “close” to the estimation day with respect to time, discharge, and season (close is defined by the window size). The selected samples are then used to build a weighted regression model, and a unique set of model coefficients is obtained (Zhang and Hirsch, 2019). Weights for each sample used in the calibration are based on differences in the values of the explanatory variables (year, day of the year, and discharge) between the estimation and sample day. WRTDS uses a bias correction factor specific to each year, day, and discharge to adjust for retransformation bias (Hirsch and De Cicco, 2015). The WRTDS load estimates were then improved by accounting for the autocorrelation structure of the model using WRTDS–K. In WRTDS–K, each estimated concentration is adjusted by the magnitude of the nearby residuals computed using equation 1; therefore, the more concentration data used to calibrate WRTDS, the more accurate the estimated WRTDS–K loads. This approach has been determined to improve short-term loads estimations (Zhang and Hirsch, 2019). WRTDS–K was implemented using the R (R Core Team, 2013) package Exploration and Graphics for RivEr Trends (more commonly known as EGRET) using all default specifications in the model.

Quantifying Trends and Changes in Concentrations and Loads

Trends and changes in TP, DP, and TSS concentrations and loads through time were examined using a time-series regression approach and using flow normalization within WRTDS. For each water-quality constituent (TP, DP, and TSS), changes in mean annual concentrations measured from WYs 1989–93 to WYs 2017–21 were computed at each site. Trends in the mean annual concentrations through time were then computed using simple time-series linear regressions between concentrations and WY, and the results are compared among sites.

To further quantify trends in water-quality concentrations and loads through time, the WRTDS modeling approach was used, which compensates for the interannual changes in hydrologic conditions and nonstationary streamflow by incorporating information on changes in the frequency distribution of daily measured streamflow (discharge) over time. This process is referred to as “flow normalization” (Hirsch and De Cicco, 2015). To compute flow-normalized loads and concentrations for each day in the period, WRTDS simulates the load given the regression model estimated for each day of the record and all the flows measured on that day of the year and then computes the mean daily load and concentration. For example, for June 1, 1995, WRTDS simulates the loads using the regression model estimated for June 1, 1995, for 33 flows (one from June 1 of each year) and then all 33 estimations are averaged together. The flow-normalization process results in a time series of flow-normalized loads (or concentrations). The flow-normalization process minimizes, but does not completely eliminate, the year-to-year variability in loads that is caused by changes in daily discharge from one year to the next. The annual time series of flow-normalized loads is similar to a time series of volumetrically weighted mean concentrations because each daily load (or concentration) has the same flows used in the computations. The 90-percent confidence intervals for the flow-normalized loads were estimated by running 100 bootstrap replicates of the WRTDS model and then interpolating the 5-percent and 95-percent quantiles with the sample cumulative distribution function (Hirsch and others, 2015). The likelihood that flow normalized load increased or decreased through the study period was estimated from the fraction of bootstrap replicates where load increased or decreased (Hirsch and others, 2015). Data for all final loads and flow-normalized loads for these three sites are available in an associated USGS data release (Diebel and Robertson, 2023).

Total Phosphorus

The overall mean annual TP concentrations for WYs 1989–2021 were 0.089 mg/L, 0.115 mg/L, and 0.128 mg/L at the Lake Winnebago outlet, De Pere, and the mouth of the Fox River, respectively (table 4). All the data assembled during this study for TP concentrations from 1974 to 2022 are shown for each site in figure 3, and the annual mean concentrations for WYs 1975–2021 are compared among sites in figure 4. From WY 1989 to WY 2021, no statistically significant change through time (p greater than [>] 0.1) in annual TP concentrations was detected at the Lake Winnebago outlet based on a linear regression between annual mean concentrations and WY, and little change through time was observed in the data before WY 1989. However, a statistically significant decrease through time in TP concentrations was detected at De Pere and the mouth of the Fox River (p<0.05; table 4). Data before the WY 1989–2021 period seem to indicate higher TP concentrations at these two sites. From WYs 1989–93 to WYs 2017–21, a decrease in TP concentrations by 0.010 mg/L was detected at the Lake Winnebago outlet, a decrease by 0.044 mg/L was detected at De Pere, and a decrease by 0.059 mg/L was detected at the mouth of the Fox River. During WYs 1989–2021, TP concentrations increased, on average, by 0.039 mg/L from the Lake Winnebago outlet to the mouth of the Fox River; most of the change was detected upstream from De Pere. This downstream increase in concentrations from the Lake Winnebago outlet to the mouth of the Fox River was much larger during WYs 1989–93 (0.058 mg/L) than during WYs 2017–21 (0.009 mg/L).

The overall mean annual TP loads computed using WRTDS–K for WYs 1989–2021 were 360 t/yr, 475 t/yr, and 557 t/yr at the Lake Winnebago outlet, De Pere, and the mouth of the Fox River, respectively (table 5). Annual TP loads at the Lake Winnebago outlet ranged from 182 t/yr in WY 1998 to 627 t/yr in WY 2019 (table 6, fig. 5). Mean annual TP loads at the Lake Winnebago outlet increased significantly through time (p<0.05) by 5.20 t/yr per year during WYs 1989–2021, and the recent mean annual loading during WYs 2017–21 was 511 t/yr. Annual TP loads at De Pere ranged from 261 t/yr in WY 2009 to 1,140 t/yr in WY 1993 (table 7, fig. 5), and mean annual loads decreased through time by 1.86 t/yr per year during WYs 1989–2021 (p>0.1—not statistically significant); the recent mean loading for WYs 2017–21 was 539 t/yr. Annual TP loads at the mouth of the Fox River ranged from 313 t/yr in WY 2009 to 1,180 t/yr in WY 1993 (table 8, fig. 5). Annual TP loads decreased through time by 0.970 t/yr per year during WYs 1989–2021 (p>0.76—not statistically significant), and the recent mean annual loading for WYs 2017–21 was 618 t/yr. During the first 5 years of the study period (WYs 1989–93), the TP load at the Lake Winnebago outlet was 52 percent of that estimated at the mouth of the Fox River, and during the last 5 years of the study (WYs 2017–21), it was 83 percent of the load estimated at the mouth of the Fox River.

WRTDS was used to further examine the trends through time in TP loads during WYs 1989–2021 after adjusting for the year-to-year hydrologic variability by computing flow-normalized loads. In other words, these are the loads that would be expected in each year if similar flows were measured in all the years being examined. Results from these analyses, with 90-percent confidence limits computed with 100 bootstrap replicates of the WRTDS model, indicated that the flow-normalized annual TP loads at the Lake Winnebago outlet changed little from WY 1989 to WY 2021 (fig. 6). The likelihood of a downward trend through time in the flow-normalized TP loads is as likely as not (likelihood = 0.530). Flow-normalized loads gradually increased until WY 2005 and then gradually decreased. During the last 5 years of the study period (WYs 2017–21), the flow-normalized annual TP load at the Lake Winnebago outlet was only 1.8 percent less than that measured during the first 5 years of the period (WYs 1989–93). However, the flow-normalized annual TP loads at De Pere and the mouth of the Fox River significantly decreased from WY 1989 to WY 2021 (fig. 6). The likelihood of a downward trend through time in the flow-normalized TP loads was highly likely (likelihood = 0.995) at De Pere and the mouth of the Fox River. During the last 5 years of the study period (WYs 2017–21), the flow-normalized TP loads at De Pere and the mouth of the Fox River were, respectively, 38 percent and 37 percent less than those measured during the first 5 years of the period (WYs 1989–93).

WRTDS was used to further examine the seasonal timing of the changes in the flow-normalized TP loading at the mouth of the Fox River (fig. 7). The decreases in flow-normalized loads were detected in all four seasons; however, the changes through time were largest in spring and summer when the loads were highest, and the changes through time were smallest during winter when the loads were smallest. During spring and summer, seasonal TP loads at the mouth of the Fox River dropped by 45.5 percent and 34.9 percent, respectively.

Dissolved Phosphorus

The overall mean annual DP concentrations for WYs 1989–2021 were 0.024 mg/L, 0.033 mg/L, and 0.036 mg/L at the Lake Winnebago outlet, De Pere, and the mouth of the Fox River, respectively (table 4). Over the entire period, about 28 percent of the phosphorus at all three sites was in dissolved forms. Throughout this period, a statistically significant increase through time (p<0.05) in annual DP concentrations was measured at the Lake Winnebago outlet; however, no statistically significant changes through time in DP concentrations were measured at De Pere and the mouth of the Fox River, based on linear regressions between annual mean concentrations and WY. From the first 5 years of this study to last 5 years, an increase of 0.013 mg/L was detected at the Lake Winnebago outlet and a decrease of 0.001 mg/L and 0.002 mg/L was detected at De Pere and the mouth of the Fox River, respectively. The statistically significant (p<0.05) increase in DP concentrations at the Lake Winnebago outlet may have been caused by the increase in DP in Lake Winnebago resulting from the lake no longer being phosphorus limited in late summer with biological processes not being able to uptake all the DP (Robertson and others, 2018b). The DP concentrations increased through time, on average, by 0.012 mg/L from the Lake Winnebago outlet to the mouth of the Fox River during WYs 1989–2021; most of the change was detected upstream from De Pere. Over the period, the amount of phosphorus in dissolved forms changed from about 13 percent during the first 5 years to 31 percent during the most recent 5 years of the study at the Lake Winnebago outlet and from about 21 percent to about 31 percent at De Pere and the mouth of the Fox River. The increase from the Lake Winnebago outlet to the mouth of the Fox River was larger during WYs 1989–93 (0.018 mg/L) and smaller during WYs 2017–21 (0.003 mg/L).

The overall mean annual DP loads computed using WRTDS–K for WYs 1989–2021 were 114 t/yr, 153 t/yr, and 162 t/yr at the Lake Winnebago outlet, De Pere, and the mouth of the Fox River, respectively (table 5). Annual DP loads at the Lake Winnebago outlet ranged from 36.2 t/yr in WY 1989 to 228 t/yr in WY 2004 (table 6, fig. 8). The mean annual loads at the Lake Winnebago outlet significantly increased through time by 2.36 t/yr per year during WYs 1989–2021 (p<0.05), and the most recent (WYs 2017–21) mean annual load was 169 t/yr. Annual DP loads at De Pere ranged from 59.6 t/yr in WY 1999 to 310 t/yr in WY 1993 (table 7, fig. 8). Mean annual loads increased through time by 0.564 t/yr per year during WYs 1989–2021 (p>0.1), and the most recent (WYs 2017–21) mean annual loading was 200 t/yr. Annual DP loads at the mouth of the Fox River ranged from 85.6 t/yr in WY 2000 to 288 t/yr in WY 1993 (table 8, fig. 8). Mean annual loads increased through time by 0.908 t/yr per year during WYs 1989–2021 (p>0.1), and the most recent (WYs 2017–21) mean annual loading was 205 t/yr. During the first 5 years of the study period (WYs 1989–93), the DP load at the Lake Winnebago outlet was 55 percent of that at the mouth of the Fox River, and during the last 5 years of the period (WYs 2017–21), it was 82 percent of the load at the mouth of the Fox River.

WRTDS was used to further examine the changes through time in DP loads during WYs 1989–2021 by computing flow-normalized loads. Results from these analyses indicate that the flow-normalized annual DP loads at the Lake Winnebago outlet increased from WY 1989 to WY 2021 (fig. 9). The likelihood of an upward trend through time in the flow-normalized DP loads was likely (likelihood = 0.668). During WYs 1997–2005, flow-normalized loads were higher than in other years. During the last 5 years of the study (WYs 2017–21), the mean flow-normalized DP load at the Lake Winnebago outlet was 11 percent greater than that measured during the first 5 years of the period (WYs 1989–93). The flow-normalized annual DP loads at De Pere and the mouth of the Fox River decreased from WY 1989 to WY 2021 (fig. 9). The likelihood of downward trends through time in the flow-normalized DP loads at these two sites was likely (likelihood = 0.807 and 0.847, respectively). During the last 5 years of the study (WYs 2017–21), the mean flow-normalized DP loads at De Pere and the mouth of the Fox River were 24 percent and 18 percent, respectively, less than those measured during the first 5 years of the period (WYs 1989–93).

Total Suspended Solids

The overall mean annual TSS concentrations for WYs 1989–2021 were 13.5 mg/L, 18.7 mg/L, and 23.9 mg/L at the Lake Winnebago outlet, De Pere, and the mouth of the Fox River, respectively (table 4). Throughout this period, no statistically significant change through time (p>0.05) in annual TSS concentrations was detected at the Lake Winnebago outlet; however, statistically significant (p<0.05) decreases in concentrations were detected at De Pere and the mouth of the Fox River. From the first 5 years of this study to last 5 years, a decrease of 7.2 mg/L, 9.2 mg/L, and 14.6 mg/L, respectively, was detected based on linear regressions between annual mean concentrations and WY. TSS concentrations increased through time, on average, by 10.4 mg/L from the Lake Winnebago outlet to the mouth of the Fox River during WYs 1989–2021; the change was about equally split between the Lake Winnebago outlet and De Pere and between De Pere and the mouth of the Fox River. This increase from the Lake Winnebago outlet to the mouth of the Fox River was larger during WYs 1989–93 (12.7 mg/L) and smaller during WYs 2017–21 (5.3 mg/L).

The overall mean annual TSS loads computed using WRTDS–K for WYs 1989–2021 were 60,400 t/yr, 86,800 t/yr, and 123,000 t/yr at the Lake Winnebago outlet, De Pere, and the mouth of the Fox River, respectively (table 5). Annual TSS loads at the Lake Winnebago outlet ranged from 29,700 t/yr in WY 1998 to 120,000 t/yr in WY 1993 (table 6, fig. 10). The mean annual WRTDS–K loads at the Lake Winnebago outlet increased through time by 223 t/yr per year during WYs 1989–2021 (p>0.1), and the recent mean annual load during WYs 2017–21 was 74,500 t/yr. Annual TSS loads at De Pere ranged from 39,700 t/yr in WY 2009 to 250,000 t/yr in WY 1993 (table 7, fig. 10). Annual loads at De Pere decreased through time by 417 t/yr per year during WYs 1989–2021 (p>0.1). The recent mean annual load during WYs 2017–21 was 98,200 t/yr. Annual TSS loads at the mouth of the Fox River ranged from 52,000 t/yr in WY 2009 to 365,000 t/yr in WY 1993 (table 8, fig. 10), and annual loads decreased through time by 1,921 t/yr per year during WYs 1989–2021 (p=0.083). The recent mean annual load during WYs 2017–21 was 116,000 t/yr. During the first 5 years of the study (WYs 1989–93), the mean annual TSS load at the Lake Winnebago outlet was 36 percent of that at the mouth of the Fox River, and during the last 5 years of the period (WYs 2017–21), it was 65 percent of the load at the mouth of the Fox River.

WRTDS was used to further examine the changes in TSS loads during WYs 1989–2021 by computing flow-normalized loads. Results from these analyses indicate that the flow-normalized annual TSS loads at the Lake Winnebago outlet decreased through time during WYs 1989–21 (fig. 11). The likelihood of a downward trend through time in the flow-normalized TSS loads is likely (likelihood = 0.876). During the last 5 years of the study period (WYs 2017–21), the mean annual flow-normalized TSS load at the Lake Winnebago outlet was 37 percent less than that measured during the first 5 years of the period (WYs 1989–93). The flow-normalized annual TSS loads at De Pere and the mouth of the Fox River decreased through time during WYs 1989–2021 (fig. 11). The likelihood of a downward trend through time in the flow-normalized TSS loads at these sites is highly likely (likelihood = 0.995 and 0.965). During the last 5 years of the study period (WYs 2017–21), the mean annual flow-normalized TSS loads at De Pere and the mouth of the Fox River were 50 percent and 65 percent less, respectively, than that measured during the first 5 years of the period (WYs 1989–93).