IntroductionThe characterization of seasonal and temporal variability of water quality is an important aspect of environmental impact statements, environmental assessments, and project plans for navigation projects such as channel deepening efforts and projects located within lock and dam impoundments. However, it is often difficult to evaluate historical trends or background quality of a river or other water body and understand their effects on navigation activities because locating, downloading, and analyzing the appropriate data can be a challenge. Additionally, with the need to accelerate and economize feasibility level assessments (USGAO 2019), there is rarely the opportunity to collect new water quality data, so the analysis of historical data is critical. This paper will discuss the use of several tools that have been developed by the USGS over the past decade or so to facilitate the identification and retrieval of water quality and streamflow data along with techniques used to create meaningful analyses and visualizations that facilitate the interpretation of these data. Many of these tools use R, a statistical software program (R Core Team 2019), and include R packages developed by the USGS (collectively referred to as USGS-R). Information about the USGS National Water Quality Network and the statistical load and trend model, weighted regressions on time, discharge, and season (WRTDS), (Hirsch et al. 2010) is presented.The Mississippi River Basin (MRB) will be used as a demonstration of these water quality analysis tools and techniques. The Mississippi River drains 41% of the contiguous United States, including parts of 31 states and 2 Canadian provinces (USACE 2021). While nutrients are critical for the maintenance of healthy ecosystems, excess nutrients may lead to eutrophication through excess algal growth resulting in reduced sunlight, loss of aquatic habitat, and dissolved oxygen deficiencies or hypoxia when the algae die and decompose (Aulenbach 2006). Knowledge of nutrient concentrations and fluxes is vitally important locally and for downstream receiving waters such as lock and dam impoundments and stretches of rivers used for recreation. Local water bodies may be regulated by rules such as total maximum daily loads (TMDLs) while nutrient fluxes are one of several factors controlling the extent and severity of the hypoxic zone that forms annually in the receiving waters of the northern Gulf of Mexico (Rabalais et al. 2001). Suspended sediments in the MRB are important for multiple reasons. In the upper basin, suspended sediment is considered a pollutant that blocks light transmission through the water column, may affect aquatic organism feeding and reproductive behaviors, degrades benthic habitat, and often carries adsorbed pollutants. In the lowermost basin, suspended sediments are a resource used to restore coastal wetlands. Throughout the basin, suspended sediments that settle, deposit, and become part of the bed material load can affect the channel geometry, channel, and floodplain geomorphology and potentially affect channel maintenance.The Mississippi River is maintained by the US Army Corps of Engineers for Deep Draft Navigation from the mouth of the river to Baton Rouge, Louisiana, approximately 412 km upstream. In this section of the river, the current navigation channel is between about 150 and 230 m wide and is about 13.7 m deep, although recent authorization has been received to deepen the channel to 15.2 m. From Baton Rouge, Louisiana, to Minneapolis/St. Paul, Minnesota, shallow draft navigation is maintained, which is generally a channel about 120 m wide and between 2.75 and 3.5 m deep (LaDOTD 2016). In total, the Mississippi Valley Division of the US Army Corps of Engineers operates and maintains over 6,760 km of navigable channels, 62 locks, 51 shallow-draft ports, and 7 deep-draft ports (USACE 2020). The management, maintenance, and alteration of this infrastructure, including channel maintenance and dredging, require basic familiarity with the water quality of the system.This study includes a unique aspect—the initial analyses were completed by graduate students as part of a river sampling methods class in the cooperative Tulane/US Army Corps of Engineers River Science and Engineering graduate certificate program. Each student was assigned a site in the MRB and throughout the semester—while learning methods for sampling hydrology, suspended sediments, water quality, and ecology—they also learned R and used it to retrieve, analyze, and model the water quality data for their site with USGS-R packages, and the WRTDS model. The final class assignment required teams of students to synthesize their results across multiple sites in the MRB, which generated the idea for this paper.Study SitesThis paper considers water quality data from 9 sites in the MRB, including 4 sites on the Mississippi River mainstem and 5 sites on its principal tributaries and distributaries, including the Iowa, Illinois, Missouri, Ohio, and Atchafalaya Rivers. The locations of the sites are shown in Fig. 1 with more details provided in Table 1. Although the period of data collection for the constituents at each of the locations varies, for this analysis, we focus on the period from 1980 through the present. These sites were selected because of their relatively long period of records and a large number of samples available for analysis.Table 1. Site and data characteristics of the nine Mississippi River basin sites and three constituents [nitrate plus nitrite (NO3), orthophosphate (OP), and suspended sediment (SSC)] used in this studyTable 1. Site and data characteristics of the nine Mississippi River basin sites and three constituents [nitrate plus nitrite (NO3), orthophosphate (OP), and suspended sediment (SSC)] used in this studySite abbreviationUSGS site numberSite nameBasin area (km2)Water quality constituentCalibration startNumber of samplesFlux bias statisticMS-CLIN05420500Mississippi River at Clinton, Iowa221,703NO319754970.049OP19824040.137SSC19913510.020IA-WAPE05465500Iowa River at Wapello, Iowa32,375NO319784650.013OP19824270.173SSC1991355−0.088IL-VALL05586100aIllinois River at Valley City, Illinois69,264NO319755660.016OP1982377−0.013SSC1997404−0.118MS-GRAF05587455bMississippi River below Grafton, Illinois443,665NO31975483−0.003OP19823970.004SSC1993224−0.022MZ-HERM06934500Missouri River at Hermann, Missouri1,353,269NO319755970.043OP1982511−0.007SSC19756390.051MS-THEB07022000Mississippi River at Thebes, Illinois1,847,180NO31975595<0.001OP19824540.011SSC1975473−0.001OH-GRAN03612500cOhio River at Grand Chain, Illinois526,027NO31975543−0.017OP1982457−0.001SSC19755150.021MS-STFR07373420dMississippi River at St. Francisville, Louisiana2,914,514NO31975589−0.009OP1982449−0.009SSC1993372−0.024AC-MELV07381495eAtchafalaya River at Melville, Louisiana241,687NO31978550<0.001OP1982448−0.001SSC19804710.007MethodsData retrieval, model estimation, and data summaries were completed using the R statistical software program (R Core Team 2019) with packages developed by the USGS. The dplyr package (Wickham et al. 2019) and ggplot2 package (Wickham 2016) were also used for data manipulation and plotting, respectively. The USGS collects surface water-quality data at thousands of locations across the United States and provides this information to the public via NWISweb (USGS 2020). Additional data collected by other organizations including State governments, the US Environmental Protection Agency, nonprofit groups, and researchers are available via the Water Quality Portal (NWQMC Undated). Both resources can be accessed with functions in the dataRetrieval package (De Cicco et al. 2018).The nine MRB sites used in this analysis are a part of the USGS National Water Quality Network. The National Water Quality Network consists of over 100 sites across the United States where streamflow and water quality measurements have been collected using consistent methods for decades, and annual estimates of concentration and flux for a variety of constituents are determined and reported annually. The term flux is used here to describe the mass of a river-borne constituent that passes a given point on a river over a given period of time. The sampling methods and data processing decisions implemented at these sites are described in detail in Lee et al. (2017) and on the National Water Quality Network webpage (Lee Undated). For this analysis, we retrieved the discrete water quality and daily streamflow data for each of the nine MRB sites from the corresponding National Water Quality Network ScienceBase webpage (Lee 2020) using the sbtools package (Winslow et al. 2016). We used the data prepared for the National Water Quality Network, instead of retrieving directly from NWISweb, to provide an update to the 1980–2010 trends presented in a previous publication (Murphy et al. 2013), which also uses the National Water Quality Network data. Our analysis focused on dissolved nitrate plus nitrite, filtered, as nitrogen (NO3–NO2), dissolved orthophosphate, filtered, as phosphorus (OP), and suspended sediment concentration, all particle sizes (SSC). The primary USGS parameter codes for these constituents are 00631, 00671, and 80154, respectively, though additional codes that represent dissolved or total fractions or related but slightly different constituents were also combined following a set of criteria described in Lee (Undated).Concentrations and fluxes of NO3–NO2, OP, and SSC were modeled with WRTDS (Hirsch et al. 2010), using the available period of records shown in Table 1 to compute daily estimates. Model fits were evaluated using plots of residuals, estimates, and observed values, and assessing the flux bias statistic, which is the ratio of the difference between the mean estimated flux and the mean observed flux divided by the mean estimated flux (Hirsch et al. 2010). After running WRTDS, WRTDS_K was used to improve the estimates by accounting for the serial correlation of the residuals using a first-order autoregressive model (Zhang and Hirsch 2019). The residuals are defined as the estimated concentration (or flux) minus the observed concentration (or flux) using only WRTDS (not WRTDS_K). These daily estimates of concentration and flux were aggregated by month and water year (WY). A water year (WY) represents a 12-month period ending on September 30 of the reported year and starting on October 1 of the previous year.Trends in monthly mean and annual mean concentration and flux were determined using generalized flow normalization (Choquette et al. 2019). This approach incorporates the influence of shifts and trends in the flow regime over the period of record and gives a water quality trend that has the effects of inter-annual variability of streamflow removed. Due to quasi-periodic weather patterns, there are times when higher-than-average or lower-than-average streamflow occurs for several years in a row. Typically, trend methods such as the Mann–Kendall test, assume these occurrences are a meaningful change, but in a few years’ time, streamflow returns to near-normal conditions and the apparent trend has vanished. The flow normalization process in WRTDS guards against these apparent trends by accounting for the quasi-periodic nature of streamflow. Flow normalization produces daily flow normalized estimates of concentration and flux that are aggregated by month and year; these smoothed estimates give the monthly and annual trends over time. All actual estimates (produced with WRTDS-K) and flow normalization estimates were generated using the EGRET package version 3.0.4 (Hirsch and De Cicco 2015), downloaded from GitHub (De Cicco Undated). Annual estimates were computed for water years, meaning a 12-month period beginning on October 1, ending on September 30, and labeled for the year containing the majority of the months.Additionally, the magnitude of change between the start and end of three trend periods (approximately 1980–2019, 1980–2010, and 2010–2019) was determined for all sites and constituents, where the available data allowed. Changes were converted to annualized percent changes: the percent change, relative to the starting year concentration or flux, was divided by the number of years in the trend period. The uncertainty of the direction of these changes was estimated using a block bootstrap method within the EGRETci package, version 2.0 (Hirsch et al. 2015). For each site and constituent combination, the WRTDS model was calibrated using the entire period of record (Table 1) and the results for the three trend periods were extracted from this single model run. Note the earliest trend start-year differs depending on the constituent and site. NO3–NO2 trends start in 1980 for all sites. OP trends start in 1982 for all sites. SSC trends start in 1980 for four sites in the lower MRB (AC-MELV, OH-GRAN, MS-THEB, and MZ-HERM) and in 1997 for the lowermost MRB site (MS-STFR) and four sites in the upper MRB (IA-WAPE, IL-VALL, MS-CLIN, and MS-GRAF). A 2010–2019 trend is reported for all sites and constituents. Monthly and annual estimates and trend period results were compared between sites and constituents. All results, including input data, output data and modeling scripts, are available in Murphy (2021).Results—The Mississippi River Basin ExampleSeasonal and temporal summaries of NO3–NO2, OP, and SSC for nine sites within the MRB are presented as an example of the types of data syntheses that can be completed with USGS data and tools that may be useful for navigation studies. The analyses present a few of the options, focusing on the relative location of sites, seasonality, and changes over time.Concentrations of NO3–NO2, OP, and SSC by Month and Location in the BasinFig. 2 depicts box plots of monthly mean concentration estimates of NO3–NO2, OP, and SSC for the period of record, at the nine sites of interest in the MRB. Sites are ordered from upstream (left) to downstream (right) in the basin (Fig. 1). The center horizontal line in the box plots represent the median values, the boxes show the interquartile range (IQR) between the first and third quartiles (Q1 and Q3), the whiskers extend between Q1 – 1.5 * IQR and Q3 + 1.5 * IQR and the circles indicate values that are potential outliers.Although there is an established drinking water standard for nitrate of 10 mg/L (USEPA 2021), the desirable level in rivers is less certain. Generally speaking, nitrate levels below about 1.0 mg/L are considered indicative of unpolluted streams in most areas of the world while OP averages about 0.1 mg/L worldwide in unpolluted streams (Meybeck 1982). Fig. 2 shows that sites draining the intensely agricultural regions of the upper MRB (IA-WAPE, IL-VALL, MS-GRAF, and to a lesser degree MS-CLIN) have high nutrient concentrations. The highest levels occur in the small MRB tributaries, such as the Iowa River and the Illinois River. The concentrations in these tributaries lead to elevated concentrations downstream in the mainstream of the Mississippi River at Grafton, Illinois. For NO3–NO2, concentrations remain elevated in the mainstem of the Mississippi River even after the addition of the Missouri River, as indicated by higher concentrations at the Mississippi River at Thebes, Illinois, compared to the upstream Missouri River at Hermann, Missouri. NO3–NO2 and OP concentrations are lowest in the Ohio River compared to the other sites, and the Ohio River also has comparably higher volumes of streamflow. The two sites farthest downstream in the MRB (MS-STFR and AC-MELV) have NO3–NO2 and OP concentrations that are slightly higher than those in the Ohio River, with means around 2 and 0.1 mg/L, respectively. Approximately 23% of the flow of the lower Mississippi River is diverted into the Atchafalaya River (Allison et al. 2012), accounting for some of the similarity in NO3–NO2 and OP concentrations between these two sites (Fig. 1).Seasonally, NO3–NO2 concentrations tend to be highest in the spring and early summer, perhaps as snow melt, spring rains, and fertilizer applications combine to cause excess in runoff from fields in agricultural regions, and then concentrations rapidly decline during the drier mid-summer and fall (Fig. 2). For many of the sites, NO3–NO2 concentrations remain elevated during winter, especially in the upper MRB. Seasonal patterns for OP are less consistent between sites. OP concentration at mainstem sites on the Mississippi River tend to follow a sinusoidal shape with highest concentration occurring mid-summer through early fall and lowest concentrations during the winter. OP in the Iowa, Illinois, and Ohio Rivers tend to have similar pattern though shifted to later in the year with high concentrations during the winter and lower concentrations in spring and summer. OP in the Missouri River is similar to the mainstem Mississippi River sites but concentrations are fairly uniform throughout the year with a slight increase during the summer.Suspended sediment concentrations have a very different spatial pattern than nutrients do across these nine sites in the MRB. The upper Mississippi River and three tributary sites have relatively low SSC concentrations with a few outliers, likely due to large storms or floods. Conversely, the Missouri River has very high SSC concentrations with many high outliers. The contributions of the Missouri River strongly influence SSC concentration in the mainstem Mississippi River at Thebes, Illinois, but below the relatively sediment-free Ohio River, concentrations are much lower. A slight increase in SSC is seen in the Atchafalaya River, after its diversion through the Old River Control structures, suggesting a within-basin source of suspended sediment for the Atchafalaya River that is absent in the lower Mississippi River, or contributions from the Red River, which may have implications for the restoration of deltaic wetlands in coastal Louisiana (Mize et al. 2018).Fluxes of NO3–NO2, OP, and SSC by Month, by Location in the BasinIn Fig. 3, box plots of monthly mean flux estimates of NO3–NO2, OP, and SSC, as a rate of metric tons per day, are shown for the nine sites in the MRB. Due to the often-overwhelming influence of streamflow magnitude on flux estimates, the flux patterns in Fig. 3 are markedly different from the concentration patterns in Fig. 2, due largely to differences in flow volumes between the sites. The smaller tributary rivers, such as the Illinois and Iowa Rivers, which have very high nutrient concentrations, have relatively small streamflows such that their fluxes are a small portion of the overall nutrient load of the Mississippi River. However, in aggregate the upper MRB, as captured at the Mississippi River site at Grafton, Illinois, has a large nutrient flux compared to the large downstream tributaries, the Missouri and Ohio Rivers. This pattern of nutrient accumulation to the total flux continues moving downstream on the Mississippi River with additional inputs from the Missouri and the Ohio Rivers, as well as other tributaries which have less robust historical measurements between the Ohio River and lowermost study site (MS-STFR), such as the Obion River, the White River, and the Yazoo River. Despite its very large drainage basin size and high nutrient concentrations, the flow of the Missouri River is small enough that the fluxes of nutrients from the Missouri River are relatively low. Conversely, the modest nutrient concentrations in the Ohio River become a relatively large nutrient flux with the river’s large flow volumes. The seasonality of nutrient fluxes is controlled by the hydrologic seasonality of the system and likely by the application of agricultural fertilizers, both peaking in the spring and early summer.Considering flux of SSC, it is easy to understand the old adage “in the lower Mississippi River, the sediment comes from the Missouri River, but the water comes from the Ohio River.” The SSC flux from the Missouri River dominates the seasonal pattern seen at the Mississippi River at Thebes, Illinois (Fig. 3). Much less suspended sediment enters the system from the Ohio River, resulting in considerably lower fluxes at St. Francisville, Louisiana, compared to those upstream in the Missouri River or the Mississippi River at Thebes. Upstream from St. Francisville, Louisiana, about 23% of the SSC flux in the Mississippi River is diverted into the Atchafalaya distributary (Allison et al. 2012), which also contributes to some of the decline in SSC flux in the Mississippi River between Thebes, Illinois (MS-THEB) and St. Francisville, Louisiana (MS-STFR). However, given the magnitude of sediment fluxes in the Mississippi River at Thebes, Illinois, the contributions of sizable tributaries downstream (such as the Arkansas, Hatchie, and Yazoo Rivers, which are not well measured), and the influx of the Ohio River, the diversion of some of this sediment load into the Atchafalaya distributary likely does not fully account for the lower sediment flux in the Mississippi River at St. Francisville, Louisiana. This apparent loss of SSC flux could be influenced by a variety of factors, including the possible deposition of these fine sediments in the floodplain and other sinks along the lower Mississippi River (Allison et al. 2012; Little and Biedenharn 2014). One shortcoming this observation highlights is the lack of sites on the Mississippi River with long-term nutrient and SSC records between Thebes, Illinois, and St. Francisville, Louisiana (Fig. 1). A multiagency effort to establish new data collection sites and to better utilize existing data could improve the spatial coverage of data and help refine the nutrient and sediment budgets for this section of the river. If better quantified, the aggregational tendency of sediments in channel and floodplains in the lower Mississippi River might be more evident and could show potential effects on navigation, flood risk management, and coastal restoration in the future.Trends in Flow-Normalized Concentrations and Flux by Month at Selected SitesAn additional method to understand the water quality at a site is to evaluate changes over time in concentration and flux with some of the variability related to fluctuating weather and streamflow removed. In Figs. 4–6, modeled trends in monthly mean flow-normalized concentration and flux are shown for NO3–NO2, OP, and SSC for the Iowa River, Missouri River, and the lowermost site on the Mississippi River (MS-STFR). The trends, which smooth out some of the influence of variable year-to-year streamflow volumes on river water quality, are graphed for four months (February, May, August, and November), representing each of the four seasons (winter, spring, summer, and fall, respectively), to examine seasonal changes over time.Fig. 4 features the Iowa River at Wapello, Iowa, and shows high concentrations and a distinct peak in NO3–NO2 concentrations and fluxes in all seasons, in the early 2000s, followed by a gradual decline. The winter and spring show relatively stable conditions during the last part of the record, whereas the summer and fall show slight increases during this time. Since 1980, long-term changes in NO3–NO2 concentrations were distinct among seasons. A large decrease occurred during the fall, smaller decrease during the winter, a large increase during the spring, and fluctuating but little change during the summer. Changes in OP concentrations [Fig. 4(a)] over time show decreases until the mid-1990s, followed by increases in all seasons with a distinct peak in OP concentration in the mid-1990s during the winter. Early 2000s show a peak in OP flux, especially during the winter and spring [Fig. 4(b)]. SSC concentrations and flux peaked in the 2010s, especially during summer months, but have been gradually declining since that time (Fig. 4).In contrast, concentrations are nearly four times less and trends are more uniform across seasons for both NO3–NO2 and OP concentrations and flux on the Missouri River at Hermann, Missouri; however, both have been steadily increasing over time (Fig. 5). Concurrently, SSC has been basically stable during the low flow winter months since the mid-1980s and significantly declining during the spring and summer months (Fig. 5).Following a rise in NO3–NO2 concentrations and fluxes in the early 1980s (Fig. 6), the mainstem of the Mississippi River at St. Francisville, Louisiana, has been fairly stable with regard to NO3–NO2 during the winter and spring, while the summer and fall show some increases, particularly during the last 10–20 years. OP concentrations and fluxes had similar trends across seasons with declines during the 1980s, followed by increases over the following years (Fig. 6). Conversely, SSC concentrations and fluxes decreased dramatically during the several decades, particularly in the winter and spring, following the sediment trends reported by (Mize et al. 2018) for this location. Due to the extremely large flow volumes in the lower part of the Mississippi River and related dilution effects, this stability over time in the lower river for NO3–NO2 is not surprising, but even within this context, there is a clear indication of increases in OP but declines in SSC concentrations and fluxes (Fig. 6).Decadal Trends and Uncertainty in Concentration and Flux for Nine SitesWhile Figs. 4–6 allow a visual comparison of seasonal trends across the three constituents at three of the sites, Fig. 7 shows a comparison of annual trends across all nine sites. The trend estimates are reported as the annualized percent change in flow normalized concentration or flux between the start and end years of three trend periods. These trend estimates are binned (i.e., color coded on Fig. 7) using the p value from a bootstrap test for trend direction. Trend estimates with p values <0.05 were considered upwards or downwards depending on the trend direction. Trend estimates with p values ≥0.05 but <0.20 were considered to have some evidence of upward or downward trends, and trend estimates with p values >0.20 were considered to have uncertain or little change. This analysis covers roughly a 40-year period from WYs 1980 or 1982 to WY 2019. Two subperiods were also analyzed: a 30-year period (or less) from WY 1980 (or WYs 1982 or 1997, depending on the constituent and site) through WY 2010, which may be compared to the nitrate analysis in Murphy et al. (2013), and a period focused on the most recent decade from WY 2010 to 2019. For NO3–NO2, the earliest trends start in WY 1980, for OP they start in WY 1982, and for SSC it depends on the site with some not starting until WY 1997 (Table 1). Because of the varying length of the trend periods by constituent and site, Fig. 7 reports the annualized percent change.Starting at the northern most site in this analysis, the Mississippi River at Clinton, Iowa, had NO3–NO2 concentrations that were increasing at about 1.8 percent per year (%/year) over the period of record, but during the last decade changes in concentration began to level off (Fig. 7). Fig. 4(a) showed that the seasonal NO3–NO2 concentration trends varied at the Iowa River, with strong increases during the spring driving the overall increase annually over the period of record. However, there has been little change over the last decade (Fig. 7). NO3–NO2 concentrations at the Illinois River at Valley City, Illinois, have decreased throughout the period with less change happening during the most recent 10 years. The Mississippi River at Grafton, Illinois, which is the first site downstream from the confluence of the Illinois and Mississippi Rivers, exhibits relatively stable conditions over time, integrating conditions from the upper Mississippi River and the Illinois River. As with Fig. 5, the Missouri River at Hermann, Missouri, exhibits markedly different temporal patterns from the other sites showing increasing NO3–NO2 concentrations, at a rate of 2.5%/year throughout the study period, with particularly strong increases in the last decade (4.3%/year). The volume of the Mississippi River moderates the influence of the Missouri River, so NO3–NO2 concentrations at the Mississippi River at Thebes, Illinois, have remained fairly stable. NO3–NO2 concentrations have gradually declined at the mouth of the Ohio River, though the direction of change is uncertain, and concentrations are stable to decreasing in the lower Mississippi River and the Atchafalaya River, with some uncertainty.OP concentrations and flux in the MRB are largely increasing at most of the nine sites (Fig. 7). While total phosphorus is commonly associated with fine sediments, OP, which is defined as the dissolved, bioavailable component of phosphorus, might be expected to be more similar to NO3–NO2. However, while NO3–NO2 trends varied depending on the site, OP increased at many of the sites for the first 30 years of record. OP continued to increase at some of these sites during the most recent 10 years, and at other sites, new increases occurred during this time that were not evident previously in the record. The exception is the Illinois River, where after a long period of increases in OP concentration, the last decade has shown some evidence of a decrease, and this trend is reflected downstream in the Mississippi River at Grafton, Illinois, though with less certainty. In the case of the Illinois River, NO3–NO2 and OP concentrations changed little or decreased, respectively, from WY 2010 to 2019, suggesting efforts to moderate nutrient inputs to the river may have been successful. Conversely, the Missouri River at Hermann, Missouri, showed large likely increases in NO3–NO2 and OP concentrations and flux. While not exhibiting significant NO3–NO2 trends for most of the trend periods, the Ohio River and the two sites in the lower MRB, show increasing OP concentrations and flux over the same period, but with much of the change occurring during the last decade.Concentration and flux trends for SSC for the time periods available show widespread decreases across the sites (Fig. 7). Almost all nine sites show decreases in SSC concentration and flux from WYs 1980 or 1997 to WY 2019, typically at a rate of around −1.5%/year. Changes in SSC may have leveled off or slowed in the last 10 years in the Illinois River, with an annualized change of about −0.5%/year, and the Missouri River, with an annualized change of about 0.04%/year. Interestingly, these downward trends in SSC are evident at the lowermost Mississippi River site at St. Francisville but are not as strong in the Atchafalaya River at Melville, Louisiana.Comparisons of Concentrations over Time and with StreamflowFigs. 8 and 9 are an illustration of a useful water quality analysis graphic generated using EGRET, the R package used to run WRTDS. These contour plots illustrate the expected (modeled) concentration of a constituent over time, in relation to the range of streamflow volumes that occurred at that site over the period of record. The black lines on the plots indicate the 5th and 95th percentiles of streamflow and their periodic oscillation shows how these magnitudes of streamflow vary over the course of a year with lower flows in the winter, increases in flow in the spring and summer, and decreases in the fall. From Fig. 8, we can ascertain that the highest NO3–NO2 concentrations in the Iowa River during the last 40 years occurred during the 1980s at moderate to high flow conditions during the winter. Concentrations in the Iowa River are still high but have decreased particularly at moderate and high flows, as illustrated by the gradual lightening of color across the x-axis. High concentrations primarily occur during high flows. Conversely, the highest NO3–NO2 concentrations on the Missouri River at Hermann, Missouri (Fig. 9), are nearly four times lower than the Iowa River, and the highest concentrations have occurred during the most recent five-year period at moderate and low flows. This plot supports the trends from Fig. 6 showing increases in NO3–NO2 concentrations over time and indicates that some relatively high concentrations occur during low to moderate flows, particularly during the winter and spring. A comparison of these two plots suggests that the source of nitrate for these two rivers is very different and that different solutions for improving nutrient levels are likely necessary.Discussion and ConclusionsThis paper illustrates the utility of several tools and analyses that could be used to understand and characterize water quality for assessments for navigation projects. These analyses utilize historical data collected, preprocessed, and archived by the USGS’s National Water Quality Network on ScienceBase (Lee 2020). The use of NWISweb (USGS 2020) allows users to identify potentially useful sites beyond those presented here and the Water Quality Portal further expands the possibilities by making water quality data from other state and federal agencies available to researchers. While training is necessary to accomplish this level of analysis, it is not unattainable and there are several online training resources such as an extensive R training curriculum provided by the USGS for USGS-R (USGS 2016) and numerous inexpensive or free online R training resources.With regard to the MRB, examining spatial and temporal patterns of concentration and flux for important constituents is valuable for multiple purposes. Overall, the tributary rivers in the northern part of the basin, which drain intensively agricultural areas show both NO3–NO2 and OP concentrations above desirable levels. The concentrations are diluted by the increased flow and lower concentration from the Ohio River yielding lower concentrations in the lower MRB. The data examined here illustrate nutrient and suspended sediment trends throughout the MRB, showing that while some sites have achieved stable NO3–NO2, only the Illinois River seems to have shown a sustained decrease in NO3–NO2 concentrations and flux through the last decade, while the Missouri River seems to have a clearly upward trend. Many sites show increases in OP during the last decade. While the concentrations are not extremely high, the upward trend is unexpected given the efforts to decrease nutrient inputs to rivers. These increases in OP are particularly relevant given that phosphorus, along with nitrates, likely play a role in Gulf hypoxia (Tao et al. 2014).It is important to note that both SSC concentrations and fluxes have shown a general downward trend throughout the period WY 1997–2020 at the three sites featured in Figs. 4–6 and at the other six sites shown in Fig. 7. These decreases have occurred long after the construction of the major river works in the MRB. The dams in the Missouri River system were completed by 1964 (MCE 2019a), the locks and dams in the upper Mississippi River were largely completed during the 1930s (MCE 2019b), and almost all of the seriously eroding banks in the lower Mississippi River were covered with concrete mats or revetments by the end of the 1970s. So, this decrease in SSC continues well beyond the timeframe expected if the decrease was exclusively related to these river management activities. That is, the decrease in SSC cannot be totally explained by river management activities, which is consistent with what was found in Allison et al. (2017) and Murphy (2020). It is possible that efforts to control erosion and minimize sediment transport from erosional areas in urban and agricultural settings have been at least somewhat successful. However, this decrease in SSC concentrations and fluxes may impact alluvial wetland development and restoration in coastal Louisiana because suspended sediment availability may be less than desired for coastal restoration.References Allison, M. A., D. S. Biedenharn, and C. D. Little. 2017. Suspended sediment loads and tributary inputs into the Mississippi River below St. Louis, MO, 1990–2013: A comparison with the Keown et al. (1981) report. Washington, DC: United States Army Corps of Engineers, Mississippi Valley Division. Allison, M. A., C. R. Demas, B. A. Ebersole, B. A. Kleiss, C. D. Little, E. A. Meselhe, N. J. Powell, T. C. Pratt, and B. M. Vosburg. 2012. “A water and sediment budget for the lower Mississippi–Atchafalaya River in flood years 2008–2010: Implications for sediment discharge to the oceans and coastal restoration in Louisiana.” J. Hydrol. 432–433: 84–97. https://doi.org/10.1016/j.jhydrol.2012.02.020. Aulenbach, B. T. 2006. Annual dissolved nitrite plus nitrate and total phosphorous loads for the Susquehanna, St. Lawrence, Mississippi-Atchafalaya, and Columbia River basins, 1968–2004. Open-File Rep. 2006-1087. Reston, VA: USGS. Choquette, A. F., R. M. Hirsch, J. C. Murphy, L. T. Johnson, and R. B. Confesor Jr. 2019. “Tracking changes in nutrient delivery to western Lake Erie: Approaches to compensate for variability and trends in streamflow.” J. Great Lakes Res. 45 (1): 21–39. https://doi.org/10.1016/j.jglr.2018.11.012. De Cicco, L., R. Hirsch, D. Lorenz, and W. Watkins. 2018. Dataretrieval: R packages for discovering and retrieving water data available from US federal hydrologic web services. Reston, VA: USGS. Hirsch, R. M., and L. A. De Cicco. 2015. User guide to Exploration and Graphics for RivEr Trends (EGRET) and dataRetrieval: R packages for hydrologic data. Reston, VA: USGS. Hirsch, R. M., D. L. Moyer, and S. A. Archfield. 2010. “Weighted regressions on time, discharge, and season (WRTDS), with an application to Chesapeake Bay river inputs.” J. Am. Water Resour. Assoc. 46 (5): 857–880. https://doi.org/10.1111/j.1752-1688.2010.00482.x. LaDOTD (Louisiana Department of Transportation and Development). 2016. Louisiana’s marine transportation system: Catalogue of Louisiana’s waterways.” Baton Rouge, LA: LaDOTD. Lee, C. 2020. Nutrient and pesticide data collected from the USGS National Water Quality Network and previous networks, 1963–2019. Reston, VA: USGS Lee, C. Undated. Tracking water quality in U.S. streams and rivers: USGS National Water Quality Network data, water-quality loads, and trends. Reston, VA: USGS. Lee, C. J., J. C. Murphy, C. G. Crawford, and J. R. Deacon. 2017. Methods for computing water-quality loads at sites in the US Geological Survey National Water Quality Network. Open-File Rep. 2017-1120. Reston, VA: USGS. Little, C. D., and D. S. Biedenharn. 2014. Mississippi River Hydrodynamic and Delta Management Study (MRHDM)-geomorphic assessment. Vicksburg, MS: Coastal and Hydraulics Laboratory, Engineer Research and Development Center. Mize, S. V., J. C. Murphy, T. H. Diehl, and D. K. Demcheck. 2018. “Suspended-sediment concentrations and loads in the lower Mississippi and Atchafalaya rivers decreased by half between 1980 and 2015.” J. Hydrol. 564: 1–11. https://doi.org/10.1016/j.jhydrol.2018.05.068. Murphy, J. C. 2020. “Changing suspended sediment in United States rivers and streams: Linking sediment trends to changes in land use/cover, hydrology and climate.” Hydrol. Earth Syst. Sci. 24 (2): 991–1010. https://doi.org/10.5194/hess-24-991-2020. Murphy, J. C. 2021. Data to incorporating water quality analysis into navigation assessments as demonstrated in the Mississippi River basin. Reston, VA: USGS. Murphy, J. C., R. M. Hirsch, and L. A. Sprague. 2013. Nitrate in the Mississippi River and its tributaries, 1980–2010: An update. Reston, VA: USGS. R Core Team. 2019. R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing. Tao, B., H. Tian, W. Ren, J. Yang, Q. Yang, R. He, W. Cai, and S. Lohrenz. 2014. “Increasing Mississippi river discharge throughout the 21st century influenced by changes in climate, land use, and atmospheric CO2.” Geophys. Res. Lett. 41 (14): 4978–4986. https://doi.org/10.1002/2014GL060361. USACE. 2020. Navigation. Washington, DC: USACE. USACE. 2021. The Mississippi drainage basin. Washington, DC: USACE. USEPA. 2021. National primary drinking water regulations. Washington, DC: USEPA. USGAO (United States Government Accountability Office). 2019. Water resources projects: Army Corps of Engineers can further enhance acceleration of feasibility studies. Report to Congressional Committees. Washington, DC: USGAO. https://www.gao.gov/assets/710/700620.pdf. USGS. 2016. Welcome to USGS-R. Reston, VA: USGS. USGS. 2020. National Water Information System data available on the World Wide Web (USGS Water Data for the Nation). Reston, VA: USGS. Wickham, H. 2016. Ggplot2: Elegant graphics for data analysis. Berlin: Springer. Winslow, L. A., S. Chamberlain, A. P. Appling, and J. S. Read. 2016. “Sbtools: A package connecting R to cloud-based data for collaborative online research.” R J. 8 (1): 387. https://doi.org/10.32614/RJ-2016-029. Zhang, Q., and R. M. Hirsch. 2019. “River water-quality concentration and flux estimation can be improved by accounting for serial correlation through an autoregressive model.” Water Resour. Res. 55 (11): 9705–9723. https://doi.org/10.1029/2019WR025338.