Controls of point and diffuse sources lowered riverine nutrient concentrations asynchronously, thereby warping molar N:P ratios

The input of nitrogen (N) and phosphorus (P) into rivers has been reduced in recent decades in many regions of the world to mitigate adverse eutrophication effects. However, legislation focused first on the reduction of nutrient loads from point sources and only later on diffuse sources. These reduction strategies have implications on riverine N:P stoichiometry, which potentially alter patterns of algal nutrient limitation and the functions or community structure of aquatic ecosystems. Here, we use a dataset spanning four decades of water quality for the Ruhr River (Germany) to show that the asynchronous implementation of point and diffuse source mitigation measures combined with time lags of catchment transport processes caused a temporally asynchronous reduction in dissolved inorganic nitrogen and total phosphorus concentrations. This asynchronous reduction increased the molar N:P ratios from around 30 to 100 in the river sections dominated by point sources, reducing the probability of N limitation of algae in favor of P limitation. The Ruhr River catchment and the environmental policies implemented here illustrate the unintended effects of nutrient control strategies on the ecological stoichiometry at the catchment scale. We urge to assess systematically, whether unintentionally warped macronutrient ratios are observable in other managed river systems and to evaluate their environmental impacts.


Introduction
Large-scale mitigation efforts reduced nutrient loads to many rivers in recent decades (Kelly and Wilson 2004, Ibisch et al 2016, Meybeck et al 2018, Westphal et al 2019 to mitigate eutrophication impacts in aquatic ecosystems (Dudgeon et al 2006) and its adverse social and economic consequences (Withers et al 2014). The most critical nutrients of concern are nitrogen (N) and phosphorus (P), which have been excessively released to terrestrial and aquatic ecosystems by intense agriculture and industrial or domestic wastewater inputs (Ibisch et al 2016). The reactive forms of these two nutrients control primary production and thus the eutrophication processes in freshwater and marine ecosystems (Rabalais et al 2009, Dodds andSmith 2016).
At first, targeted nutrient mitigation measures aimed at reducing P inputs from point sources, starting already in the 1970s (Schindler et al 2016). Only later N and P loads from diffuse sources, especially from agriculture, were addressed by policy measures (Ibisch et al 2016). This asynchronous implementation of measures targeting different nutrient sources, but also shifts in their effectiveness altered not only absolute instream N and P concentrations or loads (Paerl 2009) but likely also their stoichiometric proportions.
The stoichiometric N:P ratio is of great importance for nutrient limitation of algal growth (Guildford and Hecky 2000, Keck and Lepori 2012) and potentially impacts the function and community structure of aquatic ecosystems (Peñuelas et al 2013, Meunier et al 2017. The first and most well-known N:P ratio is the Redfield ratio of 16N:1P, which describes the surprisingly consistent ratio of nutrients in marine phytoplankton (Redfield 1960). Various other critical N:P ratios have been proposed in recent years to better predict nutrient limitations (Bergström 2010, Keck andLepori 2012). Knowing the limiting nutrient of in-stream algal growth is crucial to implement effective measures to mitigate adverse environmental impacts.
Alterations in aquatic N:P ratios by human activities were shown at different spatial scales (Penuelas et al 2020) and for long-term time series (Justić et al 1995, Radach and Pätsch 2007, Grizzetti et al 2012, Minaudo et al 2015, Burson et al 2016, Oelsner and Stets 2019, Penuelas et al 2020. However, altered N:P ratios of river ecosystems are rarely studied and even less in a Driver-Pressure-State-Impact perspective, although it constitutes a crucial step towards understanding the effects of policy measures on biologically relevant N:P shifts in river ecosystems. We, therefore, analyzed the entire cause-effect-response-chain in a long-term nested record from a central European river catchment (Ruhr River, Germany). The proxies comprise quantitative N and P inputs triggered by sets of policy measures including source control, end-of-pipe treatment and shifting land-use intensities, resulting riverine N and P concentrations, and N:P ratios.
The data record is based on weekly samples from 1989 to 2013 provided by the local river basin management organization and covers a ruralurban impact gradient from upstream to downstream reaches situated at low-mountain to lowland elevations. The time series represents a period of intense, long-term efforts to manage N and P loads and allowed to test the following hypothesis: Asynchronous mitigation measures targeting nutrients from point and diffuse sources led to sequential reduction of P and N loadings and instream concentrations in the Ruhr River. At the same time, N:P ratios exhibit asynchronous temporal changes of biological relevance.
To test this hypothesis, we focused on three objectives: (1) We evaluated the effects of N and P mitigation measures over time and along the river gradient by using annual average concentrations and concentration-discharge relationships. (2) We determined the temporal and spatial interplay of asynchronous source reduction on N and P concentration (3) We evaluated whether temporal patterns or spatial magnitude of reductions in N and P concentrations have the potential to affect eutrophication via nutrient limitations in relation to molar N:P thresholds as proxies.

Study site
The Ruhr River in western Germany (figure 1) has a length of 219 km and a catchment area of 4478 km 2 . It

Data sources
The river basin management organization for the Ruhr River (Ruhrverband) has carried out weekly routine water quality measurements at seven monitoring sites along the main river from upstream to downstream reaches (figure 1). Westphal et al (2019) showed that the monitoring sites of the upper, middle and lower river reaches divide into three distinct water quality types mainly due to increasing proportions of wastewater related to increasing shares of urban areas (table 1), resulting in corresponding effects on water quality along the catchment gradient. Thus, we selected three representative sites for this study: Wildshausen (UP), Westhofen (MID), and Duisburg (DOWN). UP is the most upstream site, whereas DOWN is located just before the confluence with the Rhine River. MID represents the middle course of the Ruhr River, having a medium anthropogenic impact with only a few medium-sized cities upstream. Since 1966, NH 4 -N and TP concentrations, as well as discharge, were measured every week. Since 1989, also nitrate (NO 3 -N) has been measured at all monitoring sites. Measurements for nitrite (NO 2 -N) are only available from 2000 to 2013. We consider them of minor importance due to their low average proportion of 1.6% for NO 2 -N relative to the sum of NH 4 -N, NO 2 -N, and NO 3 -N. This proportion was derived based on annual water quality reports published by the Ruhrverband (Ruhrverband 2001(Ruhrverband , 2014. We, therefore, refer to dissolved inorganic nitrogen (DIN) as the sum of NO 3 -N and NH 4 -N in the context of this study. TP,and NO   . This data set balances N inputs (different fertilizer types, atmospheric deposition, N-binding by legumes, externally produced feed, and co-substrates) and N outputs (plant and animal market products) on an annual basis. Analogous time series data for agricultural P surplus are not available. However, point sources dominate the P transported to the Ruhr River, and agricultural, diffuse P sources are of minor importance (Westphal et al 2019). We, therefore, reconstructed time series for per-capita TP loads of domestic wastewater (national level) and the share of population equivalents (PE) from different WWTP steps (catchment level) based on archive and literature information. Further references can be found in Westphal et al (2019).

Data analysis
In order to address objective (1) we characterized the temporal evolution of annual median concentrations for NO 3 -N, NH 4 -N, and TP, and linked them to data on N-surplus, construction and operation of WWTPs and per-capita changes in TP loads. We characterized the long-term evolution of export patterns by determining concentration-discharge relationships using annual slopes b of linear log C-log Q relationships for NO 3 -N, NH 4 -N, DIN, and TP, at each monitoring site. These relationships allow to classify into dilution-driven, mobilization-driven or constant export patterns, which link to potential nutrient sources (point and diffuse sources) (Musolff et al 2015): A slope b of −1 indicates that the concentration of the solute varies inversely with discharge, which implies a dilution-controlled behavior in which a constant flux of the solute is diluted by varying discharge (Godsey et al 2009) as observed for wastewater point sources (Greene et al 2011). A positive slope is indicative of the mobilization of solutes with higher discharge. However, it can also point to concentration-discharge independence attributed to high reactivity that reduces low flow concentrations more efficiently or a threshold-driven transport of the constituents, mainly related to diffuse sources (Musolff et al 2015). If solute concentrations do not vary with discharge (−0.2 < b < 0.2, Godsey et al 2009), a chemostatic export pattern prevails that is assumed to originate from diffuse sources (Greene et al 2011). Thompson et al (2011) and Musolff et al (2017) additionally define a chemostatic export pattern by the ratio of the coefficients of variations of concentrations and discharge (CV C /CV Q < 0.5) since a low slope b does not necessarily imply that concentrations are invariant.
To address objective (2) we visualized the relative importance of both nutrients by plotting the weekly measured TP against DIN and by fitting a segmented regression to the data, which also required a breakpoint analysis, using the 'segmented' package version 1.1-0 (Muggeo 2008) of the software R (R Core Team 2018). We further determined the median annual molar N:P ratios over time and for each monitoring site. We based stoichiometric calculations on DIN and TP because this ratio is the most reliable predictor for nutrient limitation at least in lake and marine systems (Morris and Lewis 1988, Bergström 2010, Kolzau et al 2014. In the case of N, the largest bioavailable pool is DIN, whereas the best proxy of P bioavailability is TP since this fraction encompasses both externally available dissolved P and internal reserves of particular P derived from luxury consumption (Ptacnik et al 2010).
Additionally, we conducted trend characterization of the N:P ratios by employing Kendall's Tau Test based on weekly data and for every monitoring site. Using a Mann-Whitney U-Test, we tested for the equality of molar N:P ratios among monitoring sites, based on weekly available data. Our data met all assumptions of the statistical tests.
For discriminating between N and P limitation, we used a range of N:P thresholds introduced by Keck and Lepori (2012), who found that predictions of N or P limitation were highly uncertain except at extreme N:P molar ratios. N:P ratios of ≤1:1 are associated with a high probability of N limitation of microphytobenthos, whereas N:P ratios ≥100:1 are associated with a higher probability of P limitation.
We performed all statistical analyses with the software R, version 3.5.2. (R Core Team 2018).

Temporal evolution of nutrient concentrations
Annual median DIN concentrations were highest with 4.51 mg L -1 , 5.03 mg L -1 , and 5.63 mg L -1 in 1991, 1996, and 1991 at UP, MID, and DOWN, respectively. Concentrations then decreased by 38%, 43%, and 54% (figure 2(a), tables 2 and 3). Note that NO 3 -N mainly dominates this pattern due to its significant contribution to DIN (figure S1(a) (available online at stacks.iop.org/ERL/15/104009/mmedia), tables 2 and 3). Agricultural N-surplus rates, likely the primary source for DIN, were highest in 1983 with 146 kg ha −1 , and exhibit a clear downward trend since 1987 (figure 2(b)). In total, surplus rates decreased by 26% until 2013 and stabilized since the mid-1990s around values of about 100 kg ha −1 yr −1 (table 2). Annual DIN concentrations and N surplus rates weakly correlated (figure S7).
Concentrations of NH 4 -N decreased in the 1970s by 96%, 97%, and 97% for UP, MID, and DOWN, respectively (figure S1(b), table 2). This reduction is in line with the proportion of biological treatment that considerably increased since 1976 in the Ruhr catchment. Since 1996, the proportion of nitrification and denitrification stages in WWTP has also increased sharply (figure S1(d)), efficiently removing NH 4 -N from wastewater.
Annual median TP concentrations have peaked at all sites in 1976 with maximum annual median concentrations of 0.42 mg L −1 , 0.59 mg L −1 , and 1.04 mg L −1 at UP, MID, and DOWN (figure 2(c), tables 2 and 3). Since the mid-1990s, annual median TP concentrations have stabilized at <0.1 mg L −1 . Per-capita TP loads in domestic wastewater, which are linked to instream TP concentrations, were highest in 1975 with 5 g TP cap −1 d −1 and then decreased by 64% until 2013 (figure 2(d), tables 2 and 3). Since the mid-1990s, per-capita TP loads have stabilized around 1.8 g TP cap −1 d −1 . Annual median TP concentrations and annual per-capita loads strongly correlated ( figure S8).
Besides the per-capita TP loads, the number of WWTPs in the Ruhr catchment also increased from the mid-1970s to the early 1990s (figure S2). Accordingly, the number of people-expressed as population equivalent (PE)-whose wastewater has been biologically treated increased by a factor of 2.3 since the mid-1970s (figure 2(e)). Since the beginning of the 1990s, also the number of people whose wastewater has been treated with P removing treatment steps such as precipitation or biological P elimination increased 3.4 fold (figure 2(e)).

Temporal evolution of export patterns
For all three monitoring sites, the C-Q regression slopes for DIN (b DIN ) were mostly in the range of −0.2 and 0.2 (figure 3), pointing to a chemostatic export pattern. However, the Mann-Kendall test revealed a significant positive trend over time for MID and DOWN, but not for UP (table S1), indicating a change of export patterns from chemostatic towards mobilization for the mid and downstream reach. The temporal development for b DIN is consistent with the development of CV DIN /CV Q ratios, remaining below 0.5 (figures S3 and S4). For NO 3 -N, the slopes b NO3 and CV NO3 /CV Q corresponds with DIN (figures S3-S5). Export patterns for NH 4 -N were dynamic throughout the time record at all sites with initially dilution-driven (b NH4 ≤ −0.2), then chemostatic (−0.2 ≤ b NH4 ≤ 0.2) and finally mobilization-driven phases (b NH4 ≥ 0.2) (figure S5). Annual CV NH4 /CV Q was also variable and increased over time and amounted to ≥0.5 (figures S3 and S4).
For TP, slopes b TP varied in time at all sites, beginning with a phase of a more dilution-driven export pattern (b ≤ −0.2) and then changing at the beginning of the 1980s towards a chemostatic export pattern with b TP being within a range of −0.2 and 0.2 (figure 3-TP) and CV TP /CV Q ratios below 0.5 (figures S3 and S4).

Resulting nutrient stoichiometry and limitation
The N:P ratios showed a temporally dynamic behavior at all sites with partly synchronized phases of increasing and decreasing trends in 1994, 2002, and 2009. (figure 4(d), table 4, figure S6). While MID and DOWN sites showed an increase in N:P ratios from an initial low level of N:P ratios between 1988  1975 (1989) and 1994, UP started already from a higher level of N:P ratios (  4(a)). In total, annual median N:P ratios have increased from 1989 to 2013 by 19% for UP, 105% for MID and 25% for DOWN (table 2). Overall mean N:P molar ratios decreased along the river gradient from UP (N:P UP = 83 ± 19), over MID (N:P MID = 79 ± 19) to DOWN (N:P DOWN = 69 ± 16). A Mann-Whitney U-test also revealed that UP and MID have significantly higher N:P ratios than DOWN (table S4).
Considering the thresholds defined by Keck and Lepori (2012), P limitations (N:P ratios ≥ 100) were likely to occur in 33% of the time in UP, 28% of the time in MID and 21% of the time in DOWN based on weekly data. Especially since the beginning of 2000, N:P ratios have been very close to or above 100, especially for UP and MID (table 4). N limitation with N:P ratios < 1 did not occur.

Temporal evolution of point-and diffuse-source nutrient contributions
We found significant declines in concentrations of N and P components due to nutrient mitigation measures, mostly related to policies introduced in the 1970s and 1980s in order to control eutrophication.
At all sites, annual slopes b and CV C /CV Q ratios for NO 3 -N and DIN indicated a chemostatic export pattern for most of the time, suggesting that diffuse sources dominated instream NO 3 -N and DIN over the entire time record and hence mitigation measures targeting diffuse sources also caused a decline in NO 3 -N and DIN concentrations. Hence, from 1987 onwards, the reduction of agricultural N-surplus rates led to a reduction of DIN and NO 3 -N concentrations. Interestingly, the decline in N surplus rates started before the adoption of fertilizer limiting regulations such as fertilizer acts (BGBl. I S. 2134 1977, BGBl. I S.1435 1989) or the EU Nitrate Directive (Directive 91/676/EEC 1991) which may also be related to changed land-use practices (Blesh and Drinkwater 2013).
For NH 4 -N, the alterations of C NH4 -Q shapes over time indicate a fundamental shift in sources, as evidenced by the temporal evolution from a  more dilution-driven to a more chemostatic export pattern. High slopes and high CV NH4 /CV Q ratios since mid-1990s point to uncoupling effects of concentrations from discharge related to changes in NH 4 -N sources. This uncoupling is likely due to fast NH 4 -N turnover in the stream that dominates over mobilization processes (see also Musolff et al 2015).
The reduction of NH 4 -N concentrations likely links to reduced N loads originating from point sources induced by the increased availability of biological wastewater treatment and the technical upgrading of WWTPs with additional nitrifying and denitrifying treatments steps in the mid-1990s. We attribute these changes to various national and European policies in the 1970s to 1990s (e.g. amend-  Table 4. Site-specific mean N:P ratios, standard deviation (sd), and trend statistics (Kendall's τ and p-value) for four different periods defined by the years of trend change according to figure 4. Computations are based on weekly data. 1989-1994 1994-2002 2002-2009 2009- (BGBl. I S. 664 1980) and technical upgrades of WWTPs due to various administrative regulations in the 1970s as the primary reason for significant concentration reduction. These measures led to a fundamental shift in P sources or transport and retention mechanisms, altering C TP −Q shapes.

Synchronicity and magnitude of changes of riverine N and P concentrations
The riverine response to the implementation of measures for addressing point sources (to control TP and NH 4 -N) and diffuse sources (to control NO 3 -N) took place at different times with a time lag of at least 10 years: for point sources mostly in the mid-1970s, and for diffuse sources not before 1987 as the main driver for NO 3 -N concentration, N-surplus rates, began to decline only then.
In addition to the asynchronous timing of the legislation addressing point and diffuse sources, there is apparent asynchronicity in the response of riverine concentration to mitigation measures. Whereas TP concentrations showed a direct response to reduced per-capita TP loads and enhanced wastewater treatment, DIN concentrations reacted with a time lag to reduced N-surplus rates. Taking the observed maximum annual DIN concentrations, we found a time lag of 4 years (UP and DOWN) to 9 years (MID). However, the determination of turning points is somewhat imprecise, as no measurements are available before 1989. Riverine concentrations often react with a delayed response to improved N management (Ehrhardt et al 2018), which is related to various reasons such as long groundwater residence times, that leads to long flush-out times of nitrate from groundwater (Hamilton 2012). Additionally, the long-term accumulation of soil N stocks (Worrall et al 2015, Dupas et al 2018 and shifts in precipitation and discharge patterns (Dupas et al 2016) influence the instream-response to improved N management.
Besides asynchronous timing in N and P concentration, there were also differences in the extent and magnitude of N and P reduction. The management of point sources (TP and NH 4 -N) was more effective than the treatment of diffuse sources (NO 3 -N). The percentage decrease has been much higher for TP and NH 4 -N (∼86% to 97%) than for NO 3 -N (∼31% to 44%) (table 2). This imbalance in the reduction of N and P was also reported for other riverine systems

Magnitude and biological relevance of shifts in molar N:P stoichiometry
At the beginning of the time series (figure 4(d), figure S6), we found lower N:P ratios at the MID and DOWN sites (∼30 to 40) than for the UP site (∼80). The critical difference lies in the level of TP concentrations, which were already at a considerably lower level in UP compared to MID and DOWN, with DIN concentrations being at a similar level in all sites. The rapid decrease in TP concentrations in direct response to the measures described above created a sudden increase of stoichiometric ratios associated with a doubling of molar N:P ratios in MID and DOWN. High TP concentrations before 1989 likely led to extremely low N:P ratios, which probably dropped to a minimum in the mid-1970s when TP concentrations reached its maximum. Therefore, N:P ratios were probably subject to even stronger fluctuations in the course of various eutrophication measures than can be shown here due to the limited length of DIN time series. Similar extremely low N:P ratios between 1960 and 1990 and a sudden shift in the early 1990s were also shown by Wentzky et al (2018) and Radach and Pätsch (2007). The increase of N:P ratios in their study coincides with the increase of ratios in our study and showed similar trend reversals in mid-1990, early 2000 and again around 2010, suggesting similar drivers across Germany. Here, increasing N:P ratios at the beginning of the time series are likely due to non-simultaneous reductions of N and P loads. These are contrasted by declining N:P ratios at the end of the time series, best explained by now constant TP concentrations but decreasing DIN concentrations.
However, shifted discharge patterns may also have played a role for sudden trend reversals, as Green et al (2007) and Green and Finlay (2010) showed that N:P ratios consistently decline during high flows. At the Ruhr River, extreme rainfall in February caused a significant flood in 2002 across monitoring sites, potentially causing a decline in N:P ratios. The significant difference in N:P ratios between sites can be best explained by the general lower TP concentrations at UP and MID due to lower urban land use contribution, compared to DOWN with a higher contribution of urbanized areas in the catchment (table 1). Other studies also found such longitudinal urbanization effects due to WTTP influence (Choi et al 2015, Yan et al 2016, Yun and An 2016 and decreasing N:P ratios along the river gradient were also described for other locations in Europe and Asia (Yin et al 2004, Dupas et al 2017).
The shift of nutrient sources in the Ruhr River not only caused substantial changes in stoichiometry but has reduced the probability of N limitation and increased the probability of P limitation over time, in particular from the beginning of the 2000s until 2013 and especially in the up and midstream reaches, where quantitative P limitation occurred in about 30% of the weekly measurements. Besides relative concentrations, absolute concentrations are also relevant as they determine the magnitude of the nutrient effects on biomass (Dodds 2003, Keck andLepori 2012). Data from more than 200 temperate streams indicated breakpoints at 40 µg TN L −1 and 30 µg TP L −1 , above which chlorophyll concentrations were substantially higher (Dodds et al 2002). These results are corroborated by Withers and Bowes (2018), which specify values below 0.03 to 0.1 mg SRP L −1 for limited algae growth.
In the Ruhr River, N and P concentrations were well above these limits throughout the complete time record despite significant efforts to control eutrophication. Thus, the N:P ratios may have had less influence on the extent of algae growth (see also figure S9), but likely influenced riverine community structure and species diversity (Bowmann et al 2005, Gafner and Robinson 2007, Li et al 2010, Penuelas et al 2020, given the magnitude of the inter-annual fluctuations of N:P ratios in all three sites of the Ruhr River. Phytoplankton diversity may have been depleted, as very few species can compete for the limiting P (Elser et al 2009). Increased P-limitation of phytoplankton also may have changed the performance of food webs, as P-limited algae are a poor food source for zooplankton due to an unsuitable biochemical composition (Müller-Navarra 1995) or low P content (Elser et al 2001). Besides structural changes, also functional changes associated with altered N:P ratios were described (Artigas et al 2008, Schade et al 2011. However, these structural and functional changes related to altered N:P ratios are complex and not well understood yet (Yan et al 2016). We, therefore, urge to systematically assess whether similar asynchronous developments and unequal reductions in N and P, and similar shifts in N:P ratios and a higher probability of P limitation, can be observed in other river systems due to similar changes in environmental policy. It is essential to assess the extent of the environmental impact of such and similar sequences of the described environmental policy decisions in order to determine whether specific stoichiometricbased freshwater management is required, e.g. that synchronize of N and P measures in time and acknowledge the natural constraints of catchments leading to substantial time lags in nutrient transport. Synchronous N and P reduction measures will be of particular benefit for those regions that experience rapid economic growth without adequate environmental protection, leading to significant degradation of water quality in freshwaters, and which will soon address these problems with more stringent environmental legislation.
Our study also highlights the need for efficient controls of diffuse nutrient sources beyond current practices and the avoidance of further N accumulation in anthropogenic landscapes (Paerl 2009, Ehrhardt et al 2018, Ellermann et al 2018. Our findings are also relevant for the freshwaterto-marine continuum, as upstream nutrient management actions in the form of a P-only reduction strategy leads to an intensification of downstream Ncontrolled phytoplankton blooms in coastal waters, particularly in those with increased N loadings (Paerl 2009).

Acknowledgments
We thank the Ruhrverband for providing the data set and the anonymous reviewers for their helpful comments on improving the manuscript. The authors would also like to thank Olaf Büttner for his support with GIS-based analyses.

Data availability statement
The data that support the findings of this study are available from the corresponding author upon reasonable request.