Improvement in municipal wastewater treatment alters lake nitrogen to phosphorus ratios in populated regions

In the surveyed lakes in 2017, mean TN concentration was 996 (95% confidence interval [CI], 380 to 2,723) μg/L (n = 111), and mean TP concentration was 32 (95% CI, 7 to 114) μg/L (n = 111; Fig. 1, SI Appendix, Fig. S1, and Dataset S1). Based on the Chinese surface water quality standards (SI Appendix, Table S1), half of all the surveyed lakes in 2017 (55 of 111) had TN concentrations higher than the grade III limit (1,000 μg/L), above which lakes are considered to be polluted in China (Fig. 1A). In contrast, only 27 lakes had TP concentrations higher than the grade III limit (50 μg/L; Fig. 1B). The mean TP concentration (∼30 μg/L) across all the lakes in this dataset is similar to the levels in Europe (∼25 μg/L) (29) and the United States (20 to 40 μg/L) (30). However, the TN concentration (∼1,000 μg/L) is much higher than those in Europe (∼650 μg/L) (29) and the United States (∼600 μg/L) (31). TN/TP mass ratio was 31.6 (95% CI, 12.5 to 128.2) in the surveyed lakes in 2017 (Fig. 1C). In freshwater ecosystems, when algae is competing for limited nutrients, limitation of algal growth is mainly determined by N when the water TN/TP mass ratio is lower than 9, while P is the main limiting factor when this value is higher than 23 (32). Based on these criteria, more than two-thirds of the 111 surveyed lakes had TN/TP mass ratios higher than 23 in 2017 (Fig. 1C), indicating widespread TN enrichment relative to TP. The mean lake TN/TP mass ratio in 2017 in China approaches the level for European lakes (31; 95% CI, 5–96) (29), but is much higher than that in the United States (21; 95% CI, 4 to 98) (31).

Available monthly monitoring data for 46 lakes during 2008 to 2017 further indicate that TN and TP concentrations are declining at different rates in lakes (SI Appendix, Fig. S2), and that increases of TN/TP mass ratios are widespread among Chinese lakes (Fig. 2 and SI Appendix, Tables S2 and S3). During 2008 to 2017, TP concentrations decreased in 22 of the 46 lakes. The decreasing rate is indicated by slope of the regression line of concentration (logarithm scale) versus time (Methods). The decreasing rate for TP concentration is 0.003 to 0.026, resulting in a half-life of 2.22 to 20.6 y. Only 8 lakes, however, showed increases in TP concentrations (SI Appendix, Fig. S2). In contrast, TN concentrations increased in 22 lakes, with a rate of 0.001 to 0.034, whereas TN decreased in only 12 lakes, with a rate of 0.003 to 0.028 (SI Appendix, Table S2). As a consequence, 24 of 46 lakes showed increases in TN/TP mass ratios, with a rate of 0.0022 to 0.0453 per month and a doubling time of 1.3 to 26.3 y, while only 3 lakes showed decreases in TN/TP mass ratios (Fig. 2).

Shifts in anthropogenic nutrient discharges from the surrounding watershed are usually responsible for alterations of lake nutrient conditions (8, 33, 34). We therefore characterized temporal changes in N and P discharges of the five anthropogenic sources listed here with a spatial resolution of 1 km² during 2008 to 2017 in China (SI Appendix, Supplementary Text S1). Results show that anthropogenic nutrient discharges are much larger in more populated Eastern China (east of the so-called Hu Huanyong Line, with a population density of ∼300 persons/km²) (35) than in less-populated Western China (∼20 persons per km²; SI Appendix, Fig. S3). However, dominant anthropogenic sources in different subregions varied distinctly due to spatial variability in types of human activities (SI Appendix, Fig. S4). Due to disproportionate changes in N and P discharges, TN/TP mass ratios in the total anthropogenic discharge have significantly increased in eastern China during 2008 to 2017 (Fig. 2C). In particular, in the lower reaches of the Yangtze River and Pearl River basins, TN/TP mass ratios of the total anthropogenic discharge were as much as 30% higher in 2017 relative to 2008 (Fig. 2C). However, small increases or even decreases in the ratios were observed in the large, yet less populated, regions of western China.

We took a closer look at 46 lakes with continuous nutrient monitoring data and found that lakes located in more densely populated regions usually had significant increases in their TN/TP mass ratios during 2008 to 2017 (Fig. 3A). Twenty-four lakes with significant increases in TN/TP mass ratios had larger population densities (P < 0.01, t test) in their surrounding watersheds, with a median level of 299 (95% CI, 92 to 743) people/km². In comparison, 22 lakes with no changes or decreases in TN/TP mass ratios had a median population density of 51 (95% CI, 1 to 477) people/km² in their watersheds. Municipal wastewater discharge in the watersheds where lakes experienced increases in TN/TP mass ratios was much higher than that for the other lakes (P < 0.01, t test). As a consequence, N and P discharges contributed by municipal wastewater in the total local nutrient discharge were also higher in the watersheds where lakes had the increasing TN/TP ratios (Fig. 3A).

Linear correlation analysis (36) was performed to examine the direct relationships between lake nutrients (e.g., TN, TP concentrations, TN/TP mass ratios) and the explanatory variables (e.g., discharge from different sources, climate variables in lake catchments; Tables 1 and 2). Among all the variables, the strongest correlations for lake nutrient concentrations and TN/TP mass ratios were with the municipal wastewater discharge in the surrounding watershed (TN: R² = 0.44 [P < 0.01; n = 46]; TP: R² = 0.29 [P < 0.01; n = 46]; TN/TP: R² = 0.17 [P < 0.01; n = 46]; Fig. 3B). A generalized linear model (GLM) was further performed to quantify the relative importance of each explanatory variable in variations of lake nutrients. GLM results further confirmed that the dominant factor responsible for variations of TN/TP mass ratio in lakes was municipal wastewater discharge in the watersheds, accounting for ∼39% of the variations (Tables 1 and 2). Apart from municipal wastewater discharge, we found that other sources could also influence lake nutrient concentrations. For instance, rural wastewater discharge was also an important factor in explaining variations of TN and TP concentrations in lakes (Tables 1 and 2)

The apparent importance of municipal wastewater discharge in influencing in-lake nutrient dynamics motivated us to investigate the nutrient discharge records collected in influents and effluents of 4,960 WWTPs during 2008 to 2017 in China (details in Dataset S2). To mitigate serious water pollution, treatment of municipal wastewater has advanced rapidly in China over the past decade. The number of WWTPs had increased rapidly from 1,535 units in 2008, with a volume of treated wastewater of 22 × 10⁹ m³, to 4,960 units in 2017, with a volume of treated wastewater of 57 × 10⁹ m³ (SI Appendix, Fig. S5). Average effluent TN and TP concentrations decreased from 16.4 and 1.0 mg/L in 2008 to 10.7 and 0.42 mg/L in 2017, respectively (Fig. 4A). By comparing nutrient concentrations in influent and effluent, it is apparent that existing wastewater treatment technologies in China have differential removal efficiencies for TN and TP from the municipal wastewater (Fig. 4A). Wastewater ammonium is microbially converted to N₂ by nitrification and denitrification, a process that requires careful tuning of facility operations (37). However, removing P from municipal wastewater is a simpler and more effective process involving coagulation, flocculation, and sedimentation (38). Thus, average removal efficiency from municipal wastewater by WWTPs is as high as 90% for TP, but less so (60% to 70%) for TN, resulting in the WWTP effluent with much higher TN/TP mass ratios (median, 23.2; 95% CI, 14.7 to 37.0) than the influent (median, 8.9; 95% CI, 7.0 to 11.1; P < 0.01, t test; SI Appendix, Fig. S6). As the percentage of treated municipal wastewater in the total wastewater discharge increases, more P than N is effectively removed from municipal wastewater discharge. Consequently, TN/TP mass ratio in the total municipal wastewater discharge has increased sharply from a median of 10.7 (95% CI, 7.6 to 15.1) in 2008 to 17.7 (95% CI, 13.2 to 27.2) in 2017 (Fig. 4 B and C).

To evaluate how changes may be responsible for observed increases on lake TN/TP mass ratios, we created a set of simple null models for TN and TP concentrations and TN/TP mass ratios based on quantitative relationship model revealed in GLM analysis (SI Appendix, Table S4; see details in Methods and SI Appendix, Supplementary Text S2). Two typical scenarios were simulated: null scenario (NUS) and real situation scenario (REA). Under NUS, we assumed that no new WWTPs were constructed after 2008, and the percentage of wastewater being treated remained unchanged during 2008 to 2017. Under REA, the development of WWTPs followed the observed trajectory (Dataset S2). The results showed that estimated nutrient concentrations under NUS were statistically different from the results under REA (SI Appendix, Tables S5 and S6). Under NUS, the estimated TN and TP concentrations in lakes would be 2.4 and 1.4 times, respectively, higher than those under REA. However, average TN/TP mass ratio in lakes under NUS was ∼25% lower than under REA (SI Appendix, Fig. S7). These impacts on TN/TP mass ratios in lakes might be alleviated by increasing N removal efficiency by WWTPs (SI Appendix, Supplementary Text S3). A simple prediction based on the quantitative relationship in SI Appendix, Table S4 indicates that if TP removal efficiency remained at the 2017 level and TN removal efficiency had a 3% increase per year, lake TN/TP mass ratios would return to 2008 levels within 10 y (SI Appendix, Fig. S8). This suggests that TN removal efficiency by WWTPs should be improved by 27% over 10 y.

In 2015, the United Nations Development Program set a target to halve the proportion of untreated wastewater in the world by 2030 (1). Encouraged by this target, China and many other developing countries worldwide (e.g., India, Indonesia, and Nigeria) have built or are constructing numerous WWTPs (39). Therefore, the challenge of imbalanced lake TN/TP ratios caused by improved sanitation is not confined to China, but also likely occurs in other countries. To achieve coherent and sustainable wastewater management, potential ecological consequences induced by N overenrichment relative to P in aquatic ecology will require consideration in future sanitation approaches in which N and P removals are considered in a holistic way. Possible short-term strategies could include refining operations of existing facilities (40), developing more efficient N-removal technologies (37), and introducing new standards setting TN/TP ratio targets for effluent discharge (38). In the longer term, increasing nutrient recovery from municipal wastewater along with source separation of human excrement may also be promising (2–4, 41).

In summary, consistent with recent changes in anthropogenic nutrient discharges, numerous Chinese lakes in highly populated regions are showing increasing TN/TP mass ratios, which can bring some unexpected consequences for the functioning of aquatic food webs, such as the increasing prevalence of non-N₂ fixing cyanobacteria (Microcystis, Planktothrix) in the eutrophic lakes (23–28). Promoting a better balance between N and P supplies will allow a greater diversity of taxa as well as improved food web control of phytoplankton biomass. These modern perspectives on nutrient stoichiometry in lakes imply that more effort should be placed on N removal from municipal wastewater in the future to achieve desired water quality outcomes as human society seeks to achieve SDGs for sanitation, water, and aquatic ecosystem health. Our findings document an unappreciated impact of wastewater treatment on aquatic ecosystems and also provide a science-based argument for more effectively managing water quality by balancing N and P removals from wastewater.

We obtained an extensive lake nutrient monitoring dataset during 2008 to 2017 from the water quality monitoring program performed by the Ministry of Ecology and Environment, China. As in refs. 42 and 43, we did not differentiate natural lakes or artificial reservoirs, since both waterbody types can be influenced by anthropogenic nutrient discharges in surrounding watersheds. The dataset consists of two parts: nutrient data revealing the condition of lakes in 2017, and nutrient data revealing the temporal trend. In dataset 1, a total of 111 freshwater lakes distributed in China’s 30 provinces were included. Measured TN and TP concentrations, locations, surface areas, and water volumes for these lakes were provided in Dataset S1. In the dataset 2, 46 of the 111 lakes with continuous monitoring data (generally with >40 monthly and paired TN and TP monitoring data during 2008 to 2017) were selected to assess temporal trends in nutrient concentrations and ratios. These 46 lakes were widely distributed in China, and had differences in lake characteristics (e.g., area, water volume) and human activity types in their watersheds (SI Appendix, Table S3). Changes in lake nutrient concentrations are a first-order process (43, 44). Therefore, we used ln-transformed TN (or TP concentrations or TN/TP mass ratio) in regression analysis between monthly nutrient data and corresponding sampling time. If P < 0.05, then the slope (k) in the regression was defined as the monthly increase rate (if k is >0, month⁻¹) or decrease rate (if k is <0, month⁻¹) during the study period; if P > 0.05, it was defined as no significant change in nutrient concentration or ratio during 2008 to 2017. Half-life (T_1/2) or doubling time (T₂) of nutrient concentrations or ratios in lakes was estimated as follows: T_{2 or 1/2} = ln (2)/k. TN or TP concentration, TN/TP mass ratio, increase or decrease rate, hydrological information and location for each lake is provided in SI Appendix, Table S3. Information about surface areas and water volumes in the surveyed lakes was derived from the HydroLAKES database by Global HydroLab (45), and Wang and Dou (46).

Procedures for selecting sampling sites in lakes and collecting water samples were generally consistent throughout the study period and were based on the Technical Specifications Requirements for Monitoring of Surface Water and Waste Water in China (47). A depth-integrated sample was collected at each sampling site. For lakes with a depth greater than 5 m, water samples at 0.5 m below surface were collected. For lakes with depths between 5 and 10 m, water was collected at 0.5 m below the surface and 0.5 m above the bottom and then mixed to make a composite sample. The sampling site was usually set at the center of the lake. However, for the large lakes, multiple sampling sites were established at the center of the lake, the deep zone, and the shore zones, respectively. Standard methods (shown here) were used for measuring TN and TP concentrations. An unfiltered aliquot of water was prepared from each bulk sample. TN concentration was determined by persulfate digestion, followed by automated colorimetric analysis (N-[1-naphyl] ethylene diamine dihydrochloride spectrophotometry), with a method detection limit (MDL) of 50 μg/l-N (47). TP concentration was determined by persulfate digestion, followed by automated colorimetric analysis (ammonium molybdate and antimony potassium tartrate under acidic conditions), with an MDL of 10 μg/l-P (47). All the TN and TP concentrations lower than the MDL were set to 1/2 of the MDL in the subsequent data analyses.

To identify the driving forces for alterations of nutrient concentrations and their ratios in lakes, we analyzed the temporal dynamics of major anthropogenic sources for N and P discharges during 2008 to 2017 in China by applying a hierarchical model of national nutrient cycling updated from Cui et al. (48) and Liu et al. (49). In China, human activities have contributed to the bulk of N and P discharges into water since the national economic reforms were implemented in 1978 (48, 49). Thus, in this study, five major and independent human-induced N and P discharge sectors were selected: municipal wastewater, rural wastewater, livestock operation, crop farming, and aquaculture. Data on anthropogenic activity were collected from a number of national statistical databases, literature, and government reports, while most parameters were derived primarily from studies based on the field experiments (SI Appendix, Supplementary Text S1). By emphasizing impacts of increasing adoption of WWTPs, estimates of N and P discharges from municipal wastewater were based on a national-scale investigation during 2008 to 2017. Total N and P discharges from WWTPs were further aggregated from each of the 4,960 WWTPs. A detailed description of the temporal changes of number of WWTPs, volumes of treated and untreated wastewater, TN and TP concentrations in influent and effluent, and N and P discharges through wastewater during 2008 to 2017 in China’s 31 provinces was provided in Dataset S2.

Forty-six lakes with the continuous nutrient monitoring data during 2008 to 2017 were selected to identify the specific relationships between lake nutrients concentrations (or ratios) and temporal anthropogenic discharges in surrounding watersheds. To address spatial variabilities in human activities in China, a raster dataset based on the population distribution (35), livestock distribution (50), and land use distribution (35) was developed. The result was a high-resolution map (1 km × 1 km grid) of annual anthropogenic nutrient discharges incorporating all of the five major sources during 2008 to 2017. By using the HydroSHEDS database, watershed boundaries for the 46 lakes were delineated by using the ArcMap 10.2.2 software (51, 52).

We conducted a GLM analysis (53) to quantify the relative associations of each explanatory variable with variations in TN and TP concentrations and TN/TP mass ratios in lakes (SI Appendix, Supplementary Text S2). Explanatory variables included the five anthropogenic discharge variables (discharges from five different sources) and two climate variables (precipitation and temperature in watersheds) (36). Based on the Bayesian information criterion (36, 54), the best models revealing the quantitative relationship between lake nutrients and anthropogenic discharges were established (results in SI Appendix, Table S4). These models were further applied in the simple scenario analysis. Scenario analysis was divided into two parts in this study. First, to quantify the impacts of WWTPs on temporal change of water quality during 2008 to 2017, we created simple null models for TN and TP concentrations, and TN/TP mass ratios in lakes individually. The null hypothesis is that the construction of WWTPs had caused no significant changes of lakes nutrients during 2008 to 2017. Two typical scenarios were compared for the null hypothesis test: NUS, representing the temporal trends of nutrient concentrations and their ratio if wastewater treatment capacity and efficiency remained at the level of 2008 throughout the study period; and REA, representing the temporal trends of nutrient concentrations and their ratios in which the development of WWTPs in watersheds followed the observed trajectory (SI Appendix, Supplementary Text S2). A single-factor ANOVA test was used to verify the null hypothesis. In addition to the null test, we also estimated the potential response of TN/TP mass ratio in lakes under different future scenarios with improvement of TN removal efficiency by WWTPs, while TP removal efficiency remained at the same level in 2017 (SI Appendix, Supplementary Text S3). Three typical scenarios were simulated where TN removal efficiency by WWTPs was improved by 1%, 2%, and 3% per year, respectively, in the next decade. The statistical analyses and data processing were carried out by using Excel 2010 (Microsoft Corporation), SPSS 16 (International Business Machines Corporation), lm and relaimpo functions (55) in R software (R Core Team), and ArcMap 10.2.2 (Environmental Systems Research Institute).

The authors declare that all the data supporting the findings of this study are available within the article and its SI Appendix.