Eutrophication will increase during the 21st century as a result of precipitation changes

See allHide authors and affiliations

Science  28 Jul 2017:
Vol. 357, Issue 6349, pp. 405-408
DOI: 10.1126/science.aan2409

More rain means more pollution

Nitrogen input from river runoff is a major cause of eutrophication in estuaries and coastal waters. This is a serious problem that is widely expected to intensify as climate change strengthens the hydrological cycle. To address the current lack of adequate analysis, Sinha et al. present estimates of riverine nitrogen loading for the continental United States, based on projections of precipitation derived from climate models (see the Perspective by Seitzinger and Phillips). Anticipated changes in precipitation patterns are forecast to cause large and robust increases in nitrogen fluxes by the end of the century.

Science, this issue p. 405; see also p. 350


Eutrophication, or excessive nutrient enrichment, threatens water resources across the globe. We show that climate change–induced precipitation changes alone will substantially increase (19 ± 14%) riverine total nitrogen loading within the continental United States by the end of the century for the “business-as-usual” scenario. The impacts, driven by projected increases in both total and extreme precipitation, will be especially strong for the Northeast and the corn belt of the United States. Offsetting this increase would require a 33 ± 24% reduction in nitrogen inputs, representing a massive management challenge. Globally, changes in precipitation are especially likely to also exacerbate eutrophication in India, China, and Southeast Asia. It is therefore imperative that water quality management strategies account for the impact of projected future changes in precipitation on nitrogen loading.

Nutrient enrichment of water bodies, or eutrophication, is a growing global problem. Whereas phosphorus is the leading concern for freshwater systems, excessive nitrogen is the primary cause of eutrophication in estuaries and coastal waters (1). Associated water quality impacts, including but not limited to the occurrence of harmful algal blooms (2, 3) and hypoxia (4, 5), have been widely documented and are on the rise (4, 6). Ecosystem and human impacts are severe (7, 8). Population growth and changes in land management practices are projected to further increase total nitrogen export globally (9) and for the continental United States (10), as is anticipated agricultural adaptation to climate change (11). Various reports have suggested that the water quality impacts of nitrogen loading may also increase in frequency and intensity as a result of future changes in precipitation (5, 12, 13).

Clear evidence substantiating concerns about the role of precipitation is lacking, however, because very little is known about the impact of changes to the physical climate itself—especially the impact of precipitation—on nitrogen export and therefore on eutrophication. This is the case even though precipitation amount, frequency, and intensity are major controls on riverine nitrogen load (1417). The impacts of changes in future precipitation patterns on nitrogen loading have been examined only for individual watersheds, and such analyses have relied on only one to three global climate models (14, 1820) or a single average across an ensemble of climate models (2124). These studies therefore do not provide a basis for understanding impacts at regional to continental scales, or for examining the robustness of conclusions to uncertainty in future climate. At the same time, emerging strategies aimed at managing eutrophication focus on setting nutrient-loading targets (25, 26). Because loading is most directly influenced by nitrogen inputs and by precipitation patterns, it is imperative to understand how changes in precipitation might in turn affect loading (27), thereby confounding management efforts.

Here, we fill this knowledge gap by providing spatially extensive and contiguous estimates of changes in riverine total nitrogen loading (henceforth, nitrogen loading) for the continental United States. These estimates are derived from anticipated changes in precipitation as projected by 21 different CMIP5 (Climate Model Intercomparison Project Phase 5) models, for three climate scenarios (the RCP2.6 “mitigation” scenario, the RCP4.5 “stabilization” scenario, and the RCP8.5 “business-as-usual” scenario), for two future time periods [the “near future” (2031–2060) and the “far future” (2071–2100)], and for 2105 different subbasins within the continental United States. We use bias-corrected and spatially downscaled (1/8°) climate model projections (28) and report changes at scales ranging from the eight-digit hydrologic unit (HUC8) “subbasin” scale (henceforth, watershed scale; fig. S1) to the continental United States. The analysis is made possible by a recently developed empirical model linking several key variables—net anthropogenic nitrogen inputs into a watershed (e.g., fertilizer application), total annual precipitation, extreme springtime precipitation, and land use—to annual nitrogen flux (17) (see supplementary materials). Although we recognize that a number of factors will have impacts on future riverine nitrogen fluxes, we focus here specifically on impacts of changes in precipitation in the absence of other concurrent changes, because these impacts cannot be avoided through management within the affected regions. We therefore keep net anthropogenic nitrogen inputs (henceforth, nitrogen inputs) and land use constant at existing levels throughout the analysis (2007 and 2006, respectively; see supplementary materials). We use the approach proposed by Tebaldi et al. (29) to assess the significance of observed changes and their consistency across the CMIP5 models; we use the term “robust” to denote results where at least 80% of models are consistent on the direction of change and where the change is statistically significant (P < 0.05) for at least 50% of models.

We find that anticipated changes in future precipitation patterns alone will lead to large and robust increases in watershed-scale nitrogen fluxes by the end of the century for the business-as-usual scenario (stippling in Fig. 1C), especially within the Upper Mississippi Atchafalaya River Basin, the Northeast, and the Great Lakes basin. Watersheds across much of the Northeast show a robust increase even under the stabilization scenario by the end of the century (fig. S2D). These spatial patterns are especially noteworthy because these regions also have high historical nitrogen fluxes (Fig. 1A) and because they discharge to coastal regions with documented water quality impairments resulting from eutrophication (7, 30). We further find that at the watershed scale, only a small fraction of areas will experience a robust increase in fluxes in the near future for any of the examined scenarios (Fig. 1B and fig. S2, A and C); this can be attributed to intermodel differences and internal climate variability (i.e., natural climate fluctuations that arise even in the absence of changes in radiative forcing).

Fig. 1 Projected changes in mean total nitrogen flux for watersheds within the continental United States for the RCP8.5 “business-as-usual” emission scenario.

(A) Total nitrogen flux for the historical period (1976–2005), averaged across 30 years and 21 CMIP5 models. (B and C) Projected change in mean total nitrogen flux for the near future (2031–2060) and far future (2071–2100) relative to the historical period. For (B) and (C), stippling highlights watersheds with a robust change in total nitrogen flux (i.e., more than 50% of the models show a significant change and more than 80% of the models agree on the sign of change). Watersheds with inconsistent projections (i.e., more than 50% of the models show significant change but fewer than 80% of the models agree on the sign of change) are shown in white. Remaining watersheds are shown in color without stippling. The black outlines highlight the upper Mississippi Atchafalaya River Basin and the Northeast region (Fig. 2).

For large aggregated regions (see supplementary materials) including the continental United States as a whole, models agree with high consistency (>80%) that nitrogen loading will increase across all three examined climate scenarios and for both the near- and far-future periods (Fig. 2 and table S1), with the only exception being the lower Mississippi Atchafalaya River Basin. These changes are robust for the far-future periods and the mitigation and business-as-usual scenarios, with significant changes observed for the majority of models and for most regions including the continental United States as a whole (filled box plots in Fig. 2). Although the projected changes in nitrogen flux at the watershed scale for the mitigation scenario are within the range of natural variability (colored regions with no stippling in fig. S2), aggregation to large regions yields a robust increase. For the stabilization scenario, a smaller projected overall increase in total precipitation relative to the other scenarios leads to less robust changes in nitrogen loading for most regions. In the near future, high interannual variability and the smaller projected magnitude of change lead to the observed consistent but not robust increases across scenarios. For the remainder of the discussion, we focus primarily on the far-future period under the business-as-usual scenario.

Fig. 2 Percent changes in mean total nitrogen load within large regions within the continental United States for the RCP2.6 “mitigation,” RCP4.5 “stabilization,” and RCP8.5 “business-as-usual” emission scenarios.

For a given model, total nitrogen load is first averaged for each 30-year period (historical, near future, far future), each scenario, and each region (using an area-weighted average of contributing watersheds), and these values are then expressed as a percent change in projected total nitrogen load within a given region, period, and model. Box plots represent the spread across the examined models (16 for RCP2.6, 20 for RCP4.5, and 21 for RCP8.5) for specific periods and scenarios, with outliers marked as dots. Filled box plots highlight regions with a robust change in total nitrogen load (i.e., more than 50% of the models show a significant change and more than 80% of the models agree on the sign of change). Gray outlines show the two-digit hydrologic unit (HUC2) regions for reference (fig. S1).

The across-model mean projected increase in nitrogen loading within the continental United States is 19% (Fig. 2), with the Northeast (28%), the upper Mississippi Atchafalaya River Basin (24%), and the Great Lakes basin (21%) experiencing the largest increases (Fig. 2). To put these numbers in context, the U.S. Environmental Protection Agency recently set a 20% load reduction target relative to 1980–1996 levels for the Mississippi Atchafalaya River Basin as a whole (26), with the aim of reducing the size of the massive annual hypoxic zone in the Gulf of Mexico (31). We find here that precipitation changes alone will instead lead to an 18% increase in loading within the Mississippi Atchafalaya River Basin as a whole. Offsetting this increase in loading would require a 30% reduction in nitrogen inputs for the region, whereas achieving a 20% loading reduction in light of the confounding effect of precipitation changes would require a 62% reduction in nitrogen inputs (see supplementary materials). For the continental United States, a 33% reduction in nitrogen inputs would be required to offset the 19% nitrogen load increase attributable to changes in precipitation.

The large spread across models indicates that the magnitude of the change in nitrogen load is uncertain, presenting an additional risk for management (Fig. 2). Across-model differences in precipitation projections translate into large uncertainties in the magnitude of nitrogen load change. In addition, we find that a large fraction of this uncertainty is due to internal climate variability (see supplementary materials) (fig. S3) and therefore represents irreducible uncertainty. For large portions of the continental United States, internal climate variability explains more than half of the total intermodel spread for both time periods and for all emission scenarios (see fig. S3, C and D, for results under the business-as-usual scenario). Because current global climate models have been shown to underestimate internal climate variability (32), the actual contribution may be even greater. Furthermore, precipitation downscaling of projected future climate is based on an assumption of climate stationarity (see supplementary materials), the limitations of which represent an additional uncertainty. This result implies that nitrogen loads are expected to increase but that the magnitude of the increase is quite uncertain. For the far future under the business-as-usual scenario, the spread between the first and third quartiles for the continental United States represents increases ranging from 9% to 24%, whereas for the Northeast this range spans an 18% to 39% increase. The full range is broader still (Fig. 2).

We further find that the magnitude of predicted changes in the nitrogen flux is explained by the compounding impacts of changes in the total annual and springtime extreme precipitation, although only the changes due to total precipitation are robust on their own (Fig. 3). The spatial patterns of change in future nitrogen flux (Fig. 1C) are comparable to those that would result only from future changes in total annual precipitation (see supplementary materials) (Fig. 3A). Conversely, accounting only for projected changes in springtime extreme precipitation or changes to the correlation between total annual and springtime extreme precipitation does not lead to robust changes in future nitrogen flux at the watershed scale (Fig. 3, B and C). This conclusion holds true even at regional scales, including for the continental United States, where the magnitude of change is explained by changes in both total and extreme precipitation, with the change in total annual precipitation having the largest impact and leading to a robust increase on its own for most regions (fig. S4). The larger contribution of change in annual precipitation to the change in mean annual nitrogen flux is attributable to the robustness of the projected changes in annual precipitation (fig. S5B) and the larger sensitivity of nitrogen flux to total annual precipitation relative to extreme precipitation (see supplementary materials).

Fig. 3 Contributions to change in total nitrogen flux.

(A to C) Contribution of total annual precipitation (A), extreme springtime precipitation (B), and correlation between annual and extreme precipitation (C) to projected changes in mean total nitrogen flux for watersheds within the continental United States for the business-as-usual emission scenario, averaged across 21 CMIP5 models (Fig. 1C). The individual contributions of these three factors were calculated by eliminating the contribution of the two other factors to the total change in the total nitrogen flux (see supplementary materials). Because of the nonlinearity of the total nitrogen flux model, the contributions are not additive (see supplementary materials). Colors and stippling are as defined in Fig. 1.

Overall, we find that regions with high historical loading (which correspond to regions with high nitrogen inputs and high precipitation) and a robust projected increase in precipitation are most likely to experience a large and robust future increase in nitrogen loading, at both the watershed and regional scales. The empirical model used here to relate nitrogen inputs, land use, and precipitation statistics to nitrogen flux is specific to the continental United States, precluding its direct application to other regions of the globe. We may, however, seek analogs in other regions that meet certain criteria and use those as heuristics to identify other regions where similar conditions exist and similar outcomes may be expected. If large increases in nitrogen load are expected for regions with (i) high nitrogen inputs, (ii) high precipitation, and (iii) a robust projected increase in precipitation within the continental United States, the same is likely to be true in other parts of the world. We therefore reexamined the business-as-usual, far-future precipitation projections across the 21 available CMIP5 models globally (bias-corrected and spatially downscaled to 1/4°) to identify regions that exhibit all three risk factors (see supplementary materials). We find that identifying regions with robust projected precipitation increases (fig. S6A) and high historical total annual precipitation (>75th percentile globally; 656 mm year−1; fig. S6B), combined with data on historical fertilizer application rates (as a proxy for nitrogen inputs) (fig. S6C), provides a good approximation of the regions within the continental United States that are likely to experience a large and robust increase in nitrogen flux (stippled region in Fig. 1C versus continental U.S. area in Fig. 4).

Fig. 4 Global regions most likely to experience large increases in total nitrogen flux.

The map shows 2015 fertilizer application rate for regions with historical (1976–2005) annual precipitation rates above the 75th percentile (averaged over a 30-year period and 21 CMIP5 models) and projected robust increases in annual precipitation by the far future (2071–2100) for the business-as-usual emission scenario. Global regions in dark orange and red therefore exhibit all three risk factors for increased future loading. Regions in yellow and light orange meet the precipitation criteria but have low nitrogen inputs; hatched regions do not meet one (diagonal hatching) or both (cross-hatching) of the precipitation criteria. The black outlines highlight the continental United States and South, East, and Southeast Asia.

Applying this heuristic approach globally makes it possible to identify other regions where changes in precipitation are likely to engender substantial increases in nitrogen load (Fig. 4). We find that large portions of East, South, and Southeast Asia, including India and eastern China, exhibit conditions that are directly analogous to those in the upper Mississippi Atchafalaya River Basin, Northeast, and Great Lakes regions of the continental United States, and these regions are therefore likely to undergo large increases in nitrogen load as a result of projected changes in precipitation. These regions are also home to more than half of the world’s population (33) and are heavily dependent on surface water supplies (34). As a result, increased eutrophication would have widespread impacts. Among countries in this region, India is especially noteworthy because it exhibits all three risk factors across more than two-thirds of its area, is one of the fastest-developing countries in the world, and has one of the fastest-growing populations (33). The precipitation projections in this region are also highly sensitive to aerosol emission trajectories (35), which are themselves uncertain (36). Portions of Europe (e.g., Italy, southern France, Denmark, northern Germany) also display all three risk factors. Other highly agricultural regions (e.g., central Europe, eastern South America, southern Australia) have comparable fertilizer application rates (fig. S6C) but have either lower historical precipitation or less robust projected precipitation changes. In general, this heuristic approach identifies global agricultural regions that are particularly susceptible to the impacts of precipitation changes.

We conclude that changes in precipitation patterns will have substantial impacts on nitrogen loading within the continental United States. These trends will compound changes due to anticipated intensification of land use (9, 10) or they may negate the benefits of strategies aimed at load reductions (9, 10), thereby exacerbating water quality impairments (37). The same scenario is likely to play out in East, South, and Southeast Asia—in particular, in India and eastern China, which have high precipitation and fertilizer application rates and are projected to experience future precipitation increases. Our findings imply that strategies aimed at managing eutrophication and associated water quality problems must account for the impact of changing precipitation patterns on nutrient loading.

Supplementary Materials

Materials and Methods

Figs. S1 to S6

Tables S1 to S3

References (3850)

References and Notes

Acknowledgments: Supported by NSF grant 1313897 (E.S. and A.M.M.) and by the Cooperative Institute for Climate Science, Princeton University, under NOAA grant NA08OAR4320752 (V.B.). We thank K. Findell, J. Ho, M. Lee, Y. Shiga, and three anonymous reviewers for incisive comments on the manuscript and analysis. We acknowledge the World Climate Research Programme’s Working Group on Coupled Modelling, which is responsible for CMIP, and we thank the climate modeling groups (listed in table S2) for producing and making available their model output. For CMIP, the U.S. Department of Energy’s Program for Climate Model Diagnosis and Intercomparison provides coordinating support and led development of software infrastructure in partnership with the Global Organization for Earth System Science Portals. For the global analysis, global climate scenarios used were from the NEX-GDDP data set, prepared by the Climate Analytics Group and NASA Ames Research Center using the NASA Earth Exchange, and distributed by the NASA Center for Climate Simulation. Data used in this study are freely available online, as listed in table S3.

Stay Connected to Science

Navigate This Article