Near-record dissolved N fluxes combined with high cumulative discharge (i.e. the volume of water moving through the river; Fig. 1) provide overwhelming support for our hypothesis (droughts store reactive N in soil and floods flush reactive N into streams) and provide a unique insight into how climate variability creates extremes in N loading. The source of high 2013 N loads can be discerned by comparing the cumulative discharge (Fig. 1a) and cumulative nitrate load (Fig. 1b). Beginning at day of year (doy) 105 in 2013 (Fig. 1a), cumulative discharge climbed steeply, driven by precipitation including two storms that raised mean daily discharge above the 99th percentile (Fig. S1). Despite periods of intense precipitation, the 2013 cumulative discharge remained largely within the 95th percentile of the 40 year record (1970–2009, grey shading, Fig. 1a). However, discharge alone does not explain the extreme N loading in 2013. Rather, the interannual contrast among cumulative nitrate flux (Fig. 1b) suggests that antecedent drought conditions (2012) stored reactive N in the soil and then this excess N was mobilized during spring runoff. Departures between cumulative nitrate flux and cumulative discharge in 2013 support our hypothesis (Fig. 1b). The intense precipitation that occurred in the early spring of 2013 (~doy 110–150; April and May) corresponds to the fastest increase in nitrate flux in the available record (2010–2015; Fig. 1b). The combined effects of elevated discharge and high nitrate concentrations resulted in a cumulative nitrate flux that was 118% greater than the average of the other five years resulting in a 34% increase (10.5 vs. 7.8 mg N L−1) in the flow-weighted mean annual nitrate concentration in 2013 compared to the average over that same period.
The transition from drought conditions in 2012 to spring 2013 was abrupt; many UMRB areas flipped from precipitation deficits >250 mm to surpluses in excess of 250 mm in less than three months (i.e. over 500 mm gain). The popular media dubbed this “weather whiplash” (O’Hanlon). We quantify a weather whiplash index (WWI) as the total precipitation from January to June (2013) minus the total precipitation from July to December (2012), divided by the total precipitation over that entire period. Positive WWI indicates switching from dry to wet conditions during the twelve-month period; the magnitude of WWI indicates the intensity of that change during the same period. The 2012–2013 whiplash cycle was historically extreme (Fig. 2a) and spatially extensive (Fig. 2b). The 2012 U.S. drought was among the most severe, extensive and costly for the UMRB (Peterson et al. 2013), which includes four of the top states for maize and soy production (Illinois, Iowa, Indiana and Missouri), the U.S.’s two most valuable agricultural commodities (Hatfield et al. 2013). These four UMRB states contribute 48% of N loading to the Mississippi River (Alexander et al. 2008).
Examining the WWI of climate models indicates that weather whiplash in the UMRB will increase in frequency and intensity as climate changes (Fig. 3). Moreover, average trends in weather whiplash predicted by 30 future climate models (Fig. 3 black line) are conservative compared to the observed changes (Fig. 3 green dashed line) in weather whiplash in the Iowa River basin (1978–2015). We compared 30 downscaled precipitation projections (each denoted by a line) from the 30 models used in the CMIP5 (see details in Methods) to project future whiplash scenarios. Of these 30 models, 19 predict an increase in weather whiplash over time (orange lines, Fig. 3) and 11 predict no trend in weather whiplash over time (grey lines, Fig. 3a). Variance in modeled whiplash (Fig. S2) approximates the variance in observed weather whiplash from the Iowa River basin (Fig. S2, green box). Matching the modeled and observed variance in weather whiplash is a critical component to understanding the probability of extreme events, including high riverine nitrate concentrations that may cause exceedance of the EPA’s drinking water standards. Cyclic patterns in the observed or climate model predicted WWI were not evident; therefore, the deviation between modeled and observed weather whiplash (Fig. 3) is due to either short-term variability (37 years of data are available) or an under estimation of the precipitation changes by the climate models. If the observed pattern of rapid changes in weather whiplash persist, this would further exacerbate related issues including flood prediction, crop productivity and environmental quality.
Weather whiplash strongly influences spring nitrate concentrations in long-term monitoring data from agricultural watersheds in the UMRB (US EPA) (Fig. 4a, S3). Across the UMRB, dry springs following wet autumns result in the lowest spring nitrate concentrations; wet springs following dry autumns result in the highest spring nitrate concentrations (Fig. 4a). Our hierarchical regression model used to describe this relationship explains 81% (conditional R2) of the variation in spring nitrate concentrations with approximately half of this variation explained by weather whiplash (i.e., random slopes effects) and the other half by site effects (i.e. random intercepts caused by e.g. land use intensity, topography, etc.). By combining the projected weather whiplash from all 30 climate models (Fig. 3) and the relationship between spring nitrate and weather whiplash (Fig. 4a), we project spring riverine nitrate concentrations will continue to climb through 2100 (Fig. 4b). Extreme events (indicated by the 97.5% credible interval, upper dashed green line for the Iowa River basin) rise even faster than the mean nitrate concentration (solid green line; Fig. 4b). The faster increase in extreme events results in an increasing frequency of spring nitrate concentrations exceeding the E.P.A. drinking water standard (Fig. 4c). Again, our projected exceedance estimates are low relative to the observed exceedance (Fig. 4c), which we attribute to the conservative nature and uncertainty of our weather whiplash model projections being derived from the entire hindcast model period (1951–2015) as opposed to the period of spring nitrate data for the Iowa River (1978–2015) (Fig. 3).
Scientists are beginning to investigate how climate change will interact with land management to affect surface water quality (Howarth et al. 2012; Baron et al. 2012; Kaushal et al. 2014). Connections between weather variation and water quality have been noted for single drought-flood events (Kaushal et al. 2008), long-term data in a limited number (≤3, all within the same state) of watersheds (David et al. 1997; Royer et al. 2006) or hypothesized from modeling exercises (Donner et al. 2002). However, to our knowledge, this study is the first to empirically demonstrate the connection between increased long-term weather variation due to changing climate and the subsequent effects on water quality across multiple decades in an extensive agricultural region. Our data expands on previous work (David et al. 1997; Royer et al. 2006; Kaushal et al. 2008; David et al. 2010) to suggest that the spring 2013 pulse of riverine nitrate export is not a unique episode, but rather a normal, widespread, and recurring event sensitive to changes in seasonal precipitation. We show that antecedent climate can poise soil conditions for greater in riverine nitrate fluxes (Figs. 1, 2). Furthermore, climate change will likely result in a stronger weather whiplash with frequent summer droughts coupled to increasingly wet springs (Fig. 3) (Hatfield et al. 2013). Increased weather whiplash will bring about increased spring stream nitrate concentrations and associated challenges in managing surface waters for drinking water quality (Fig. 4). While our analysis clearly indicates weather whiplash is connected to the magnitude of N loss, we do not evaluate how shifting patterns in weather whiplash will affect the timing of loading, which is an important consideration for understanding coastal hypoxia development. Demonstrating the connection between climate variability and water quality leads us to posit that climate change will amplify water quality problems in the agricultural Midwest unless substantial changes are made in management.
The UMRB is beginning to show improvements in water quality (Murphy et al. 2013) after decades of decline (Sprague et al. 2011). Unfortunately, increasingly variable weather may counteract these improvements by enhancing N loading to streams and rivers. Currently, farmers are advised to add supplemental N fertilizer during wet springs to account for early season losses (Fernandez 2009). As weather whiplash increases in this region (Fig. 3), it is likely that land managers will respond to wetter springs by applying more N fertilizer (Hatfield et al. 2013). Without future changes in land management, the nascent water quality improvements in the region (Murphy et al. 2013) may quickly dissipate due to unforeseen interactions between climate and agriculture. This may further increase the economic damage associated with a changing climate as more municipalities construct and operate nitrate removal systems to meet drinking water standards. Currently, the Des Moines Water Works (Iowa, USA) operates a large nitrate removal facility in order to comply with the E.P.A. drinking water standard. The facility cost $4.1 million to build and $7000 USD/day to operate. In 2015, the city operated the facility for a record 177 days at a cost of ~$1.5 M USD and requires $80 M in upgrades in the near future (Des Moines Water Works 2016b). As weather whiplash (Fig. 3) and the associated increase in spring nitrate concentrations (Fig. 4a, b) combine to increase the likelihood of exceeding the EPA safe drinking water standard (Fig. 4c), more local municipalities in agricultural regions will be forced to invest in nitrate removal systems to meet their drinking water needs.
Current economics are driving agricultural intensification in the U.S. and across the globe (Donner and Kucharik 2008; Secchi et al. 2008). In the Midwestern US, this intensification is interacting with climate change to affect water quality. Unchecked, it is possible that weather whiplash and agricultural activities will combine to form a positive feedback loop that motivates farmers to apply more fertilizer to offset excess losses resulting from wetter springs, a practice that is currently being suggested by local managers (Fernandez 2009). Unfortunately, this potential for amplification of water quality problems occurs at a time when the need to reign in the environmental impacts of excessive fertilizer use is becoming widely recognized (Force 2013). Combined, our observations illustrate a harbinger of a future in which increased climatic variation amplifies negative trends in water quality in a region already grappling with impairments (Paulsen et al. 2006).