Journal list menu

Volume 50, Issue 3 p. 627-638
Open Access

Long-term impacts of drain spacing, crop management, and weather on nitrate leaching to subsurface drains

Eileen J. Kladivko

Corresponding Author

Eileen J. Kladivko

Dep. of Agronomy, Purdue Univ., 915 W. State St., West Lafayette, IN, 47907 USA


Eileen J. Kladivko, Dep. of Agronomy, Purdue Univ., 915 W. State St., West Lafayette, IN 47907, USA.

Email: [email protected]

Contribution: Conceptualization, Data curation, Formal analysis, Funding acquisition, ​Investigation, Methodology, Project administration, Resources, Supervision, Validation, Visualization, Writing - original draft, Writing - review & editing

Search for more papers by this author
Laura C. Bowling

Laura C. Bowling

Dep. of Agronomy, Purdue Univ., 915 W. State St., West Lafayette, IN, 47907 USA

Contribution: Data curation, Formal analysis, Methodology, Software, Writing - original draft, Writing - review & editing

Search for more papers by this author
First published: 09 March 2021
Citations: 8

Assigned to Associate Editor Anthony Buda.


Subsurface drainage is an essential water management practice for many poorly drained soils in the U.S. Midwest, but this practice also contributes nitrate-N loads to surface waters. This paper summarizes results from Years 16–31 of a long-term drainage research project in southeastern Indiana and compares results with the first 15 yr of the study. The study compared three drain spacings (5, 10, and 20 m) managed with a no-till corn (Zea mays L.)–soybean [Glycine max (L.) Merr.] rotation, with cover crops in about half of the years. Drainflow and nitrate-N losses per unit area were greatest for the 5-m spacing and lowest for the 20-m spacing. Nitrate-N concentrations did not vary with drain spacing and were generally in the range of 4–9 mg L–1. Annual nitrate-N loads were linearly correlated with annual flow volumes, reflecting the relatively constant concentrations over the 16-yr period. Whereas nitrate-N concentrations were relatively constant throughout the year, short-term concentration spikes occurred for nitrate-N during June–July of corn years. About 70% of annual drainflow and N loads occurred during the fallow season of November–April. The results underscore the interacting effects of drainage design, crop management, and weather in determining the magnitude of N loss from drained agricultural fields.


Subsurface drainage is a common agricultural water management practice for poorly drained soils throughout the world. The practice of subsurface drainage provides many benefits for crop production and soil and water quality compared with similar soils without subsurface drainage (Skaggs & Van Schilfgaarde, 1999), including increased infiltration, reduced surface runoff and erosion, and more reliable crop growth and yield. Subsurface drains have generally been found to reduce losses of sediment and phosphorus but to increase losses of nitrate-N due to greater water flow through the soil profile compared with similar undrained soils (Gilliam et al., 1999). As greater attention has been focused on hypoxic zones in water bodies throughout the world, there have been increased efforts to better understand nutrient losses from drained agricultural lands and to devise management practices and systems that maintain the benefits of drainage while minimizing deleterious effects on water quality.

Studies of nitrate leaching into subsurface drains have been conducted for many years (Baker & Johnson, 1981; Logan et al., 1980). Many different factors affect the nitrate concentrations and loads in the drains, including fertilizer N rates and management, cropping system and cash crop yield, cover crop growth, yearly weather variations, drainage intensity, and water table control practices (David et al., 2010; Dinnes et al., 2002; Drury et al., 2014; Gentry et al., 2000; Jaynes et al., 1999, 2001; Kaspar et al., 2012; Kladivko et al., 1999; Randall & Mulla, 2001; Ruffatti et al., 2019; Sands et al., 2008). Given the importance of reducing nutrient loads to the Gulf of Mexico (Mississippi River/Gulf of Mexico Watershed Nutrient Task Force, 2008), major efforts have been made to compile, interpret, and rank different management practices for their effectiveness across different states in the U.S. Midwest. A regional Extension publication highlights 10 of these practices (Christianson et al., 2016), three of which have been part of our long-term study in southeastern Indiana (Kladivko et al., 2004). Reducing N fertilizer rates (Dinnes et al., 2002; Jaynes et al., 2001; Lawlor et al., 2008) and growing cover crops (Kaspar et al., 2012; Ruffatti et al., 2019) are two in-field practices that generally reduce nitrate concentrations or loads from subsurface drains. The magnitude of these reductions is site- and year specific, pointing to the need for more field studies coupled with modeling work to develop site-specific recommendations.

A third practice highlighted by Christianson et al. (2016) that has not been adequately studied is the original design of the drainage system. Farmers must make a decision about their desired drainage intensity (drain spacing and depth) when they install a new drainage system in a field. This choice has traditionally been made to optimize economic return based on crop yield (Skaggs et al., 2006). Few field studies have looked at the impact of this decision on nitrate losses. Our earlier work (Kladivko et al., 2004) showed greater nitrate losses with narrower drain spacings. Skaggs et al. (2005) showed, with experimental data and modeling, that an increase in nitrate load with increased drainage intensity would likely be common due to the greater water flow from the drainage system. Other experiments in Minnesota (Nangia et al., 2010a; Sands et al., 2008) and Indiana (Hofmann et al., 2004) have also found greater nitrate loads with narrower drain spacings. These few studies suggest that drainage intensity is an important consideration for water quality, and more work is needed to develop approaches to optimize drainage design for both crop yield and water quality purposes.

One of the factors that has received increased attention recently is changes in the climate, primarily related to increased precipitation and therefore increased drainage flows. The effects of year-to-year variations in precipitation on drainage and nitrate loads have been discussed for decades (e.g., Randall & Goss, 2001; Randall & Mulla, 2001). Recently, however, there is greater attention being paid to long-term increases in precipitation amounts and variability and the contribution of these trends to increased drain flows and nitrate losses from drained agricultural lands (Baeumler & Gupta, 2020; Nangia et al., 2010b). These climate changes give even more impetus to develop integrated management systems that are resilient under varying conditions. Martinez-Feria et al. (2019) applied the concept of “portfolio effect” theory from financial asset management to evaluate the potential for multiple N management practices to produce synergistic

Core Ideas

  • Narrower drain spacings lose more nitrate per hectare than wider drain spacings.
  • Cover crops reduce nitrate losses in subsurface drains.
  • Even with low concentrations, higher rainfall and drainage still give large loads.
  • Weather, drainage design, and crop management combined control nitrate losses.

effects compared with any one practice individually. Sharpley et al. (2019) emphasized the complexity of the challenge for managing crop nutrients for water quality improvement, requiring site-specific solutions for successful systems.

Given the many factors that affect nitrate losses from agricultural fields to subsurface drainage waters, there is a need for long-term (>10 yr) studies to evaluate the relative effects of cropping system management, drainage system design, and year-to-year weather variations. Our study is a unique 31-yr project in southeastern Indiana, encompassing three different drainage intensities and changes in management practices over time during the study. The first 15 yr of drainage and nitrate data were presented in Kladivko et al. (2004), which evaluated changes in cropping system and management, including addition of cover crops and changes in N fertilization rates. The objectives of this paper are to extend the inferences from the earlier paper over the succeeding 16-yr period, evaluating (a) the effect of three different drain spacings on drainage flow and nitrate leaching and (b) the continued effects of the no-till corn (Zea mays L.)–soybean [Glycine max (L.) Merr.] system with cover crops on nitrate leaching. A detailed time series analysis of the complete 31-yr period is beyond the scope of this paper, but general comparisons of precipitation and flow between the two time periods will be made.


A subsurface drainage research project was begun in 1983 at the Southeast Purdue Agricultural Center (39o01′33″ N, 85o32′24″ W) in southeastern Indiana, USA. The site has been described in detail by Kladivko et al. (1991, 1999, 2004). The soil at the site is a Cobbsfork silt loam, formerly called Clermont (fine silty, mixed, superactive, mesic Typic Glossaqualf), and is typical of extensive areas of similar soils across southern Ohio, Indiana, and Illinois. The soil was formed in 50–120 cm of loess over glacial till. The surface soil at the study site is a light gray, low-organic-carbon (0.7%) silt loam containing 66% silt, 22% sand, and 12% clay. The soil is slowly permeable and has a borderline fragipan at about 120 cm depth that severely restricts further downward drainage.

The field experimental site had drains (10 cm in diameter) installed at spacings of 5, 10, and 20 m at an average depth of 75 cm and a slope of 0.4%. Three drain lines (225 m in length) were installed at each spacing, with the outside drain lines on each spacing acting as border drains between treatments. Each spacing was replicated in two blocks separated by 40 m (Figure 1). The center drains of the 5-, 10-, and 20-m plots discharged into observation culverts at the lower end of the field before being routed to the main drain. Subsurface drainflow volumes were monitored continuously with tipping-bucket flow gauges connected to a datalogger. Flow-proportional drainflow samples were continuously collected with automatic water samplers (Isco) during all time periods in which there was flow, with the samples being retrieved every weekday. Water samples were frozen until subsequent laboratory analyses by commercial laboratories. Nitrate-N mass losses (loads) were calculated as the product of concentration and water flow volume for each day and were expressed on a per hectare basis, assuming that each drainline collects water from midplane to midplane. A linear interpolation of nitrate-N concentrations was used to estimate concentrations on days between measurement points, with a few exceptions where averages from other drains or adjacent time periods were used as estimates when there were limited measurement points. Rainfall was measured at the field site from 2000 to 2007. Rainfall data from a weather station 10-km distant (North Vernon 2 ESE; 39o02′ N, 85o36′ W) were used for 2008–2015.

Details are in the caption following the image
Layout of subsurface drain spacing experiment at the Southeast Purdue Agricultural Center. Surface elevation contours are shown in meters

During this time period of the study (post-1999), high drainflow events sometimes flooded the measurement culverts and submerged the tipping buckets, resulting in no data on the flow amounts for a period of 1–3 h, on average. Such flooding events occurred 3–11 times per year post-1999. The flooding events were automatically identified with a C program and filled using either data from the replicate field block or linear interpolation between the first and last observed values on either side of the missing period. Overall the number of missing (6-min) time periods from 2000 to 2015 to be interpolated varied from 3.5% of the entire time series in Drain 4 to 11% of the time periods in Drain 5, and 14–48% of the total drainage volume was estimated. The majority of flow estimation was based on correlation with other drains, so only about 14% of the flow volume was estimated by interpolation. For the 6-min time step of data collection, the drain-spacing pairs are very highly correlated (r = .945, .915, and 0.967 for the 5-, 10-, and 20-m pairs, respectively), so there is high confidence in data filling by correlation.

Corn or soybeans were no-till planted each year (Table 1). Cover crops were drilled after corn harvest in most corn years preceding soybeans (Table 1), but starting in 2009 aerial seeding of mixtures including oats (Avena sativa L.) and daikon radish (Raphanus sativus L.) was done into the standing soybeans near physiological maturity. Total fertilizer N rates for corn were 200–233 kg N ha−1, with the majority being applied as sidedress urea ammonium-nitrate liquid (injection between rows ∼5–10 cm deep) and 28–45 kg N ha−1 applied as liquid starter in a 5 cm by 5 cm band with the planting operation (Table 1). Cash crop yields for the 16-yr period are provided in Table 1.

TABLE 1. Field management practices and crop yields
Year Crop Preplant N Starter N Sidedress N Total N applied Fall-seeded cover crop Cash crop yielda
5 m 10 m 20 m
kg ha−1 Mg ha−1
2000 soybean nab 0 0 0 4.6 4.2 4.3
2001 corn 177c 28 na 205 wheat 9.9 9.8 9.4
2002 soybean na 0 0 0 3.2 3.0 3.0
2003 corn na 28 177 205 wheat 7.5 7.5 7.4
2004 soybean na 0 0 0 4.7 4.6 4.6
2005 corn na 28 177 205 wheat 10.0 9.7 11.2
2006 soybean na 0 0 0 4.1 3.8 3.9
2007 corn na 28 183 211 13.1 13.0 13.0
2008 corn na 28 200 228 wheat 11.6 10.8 9.8
2009 soybean na 0 0 0 oat/radish 3.9 3.7 3.9
2010 corn na 28 177 205 12.5 11.8 11.2
2011 corn na 28 172 200 wheat 9.0 8.8 8.8
2012 soybean na 0 0 0 oat/radish 3.8 3.9 4.0
2013 corn 62d 45 126 233 cereal rye 13.5 12.6 12.7
2014 soybean 0 0 0 0 O/R/Ry/Cle 3.4 3.4 3.6
2015 corn 62d 45 126 233 cereal rye 14.9 14.3 14.0
  • a Average of both blocks.
  • b Not applied.
  • c As injected anhydrous ammonia.
  • d As broadcast diammonium phosphate granules.
  • e Oats/radish/cereal rye/crimson clover.

Annual drainflow per unit area and N loads were calculated for each drain for each of the 16 yr. A three-way ANOVA was first used to evaluate the impact of eastern vs. western block, drain spacing, and year on drainflow and N load. Because the differences in flow and N load between the two blocks were large (p < .01), tests were subsequently performed separately for each block, with a significance level of .05, similarly to Kladivko et al. (2004). A two-way ANOVA without replication (repeated measures design) with spacing as the within-subjects factor was used to test for differences in annual drainflow and N loads between spacings. The pairwise paired t test was used rather than Tukey's multiple comparisons test to identify differences between drain spacings because it takes into account the blocking structure of the repeated-measures design (Helsel et al., 2020).

To evaluate if the differences in flow rate between spacings decreased over the 16-yr period, the ratios of the 10- and 20-m spacings to the 5-m spacing were tested using a simple linear regression with time (t test) to determine if the slope was significantly different than zero, with a significance level of .05.

Spectral analysis was performed on monthly time series of flow-weighted mean concentration and drainflow to identify the dominant periods of variation. Data from 2000 to 2007, the period when a 2-yr corn–soybean rotation was used, were standardized by subtracting the mean and dividing by the standard deviation before calculating the one-sided periodogram (MATLAB, 2019). The periodogram provides an estimate of the fraction of total time series variance vs. frequency that are compared between the different series (Bras & Rodriguez-Iturbe 1993).


Most results are presented first as annual averages, comparing across the three drain spacings. The 2000–2015 period is then compared with the earlier 15-yr period of 1985–1999 (Kladivko et al., 2004). Monthly and seasonal patterns for hydrology, concentrations, and loads are then discussed.

3.1 Hydrology

Annual rainfall over the 16-yr period ranged from 906 mm in 2010 to 1,536 mm in 2008 (Table 2), with the 16-yr average (1,250 mm) being 5% higher than the 30-yr “normal” of 1,186 mm (1981–2010). This represents a 12% increase (132 mm) from the previous 15-yr period annual average rainfall of 1,118 mm. Drainflow varied with annual rainfall and with timing of rainfall within each year as well as with drain spacing. As expected and as found in the earlier 15-yr period (Kladivko et al., 2004), more drainflow per unit area occurred for the more closely spaced drains (Table 2). Repeated measures ANOVA indicated that there were statistically significant differences in drainflow due to spacing for both the eastern and western block (p < <.01). Data were log-transformed for the eastern block to maintain normality of the residuals. Paired t tests showed significant differences (p < .05) in drainflow per unit area in all drain pairs (5 vs. 10, 10 vs. 20, 5 vs. 20 m) within each block when considering the 16-yr period as a whole. Averaged across both blocks, drainflow varied from 19.4 cm for the 20-m spacing in the 2012 drought year to 93.1 cm in the 5-m spacing in 2011. April 2011 was extremely wet, with nearly three times the 30-yr normal rainfall for April of 114 mm, followed by a wet May and June. Due to the nearly saturated soils in April, nearly 85% of the 327 mm of rainfall ended up as drainflow in the 5-m spacing plots.

TABLE 2. Annual precipitation, percentage of “normal” precipitation, and drainflow as affected by drain spacing
Year Precipitation East block West block Average of both blocks
Amount % of normala 5 m 10 m 20 m 5 m 10 m 20 m 5 m 10 m 20 m
mm % cm
2000 1,464 123 24.6 17.3 17.5 31.8 29.2 21.6 28.2 23.2 19.6
2001 1,302 110 36.8 27.5 25.4 49.1 39.9 30.6 42.9 33.7 28.0
2002 1,390 117 45.4 33.8 32.6 58.1 45.6 30.4 51.7 39.7 31.5
2003 1,207 102 54.9 40.0 36.6 64.9 52.2 44.0 59.9 46.1 40.3
2004 1,230 104 53.5 40.8 33.1 69.0 49.3 41.6 61.2 45.1 37.3
2005 1,118 94 44.6 36.6 29.1 52.1 42.2 31.9 48.3 39.4 30.5
2006 1,420 120 67.5 45.6 46.4 79.5 63.4 52.7 73.5 54.5 49.5
2007 1,054 89 32.0 23.7 26.2 36.6 32.4 27.1 34.3 28.0 26.7
2008 1,536 130 74.1 51.8 51.1 85.8 70.2 52.4 80.0 61.0 51.7
2009 1,268 107 56.2 43.3 40.7 68.4 59.8 47.5 62.3 51.5 44.1
2010 906 76 37.9 31.0 27.6 49.0 39.1 33.2 43.5 35.0 30.4
2011 1,440 121 89.0 72.4 66.2 97.2 97.3 68.7 93.1 84.9 67.4
2012 939 79 21.6 18.8 17.4 20.4 26.4 21.4 21.0 22.6 19.4
2013 1,112 94 41.0 32.0 27.9 34.8 42.8 32.3 37.9 37.4 30.1
2014 1,197 101 39.8 28.9 23.9 37.4 41.3 31.3 38.6 35.1 27.6
2015 1,417 119 42.4 29.3 25.5 36.3 39.1 27.6 39.4 34.2 26.6
16-yr avg. 1,250 105 47.6b 35.8b 33.0b 54.4b 48.1b 37.1b 51.0 42.0 35.0
  • a Thirty-year normal precipitation (1981–2010) from North Vernon, IN.
  • b Paired t-tests showed significant differences (p < .05) in drainflow in all drain pairs (5 vs. 10, 10 vs. 20, 5 vs. 20 m) within each block when considering the 16-yr period as a whole.

There are multiple lines of evidence suggesting development of flow paths in the soil and “maturation” of the drainage system with time. Staff at the research site (J. Wahlman, personal communication) have noted faster responsiveness of drainflow to rainfall events over the history of the project, even with smaller events that previously had not generated flow. Although rainfall in 2000–2015 was 12% higher than in 1985–1999, the drainflow was 120–160% higher, depending on the drain spacing. Bowling and Kladivko (2016) performed trend analysis of time series data from 1996 to 2014 from this site and found statistically significant increasing trends in event drainage volume and peak flow rates and a decreasing trend in event lag time. The relative differences in flow among the three drain spacings have also become smaller with time. Over the current 16-yr period, the ratio of flow from the 10-m spacing to the 5-m spacing, averaged across blocks, increased 1.3% per year (R= .44; p < .01), and the ratio of the 20-m to 5-m spacing flow increased 0.9% per year (R= .27; p = .037).

Our hypothesis is that the changes in flow patterns with time are due in part to development in soil properties resulting from long-term drainage installation (installed in 1983) as well as long-term no-till (begun in 1994). Several other researchers have found changes in soil morphology and water flow properties with time after installation of drainage. Kapilevich et al. (1991) found formation of cracks and increased water flow on glacial-lacustrine clays in Belarus from between 5 and 18 yr after drainage installation, depending on the soil. The Southeast Purdue Agricultural Center site does not have as much clay as the Belarus sites and would not be expected to develop as strong of a crack pattern, but preferential flow has been measured at this site in previous studies (Kladivko et al., 1999; Kung et al., 2000). Any developing flow pathways that extend further from the drain than 2.5 m (midplane position of 5-m spacing) would contribute to relative increases in flow from the 10- and 20-m spacings compared with the 5-m spacing. Montagne et al. (2009) reviewed drainage studies that measured hydraulic or morphological characteristics of drained fields, looking for evidence of recent soil evolution resulting from the drainage. Although few studies had direct measurements of changes with time, they argued that changes in water flow within the soil were affecting eluviation processes and iron and manganese dynamics sufficiently to be an agent of soil evolution over time scales of several decades. Welage (2020) and Mitchell (2020) found few significant differences in soil physical properties between the narrowest drain spacing (5 m) and the undrained “control” (40 m) at this site, with standard techniques of measuring bulk density, water retention, penetration resistance, and hydraulic conductivity. However, Welage (2020) did observe greater numbers of earthworm (Lumbricus terrestris) channels in pits dug in the 5-m vs. 40-m spacings, which would also contribute to greater flow. The soil pits avoided the area containing the old tile trench, so the flow is not just a result of disturbance from tile installation. The possibility of changes in soil flow patterns over decadal time scales as a result of subsurface drainage is an interesting and potentially important topic, considering the vast land areas in humid regions that are subsurface drained for optimum agricultural use.

3.2 Nitrate-N concentrations

Annual flow-weighted nitrate-N concentrations were consistently below the 10 mg L−1 drinking water standard and were more commonly in the 4–8 mg L−1 range (Figure 2). Concentrations did not vary consistently among the three drain spacings. These low nitrate-N concentrations are similar to the 6–10 mg L−1 range of the last 4 yr (1996–1999) of the earlier 15-yr period (Kladivko et al., 2004) and are similar to other Midwestern studies with cover crops showing nitrate-N concentrations in the 5–8 mg L−1 range for cereal rye (Secale cereale L.) (Kaspar et al., 2012). There was a weak trend for a continued small decrease in nitrate-N concentrations over time during this 16-yr period (2000–2015); this trend was statistically significant for the 10-m and 20-m spacing (p = .01) (Figure 2). One possible explanation could be a small cumulative effect of continued cover crop use on this soil. In the years when cover crop was grown, cover crop biomass in spring generally ranged between 1.0 and 2.0 Mg ha−1, depending on the termination timing and growing season weather that particular year, and the resulting N content in biomass was between 10 and 20 kg N ha−1 (8–16% of applied N averaged over all years). Another possible explanation is the slight increase in corn grain yields over the 16-yr period (about 3% increase per year), which would have removed more N in corn grain, leaving less residual N to be leached. Another potential explanation is a profile flushing effect (Lawlor et al., 2008; Randall & Mulla, 2001) resulting from many sequential years of much greater drainage than in the previous 15-yr period, as discussed above. Other potential contributing factors might be a dilution of nitrate-N concentrations with the greater flow or greater denitrification with the overall wetter conditions. Annual variations in concentrations reflect a combination of cover crop growth, cash crop growth and yield, fertilization, and other factors and are discussed in more detail in the section on monthly and seasonal patterns.

Details are in the caption following the image
Annual flow-weighted nitrate-N concentration (mg L−1) for three drain spacings, 2000–2015. The slopes of the 10-m and 20-m spacings are statistically different from zero (= .01 for both), whereas for the 5-m spacing it is not (p = .19)

3.3 Nitrate-N loads

Annual nitrate-N loads varied primarily with drainflow, as is evidenced by the variance in daily drainflow being much greater than the variance in daily concentration. The CV for drainflow varied from 1.5 to 4.6, whereas the CV for concentration was between 0.4 and 0.6 for different drains. Averaged across both blocks, nitrate-N loads varied from 12.7 kg ha−1 for the 20-m spacing in 2012 to 65.1 kg ha−1 from the 5-m spacing in 2011 (Table 3). The highest and lowest nitrate-N loads corresponded with the highest and lowest drainflows. Over the 16-yr period, differences in annual nitrate-N loads across all three spacings were well explained by the differences in flow (p < .001) (Figure 3). Therefore, even though nitrate-N concentrations remained low throughout this 16-yr period, there were still some years with high loads comparable to the earlier 15-yr period (1985–1999) (Kladivko et al., 2004) with its overall higher concentrations but lower flows. These results underscore the importance of such noncontrollable factors as weather in determining nitrate-N losses from cropped fields. Baeumler and Gupta (2020) also concluded that precipitation variations were the main driver for variations in river N loads in the U.S. Midwest.

TABLE 3. Annual nitrate-N loads in drainflow as affected by drain spacing
Year Crop Nitrate-N load
East block West block Average of both blocks
5 m 10 m 20 m 5 m 10 m 20 m 5 m 10 m 20 m
kg ha−1
2000 soybean 19.2 14.1 12.7 25.6 25.5 20.4 22.4 19.8 16.6
2001 corn 31.2 27.3 19.9 45.2 37.3 32.7 38.2 32.3 26.3
2002 soybean 19.3 14.2 13.9 33.9 25.8 22.0 26.6 20.0 17.9
2003 corn 37.2 28.0 24.6 51.4 37.4 40.8 44.3 32.7 32.7
2004 soybean 21.3 16.1 12.7 36.1 23.5 28.1 28.7 19.8 20.4
2005 corn 42.5 38.3 22.8 38.1 31.3 28.9 40.3 34.8 25.8
2006 soybean 56.7 39.3 25.1 40.4 29.9 32.7 48.5 34.6 28.9
2007 corn 21.8 17.5 16.6 24.1 20.2 23.1 23.0 18.9 19.8
2008 corn 55.3 40.1 35.2 63.7 47.5 48.8 59.5 43.8 42.0
2009 soybean 30.5 21.7 21.2 41.6 37.5 33.0 36.1 29.6 27.1
2010 corn 26.5 23.6 20.4 31.5 28.4 24.7 29.0 26.0 22.5
2011 corn 61.7 39.4 35.9 68.5 59.5 43.6 65.1 49.4 39.8
2012 soybean 20.0 10.9 10.4 13.1 15.2 15.1 16.6 13.0 12.7
2013 corn 25.2 14.8 14.9 26.8 22.3 18.4 26.0 18.5 16.7
2014 soybean 15.1 9.3 7.5 16.3 14.3 14.9 15.7 11.8 11.2
2015 corn 20.6 12.7 11.1 22.6 19.3 17.3 21.6 16.0 14.2
16-yr avg.   31.5a 23.0a 19.1a 36.2a 29.7a 27.8a 33.8 26.3 23.4
  • a Paired t tests showed significant differences (< .05) in nitrate-N load in all drain pairs (5 vs. 10, 10 vs. 20, 5 vs. 20 m) within each block except the 10 vs. 20 m in the West block (= .14) when considering the 16-yr period as a whole.
Details are in the caption following the image
Relationship between annual drainflow and annual nitrate-N load. Each point represents the annual load and drainflow for a different year and drain spacing for 2000–2015 (p < .001)

Repeated-measures ANOVA indicated that there were statistically significant differences in nitrate-N load due to spacing for both the eastern and western block (p < .01). Data were log-transformed for the eastern block to maintain normality of the residuals. Nitrate-N loads were greater for the narrower drain spacings than the wider spacings due to the greater flow in the narrower drain spacings. Paired t tests showed significant differences (< .05) in nitrate-N load in all drain pairs (5 vs. 10, 10 vs. 20, 5 vs. 20 m) within each block except the 10 vs. 20 m in the West block (= .14), when considering the 16-yr period as a whole. Sands et al. (2008) and Hofmann et al. (2004) also found greater nitrate-N loads from systems with greater drainage intensity. Drain spacing is one of the controllable factors that affect N loss, but obviously the decision about drain spacing is generally made once, at the time of initial installation of the system. In addition, drain spacing is sometimes intensified in fields in which the older drainage system is deemed inadequate by the producer by installing new drains at the midplanes between current drains. Although a decision to reduce drainage intensity cannot be made after the drains have been installed, there are practices that can reduce the effective drainage intensity during the fallow season when full drainage capacity may not be needed, such as drainage water management (controlled drainage). Drainage water management has been shown to decrease nitrate-N loads from subsurface drains, primarily by reducing water flow from the drains (Drury et al., 2014; Thorp et al., 2008; Youssef et al., 2018).

3.4 Monthly patterns of flow, load, rain, and concentrations

The monthly distributions of flow, nitrate-N load, and rainfall as a percentage of the annual totals are shown in Figure 4. Rainfall was distributed relatively uniformly throughout the year, on average. Drainflow, however, varied much more over the year due to higher evapotranspiration during the growing season. As per the first 15 yr of the study, drainflow and nitrate-N load corresponded very closely together (Figure 4), reflecting the relatively small differences in concentrations across months. Drainflow and nitrate-N load peaked in March, April, and May, prior to planting, and reached a minimum (often zero) in August and September. Drains continued flowing all winter, with a secondary peak in December. On average, about 70% of the annual flow and load occurred during the fallow season at this site. This result is consistent with many findings of drainflow and nitrate-N loss occurring primarily in late fall, winter, and early spring in climates where the soil is not frozen all winter (Gilliam et al., 1999).

Details are in the caption following the image
Boxplots of annual rainfall, drainflow, and nitrate-N load for the 20-m West plot (Drain 6) for 2000–2015. The box height shows the interquartile range; the whiskers extend to the 5th and 95th percentile values. The median is shown by the horizontal line

Although nitrate-N concentrations were relatively stable across months, the monthly flow-weighted nitrate-N concentrations showed a tendency for short-term higher concentrations in June or July of years growing corn. An analysis of periodicity for the 8-yr subset (2000–2007) of these years in a consistent 2-yr soybean–corn rotation found a strong 2-yr periodicity for nitrate-N concentration but 1-yr and 6-mo periodicity for the flow (Figure 5). The 1-yr and 6-mo periods in drainflow explained 23–36% and 21–32%, respectively, of the drainflow time series variance and reflected the climatological controls described above that resulted in peak drainflow in April and December (Table 4). In contrast, these cycles only explained 1–10% and 5–20%, respectively, of the variance in concentration. Drains 1 and 2 were affected by a herbicide spill in 2005. This one-time spike in concentration resulted in the appearance of enhanced low-frequency variation. Between 28 and 49% of the variance in concentration is associated with the 2-yr cycle of crop management practices, including growing corn and sidedress fertilizer application of urea ammonium-nitrate in June or July of corn years from 2003 onward.

Details are in the caption following the image
Spectral density for monthly drainflow (top) and concentration (bottom) estimated using the one-sided periodogram for 2000–2007. Drainflow has strong periodic cycles with periods of 6 and 12 mo (one and two cycles per year), whereas concentration only has a 2-yr cycle. Drains 1 and 2 were affected by a herbicide spill in 2005; this one-time spike in concentration resulted in the appearance of enhanced low-frequency variation
TABLE 4. Percent variance explained by the three dominant periodic cycles of 6 mo, 1 yr, and 2 yr
Period Drain 1 Drain 2 Drain 3 Drain 4 Drain 5 Drain 6
6 mo 23.9 24.3 22.8 24.3 22.6 26.1
1 yr 22.5 21.4 25.2 24.2 26.8 31.8
2 yr 5.6 5.1 5.8 5.7 6.2 5.0
Nitrate-N concentration
6 mo 1.6 1.3 3.9 9.5 5.6 3.0
1 yr 5.9 6.8 5.0 19.5 12.0 10.1
2 yr 40.4 41.3 43.9 27.6 31.7 49.2

The sparse flow and relatively few water samples from the June–July periods preclude a more detailed assessment of time lags between sidedressing and nitrate-N concentration increases. Additionally, the main N fertilizer in 2001 was still preplant anhydrous ammonia, applied on 9 April, so the sidedress hypothesis does not apply to that year. The higher June–July nitrate-N concentrations may also be due in part to the lack of cover crop being grown in the spring before corn planting, so there was nothing actively taking up N as it became available during late spring microbial activity. The lower nitrate-N concentrations following corn harvest and into the soybean years were a result of both the cover crop growth after corn before soybeans and the lack of N fertilization in the soybean years. Although the nitrate-N concentrations were higher in June or July in most corn years, the impact on loads was minimal due to lower drainflow rates during June and especially July than in the months prior to June.

These results differed from the 1984–1999 time period, when there was little or no monthly variation in nitrate-N concentration (Kladivko et al., 2004), likely because the main N fertilizer during that earlier period was pre-plant anhydrous ammonia applied in April while the soil was still cold. Thus, nitrification to nitrate-N had to occur before the N was subject to leaching. In addition, most of those earlier years also had a nitrification inhibitor applied with the ammonia, slowing the conversion process even further.

3.5 Drainage system response to management, weather, and time

Nitrate-N losses to subsurface agricultural drains remain a concern in many highly productive agricultural regions with poorly drained soils. Many different management practices and systems have been proposed and are being tested for their potential effectiveness in reducing N loads. This unique 31-yr dataset provides an opportunity to observe the relative responses of N loads to changes in management practices, variations in weather, differences in the original drainage design factor of drain spacing, and potential changes in soil with increasing age of the drainage system. Although all of these factors are interacting at this site, we can still glean some important insights.

The controllable factor of cropping system management is important. The combination of cover crops and reduced N fertilizer rate, along with the change to a corn–soybean rotation from continuous corn, reduced nitrate-N concentrations and loads greatly at this site during the previous 15-yr period (Kladivko et al., 2004). The continued use of these practices maintained the lower nitrate-N concentrations throughout the current 16-yr period. This reduction in nitrate-N concentration with cover crops and the resulting reduction in loads has been found consistently throughout the Midwest and other regions (Hanrahan et al., 2018; Kaspar et al., 2012; Ruffatti et al., 2019).

However, the importance of the uncontrollable factor of climate and weather underscores that cropping system management does not control everything. Weather is still the overriding factor controlling how much drains flow and therefore how much water is available to leach the soil nitrate. Nitrate-N concentrations in 2001–2011 remained as low as in the last years (1997–1999) of the earlier 15-yr period (Kladivko et al., 2004), but nitrate-N loads were greater due to the greater rainfall and drainflow. The strong relationship between nitrate-N load and flow was illustrated with annual data (Figure 3) and with the monthly distributions of flow and load being very similar (Figure 4).

The controllable factor of drain spacing is an important choice. Because closer drain spacings are intended to remove more water faster, it makes sense that they also remove more nitrate-N from the field. If a system is designed for greater drainage intensity than before, then other management practices should also be intensified to reduce nitrate-N losses, such as with cover crops, controlled drainage, or bioreactors. Recent work by Castellano et al. (2019) also suggests that greater N use efficiency might be possible with greater drainage intensity as N fertilizer rate might be able to be reduced, but this hypothesis needs further testing.

Drains respond quickly to changes in management or crop yield but not instantaneously, and there is a time-averaging effect or lag in response to changes in N additions (Figure 6). In Crop Year 2005, two of the plots had a herbicide tank leak that killed the corn crop in a substantial portion of the plots. The resulting lower average plot yield (∼3 Mg ha−1 lower than the adjacent plots’ yields of 9 Mg ha−1) meant that ∼37 kg ha−1 less N was removed from the plots in corn grain. The higher residual N was evident in drainflow concentrations already in August of that year and persisted through the fall, winter, and spring. Nitrate-N concentrations were not fully back to the level of the other plots until late August of 2006, nearly 1 yr and 37–54 cm of flow later. These data illustrate the lag time for changes in management to be completely expressed in drainage waters as the drains integrate flow from near and far from the drain (Jury, 1975a, 1975b; Kladivko et al., 1991). Thus, changes from cover crops and phases of the cash crop rotation will be averaged over at least a year from the time the particular practice or crop is established.

Details are in the caption following the image
Nitrate-N concentrations in 2005 and 2006, illustrating the divergence in concentrations for Drain 1 (10 m east) vs. Drain 4 (10 m west) due to accidentally killed corn crop and therefore more residual N available to be leached (see text for full explanation). Concentrations in Drain 1 returned to the same level as other drains after one full year of drainage. The points are measured concentrations, and the lines are trend lines for illustration purposes

The age of the drainage installation may affect the flow behavior and therefore the N loads from the drains. This site has some evidence of quicker flow responses to rainfall events (J. Wahlman, personal communication) and greater flashiness and event flow volumes (Bowling & Kladivko, 2016). Recent assessments of the site have not been able to document physical property changes among the different drain spacings and the undrained control (Mitchell, 2020; Welage, 2020), but the repeated observations of the research farm staff suggest that changes have occurred over the 31-yr period. These potential changes with age of the system cannot be completely separated from changes in weather and management over the years.

In conclusion, nitrate-N losses in subsurface drains are controlled by many interacting factors, and therefore no single practice will be able to reduce the losses to the level desired for water quality. Drainage system design, drainage system age, weather/climate, and crop management practices (e.g., cover crops, fertilizer N management, and crop rotation) all affect nitrate-N loads through their effects on drainage flow volumes and/or nitrate-N concentrations. Continued efforts to improve the design and management of systems approaches for both crop yield and water quality are urgently needed.


The authors thank the many people who have worked on this project over the years, including the Southeast Purdue Agricultural Center farm crew, graduate students and post docs, faculty colleagues, and NRCS colleagues. This project was supported in part by the Purdue Agricultural Research Programs and the USDA National Institute of Food and Agriculture, Hatch project 87887.


    Eileen J. Kladivko, Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Resources, Supervision, Validation, Visualization, Writing-original draft, Writing-review & editing; Laura C. Bowling, Data curation, Formal analysis, Methodology. Software, Writing-original draft, Writing-review & editing.


    The authors declare no conflict of interest.