The two study watersheds, Kirkpatrick Ditch Watershed (KDW; 26.3 km2) and Shatto Ditch Watershed (SDW; 13.3 km2), both located in northern Indiana (Figure 1), are predominantly planted with a corn (Zea mays L.) and soybean (Glycine max L.) rotation with approximately equal amounts of each crop each year. In KDW, soils are primarily Mollisols with a silty-clay texture while the SDW has soils that are primarily Alfisols with a texture that ranges from sandy loams to loams and muck (Christopher et al. 2021). The fields in each watershed are “working lands”, managed by independent agricultural producers and are representative of typical agricultural practices in the midwestern U.S., including extensive subsurface tile drainage systems (Gökkaya et al. 2017, Trentman et al. 2020; Table 1) and N fertilization of corn at a rate of about 150 kg ha-1 yr-1.
In both watersheds, farmers were reimbursed for costs associated with use of winter cover crops through the U.S. Department of Agriculture Regional Conservation Partnership Program. Farmer participation was voluntary and farmers made all decisions related to selection of cover crop species, timing of planting and termination, and all other management options. In KDW, cover crops were maintained at 12-32% of the total tillable acres during our study period, while in SDW ranged from 22-68% (Table 1).The most common cover crop species planted by producers in both KDW and SDW were annual rye-grass (Lolium multiflorum Lam.) and cereal rye (Secale cereale L.)
Sampling design and protocol
We monitored sixteen tile drains and the watershed outlet in KDW, and only the watershed outlet for SDW, and we sampled twice monthly throughout the 2018-2020 water years (Oct 1- Sept 30). For both tile drains and stream sites, we measured instantaneous discharge and sampled for dissolved inorganic nitrogen (nitrate and ammonium; hereafter DIN), soluble reactive phosphorus (SRP), and DSi. Additional results for DIN and SRP are reported in companion studies by Trentman et al. (2020, 2021) and Speir et al. (2020, In Review).
Instantaneous discharge was measured at tile drains using a timer and graduated receptacle; to ensure accurate volume and time values, discharge was measured multiple times until three consecutive measurements were all within 10%. For larger tile drains (diameter > 0.5m) or when any tile flow was > 10 L s-1, we used an electromagnetic water velocity meter (March-McBirney Model 2000 Flo-Mate) and a wading rod to calculate discharge (Q) as:
drain (m), and v is water velocity (m s-1). Stream discharge was monitored by U.S. Geological Survey gages at the KDW (station #05524546) and SDW (station #03331224) watershed outlets.
We collected water samples for nutrient analyses collected directly from the tile drains or stream sites using a 60-mL syringe that was rinsed with sample water before collection. We collected separate samples for DIN/SRP and DSi analysis and we filtered all DIN and SRP samples immediately upon collection using glass fiber filters (Whatman GF/F). For DSi, we used cellulose filters (0.45µm pore size; Fisherbrand) to prevent contamination from glass. We transported samples on ice, froze them until analysis, and colorimetrically analyzed all samples using a Lachat QuikChem flow injection analyzer (Hach Company). We analyzed samples for SRP using the ascorbic acid method (Murphy and Riley 1962), for nitrate using the cadmium reduction method (APHA 2012), for ammonium using the phenol-hypochlorite method (Solórzano 1969), and for DSi using the heteropoly blue method (Sultan 2014). For all nutrient analyses, we ran a certified standard to validate the standard curve and routinely calculated the method detection limit. The dataset analyzed here includes 1,108 individual tile drain measurements and 72 outlet measurements from KDW along with 75 outlet measurements from SDW. We monitored the same tile drains every year in KDW; however, due to changes in field management, the number of tiles draining fields with and without cover crops varied each year, with cover crop tiles ranging between 14-71% of the total monitored field tile drains. We found samples that were below detection were primarily from KDW tile drains; in total less than 1% of DIN, ~10% of SRP, and no DSi samples were below detection.
We conducted all data and statistical analyses using R (The R Foundation for Statistical Computing, Version 4.0.5, 2021). To characterize the seasonal pattern in DSi concentrations and stoichiometry, we divided data into four seasons: autumn (October-December), winter (January-March), spring (April-June), and summer (July-September) that correspond to crop planting and harvest and distinct temperature and hydrologic conditions affecting vegetation growth and nutrient transport in midwestern agricultural systems (Williams et al. 2015; Hanrahan et al. 2018).
In order to evaluate the effect of tile drain and stream discharge on DSi concentrations, we modeled the relationship between DSi concentrations and discharge using a power-law equation, expressed as C = aQb, where C is concentration, Q is discharge, and b is a constant representing the slope of the relationship. The sign of the slope indicates whether a solute exhibits enrichment (positive slope), dilution (negative slope), or chemostatic (zero slope) behavior with discharge (Godsey et al. 2009; Bieroza et al. 2018). The value of the slope is used to evaluate a solute’s response to discharge, where values close to one represent a proportional change in concentration with discharge (Leong et al. 2014).
We analyzed the effects of winter cover crops on DSi concentrations, loss, and stoichiometry using data collected from fourteen unique tiles draining specific fields (field tile drains) and we removed two “county” drains which aggregate a larger drainage area across multiple fields because these drains incorporate multiple fields and cannot be classified on the basis of cover crop use. In all other analyses, we included the county drains.
Within each water year and season, we quantified the effect of winter cover crops on KDW field tile drain DSi concentrations, loss, and stoichiometry using a Hierarchical Regression Model (HRM) and pair-wise comparisons. The HRM tests for differences between cover crop and no cover crop treatments while accounting for the random effects associated with tile drain location and the fixed effects of cover crop treatment, year, and season. We then used a post-hoc pairwise test to identify which years and seasons showed a significant (p<0.05) response to cover crop planting. We conducted these tests using the lme4 (Bates et al. 2015), lmerTest (Kuznetsova et al. 2017), and emmeans (Lenth et al. 2021) packages in R.
We calculated nutrient ratios between DIN, SRP, and DSi using molar concentrations of each sample value, and we based predictions for nutrient limitation on a modified Redfield ratio for freshwater diatoms (hereafter “freshwater Redfield ratio”) of 106C:16N:1P:40Si (Redfield 1963; Brzezinski 1985). We also modeled daily loads of DIN, SRP, and DSi from each watershed outlet using Loadflex, an R package which estimates solute loads using a composite model that combines a regression model with a residuals correction for a more accurate estimation of nutrient loads (Appling et al. 2015); we then used estimated loads to calculate daily yields of nutrients based on the area of each watershed.
We used estimated daily yields to calculate the Indicator of Freshwater Eutrophication Potential (IFEP) which predicts the potential growth of non-siliceous algae based on the amount of N or P in excess relative to the N:P:Si stoichiometric demand of diatoms expressed in the freshwater Redfield ratio (Garnier et al. 2010; Dupas et al. 2015). The IFEP is expressed as kg C area-1 time-1 and is calculated as follows:
Positive IFEP values represent potential DSi limitation of diatoms, and thus the potential for growth of non-siliceous taxa, including cyanobacteria. Conversely, negative values indicate DSi is abundant relative to N and P, and thus conditions are favorable for diatom growth. When comparing N- and P-IFEP values, the lower of the two indicates whether N or P is likely the limiting nutrient. Using one-sample t-tests (x̄=0, α=0.05), we assessed whether monthly averages of daily IFEP values were significantly greater than zero, indicating the stoichiometry of the nutrient loads favored non-siliceous, and possibly harmful, algal growth.