George M. Chescheir

Nitrogen Removal in Streams of Agricultural Catchments—A Literature Review

Excess nutrient loads have been recognized to be the major cause of serious water quality problems recently encountered in many estuaries and coastal waters of the world. Agriculture has been recognized in many regions of the world to be the largest single source of nitrogen emissions to the aquatic environments, and best management practices have been proposed to reduce nutrient losses at the field edge. As a result, there is growing awareness that nutrient management must be handled at the watershed scale. However, the key to nutrient management at the watershed scale is the understanding and quantification of the fate of nutrients both at the field scale and after they enter the aquatic environment. There has been widespread evidence since the late 1970s that nitrogen can be removed from water during its downstream transport in watersheds or basins. Although this information is becoming crucial, no overview has been proposed, so far, to qualitatively as well as quantitatively summarize available information in the literature. For this reason, we propose a review on the biogeochemical processes involved in nitrogen removal in streams, the rates of removal reported, and the factors influencing those rates. Nitrogen removal rates in agricultural streams should be expected to vary between 350 and 1250 mg N m−2 day−1. Mass transfer coefficients (coefficient evaluating intrinsic ability of a stream to remove nitrogen) values in agricultural streams may vary between 0.07 and 0.25 m day−1, although these values correspond to values obtained from reach scale studies. Reviewing values obtained from different measurement scales has revealed that results from incubations or experiments performed in the laboratory clearly underestimate mass transfer coefficients compared to those reported at the reach scale, from severalfold to more than one order of magnitude. Nitrogen removal rates and efficiency in streams are the highest in the summer, and this is critical for receiving ecosystems, which are most sensitive to external inputs at this period of the year. Removal efficiency is the lowest in winter in temperate climates due to high flow and loading combined with lowest removal rates. In-stream processes, on an annual basis, may remove at the watershed scale as much as 10 to 70% of the total N load to the drainage network.

THE NITROGEN SIMULATION MODEL, DRAINMOD-N II

DRAINMOD-N II is a field-scale, process-based model that was developed to simulate nitrogen dynamics and turnover in the soil-water-plant system under different management practices and soil and environmental conditions. It is an enhanced version of the nitrogen (N) simulation model, DRAINMOD-N, that simulates a more complete N cycle, adds a carbon (C) cycle, and operates at different levels of complexity. Processes considered in the model include atmospheric deposition, application of mineral N fertilizers including urea and anhydrous ammonia (NH3), soil amendment with organic N (ON) sources including plant residues and animal waste, plant uptake, organic C (OC) decomposition and associated N mineralization/immobilization, nitrification, denitrification, NH3 volatilization, and N losses via subsurface drainage and surface runoff. Nitrogen pools considered in the model are nitrate-nitrogen (NO3-N), ammoniacal nitrogen (NHx-N) and ON. DRAINMOD-N II includes a submodel that simulates C dynamics in the soil-plant system using a C cycle similar to that of the CENTURY model. A simplified approach is used to simulate temporal changes in soil pH; consequently, the model determines the composition of the NHx-N pool and, if necessary, changes its operation mode. DRAINMOD-N II simulates N reactive transport using a finite difference solution to a multiphase form of the one-dimensional advection- dispersion-reaction equation. Model output includes daily concentrations of NO3-N and NHx-N in soil solution and drain flow, daily OC content of the top 20 cm soil layer, and cumulative rates of simulated N processes.

DRAINMOD: Model Use, Calibration, and Validation

DRAINMOD is a process-based, distributed, field-scale model developed to describe the hydrology of poorly drained and artificially drained soils. The model is based on water balances in the soil profile, on the field surface, and, in some cases, in the drainage system. This article briefly describes the model and the algorithms that are used to quantify the various hydrologic components. Inputs for soil properties, site parameters, weather data, and crop characteristics required in the application of the model are presented and discussed with respect to their role in calibration. Methods for determining field effective values of key inputs to the model, either independently or as a part of the calibration process, are demonstrated in a case study. The case study involved calibrating DRAINMOD with two years of field data for a subsurface drained agricultural field in eastern North Carolina, followed by testing or validation of the model with two additional years of data. Performance statistics indicated that the model with calibrated input data accurately predicted daily water table depths with Nash-Sutcliffe modeling efficiency (EF) values of 0.68 and 0.72, daily drainage rates (EF = 0.73 and 0.49), and monthly drainage volumes (EF = 0.87 and 0.77) for the two-year validation period.

DRAINMOD MODIFICATIONS FOR COLD CONDITIONS

The field hydrology model DRAINMOD was modified to include freezing and thawing, and snowmelt components. Based on daily hydrologic predictions of the original model, the modified DRAINMOD numerically solves the heat flow equation to predict soil temperature. When freezing conditions are indicated by below zero temperatures, the model calculates ice content in the soil profile and modifies soil hydraulic conductivity and infiltration rate accordingly. Recorded precipitation is separated as rain and snow when daily average air temperature is above or below a rain/snow dividing base temperature. Snow is predicted to accumulate on the ground until air temperature rises above a snowmelt base temperature. Soil surface temperature is recalculated when snow cover exists. Daily snowmelt water is added to rainfall, which may infiltrate or run off depending on soil freezing condition. The modified DRAINMOD predictions of soil temperature agreed well with field observations at Plymouth, North Carolina, Truro, Nova Scotia, and Lamberton, Minnesota. Assuming air temperature as the soil surface boundary condition increased the variability of soil temperature predictions at shallow depths, agreement with field measurements was still good. The method of using average air temperature as an indicator to separate snow and rain worked very well for Carsamba, Turkey. At Truro, Nova Scotia, however, the method was not as successful, and several snow events were predicted as rainfall and vice versa. Compared with the original version of DRAINMOD, the modified version predicts fewer drainage flow events in winter months because of snow accumulation on the surface. Subsurface drainage and/or surface runoff resulting from snowmelt are predicted when air temperature rises, the snow melts, and the soil begins to thaw.

Effects of Subsurface Drain Depth on Nitrogen Losses from Drained Lands

A simulation study was conducted to determine effects of drain depth on nitrogen (N) loss in drainage water. Simulations were conducted for drain depths of 0.75, 1.0, 1.25, and 1.5 m for a Portsmouth sandy loam at Plymouth, North Carolina. A wide range of drain spacing was considered for each depth. Corn yields were predicted and an economic analysis was conducted to determine the drain spacing giving maximum predicted profit for each depth. Results showed that nitrogen losses from subsurface drains can be reduced by placing the drains at shallow depths. In order to satisfy agricultural production requirements, shallow drains must be placed closer together than deeper drains. While predicted agricultural profits for the shallow drains are reduced somewhat compared to the deeper drains, overall profits are substantially increased when the cost of removing N from drainage water is considered.

Nutrient and sediment removal in forested wetlands receiving pumped agricultural drainage water

The effectiveness of two forested wetland buffer areas at removing sediment and nutrients from pumped agricultural drainage water was evaluated in a two-year field study. The movement of these potential pollutants during pumping events was determined by sampling water quality at 36 stations distributed over each wetland. Automatic water sampling continued after pumping events to determine the nutrient and sediment removal rates in the water left standing on the wetlands. The total volume of water pumped during small events was effectively stored on the wetland until displaced by subsequent pumping or removed by evapotranspiration. Nutrient and sediment concentrations of this stored water, which had several days of residence time on the wetland, were near background levels before leaving the wetlands. Nutrient concentrations leaving the wetlands during these small events ranged from 0.03 to 0.04 mg/L for total phosphorus (TP) and from 0.0 to 0.1 mg/L for nitrate nitrogen (NO3-N). Pumped water completely traversed the wetland during the less-frequent, larger pumping events. Nutrient and sediment concentrations at the wetland outlet were often higher than background concentrations, ranging from 0 to 70 percent of the inflow concentrations. Nutrient concentrations leaving the wetlands during these larger events ranged from 0.06 to 0.12 mg/L for TP and from 0.0 to 4.7 mg/L for NO3-N. Sediment, TP, and NO3-N were removed from drainage water standing on the wetland. This nutrient and sediment removal was described using a first order decay model. Deposition of sediment was observed only within 800 m of the pumps, and was not resuspended during subsequent large pumping events.

Hydrology and Water Quality of Forested Lands in Eastern North Carolina

More than 100 site years of hydrology and water quality data spanning 25 years (1976-2000) were compiled from research and monitoring studies on forest stands with natural vegetation and tracts managed for timber production. A total of 41 watersheds located on poorly drained to very poorly drained soils on flat divides between coastal streams were included ranging in area from 7.3 to 6070 ha. Hydrology and nutrient concentration data from the study sites are used to examine how variation among sites may be related to soil type, drainage intensity, vegetation, and physiographic setting. The median annual hydrologic response (outflow as a percentage of precipitation) among the sites was 31%, with an interquartile range of 26-35%. Nutrient concentrations in forest outflow were generally low for most study sites compared with typical values for other land uses. Mean seasonal concentrations of nutrient fractions in drainage from 75% of the study sites were <1.8 mg/L for total N (TN), <0.08 mg/L for total P (TP). Concentrations of Org-N, TN, and TP were all consistently higher in drainage from organic soils compared with mineral soils for both paired comparisons and the overall data base. TN exports from 75% of the study sites were less than 6.5 kg/ha/yr, predominantly as Org-N and TP exports from all forest sites was <0.36 kg/ha/yr.

DRAINMOD-FOREST: Integrated modeling of hydrology, soil carbon and nitrogen dynamics, and plant growth for drained forests

We present a hybrid and stand-level forest ecosystem model, DRAINMOD-FOREST, for simulating the hydrology, carbon (C) and nitrogen (N) dynamics, and tree growth for drained forest lands under common silvicultural practices. The model was developed by linking DRAINMOD, the hydrological model, and DRAINMOD-N II, the soil C and N dynamics model, to a forest growth model, which was adapted mainly from the 3-PG model. The forest growth model estimates net primary production, C allocation, and litterfall using physiology-based methods regulated by air temperature, water deficit, stand age, and soil N conditions. The performance of the newly developed DRAINMOD-FOREST model was evaluated using a long-term (21-yr) data set collected from an artificially drained loblolly pine ( L.) plantation in eastern North Carolina, USA. Results indicated that the DRAINMOD-FOREST accurately predicted annual, monthly, and daily drainage, as indicated by Nash-Sutcliffe coefficients of 0.93, 0.87, and 0.75, respectively. The model also predicted annual net primary productivity and dynamics of leaf area index reasonably well. Predicted temporal changes in the organic matter pool on the forest floor and in forest soil were reasonable compared to published literature. Both predicted annual and monthly nitrate export were in good agreement with field measurements, as indicated by Nash-Sutcliffe coefficients above 0.89 and 0.79 for annual and monthly predictions, respectively. This application of DRAINMOD-FOREST demonstrated its capability for predicting hydrology and C and N dynamics in drained forests under limited silvicultural practices.

FIELD TESTING OF DRAINMOD-N

This study was conducted to evaluate the performance of DRAINMOD-N, a nitrogen fate and transport model for artificially drained soils, based on a comparison between predicted and observed hydrologic and nitrogen variables for an experimental site in eastern North Carolina. The site consisted of six plots drained by subsurface drain tubes 1.25 m deep and 23 m apart. Each plot was instrumented to measure water table depth, subsurface drainage, surface runoff and subirrigation rates. There were two replications of three water management treatments: conventional drainage, controlled drainage and subirrigation. Crops were winter wheat followed by soybean. Results showed the model did a good job in describing the hydrology of the site. On average the predicted daily water table depths were within 0.13 m of observed during the 14-month study period. Differences between predicted and observed cumulative subsurface drainage and surface runoff volumes were less than 0.10 and 0.09 m, respectively, for all treatments. Predictions for the movement and fate of nitrogen were also in good agreement with measured results. Simulated nitratenitrogen (NO3-N) losses in subsurface drainage water were within 1.5 kg/ha of the observed values for the 14-month period. Differences between simulated and observed total NO3-N losses (subsurface drainage plus surface runoff) were within 3.0 kg/ha. Results of this study indicated DRAINMOD-N could be used to simulate nitrogen losses in poorly drained soils with artificial drainage. The model, however, needs to be tested for longer periods of time and under different climatic conditions and soil types, before it can be recommended for general use.

WATGIS: A GIS-Based Lumped Parameter Water Quality Model

A Geographic Information System (GIS)-based, lumped parameter water quality model was developed to estimate the spatial and temporal nitrogen-loading patterns for lower coastal plain watersheds in eastern North Carolina. The model uses a spatially distributed delivery ratio (DR) parameter to account for nitrogen retention or loss along a drainage network. Delivery ratios are calculated from time of travel and an exponential decay model for in-stream dynamics. Travel times from any point in the drainage network to the watershed outlet are obtained from simulations using a combined physically based field hydrology and drainage canal routing model (DRAINMOD-DUFLOW). Nitrogen load from contributing areas in the watershed delivered to the main watershed outlet is obtained as the product of field export with the corresponding delivery ratio. The total watershed load at the outlet is the combined loading of the individual fields. Nitrogen exports from source areas are measured. The lumped water quality model is integrated within a GIS framework with menu interface, display options, and statistical procedures. Within this framework, the model can be used as a screening tool to analyze the effects of different land and water management practices on downstream water quality. A description of the model is presented along with the results from the evaluation of the model to characterize the seasonal and annual export of nitrogen from a drained forested watershed near Plymouth, North Carolina. Results of the study showed that the lumped parameter model can reasonably predict the loads at the outlet of the watershed. Predicted loads for 1997 were highly correlated with the observed loads (correlation coefficients of 0.99, 0.90, and 0.96 for nitrate-nitrogen, TKN, and total nitrogen respectively). Sensitivity and uncertainty analyses indicated that predicted outlet loads were sensitive to field flow predictions and export concentrations. Overall, the results indicate that the lumped parameter model can be an effective tool for describing the monthly nitrogen loads from a poorly drained coastal plain watershed.