Dan B. Jaynes

Predicting Water, Sediment and NO3-N Loads under Scenarios of Land-use and Management Practices in a Flat Watershed

Changes in land-use or management practices may affect water outflow, sediment, nutrients and pesticides loads. Thus, there is an increasing demand for quantitative information at the watershed scale that would help decision makers or planners to take appropriate decisions. This paper evaluates by a modeling approach the impact of farming practices and land-use changes on water discharge, sediment and NO3-N loads at the outlet of a 51.29 km2 watershed of central Iowa (Walnut Creek watershed). This intensively farmed (corn-soybean rotation) watershed is characterized by a flat topography with tiles and potholes. Nine scenarios of management practices (nitrogen application rates: increase of current rate by 20, 40%, decrease of current rate by 20, 40 and 60%; no tillage) and land-use changes (from corn-soybean rotation to winter wheat and pasture) were tested over a 30 yr simulated period. The selected model (Soil and Water Assessment Tool, SWAT) was first validated using observed flow, sediment and nutrient loads from 1991 to 1998. Scenarios of N application rates did not affect water and sediment annual budgets but did so for NO3-N loads. Lessening the N rate by 20, 40 and 60% in corn-soybean fields decreased mean NO3-N annual loads by 22, 50 and 95%, respectively, with greatest differences during late spring. On the other hand, increasing input N by 20 and 40% enhanced NO3-N loads by 25 and 49%, respectively. When replacing corn-soybean rotation by winter wheat, NO3-N loads increased in early fall, immediately after harvest. Pasture installation with or without fertilization lessened flow discharge, NO3-N and sediment delivery by 58, 97 and 50%, respectively. No-tillage practices did not significantly affect the water resource and sediment loads. Finally, such realistic predictions of the impact of farming systems scenarios over a long period are discussed regarding environmental processes involved.

Nitrogen fertilizer effects on soil carbon balances in Midwestern U.S. agricultural systems

A single ecosystem dominates the Midwestern United States, occupying 26 million hectares in five states alone: the corn-soybean agroecosystem [Zea mays L.-Glycine max (L.) Merr.]. Nitrogen (N) fertilization could influence the soil carbon (C) balance in this system because the corn phase is fertilized in 97-100% of farms, at an average rate of 135 kg N x ha(-1) x yr(-1). We evaluated the impacts on two major processes that determine the soil C balance, the rates of organic-carbon (OC) inputs and decay, at four levels of N fertilization, 0, 90, 180, and 270 kg/ha, in two long-term experimental sites in Mollisols in Iowa, USA. We compared the corn-soybean system with other experimental cropping systems fertilized with N in the corn phases only: continuous corn for grain; corn-corn-oats (Avena sativa L.)-alfalfa (Medicago sativa L.; corn-oats-alfalfa-alfalfa; and continuous soybean. In all systems, we estimated long-term OC inputs and decay rates over all phases of the rotations, based on long-term yield data, harvest indices (HI), and root:shoot data. For corn, we measured these two ratios in the four N treatments in a single year in each site; for other crops we used published ratios. Total OC inputs were calculated as aboveground plus belowground net primary production (NPP) minus harvested yield. For corn, measured total OC inputs increased with N fertilization (P < 0.05, both sites). Belowground NPP, comprising only 6-22% of total corn NPP, was not significantly influenced by N fertilization. When all phases of the crop rotations were evaluated over the long term, OC decay rates increased concomitantly with OC input rates in several systems. Increases in decay rates with N fertilization apparently offset gains in carbon inputs to the soil in such a way that soil C sequestration was virtually nil in 78% of the systems studied, despite up to 48 years of N additions. The quantity of belowground OC inputs was the best predictor of long-term soil C storage. This indicates that, in these systems, in comparison with increased N-fertilizer additions, selection of crops with high belowground NPP is a more effective management practice for increasing soil C sequestration.

DEVELOPMENT AND APPLICATION OF SWAT TO LANDSCAPES WITH TILES AND POTHOLES

SWAT (Soil and Water Assessment Tool) is a watershed model that has been incorporated into USEPA’s modeling framework called BASINS used for total maximum daily load (TMDL) analysis. It is thus important that SWAT realistically simulate tile flow and pothole landscapes that are common in much of the Corn Belt and Great Lakes states. In this study, SWAT was modified to simulate water table dynamics and linked with a simple tile flow equation. Algorithms were also added to SWAT to include simulation of potholes (closed depressions), surface tile inlets, and aeration stress on plants. Flow interaction between HRUs was introduced in order to simulate pothole water. The modified SWAT model (SWAT-M) was evaluated using eight years of measured flow data from Walnut Creek watershed (WCW), an intensively tile-drained watershed in central Iowa. The model was calibrated during the period of 1992 to 1995 and validated during the period of 1996 to 1999. In addition, comparisons between the enhanced version (SWAT-M) and older version (SWAT2000) of SWAT were conducted. For assessing overall performance of the SWAT models, predictions were compared to data measured at stream sites approximately at the midpoint of the watershed (site 310) and at the outlet (site 330). Nash-Sutcliffe E values for the simulated monthly flows during the calibration/validation periods were 0.88/0.82 and 0.84/0.72 at sites 310 and 330, respectively. The relative mean errors (RME) of the simulated monthly flows during the calibration/validation periods were -2% / -1% and -18% / 10%, respectively, for the same two sites. These statistical values indicate that SWAT-M estimated both pattern and amount of the monthly flows reasonably well for the large flat landscape of WCW containing tile drains and potholes. SWAT-M needs to improve in its daily prediction because of its lower E values (-0.11 to 0.55), compared to the monthly results. In applying the model to a third site (site 210) that was predominantly influenced by tile drainage, it was concluded that the pattern and amount of simulated monthly subsurface flows (E values of 0.61 and 0.70 and RME values of 10% and -9% for the calibration and validation periods, respectively) were relatively close to the measured values. Nevertheless, SWAT-M simulation of daily subsurface flows was less accurate than monthly results. In general, the pattern and amount of monthly flow and subsurface tile drainage predicted by SWAT-M has been greatly improved as compared to SWAT2000.

Denitrification in Wood Chip Bioreactors at Different Water Flows

Subsurface drainage in agricultural watersheds exports a large quantity of nitrate-nitrogen (NO(3)-N) and concentrations frequently exceed 10 mg L(-1). A laboratory column study was conducted to investigate the ability of a wood chip bioreactor to promote denitrification under mean water flow rates of 2.9, 6.6, 8.7 and 13.6 cm d(-1) which are representative of flows entering subsurface drainage tiles. Columns were packed with wood chips and inoculated with a small amount of oxidized till and incubated at 10 degrees C. Silicone sampling cells at the effluent ports were used for N(2)O sampling. (15)Nitrate was added to dosing water at 50 mg L(-1) and effluent was collected and analyzed for NO(3)-N, NH(4)-N, and dissolved organic carbon. Mean NO(3)-N concentrations in the effluent were 0.0, 18.5, 24.2, and 35.3 mg L(-1) for the flow rates 2.9, 6.6, 8.7, and 13.6 cm d(-1), respectively, which correspond to 100, 64, 52, and 30% efficiency of removal. The NO(3)-N removal rates per gram of wood increased with increasing flow rates. Denitrification was found to be the dominant NO(3)-N removal mechanism as immobilization of (15)NO(3)-N was negligible compared with the quantity of (15)NO(3)-N removed. Nitrous oxide production from the columns ranged from 0.003 to 0.028% of the N denitrified, indicating that complete denitrification generally occurred. Based on these observations, wood chip bioreactors may be successful at removing significant quantities of NO(3)-N, and reducing NO(3)-N concentration from water moving to subsurface drainage at flow rates observed in central Iowa subsoil.

SPATIO-TEMPORAL ANALYSIS OF YIELD VARIABILITY FOR A CORN-SOYBEAN FIELD IN IOWA

Spatio-temporal analyses of yield variability are required to delineate areas of stable yield patterns for application of precision farming techniques. Spatial structure and temporal stability patterns were studied using 1995- 1997 yield data for a 25-ha field located near Story City, Iowa. Corn was grown during 1995-1996, and soybean in 1997. The yield data were collected on nine east-west transects, consisting of 25 yield blocks per transect. The two components of yield variability, i.e., large-scale variation (trend) and small-scale variation, were studied using median polishing technique and variography, respectively. The trend surface, obtained from median polishing, accounted for the large-scale deterministic structure induced by treatments and landscape effects. After removal of trend from yield data, the resulting yield residuals were used to analyze the small-scale stochastic variability using variography. The variogram analysis showed strong spatial structure for the yield residuals. The spatial correlation lengths were found to vary from about 40 m for corn to about 90 m for soybean. The range parameter of the variograms showed a significant correlation coefficient of –0.95 with the cumulative growing season rainfall. The total variance of 1995 corn yield was partitioned as 56% trend, 37% small-scale stochastic structure, and 7% as an interaction of both. Yield variance of 1996 corn was about 80% trend and 20% small-scale stochastic structure. Contrary to corn years, the total yield variance for soybean in 1997 was partitioned as about 25% trend and about 75% small-scale stochastic structure. The significant negative correlation of range with rainfall shows that small-scale variability may be controlled by factors induced directly or indirectly by rainfall. More years of data are required to substantiate these relationships. The lack of temporal stability in large-scale and small-scale variation suggest that longer duration yield data analyses are required to understand and quantify the impact of various climatic, and management factors and their interaction with soil properties on delineation of areas under consistent yield patterns before applying variable rate technology.

YIELD VARIABILITYWITHINACENTRAL IOWA FIELD

Technologies to support precision farming (PF) began to emerge in 1989 when the Global Positioning System (GPS) became available to a limited extent and was tested as a means for locating farm equipment within fields. Substantial PF technology is available with rapidly decreasing costs and increasing capabilities. However, one major class of information that is missing is a method for determining how much material to apply or what action to take as a result of a specific condition at any position within a field. Developing this information will require knowing the spatial and temporal variability of plant response and will most likely be obtained by measuring yield variability. This field study was designed to quantify yield variability within a 16 ha field which has had consistent practices for several years. Crop yields showed a coefficient of variation ranging from near 12% in 1989 and 1992 to over 30% in 1990 and 1993. Rankings of the long-term relative yield for 224 locations were not stable even after 6 years when recalculated each year. Many PF scenarios are based on the assumption of a stable yield pattern within a field, but only a few points in this field have exhibited such a pattern. Perhaps stable patterns will eventually emerge, but the time frame for this to occur may be quite long. Overall, this study suggests that implementation of PF practices within the Clarion-Nicollet-Webster soil association area will reveal both difficulties and opportunities.

Simulating Long-Term Effects of Nitrogen Fertilizer Application Rates on Corn Yield and Nitrogen Dynamics

Thoroughly tested agricultural systems models can be used to quantify the long-term effects of crop management practices under conditions where measurements are lacking. In a field near Story City, Iowa, ten years (1996-2005) of measured data were collected from plots receiving low, medium, and high (57-67, 114-135, and 172-202 kg N ha-1) nitrogen (N) fertilizer application rates during corn (Zea mays L.) years. Using these data, the Root Zone Water Quality Model linked with the CERES and CROPGRO plant growth models (RZWQM-DSSAT) was evaluated for simulating the various N application rates to corn. The evaluated model was then used with a sequence of historical weather data (1961-2005) to quantify the long-term effects of different N rates on corn yield and nitrogen dynamics for this agricultural system. Simulated and measured dry-weight corn yields, averaged over plots and years, were 7452 and 7343 kg ha-1 for the low N rate, 8982 and 9224 kg ha-1 for the medium N rate, and 9143 and 9484 kg ha-1 for the high N rate, respectively. Simulated and measured flow-weighted average nitrate concentrations (FWANC) in drainage water were 10.6 and 10.3 mg L-1 for the low N rate, 13.4 and 13.2 mg L-1 for the medium N rate, and 18.0 and 19.1 mg L-1 for the high N rate, respectively. The simulated N rate for optimum corn yield over the long term was between 100 and 150 kg N ha-1. Currently, the owner-operator of the farm applies 180 kg N ha-1 to corn in nearby production fields. Reducing long-term N rates from 180 to 130 kg N ha-1 corresponded to an 18% simulated long-term reduction in N mass lost to water resources. Median annual FWANC in subsurface drainage water decreased from 19.5 to 16.4 mg N L-1 with this change in management. Current goals for diminishing the hypoxic zone in the Gulf of Mexico call for N loss reductions of 30% and greater. Thus, long-term simulations suggest that at least half of this N loss reduction goal could be met by reducing N application rates to the production optimum. However, additional changes in management will be necessary to completely satisfy N loss reduction goals while maintaining acceptable crop production for the soil and meteorological conditions of this study. The results suggest that after calibration and thorough testing, RZWQM-DSSAT can be used to quantify the long-term effects of different N application rates on corn production and subsurface drainage FWANC in Iowa.

EVALUATION OF SWAT IN SIMULATING NITRATE NITROGEN AND ATRAZINE FATES IN A WATERSHED WITH TILES AND POTHOLES

We evaluated a version of the Soil Water Assessment Tool (SWAT-M) that was modified to more accurately simulate tile drainage and water flow in a landscape dominated by closed surface depressions or potholes at a watershed scale using ten years of measured nitrate-nitrogen (NO3-N) and atrazine data in stream discharge in the Walnut Creek watershed (WCW). The model was calibrated during the period of 1992 to 1995 and validated during the period of 1996 to 2001. Stream sites in the middle and outlet of the WCW were selected to assess overall performance of the model, while one drainage district drain was used for investigating chemical loads in subsurface flows. With the introduction of an independent tile drain lag time parameter, the performance of SWAT-M for daily flow simulation was improved. In comparison to our previous results, the Nash-Sutcliffe E values for the calibrated daily flow at the mid-watershed and outlet simulated by the enhanced SWAT model rose from 0.55 to 0.69 and from 0.51 to 0.63, respectively. Of special note, the E value for calibrated flow rose from -0.23 to 0.40 for the drainage district drain, which was dominated by tile and subsurface flow. Both the predicted corn yields and N uptake by corn were very similar to the measured data. The predicted yield and N uptake by soybean were relatively lower than the measured values. The monthly NO3-N loads in stream discharges at the center and outlet of the Walnut Creek watershed were accurately predicted with good Nash-Sutcliffe E values of 0.91/0.80 and 0.85/0.67 in calibration/validation, respectively. Nevertheless, the model’s simulation of the daily NO3-N loads was not as good as the monthly simulation. The good agreement between the simulated and measured monthly NO3-N loads from the drainage district site leads us to conclude that SWAT can reasonably simulate tile flow from pothole-dominated landscapes, although the model needs to be improved in the simulation of daily subsurface NO3-N fluxes. The enhanced SWAT-M model simulated the NO3-N loads in a watershed with intensive tile drainage systems much more accurately than the original SWAT2000 version. A second pesticide degradation half-life in soil was added for SWAT-M, which greatly improved the model performance for predicting atrazine losses from the watershed. Overall, SWAT-M is capable of simulating atrazine loads in the stream discharge of the WCW and is a much-improved tool over SWAT2000 for predicting both daily and monthly atrazine losses in nearly level, tile-drained watersheds.

Using Soil Attributes and GIS for Interpretation of Spatial Variability in Yield

Precision farming application requires better understanding of variability in yield patterns in order to determine the cause-effect relationships. This field study was conducted to investigate the relationship between soil attributes and corn (Zea mays L.)-soybean (Glycine max L.) yield variability using four years (1995-98) yield data from a 22-ha field located in central Iowa. Corn was grown in this field during 1995, 1996, and 1998, and soybean was grown in 1997. Yield data were collected on nine east-west transects, consisting of 25-yield blocks per transect. To compare yield variability among crops and years, yield data were normalized based on N-fertilizer treatments. The soil attributes of bulk density, cone index, organic matter, aggregate uniformity coefficient, and plasticity index were determined from data collected at 42 soil sampling sites in the field. Correlation and stepwise regression analyses over all soil types in the field revealed that Tilth Index, based upon soil attributes, did not show a significant relationship with the yield data for any year and may need modifications. The regression analysis showed a significant relationship of soil attributes to yield data for areas of the field with Harps and Ottosen soils. From a geographic information system (GIS) analysis performed with ARC/INFO, it was concluded that yield may be influenced partly by management practices and partly by topography for Okoboji and Ottosen soils. Map overlay analysis showed that areas of lower yield for corn, at higher elevation, in the vicinity of Ottosen and Okoboji soils were consistent from year to year; whereas, areas of higher yield were variable. From GIS and statistical analyses, it was concluded that interaction of soil type and topography influenced yield variability of this field. These results suggest that map overlay analysis of yield data and soil attributes over longer duration can be a useful approach to delineate subareas within a field for site specific agricultural inputs by defining the appropriate yield classes.

SIMULATING THE LONG-TERM PERFORMANCE OF DRAINAGE WATER MANAGEMENT ACROSS THE MIDWESTERN UNITED STATES

Drainage water management (DWM) has been proposed as a solution to reduce losses of nitrate (NO3) from subsurface drainage systems in the midwestern U.S.; however, tests of DWM efficacy have only been performed over short time periods and at a limited number of sites. To fill this gap, the RZWQM-DSSAT hybrid model, previously evaluated for a subsurface-drained agricultural system in Iowa, was used to simulate both conventional drainage (CVD) and DWM over 25 years of historical weather at 48 locations across the midwestern U.S. Model simulations were used to demonstrate how variability in both climate and management practices across the region affects the ability of DWM to reduce losses of NO3 in subsurface drainage. The regional average simulated reduction in drain flow was 151 mm yr-1 when using DWM instead of CVD, and the regional percent reduction over the long term was 53%. Reductions in drain flow were offset mainly by increases in surface runoff and evapotranspiration. Similarly for nitrogen (N), the regional average simulated reduction in NO3 losses through subsurface drains was 18.9 kg N ha-1 yr-1, and the regional percent reduction over the long term was 51%. Subsurface drain NO3 loss reductions were counterbalanced mainly by increases in stored soil N, denitrification, and plant N uptake. The simulations suggest that if DWM can be practically implemented throughout the region, particularly in the southern states, then substantial reductions in the amount of NO3 entering surface waters from agricultural systems can be expected.