Kyle R. Mankin

MODELING BACTERIA FATE AND TRANSPORT IN WATERSHEDS TO SUPPORT TMDLS

Fecal contamination of surface waters is a critical water-quality issue, leading to human illnesses and deaths. Total Maximum Daily Loads (TMDLs), which set pollutant limits, are being developed to address fecal bacteria impairments. Watershed models are widely used to support TMDLs, although their use for simulating in-stream fecal bacteria concentrations is somewhat rudimentary. This article provides an overview of fecal microorganism fate and transport within watersheds, describes current watershed models used to simulate microbial transport, and presents case studies demonstrating model use. Bacterial modeling capabilities and limitations for setting TMDL limits are described for two widely used watershed models (HSPF and SWAT) and for the load-duration method. Both HSPF and SWAT permit the user to discretize a watershed spatially and bacteria loads temporally. However, the options and flexibilities are limited. The models are also limited in their ability to describe bacterial life cycles and in their ability to adequately simulate bacteria concentrations during extreme climatic conditions. The load-duration method for developing TMDLs provides a good representation of overall water quality and needed water quality improvement, but intra-watershed contributions must be determined through supplemental sampling or through subsequent modeling that relates land use and hydrologic response to bacterial concentrations. Identified research needs include improved bacteria source characterization procedures, data to support such procedures, and modeling advances including better representation of bacteria life cycles, inclusion of more appropriate fate and transport processes, improved simulation of catastrophic conditions, and creation of a decision support tool to aid users in selecting an appropriate model or method for TMDL development.

The new gold rush: Fueling ethanol production while protecting water quality

Renewable fuel production, particularly grain-based ethanol, is expanding rapidly in the USA. Although subsidized grain-based ethanol may provide a competitively priced transportation fuel, concerns exist about potential environmental impacts. This contribution focuses on potential water quality implications of expanded grain-based ethanol production and potential impacts of perennial-grass-based cellulosic ethanol. Expanded grain-based ethanol will increase and intensify corn production. Even with recommended fertilizer and land conservation measures, corn acreage can be a major source of N loss to water (20-40 kg ha(-1) yr(-1)). A greater acreage of corn is estimated to increase N and P loss to water by 37% (117 million kg) and 25% (9 million kg), respectively, and measures to encourage adoption of conservation practices are essential to mitigate water quality impairments. Dried distillers grains remaining after ethanol production from corn grain are used as animal feed and can increase manure P content and may increase N content. Cellulosic fuel-stocks from perennials such as switchgrass and woody materials have the potential to produce ethanol. Although production, storage, and handling of cellulosic materials and conversion technology are limitations, accelerating development of cellulosic ethanol has the potential to reduce dependence on grain fuel-stocks and provide water quality and other environmental benefits. All alternative fuel production technologies could have environmental impacts. There is a need to understand these impacts to help guide policy and help make programmatic and scientific decisions that avoid or mitigate unintended environmental consequences of biofuel production.

Vegetative Treatment Systems for Management of Open Lot Runoff: Review of Literature

Runoff from open lot livestock systems (beef and dairy) defined as Concentrated Animal Feeding Operations (CAFO) must be controlled by systems designed and managed to prevent the release of manure-contaminated runoff for storms equal to or less than a 25-yr, 24-h design storm. This performance standard has been attained for open lot systems with some combination of clean water diversion, settling basins, runoff collection ponds, and irrigation systems (baseline system). An alternative approach is to rely on overland flow and infiltration into cropland with perennial forage or grasses for treatment of open lot runoff. Such vegetative systems have been researched since the late 1960s. This article reviews the research literature on vegetative treatment systems (VTS) for managing open lot runoff summarizing available science on system performance, design, and management. Based upon this review of the literature, the following conclusions are drawn about the application of VTS to manage runoff from open lot livestock production systems: (1) Substantial research (approximately 40 identified field trials and plot studies) provides a basis for understanding the performance of VTS. These performance results suggest that a vegetative system consisting of a settling basin and VTA or Vegetative Infiltration Basin (VIB) has the potential to achieve functional equivalency to conventional technologies. (2) The existing research targeting VTS is confined to non-CAFO applications, likely due to past regulatory limits. Unique challenges exist in adapting these results and recommendations to CAFO applications. (3) The pollutant reduction resulting from a VTS is based upon two primary mechanisms: 1) sedimentation, typically occurring within the first few meters of a VTS, and 2) infiltration of runoff into the soil profile. Systems relying primarily on sedimentation only are unlikely to perform equal to or better than baseline technologies. System design based upon sedimentation and infiltration is necessary to achieve a required performance level for CAFO application. (4) Critical design factors specific to attaining high levels of pollutant reduction within a VTS include pre-treatment, sheet flow, discharge control, siting, and sizing. Critical management factors include maintenance of a dense vegetation stand and sheet flow of runoff across VTA as well as minimization of nutrient accumulation.

Source specific fecal bacteria modeling using soil and water assessment tool model

Fecal bacteria can contaminate water and result in illness or death. It is often difficult to accurately determine sources of fecal bacteria contamination, but bacteria source tracking can help identify non-point sources of fecal bacteria such as livestock, humans and wildlife. The Soil and Water Assessment Tool (SWAT) microbial sub-model 2005 was used to evaluate source-specific fecal bacteria using three years (2004-2006) of observed modified deterministic probability of bacteria source tracking data, as well as measure hydrologic and water quality data. This study modeled source-specific bacteria using a model previously calibrated for flow, sediment and total fecal coliform bacteria (FCB) concentration. The SWAT model was calibrated at the Rock Creek sub-watershed, validated at the Deer Creek sub-watershed, and verified at the Auburn sub-watershed and then at the entire Upper Wakarusa watershed for predicting daily flow, sediment, nutrients, total fecal bacteria, and source-specific fecal bacteria. Watershed characteristics for livestock, humans, and wildlife fecal bacterial sources were first modeled together then with three separate sources and combinations of source-specific FCB concentration: livestock and human, livestock and wildlife and human and wildlife. Model results indicated both coefficient of determination (R(2)) and Nash-Sutcliffe Efficiency Index (E) parameters ranging from 0.52 to 0.84 for daily flow and 0.50-0.87 for sediment (good to very good agreement); 0.14-0.85 for total phosphorus (poor to very good agreement); -3.55 to 0.79 for total nitrogen (unsatisfactory to very good agreement) and -2.2 to 0.52 for total fecal bacteria (unsatisfactory to good agreement). Model results generally determined decreased agreement for each single source of bacteria (R(2) and E range from -5.03 to 0.39), potentially due to bacteria source tracking (BST) uncertainty and spatial variability. This study contributes to new knowledge in bacteria modeling and will help further understanding of uncertainty that exists in source-specific bacteria modeling.

SURVIVAL OF FECAL BACTERIA IN DAIRY COW MANURE

Bacterial pollution of water is impacted to a great extent by the ability of bacteria to survive in manure following excretion. We investigated the effects of environmental temperature (4°C, 27°C, and 41°C) and manure moisture content (30%, 55%, and 83%) on the survival and release of indicator bacteria in dairy cow manure. Fresh manure samples of about 60 g were packed to 12 mm depth in polystyrene dishes and held at controlled temperatures and moisture contents for up to 103 days. Supernatant from a distilled-water extraction was enumerated for fecal bacteria (fecal coliforms, Escherichia coli, and fecal streptococci) by the membrane filtration method. Bacterial populations increased as much as 2.5 log10 (over 300×) in the three days following excretion. Temperature had a significant overall effect on survival of all three fecal bacteria, whereas moisture content produced a consistent effect on fecal streptococci survival only. Fecal streptococci showed no significant die-off at any temperature or moisture content studied. In contrast, no measurable E. coli or fecal coliforms were found in supernatant water samples from the 41°C treatment after day 35. E. coli and fecal coliform populations for the 4°C treatment at lower moisture content (30% and 55%) conditions were close to the detection limits after five weeks, but significant numbers of E. coli (2.34 × 104 cfu g-1 wet manure) and fecal coliforms (3.84 × 104 cfu g-1 wet manure) remained for the 4°C treatment at 83% moisture content after 103 days. First-order die-off rate coefficients for E. coli were found to be appropriate after day 3 for about a 3-week period and averaged 0.11 d-1 at 4°C, 0.20 d-1 at 27°C, and 0.32 d-1 at 41°C. Results from this study suggest that barnyard, feedlot, and manure management practices that detain manure at higher temperatures (e.g., 41°C) will decrease the E. coli and fecal coliform populations but not those of fecal streptococci. Coliform bacterial populations tested remained viable for long periods (>3 months) particularly at moderate temperature (27°C) for any moisture level, and streptococci survived under all conditions studied.

Watershed-scale AMC selection for hydrologic modeling

The Natural Resources Conservation Service curve–number (CN) method commonly uses three discrete levels (1, 2, and 3) of antecedent moisture condition (AMC) to describe soil moisture at the time of a runoff event. However, this may not adequately represent soil water conditions for watershed modeling purposes. The objectives of this study were to evaluate the use of individual–event watershed–scale AMC values to adjust field–scale CN, and to assess which hydrologic parameters would provide the best estimate of individual–event AMC. Landsat Thematic Mapper images from 1997 and 1998 were used to obtain 10 landcover classes for Red Rock Creek watershed, Kansas. The canopy growth of crops was used to provide temporal adjustment of CNs in the watershed. Stream–flow data for 1997–1999 was collected from a U.S. Geological Survey gaging station near the watershed outlet, and base flow was separated to obtain surface–runoff amounts. Watershed–average AMC factors were estimated from measured rainfall and surface runoff amounts for each of 23 events and used to adjust CNs in the AGNPS watershed model. For individual runoff events, calibration was achieved with AMCs that averaged 1.5 and ranged from 0.9 to 2.4. Therefore, an AMC of 2, as used in many hydrologic models, would overestimate the surface runoff amounts in this sub–humid Kansas watershed. Generally, AMC increased with 5-day antecedent rainfall above 5 mm. Soil moisture and 5–day antecedent rain were found to be significantly correlated to AMC.

Escherichia coli Sorption to Sand and Silt Loam Soil

Interactions between bacteria and soil particles influence bacteria retention and transport in the soil matrix and, consequently, bacterial contamination of water resources. The objective of this study was to quantify sorption and desorption of Escherichia coli to silt loam soil and sand by evaluating suspensions with a range of sorbent and bacteria concentrations. Three sorption isotherms were assessed. Soil or sand sorbent (1, 10, or 20 g) was added to 50 mL of E. coli solution (103, 105, 106, or 107 colony forming units per milliliter, cfu/mL). The bacterial concentration differences between the centrifuged supernatant of E. coli solutions with and without sorbent were used to determine the number of E. coli cells sorbed. The samples without sorbent were used to offset the effects of bacterial die-off, sedimentation, and sorption to surfaces other than that of the sorbent. Blank NaCl solution (0.15 M) was used to wash sorbed E. coli off the sorbent three times in desorption trails. Numbers of E. coli sorbed per gram of sorbent varied with the initial bacterial concentration and quantity of sorbent, and were greater for soil (1.2 × 104 to 9.5 × 108 cfu/g) than for sand (<0 to 3.6 × 107 cfu/g). Freundlich isotherms described observed sorption data well for both sorbents (model efficiency, E = 0.94 for soil and 0.95 for sand) with a greater log Kf (3.26) and a similar nf (1.23) for soil compared to sand (log Kf = 1.68 and nf = 1.22). Langmuir isotherms also described well both sand and soil sorbent data; model efficiency for soil improved upon removal of 20 g sorbent data, perhaps due to near-saturation sorption (99.7%) for these conditions. Model efficiency for the linear isotherm improved when only data within a limited concentration range (<107 cfu/mL) were used. The estimated linear partition coefficient for soil was an order of magnitude greater than that for sand (106 versus 14.5 mL/g). E. coli sorption to both soil and sand particles was reversible, but E. coli detachment from sand was nearly 100% of attached cells after one washing, whereas a total of less than 15% of cells were detached from soil after three washings. Differences in sorption and reversibility between sand and soil will lead to different patterns of retention and transport in the environment for those two media.

Summary of Findings and Recommendations

This book responds to questions in three general areas: characterization of hypoxia; characterization of nutrient fate, transport and sources; and the scientific basis for goals and management options. In the sections below, these questions (shown in italics below) are addressed very briefly with references to those sections of this book where more detailed science on that particular question may be found.

Modeling runoff and sediment yields from combined in-field crop practices using the Soil and Water Assessment Tool

Cropland best management practice recommendations often combine improvements to both tillage and fertilizer application practices to reduce sediment losses with surface runoff. This study evaluated the impact of conventional-till and no-till management practices with surface or deep-banded fertilizer application in sorghum-soybean rotation on runoff and sediment-yield predictions using the Soil and Water Assessment Tool (SWAT) model. The model was calibrated using USDA Natural Resources Conservation Service runoff curve number for antecedent moisture condition II (CNII), saturated hydraulic conductivity, and available water capacity parameters for runoff and USLE cropping factor (Cmin) for sediment-yield predictions for three field plots (0.39 to 1.46 ha [0.96 to 3.6 ac]) with different combinations of practices and validated for three field plots (0.40 to 0.56 ha [1.0 to 1.4 ac]) over a period of 2000 to 2004. Surface runoff calibration required CNII values greater than the recommended baseline values. No-till treatments required slightly greater curve number values than the till treatment, and this difference was similar to that associated with increasing the soil hydrologic group by one classification. Generally the model underpredicted the sediment yield for all management practices. Baseline Cmin values were adequate for treatments with soil disturbance, either by tillage or fertilizer deep-banding, but best-fit Cmin values for field conditions without soil disturbance (no-till with surface-broadcast fertilizer) were 2.5 to 3 times greater than baseline values. These results indicate current model limitations in modeling undisturbed (no-till) field management conditions, and caution that models calibrated for fields or watersheds predominated by tilled soil conditions may not function equally well in testing management scenarios without tillage.

ASSESSMENT OF A GIS–AGNPS INTERFACE MODEL

Data from geographic information systems (GIS) are important for developing input datasets for watershed models. However, the process used to translate GIS data into model–input data can introduce errors. Although GIS data are routinely checked to assure their mapping accuracy, it is far less likely that the derived model–input data are verified. The objective of this project was to assess GIS interface–generated data for the AGNPS watershed model by comparison to baseline data from a manually conducted field survey for the Horseshoe Creek watershed in Kansas. Multiple descriptors were used to assess agreement between datasets. Very good agreement was found for soil parameters (soil texture and erodibility), moderate agreement for land–use parameters (crop and management practice factors), and poor agreement for topographic parameters (slope and slope length). The absence of GIS data layers pertaining to management practices, such as contouring and terracing, reduced the agreement in land–use terms. Poor estimation of slope by the GIS interface model for larger cell sizes resulted in the poor agreement in topographic values. Issues related to appropriate cell sizes, generation of land–management practice GIS coverages, accuracy of GIS coverages, and accuracy of interface algorithms must be addressed by watershed modelers using GIS data.