| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
a Swedish Univ. of Agricultural Sciences, Division of Water Quality Management, Box 7072, S-750 07 Uppsala, Sweden
b Univ. of Maryland, Biological Resources Engineering, College Park, MD 20742
* Corresponding author (Faruk.Djodjic{at}mv.slu.se)
Received for publication March 14, 2001.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
Abbreviations: BMP, best management practice • CSA, critical source area • DSS, decision support system • ES, expert system • GLEAMS, Ground Water Loading Effects of Agricultural Management Systems • GIS, geographic information systems • NPS, nonpoint source • PI, Phosphorus Index
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
The complexity of the NPS control issue at the watershed scale demands simultaneous consideration of many factors, including hydrology, topography, land use, pollutant origin and properties, management practices, and soil properties. In the case of P, regardless of the great number of factors involved in P behavior, pollution occurs only when both P source and an effective transport mechanism are simultaneously present at the same site. Beegle (2000) concluded that "a key concept to managing nutrient pollution effectively is to focus on where these two factors overlap (i.e., where high source is coupled with high potential for transport)." Accuracy in the identification of critical source areas (CSAs), as well as the BMP selection process, depends on our ability to survey and process spatially variable data. Attempting such analysis by hand is very complex (with inherent error in simplifying terms) and time consuming. For these reasons quantitative and qualitative analysis tools are used to ease the process. These tools include geographic information systems (GIS), distributed parameter hydrologic and pollutant transport models, and expert systems (ES).
The Phosphorus Index (PI) provides an opportunity to identify sensitive areas within a watershed. The USDA Natural Resources Conservation Service (NRCS) developed the PI as an assessment tool to quantify and rank the vulnerability of different fields to P losses within a watershed (Lemunyon and Gilbert, 1993). The PI is based on the above-mentioned "P source–P transport mechanism" approach and important factors for P behavior are grouped into two main sets (i.e., P source factors and P transport factors). In the original PI these two sets of factors were simply added, but in the newer, modified versions, the multiplication of these two sets of factors enables identification of the CSAs where P source and P transport factors overlap. The PI as a tool for targeting CSAs has been used in several studies. Gburek et al. (2000) applied the modified PI, augmented with the hydrologic return period concept, on a small watershed within the Susquehanna River basin in east-central Pennsylvania. Heathwaite et al. (2000) applied both P and N indices in the same subwatershed in the Chesapeake Bay basin to identify areas of high and low risk for P and N loss.
The next step after CSA identification is implementation of site-specific BMPs. The choice of BMP is dependent on a site-specific probable cause for P losses and site-specific conditions. Sharpley and Tunney (2000) suggested that BMPs can be grouped into source and transport categories. However, once the CSA is identified and the surveys of its topographic, hydrologic, and soil conditions are made in a GIS environment, we can apply our knowledge and develop sets of rules, which help us to process large series of spatially variable data. These sets of rules are also known as ES. In essence, an ES is a hierarchy of rules that describes the conditions under which a set of low-level constituent information (input data) is tested to prove or oppose a given hypothesis. In our case, the hypothesis is a certain probable cause for P losses (Fig. 1) in a diagnosis ES or BMP to reduce P losses (Fig. 2) in a prescription ES. The diagnosis ES in our case identifies the probable causes for P losses, based on field and soil properties, whereas the prescription ES recommends the most appropriate BMP for P management based on field-specific probable causes and relevant field and soil properties. A rule is a conditional statement, or a list of conditional statements about the values of variables. Rule names are arbitrary, and in this study rule names were chosen to correspond with a given probable cause (Fig. 1) or BMP (Fig. 2). Finally, the conditions are specified requirements defined for each rule that must be fulfilled to prove a given hypothesis. In Fig. 1 and 2 the conditions are preceded by question marks. Abbreviations used to denominate conditions are widely used. For instance, USLE K and LS are the erodibility and length-slope factor of the Revised Universal Soil Loss Equation (RUSLE; Renard et al., 1991).
|
|
The use of PI, ES, and GLEAMS, in conjunction with each other, forms a decision support system (DSS) for NPS pollution control planning at the watershed scale. The usefulness of a DSS is dependent upon its ability to process data into information. This ability is directly related to the amount of qualitative and quantitative knowledge incorporated into the DSS and on the quality and quantity of data that the system is able to store and process. It is important to note that a DSS is an "aid" for decision-making and that it requires understanding and reason from the decision maker(s) to generate worthwhile information, as the information is, in most cases, a reflection of the quality of the input data. Young et al. (2000) developed a DSS that predicts "the likely response of key features of the riverine environment to proposed flow management scenarios." Gilliland (1998) used a DSS including GIS, a distributed hydrologic model, and Prolog scripting language to predict NPS pollution potential in the lower Elkhorn River basin in eastern Nebraska, based primarily on bacterial loads from three different land uses. The output of their DSS compared favorably to that of field observations and to data from previously validated bacterial loading models in the watershed. A practical, computer-based system (EMA) was developed at the University of Hertfordshire to measure environmental performance by comparing the actual farm practices with what is assumed to be the best practice for a given site (Lewis and Bardon, 1998). It acts as both a decision analysis and decision support system and uses a positive–negative eco-rating scale to describe actual farm practices from an environmental point of view.
The overall objective of this study was to develop and apply a multicomponent DSS consisting of PI, diagnosis ES, prescription ES, and a hydrologic–nonpoint-source pollution model (GLEAMS) in order to (i) identify CSAs within a watershed, (ii) make a diagnosis of the probable causes, (iii) prescribe the appropriate BMP, and (iv) test the effects of prescribed BMPs.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
|
|
The Maryland Phosphorus Site Index (Coale, 2000) combines P loss potential due to site and transport characteristics and P loss potential due to management practice and source characteristics. Soil erosion, runoff class, subsurface drainage, leaching potential, distance from the edge of field to surface water, and priority of receiving water are the subcategories of site and transport characteristics used to describe the potential for P removal both by surface runoff and subsurface drainage. All subcategories, except the soil erosion subcategory, are divided into subclasses with a numerical value for each subclass. The soil erosion value is calculated with the Revised Universal Soil Loss Equation. The sum of the subcategory values represents the total site and transport value. Similarly, the management practice and source characteristics (the soil test P fertility index value, P fertilizer and/or manure application rate, and methods of application) are combined to obtain a total management and source value. Multiplying the total site and transport value by the total management and source value results in a P loss rating value (PLRV). Based on PLRV, all fields are divided into four P index rating classes: LOW potential for P movement (PLRV < 600), MEDIUM potential for P movement (600 < PLRV < 1200), HIGH potential for P movement (1200 < PLRV < 1800), and VERY HIGH potential for P movement (PLRV > 1800). IMAGINE 8.4 and Knowledge Engineer were used in this study to produce, compute, and modify all input data layers, as well as to compute and classify fields in PLRV classes. Measured data were input as much as possible and missing data were obtained from default values and the literature (Novotny and Olem, 1994).
Diagnosis ES includes seven different probable causes for P losses: high P level in soil, excessive P fertilization, high soil erodibility, low soil cover, stream proximity, moderate erodibility, and subsurface drainage (Fig. 1). The first two probable causes may be categorized as P source–related and the other five as P transport–related. It is important to emphasize that the rules in the diagnosis ES were developed in such a way that a certain probable cause would be diagnosed only if both P source and P transport criteria were met at the same area of interest (raster, field). For example, subsurface drainage as a transport-related probable cause could be identified as a probable cause only if the high P source condition was also fulfilled (Fig. 1). Similarly, high P level in soil as a source-related probable cause could be identified as a probable cause only if the efficient P transport condition was satisfied. In such a way, diagnosis ES follows the "P source–P transport mechanism" approach and refines PI calculations. "High P source" and "efficient P transport" in diagnosis ES stand for "total management and source value" and "total site and transport value" in PI, respectively.
Prescription ES consists of 10 BMPs aimed to reduce P losses: riparian buffer strips, no tillage, crop rotation, grassed waterway, reduced fertilizer application, P fertilizer incorporation, contour strip cropping, constructed wetlands, terracing, and conservation tillage (Fig. 2). The identification of the proper BMP is based on the results from the diagnosis ES and site-specific conditions such as slope. Both diagnosis and prescription ES were applied only on fields with PLRV higher than 1200 (i.e., fields that were categorized as HIGH PI class).
One of the fields categorized as HIGH PI class with a related probable cause and advised BMP was used as an example in GLEAMS simulations to determine the effectiveness of the prescribed BMP. This field is hereafter referred to as Field 16A. Field 16A is tile-drained and has an area of 3.2 ha, and an average slope of 1.3%. Based on the GIS data and according to the Swedish P-classification system (PAL), it has been classed as the highest soil-P class (>160 mg P kg-1). The soil is characterized as a sandy loam with 17 to 18% clay and 27 to 29% silt, and 2.8% organic matter in the topsoil. The GLEAMS model is a field-scale hydrological model useful in assessing the relative effects of different agricultural management practices on nonpoint-source pollution (Knisel and Davis, 1999). It consists of three submodels: hydrology, erosion and/or sediment yield, and chemical transport. The chemical transport submodel is further divided into nutrient and pesticide components. The GLEAMS model has been used in several applications around the world (Morari and Giupponi, 1997; Gottesbüren et al., 2000; Knisel and Turtola, 2000). Also, it has been successfully tested for Swedish conditions at two different locations (Shirmohammadi et al., 1998; Ulén et al., 1998). The GLEAMS hydrology submodel was calibrated with discharge measurements from a field at the Näsbygard farm, which is situated within the Vemmenhög watershed. The actual measurements of soil physical and chemical properties and management practices were used as input data for the chosen field. The actual input data for eight years (1990–1997) were used in the GLEAMS model simulations and were repeated three times (3 x 8 = 24 yr in total) to examine the long-term effects of applied BMPs. Three different simulation treatments were conducted:
(i) Business as Usual, or continuation of the management practices that were the reasons behind the identification of this field as a critical source area;
(ii) BMP, or application of the BMP from the first year; and
(iii) INTER, or combination of the first two treatments, that is, Business as Usual in the first eight years and BMP application in the following 16 years.
|
RESULTS AND DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
|
|
Diagnosis Expert System
The results of diagnosis ES computations performed on the Vemmenhög watershed for 1995 and 1996 are shown in Table 1. The results are reasonable knowing the P levels in soil, P fertilizer application rates, and the transport mechanisms for P movement in different parts of the watershed. Subsurface drainage was indicated as the largest single probable cause for P losses within the watershed. This is understandable, knowing that the total watershed area is tile drained (Kreuger, 1998). High P losses through subsurface drainage have been measured elsewhere (Gaynor and Findlay, 1995; Heckrath et al., 1995; Beauchemin et al., 1998; Stamm et al., 1998; Simard et al., 2000). However, according to PI, all tile-drained fields were classed within the same risk value class regarding soil drainage class, irrespective of the soil vulnerability to preferential flow losses of P. Stream proximity was a second transport-related probable cause. Lack of buffer zones, zones where no P application occurred, and the extensive culvert system within the Vemmenhög watershed were the main reasons for this. Excessive P fertilization may be somewhat overestimated as a probable cause. Namely, diagnosis ES assumes high fertilizer and/or manure application rates to be annual (due to annual based calculations), although, in reality the common practice is to apply P fertilizer or manure once every three to four years. The spatial distribution of probable causes for high-risk fields in the watershed, combined for both years, is shown in Fig. 5b.
Prescription Expert System
The results of prescription ES computations performed on the Vemmenhög watershed for 1995 and 1996 are shown in Table 1. The spatial distribution of recommended BMPs for high-risk fields in the watershed, combined for both years, is shown in Fig. 5c. The advised BMP pattern resembles the pattern of the probable causes (Fig. 5b). This is understandable knowing that the prescription of BMPs in the prescription ES knowledge tree (Fig. 4) was based on certain probable causes. Consequently, riparian buffer strips were recommended as a BMP for fields where the diagnosed probable cause was stream proximity.
The counter measure (BMP) for subsurface drainage losses was P fertilizer incorporation. This assumption was based on results of a lysimeter study in which P leaching losses were compared when P fertilizer was either incorporated, or applied on the surface of conventionally tilled and nontilled soils (Djodjic et al., 2002). Incorporation of P fertilizer resulted in a significant reduction of P leaching losses during the first year after P fertilizer application, compared with the other two treatments. This was explained by the fact that water percolating through macropores bypassed the soil matrix where the incorporated P fertilizer was located. According to Bergström and Shirmohammadi (1999), macropore or preferential flow was the dominant water and solute transport pathway in structured clay soils, such as the one used in the above-mentioned study. Soils within the Vemmenhög watershed commonly have a lower clay content (Fig. 1b), but preferential flow has been observed in soils with clay contents as low as 12% (Kladivko et al., 1991). Also, Laubel et al. (1999) concluded that preferential flow accounted for elevated dissolved and particulate P concentrations in drainage from a sandy loam soil with clay contents of 15 to 20%. Therefore, it is reasonable to assume that P fertilizer incorporation may have reduced P leaching losses from soils in the Vemmenhög watershed. Finally, reduced P fertilizer and/or manure rate was prescribed on fields where high P level in soils or excessive P fertilization were diagnosed as the main probable causes.
A relatively low number of the diagnosed probable causes (four out of seven) and prescribed BMPs (three out of ten) was probably due to the uniformity of the Vemmenhög watershed in soil properties and topography. It is important to stress that the ES selected the first probable cause–BMP for which all demands (rules) were fulfilled. Therefore, one should keep in mind the importance of the arrangement of the probable causes–BMPs within the ES knowledge tree, both during the development of ES and during the evaluation of the results.
GLEAMS Simulations
From the created PI, probable cause, and BMP maps, a field situated in the central part of the Vemmenhög watershed (Field 16A, Fig. 5c) was chosen for GLEAMS simulations. The hydrology submodel of GLEAMS was calibrated using the discharge measurements from the Näsbygård field (Fig. 5c), situated in the northeastern part of the Vemmenhög watershed. The GLEAMS model overestimated discharge during dry years, but the overall model efficiency, which was estimated with a method proposed by Loague and Green (1991), was high (0.94). The optimum value for model efficiency is 1, whereas negative values indicate poor fit. An explanation for the overestimation might be low ground water levels during dry periods. Part of the root-zone discharge simulated with GLEAMS would, under such conditions, be used for ground water recharge. Also, precipitation measured at the rain gauge located 4 km northeast of the watershed could occasionally be somewhat different from precipitation within the watershed (Hoffmann and Johnsson, 1999), especially during intensive summer rainfall.
Field 16A was categorized in the high PI ranking class in both years. In 1995 the diagnosed probable cause was stream proximity and in 1996, high P levels in soil. Accordingly, the prescribed BMP in 1995 was riparian buffer strip and in 1996 reduced P fertilizer and/or manure application. In this study, the 1996 BMP (reduced fertilizer and/or manure application) was tested in GLEAMS simulations. The excessive manure applications in a Business as Usual scenario were replaced in the other two scenarios (INTER and BMP) with lower rates that were based on crop demand of P (Table 2). Table 3 and Fig. 6 show the amount of P losses during the 24-yr GLEAMS simulation period. The recommended BMP reduced runoff P losses by 55% and sediment P losses by 71% compared with the Business as Usual scenario, if applied from the initial year. If applied after 8 yr (INTER), the reduced manure rate would decrease P losses in runoff by 31% and in sediment by 33%. On the other hand, there was no reduction in P leaching losses.
|
|
|
|
|
CONCLUSIONS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
However, even with the above shortcomings, the DSS is a valuable tool for overview, management, and processing of spatially variable data, and provides us with information and answers in a very short time period. Once the knowledge is developed and the necessary input data is gathered, the analysis time is significantly reduced with the use of DSS. Such a system may help us not only to find the "hot spots" and determine the proper "cure" for certain problems, but also to determine representative sites for closer monitoring. Inclusion of remote sensing and a software interface to connect different parts of DSS is probably the next step in the development. The DSS developed in this study clearly showed that using a combination GIS and NPS pollution model can help us to better allocate our resources for pollution abatement on a watershed scale.
The final test of the DSS described in this study will be its usefulness to regulators, advisors, and farmers in their everyday practice. The suggested improvements combined with a user-friendly interface should facilitate its applications.
|
ACKNOWLEDGMENTS |
|---|
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  D.E. Radcliffe, J. Freer, and O. Schoumans Diffuse Phosphorus Models in the United States and Europe: Their Usages, Scales, and Uncertainties J. Environ. Qual., August 24, 2009; 38(5): 1956 - 1967. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
Y. Chen, R. C. Viadero Jr., X. Wei, R. Fortney, L. B. Hedrick, S. A. Welsh, J. T. Anderson, and L.-S. Lin Effects of Highway Construction on Stream Water Quality and Macroinvertebrate Condition in a Mid-Atlantic Highlands Watershed, USA J. Environ. Qual., June 23, 2009; 38(4): 1672 - 1682. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
B. Kronvang, M. Bechmann, H. Lundekvam, H. Behrendt, G. H. Rubaek, O. F. Schoumans, N. Syversen, H. E. Andersen, and C. C. Hoffmann Phosphorus Losses from Agricultural Areas in River Basins: Effects and Uncertainties of Targeted Mitigation Measures J. Environ. Qual., November 7, 2005; 34(6): 2129 - 2144. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
W. J. Dougherty, N. K. Fleming, J. W. Cox, and D. J. Chittleborough Phosphorus Transfer in Surface Runoff from Intensive Pasture Systems at Various Scales: A Review J. Environ. Qual., November 1, 2004; 33(6): 1973 - 1988. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| The SCI Journals | Agronomy Journal | Crop Science | |||
| Journal of Natural Resources and Life Sciences Education |
Vadose Zone Journal | ||||
| Soil Science Society of America Journal | Journal of Plant Registrations | The Plant Genome | |||
