The data used in this study can be divided into four groups:
The data utilized in this study has been extracted from two USGS publications:
USGS Open-File Report 94-396 contains data that were collected from nine streams in five Midwestern States during late spring-early summer of 1990. Two of the nine streams (the Sangamon River and the West Fork Big Blue River) were sampled from April 1991 to March 1992. Table 3.1 contains detailed information about number of samples and the time interval during which samples were collected.
Table 3.1 Description of atrazine and nitrate plus nitrite as nitrogen samples published in USGS Open-File Report 94-396, (Scribner et al., 1994).
___________________________________________________________________________ Drainage No of samples No of samp. days _________________________________ Name Station Area Sampled Atrazine NO2+NO3 Atrazine NO2+NO3 ID km² Period (GCMS) as N (GCMS) as N ___________________________________________________________________________ W. Fork Big 6880800 3123 4/03/90-7/27/90 37 176 28 57 Blue R.,NE 3/26/91-3/09/92 106 96 64 63 Sangamon R,IL 5572000 1425 4/04/90-8/23/90 53 255 43 101 4/12/91-3/25/92 198 168 93 93 Huron R.,OH 4199000 961 3/30/90-8/20/90 59 208 53 142 Delaware R,KS 6890100 995 4/04/90-6/29/90 32 154 25 45 Roberts Cr,IA 5412100 267 4/03/90-7/21/90 22 133 20 43 Old Man's,IA 5455100 521 4/16/90-7/02/90 49 171 30 46 Cedar R,IA 5431200 12261*4/15/90-6/29/90 41 187 33 60 Silver Cr,IL 5594800 1202 4/13/90-8/17/90 34 120 29 59 Iroquois R,IL 5526000 5416 4/04/90-8/18/90 49 177 46 71 ___________________________________________________________________________ Total: 680 1845 464 780 ___________________________________________________________________________ * for analysis a drainage area estimated from DEM has been used (17409 km²)
Figure 3.1 Location of sampling sites for study of atrazine and nitrate plus nitrite as nitrogen concentrations in storm runoff; USGS Open-File Report 94-396, (Scribner et al., 1994)
Concentrations of eleven herbicides, two triazine metabolites, nutrients, streamflow, specific conductance, and pH were measured. Location of each sampling site is described by latitude and longitude. Such estimates as cropland percentage, atrazine use, and nitrogen use are also presented. Figure 3.2 presents an example of measured atrazine concentrations in the Old Mans Creek in Iowa.
Figure 3.2 Atrazine Concentrations in the Old Mans Creek, Iowa (data from USGS Open-File Report 94-396, Scribner et al., 1994).
In Open-File Report 93-457 Scribner et al. (1993) present data for selected herbicides, two atrazine metabolites, and nitrate plus nitrite as nitrogen in surface water of the Midwestern United States, 1989-90. One hundred and forty seven Midwestern streams were studied. Two to three samples per site were taken annually--before application of herbicides, during the first major runoff after application of herbicides, and in the fall during a low-flow period when most of the streamflow was derived from the ground water. Table 3.2 presents the distribution of samples within studied period, whereas Figure 3.3 shows the reconaissance study drainage areas.
Table 3.2 Temporal distribution of atrazine and nitrate plus nitrite as nitrogen samples from the Midwestern reconnaisance study; USGS Open-File Report 93-457, (Scribner et al., 1993).
_____________________________________________________ Number of samples __________________________________________ Month Atrazine (GC/MS) NO2 + NO3 as N 1989 1990 1989 1990 _____________________________________________________ MAR 64 38 95 38 APR 25 10 48 10 MAY 69 36 72 37 JUN 51 13 51 13 JUL 11 5 12 5 AUG 1 - 1 - SEP 1 - 1 - OCT 114 - 114 - NOV 30 - 32 - _____________________________________________________ Total: 366 102 426 103 _____________________________________________________
Atrazine concentrations are reported as two different (but related ) components: ELISA atrazine and GC/MS atrazine, since this chemical is usually measured by two methods: ELISA ( Enzyme-Linked Immunosorbent Assay) and GC/MS (gas chromatography/mass spectrometry). The ELISA test is easier to perform but is less precise than the GC/MS method. It is affected to varying degrees by the existence of propazine, prometon, simazine, deethylatrazine, cyazine, and deisopropylatrazine (Thurman et al., 1990). ELISA and GC/MS results can be related by regression so that results determined by ELISA can be converted to an estimate of what a GC/MS measurement of the same sample would have yielded (Thurman et al., 1992, Moody and Goolsby, 1993, Goolsby et al., 1993). In this study the atrazine concentrations measured by the GC/MS method are utilized.
Figure 3.3 Watersheds sampled during 1989 and 1990 in the midwestern United States (Battaglin, 1995).
Elevation data form the spatial framework for modeling basic hydrologic processes. In this research, a Digital Elevation Model is used not only to determine the paths of travel of agricultural chemicals, but also to divide the study region into small watersheds--modeling units, to determine the "flow" topology between the modeling units, delineate the stream network and to determine the characteristics of the drainage area. Two versions of DEM are utilized: 1-degree (3-second) DEM and 500 m (15 second) DEM.
The Digital Elevation Model (DEM) consists of a 2-D array of ground positions at regularly spaced intervals. One-Degree DEM (3x3 arc-second data spacing) are generally available throughout the US and therefore they have been used in this research. The majority of 1-Degree DEMs were produced by Defense Mapping Agency (DMA) either from cartographic (maps 1:24000 scale through 1:250000 scale) or photographic sources. Some of these DEMs were created by regriding 7.5-minute and 30-minute DEMs. The compressed (and uncompressed) DEM files are available via the Internet at ftp://edcftp.cr.usgs.gov/pub/data/DEM/250/ . Needed maps can also be selected from the map of the USA that is displayed on the screen: http://edcwww.cr.usgs.gov/glis/hyper/guide/1_dgr_demfig/index1m.html
The 500m DEM has been created by resampling 1-Degree DEM. It has been released by the USGS on the CD-ROM ( Rea and Cederstrand, 1995). In this research 500 km DEM has been applied to create a map of the drainage area of each sampling site from which measurements have been utilized for model development. Although, for some regions the quality of determined flow direction have been poor and some editing had to be done, the 500 m DEM appeared to have an optimal resolution. A the cost of precision, the Mississippi-Missouri-Ohio River basin of area almost 3*106 km² could be treated as a one processing unit, that was impossible for 100 m DEM. Since the size of this basin grid is about 90 Mb, the complexities of the analysis is restricted to the available computer power.
Figure 3.4 Digital Elevation Model of the Iowa-Cedar River basin, vertical scale enlarged 500 times. The white line represents the basin boundaries.
Reach File 1 (RF1) is a representation of streams in the conterminous United States at a scale of approximately 1:500,000. The original file was prepared by the US Environmental Protection Agency (USEPA). In 1994 it was translated on the mainframe computer into an Arc/INFO coverage. RF1 in Arc/Info export format is available via Internet: http://nsdi.usgs.gov/nsdi/wais/water/rf1.HTML Because of the size of the entire RF1 data set . (Arc/INFO Export file occupies 57 MB), it also is available in 18 separate files, each covering a 2-digital hydrologic unit code (Water Resources Region).
In this research the RF1 is applied to adjust the DEM in order to ensure that the streams delineated from the DEM are compatible with the streams from RF1. Figure 3.5 shows an example of RF1-the Cedar River above Waterloo, Iowa.
Figure 3.5 The Cedar River above Waterloo, IA; rivers selected from Reach File 1.
In 1995, the USGS published maps of herbicide and nitrogen-fertilizer use (Battaglin and Goolsby, 1995a, b). Five coverages summarizing the use of 96 herbicides were constructed from the tabular estimates of herbicide use by county and by crop published in Gianessi and Puffer (1991). Atrazine use in the Mississippi-Missouri and Ohio River basins in 1989 extracted from the USGS coverage is presented in Figure 3.6.
Figure 3.6 Atrazine use in the Mississippi-Missouri and Ohio River basins in 1989 (from Battaglin and Goolsby, 1995b)
Although the crop acreage used in the assessment were from 1987, the herbicide use estimates generally reflect 1989 usage amount. Data from such sources as surveys of weed scientists, surveys of farmers, and crop acreage, were utilized. Estimates of the number of pounds of atrazine used per square mile in 1989 are contained the Arc/Info coverage HERBICIDE1, the attribute H1980.USE.
Battaglin and Goolsby (1995a) constructed seven coverages summarizing annual nitrogen sales in U. S. counties, for the fertilizer years 1985-1991. The fertilizer year starts July 1 of the previous year and ends June 30, e. g. fertilizer year 1990 starts 07/01/89 and ends 06/30/90). The maps were prepared from estimates reported by the U. S. Environmental Agency (years 1985-89) and from Jerald Fletcher in cooperation with the National Fertilizer and Environmental Research center, Tennessee Valley Authority (years 1990 and 1991). The sales of fertilizer do not account for the use of manure. Total nitrogen-fertilizer use in tons per square mile from July, 1989 to June, 1990 is stored in attribute NTOT90 of coverage NIT90; nitrogen-fertilizer use from July, 1990 to June, 1991 is in attribute NTOT91 of coverage NIT91.
Coverages of agricultural chemical use estimates are available in Arc/Info export format from National Spatial Data Infrastructure (NSDI), National Geospatial Data Clearinghouse: http://nsdi.usgs.gov/nsdi/wais/water/herbicide1.HTML (.../nit89.HTML, .../nit90.HTML .../nit91.HTML).
The gauging station locations, drainage areas, and the daily flow data for 31,000 USGS gauging stations are available from the CD-ROM published by the Hydrosphere Data Products, INC. (Hydrosphere, 1993a). These data originate from the USGS WATSTORE system (Daily and Peak Values Files For Stream Flows).
Additional data such as daily rainfall, snowfall, maximum temperature and the minimum temperature, which were observed at 17,000 NCDC Stations, are also available on CD-ROM published Hydrosphere Data Products, INC (Hydrosphere, 1993b).
For the purpose of this research the following data have been exported Hydrosphere CD_ROMS in ASCII (text) format:
Section 3.2.1 presents the Geographic Information System software used in this research. Section 3.2.2 introduces statistical program, S-Plus, used to develop a regression model. The research was performed on SUN SPARC station IPX (operation system UNIX) and 486DX2-66 (operation system DOS).
Arc/Info and ArcView constitute the GIS software used in this research. (Arc/Info and ArcView are registered trademarks of the Environmental Systems Research Institute, Inc., Redlands, California) The Arc/Info is a spatial analysis system which represents spatial data in separate layers and it provides operators for manipulating these data. It contains three basic spatial primitives for vector data: Points, Lines, and Polygons. This software supports also three derived data structures: Grids (a rectangular mesh of points), Triangulated Irregular Networks (a TIN is a set of points connected by triangles), and Networks (a set of connected arcs with assigned flow properties). Each spatial primitive can have an associated record in a database, an Info file. The fields of this record contain user specified descriptive attributes, such as area, length, category, name, etc. The one to one correspondence between the spatial features (point, arc or polygon) and data records (Info) is the basis of the hybrid Arc/Info data model.
Besides the Arc/Info core system, two Arc/Info processors are extensively used in this research: GRID and TABLES. The GRID manipulates maps in raster format, grids. The TABLES are used to handle the data stored in the following attribute tables:
Arc/Info macro language (AML) enables the automation of complex or repeated tasks.
ArcView is a GIS software completely operated from the graphical user interface. Although it can perform only simple spatial operations on the maps in vector format, it has very powerful and convenient tools to manage attribute tables. Avenue, an object oriented script language, allows the user to build complex GIS applications. ArcView is available for workstations and for personal computers. ArcView version 2.1 was used in this research (ESRI, 1995).
The data analysis and model parameter estimation were performed by the application of two computer programs: Microsoft Excel spreadsheet version 5.0a (Microsoft and Excel are registered trademarks of Microsoft Corporation) and the statistical software, SPlus version 3.2 (S-PLUS is registered trademark of Mathsoft, Inc.). MS Excel was used mainly as a preprocessor for data editing, simple calculations such as unit conversion, and data verification. All the statistical analysis which is presented in this project, was performed using S-Plus.
S-Plus is an interpreted language that evaluates expressions. The results of the expression is an object. All input data must be one of the S-Plus objects. Two types of objects has been utilized in this research:
There are two more objects that are designed to store tables,
array and matrix, but the data frame, unlike the matrix, can have
values of different modes in different columns (MathSoft, 1993,
Venables and Ripley, 1994). A detailed description of the application
of S-Plus in this study is presented in Section 5.5.