Research Paper 2: Data Quality for Weed Control and Groundwater Protection Project Matt Stuemky Geography 587 - GPS/GIS Field Techniques Dr. John Wilson University of Southern California Geographic Information Science & Technology Program August 28, 2009 I. Introduction In the early 1990's, researchers from Montana State University conducted a GIS-based study designed to assess the likelihood that long-term application of herbicides such as picloram to weed infested areas would result in groundwater contamination (Wilson et al., 1993). Agricultural areas in Montana are often affected during the growing season with noxious weeds such as leafy spurge and spotted knapweed. The study area selected for their research was Teton County, one of the top crop-producing areas in the state of Montana. The first part of this report critiques key aspects of that original study, including the data sources selected, the methods and model used, and the final results produced. The impetus for assessing the original study is to identify some of its limitations in order to develop a proposal for a new GIS, one that will provide an alternative approach to mapping the relationship between noxious weed treatment activities and groundwater contamination throughout Montana. The new project proposes to incorporate some of the successful aspects of the earlier study but also relies on a number of new data sources and uses a completely different model. II. Original GIS project: data quality issues, potential sources of error The original research project utilized various soil, weather and chemistry related data sources as inputs into the Chemical Movement in Layered Soils (CMLS) model. The CMLS model is a one-dimensional solute transport model, which utilizes a "piston flow theory" (Wilson et al., 1993) designed to "estimate the movement of chemicals in soils in response to downward movement of water" (Oklahoma State University, Department of Plant and Soil Sciences, 2009). The output of the CMLS model runs, which estimates the depth of movement of chemicals at the end of the growing season, were overlaid in ARC/INFO with the Teton County weed maps (specifically, leafy spurge and spotted knapweed) to determine where groundwater was likely to be contaminated. The final hazard maps that were produced depicted three sets of CMLS model predictions including "worst" case (most movement), "average" case (weighted average), and "best" case (least movement) over 15 weather years. Model runs for each year were confined to Montana's growing season months only, which run from April to October of each year. Seven specific topics related to the original study conducted by Wilson et al. (1993) are critiqued: the CMLS Model, soils data (STATSGO), weather data (MAPS and WGEN), noxious weed coverage data, field collected data, roads data, and geographic scale. P a g e 1 | Matt Stuemky – Research Paper 2 – GEOG 587 (Summer 2009) CMLS Model At the time it was used in the original study, CMLS was a rather popular simulation model used to estimate the movement of chemicals through soil layers over time. The basic algorithm divides the soil into 20 layers. One of the shortcomings that seems evident, and one that is a potential source of error in the predicted results, is that the model assumes that various soil properties (e.g. texture, bulk density, organiccarbon content, etc.) to be uniform within each layer (Wilson et al., 1993). While the CMLS model had a favorable reputation at the time it was used, and it seems to be relatively easy to work with, it is a considerably simple solute transport model compared to other models due to the fewer number of inputs required. Other process-based methods such as the Generalized Preferential Flow Transport Model (GPFM) used in other studies (Sinkevich et al., 2005) use land cover data and other inputs. The results produced by the CMLS model used in the original study demonstrated that climate and soil data with a high spatial resolution are needed in order to more accurately map the locations where chemical components (such as those found in picloram) are likely to result in groundwater contamination. A modified version of the CMLS model was central to the original study, but the model seems to be limited, in part due to its allowance of "lower resolution" soil and weather input parameters (Inskeep, 1996), compared to other models that similarly predict the mean travel times and leaching behavior of chemicals through soil. Inskeep (1996) described how the results of using the LEACHAM model demonstrated the CMLS model's somewhat "oversimplification" of the solute transport process such as the temporal and soil depth effects on ET. Although LEACHAM requires more input parameters, it seemed to predict considerably slower mean travel times (solute transport) compared to the CMLS model. Primary data sources Soils data The USDA-SCS State Soil Geographic Database (STATSGO) was used in the original study. It is a statewide coverage of soil series compositions, dividing the Montana landscape into polygon map units. At a scale of 1:250,000, the STATSGO coverages were generalized from more detailed soil survey maps. Based on the results described in follow-up studies (Wilson et. al, 1996; Inskeep at al., 1996), the use of this relatively coarse scale data source in the original study produced less accurate results from the model and, as a consequence, did not always precisely map the areas where groundwater contamination risk was highest. It seems evident that one of the primary shortcomings of the original study was the decision to use the STATSGO coverage. Later research seemed to indicate that STATSGO was somewhat too coarse to adequately identify all locations where the depth of solute movement (e.g. picloram) exceeded the rooting depth (Wilson P a g e 2 | Matt Stuemky – Research Paper 2 – GEOG 587 (Summer 2009) et al. 1996). Additionally, the soil data derived from STATSGO lacked measured bulk density values for some records, which then had to be estimated. In a later studies, the data quality issue was remedied in part by using the Soil Survey Geographic (SSURGO) database (Wilson et al. 1996). SSURGO provides a finer, countylevel spatial resolution of soil survey data, which also contains soil properties such as bulk density, texture and surface organics (Inskeep et al., 1996). Based on the results produced, the use of SSURGO data seemed to more successful identify areas with a high likelihood of chemical transport below the root zone (Inskeep et al. 1996). Weather data For weather and climate data, the Montana Agricultural Potentials System (MAPS) database was used. It is a statewide coverage containing approximately 18,000 20-square kilometer cells. MAPS cell boundaries that encompassed Teton County were clipped and converted into a climate polygon coverage that consisted of 324 individual MAPS cells. Attribute tables containing mean monthly precipitation totals and temperatures were also extracted. The WGEN weather simulator and FORTRAN programs were used to produce the daily precipitation and evapotranspiration (ET) totals required by the CMLS model. It seems that the decision in the later study (Wilson et al. 1996) to use the ANUSPLIN program resulted in finer-scale (and better quality) climate data compared to the inputs created for CMLS model in the original study, which relied largely on estimated climate data for the 324 MAPS cells covering Teton County. In comparison, ANUSPLIN calculations for climate data (e.g. monthly precipitation, annual precipitation and temperature) were derived from Digital Elevation Models (DEMs) data, which factored in attributes such as latitude, longitude and elevation combined with published climate data from 1961 to 1990. The original study seemed to rely more heavily on estimated values (generated by WGEN), except for some National Weather Service data from two climate station sites, which including one in Great Falls which is located outside the study areas inside Teton County (Wilson et al., 1993). In other words, there was a higher level of uncertainty in the weather data produced from the original study. Noxious weeds coverage data In addition to the limitations of the CMLS model and the soil and weather data sources used, another problem with the original study was that the final estimations of herbicide chemical movement (solute transport) through soil layers. It only provided a "partial view" for identifying areas of potential groundwater contamination over time (Wilson et al., 1993). More specifically, the predictions did not fully account for geographic reality, such as the fact that noxious weeds like leafy spurge (among others) do not uniformly cover the ground in the same density over a given soil type. This level of uncertainty, which cannot be P a g e 3 | Matt Stuemky – Research Paper 2 – GEOG 587 (Summer 2009) eliminated completely, reduces the accuracy of predicted results identifying areas that are "best" (least movement) and "worst" (most movement) for herbicide chemical transport through soils. For the agricultural regions of Montana, perhaps the use of higher resolution soil series data (such as SSURGO), combined with remotely sensed based land use/land cover (LULC) data would provide a means to more precisely measure groundwater contamination risks in areas covered with noxious weeds. In recent years, remotely sensed imagery has been used extensively in GIS-based studies of groundwater contamination (Werz & Hötzl, 2007). Additional data sources The original study may have produced hazard maps with potentially more accurate results if additional data sources were utilized somehow, although there were undoubtedly time and money constraints to the project, as well as limitations placed by the decision to use the CMLS model as the basis for the GIS. Several additional data sources were already mentioned previously, as they were utilized in the follow-up studies conducted by Wilson and Inskeep (Wilson et al., 1996; Inskeep et al., 1996), including the use of higher resolution SSURGO soils data and Digital Elevation Model data (for latitude, longitude and elevation attributes) in conjunction with climate/weather data . The different results produced by the later studies seemed to demonstrate that the higher resolution data combined with additional inputs produced more accurate results. Field collected data It is well-established that groundwater contamination from agriculture, whether from pesticide use or other activities is a considerable problem, and that some type of direct monitoring is essential. Groundwater monitoring typically focuses on field collection activities such as taking soil samples from the ground in and around agricultural areas and by taking water samples from irrigation and other type wells. However, such monitoring is expensive and time consuming, especially when conducted in the course of obtaining data for a specific project that has both time and money constraints, or when the study area in question is particularly large (Bazimenyera & Zhonghua, 2008; Piscopo, 2001). However, one strategy is to perform monitoring activities on the highest risk areas only, thus requiring fewer observation points (e.g. water wells) for a study (Sinkevich et al., 2005). The original study conducted by Wilson did not directly incorporate any field collected data in order to evaluate the presence of pesticide chemicals within specific soil types. Although expensive, this could have been done during the course of the research project. It should be mentioned that later groundwater contamination studies did in fact make use of field data: measurements of herbicide contents in soil samples taken at a specific irrigation sites were used to compare the observed data with predicted data inputs used by P a g e 4 | Matt Stuemky – Research Paper 2 – GEOG 587 (Summer 2009) the LEACHM model (Inskeep et al., 1996). Making use of field data from a small number of carefully-selected collection points is one approach. Other approaches that could have been considered would be to make use of existing field data collected from other research projects, or published datasets derived from ongoing, longterm field collection activities performed by local or state government agencies. Of course, one of the main benefits of incorporating field data into a study is to "ground truth" the quality of the primary data sources used and to validate the results produced. Roads data In the original study, the use of roads as a data source was mentioned, to be considered for future research (Wilson et al., 1993). However, in the later studies conducted by Wilson and Inskeep it did not seem to be incorporated in any way. Upon reflection, the omission of roads data, while not critical to the outcomes produced using the CMLS model itself, nevertheless seems significant enough that it should probably have been more strongly considered for inclusion into GIS-based groundwater contamination research projects. Herbicides are commonly applied in areas along roadways and especially when considering the locations of unpaved roadways (e.g. local county roads, farm roads, etc.) which provide direct access to the ground surface. Picloram and other weed control products would likely accumulate over time as vehicles used to apply the weed control products traverse the same roads over and over again. Groundwater contamination would likely be higher in areas concentrated under and adjacent to these roads. Incorporating roadways as an input to the GIS and performing some type of spatial analysis techniques, including the use of a buffering operation along selected roadways, seems like it would provide a means to further assess the highest risk locations for groundwater contamination. Geographic scale The use of the MAPS and STATSGO databases as inputs provided a level of precision that was well suited to county-level study areas. However, it seems that a more flexible GIS was needed, one that is capable of assessing groundwater contamination risk as a result of weed control measures at multiple scales, from broad overviews that are county-wide or statewide to more focused, large-scale (1:24,000) study areas. The original study seemed limited both by the data sources used and the nature of the CMLS model itself, which prevented the GIS from being able to scale up or down much beyond what was attempted for selected areas of Teton County. This seemed to be confirmed by Inskeep when he stated that uncertainties in the quality of data sources, the variability of soil characteristics over time and location, and the "simplification of deterministic models" such as CMLS, makes it particularly challenging to apply a GIS such as this for larger geographic scales (Inskeep et al., 1996). P a g e 5 | Matt Stuemky – Research Paper 2 – GEOG 587 (Summer 2009) III. Proposal for a new GIS project Over the past 25 years, numerous approaches have been developed for assessing the vulnerability of groundwater caused by human activities at or near the surface of the land. Three broad categories have emerged for assessing groundwater contamination risk: (1) overlay and index methods, (2) process-based simulation models, and (3) statistical methods (National Research Council, 1993). The earlier studies by Wilson and Inskeep (Wilson et. al. 1993, 1996; Inskeep et al., 1996) used modified versions of CMLS and LEACHM, which are process-based simulation models that focus more on transport times of specific chemicals through the soil. Rather than use a process-based simulation model such as CMLS, this new project proposal will instead be based on an overlay and index method, using the DRASTIC model. The project described here is a GIS-based study that will be designed to map and analyze groundwater contamination risk for any county within the state of Montana that implements noxious weed control measures. For the pilot phase (first six months) of the project, the top five crop-producing counties in Montana will be selected as study areas, including Teton County. Results for Teton County specifically will be compared to the results from various earlier GIS-based groundwater assessment studies of that area, including those conducted by Wilson and Inskeep. The DRASTIC model: advantages DRASTIC is one of the most well-known and widely used groundwater vulnerability assessment methods used. DRASTIC was developed by the U.S. Environmental Protection Agency (see Aller et al. 1985, 1987). DRASTIC is an acronym that is based on seven parameters used by the model: D - depth to water table, R - recharge rate, A - aquifer media, S - soil type, T - topography, I - impact of vadose (unsaturated) zone, and C - conductivity (hydraulic) (Piscopo, 2001; Remesan & Panda, 2008). The DRASTIC model offers several advantages. First, the model uses relatively simple mathematical operations rather than relying on complex formulas to produce the index values that can be used to assess risk levels. Second, as with other overlay and index methods, DRASTIC was developed to take advantage of the availability of mapped data, with less emphasis on modeling "processes" affecting ground water contamination (such as with the CMLS and LEACHM models used by Wilson and Inskeep). DRASTIC is particularly well-suited for implementation into a GIS because it is designed to perform qualitative compilations and interpretations from mapped data (National Research Council, 1993). The abundance of spatial data from numerous state and federal government agencies as well as other organizations, including soil, geological, precipitation, land use/land cover, Digital Elevation Models (DEM), etc., allow a GIS-based implementation of the DRASTIC model to be more straightforward compared to most process-based and P a g e 6 | Matt Stuemky – Research Paper 2 – GEOG 587 (Summer 2009) statistical-oriented approaches. Third, whereas the CMLS model used by Wilson is a fairly scale-dependent model, suited for large-scale mapping, index and overlay models such as DRASTIC generally offer an easier implementation for study areas of variable mapping scales (National Research Council, 1993). Finally, the DRASTIC model allows for adaptations, that is, the use of additional (and/or alternative) "custom" parameters as data inputs, in addition to the seven primary ones. For this GIS project, which is designed to assess groundwater contamination brought on by weed control measures, two additional parameters will be used as inputs into the DRASTIC model: one for roads and one for noxious weed coverage. The DRASTIC model: disadvantages Despite the advantages offered, the DRASTIC model shares the same problems of uncertainty that affect many GIS-based groundwater contamination studies. In fact, the level of uncertainty can be higher with index and overlay methods such as DRASTIC compared to other models given the variety and amount of data to be used. By making use of data from different providers, even if the assessment method being used is valid, poor quality source data will be a considerable source of error in the final results (i.e. the risk maps produced) for the groundwater assessment. In 1993, the National Research Council published a book titled Ground Water Vulnerability Assessment in which the topic of uncertainty is addressed at great length. Regarding uncertainty related to data quality "...problems can be reduced by making sure the variability in the attributes over a study area is accurately reflected in the interpolated values of both the spatial and nonspatial attributes used" (NRC, 1993, pp. 104-105). Another element of uncertainty relates to the fact that the DRASTIC model doesn't directly provide a means to predict actual chemical concentrations and their movement through soil as do CMLS, LEACHM and other process-based models. Some researchers contend that this is one of the key failings of index and overlay models used for assessing ground water contamination (Sinkevich et al., 2005). The DRASTIC model: assumptions, index, parameter weights and ratings There are four fundamental assumptions characteristic of the DRASTIC model: 1) any contaminant (e.g. pesticide) is introduced at the ground surface, 2) precipitation is the mechanism for moving the contaminant into the ground, 3) the contaminant is able to move through the ground with the same mobility as water, and 4) the study area must be 100 acres or larger (Remesan & Panda, 2008). As indicated above, there are seven primary parameters used for the DRASTIC model. Two additional "custom" parameters will also be used for the GIS project. P a g e 7 | Matt Stuemky – Research Paper 2 – GEOG 587 (Summer 2009) Weight values from 1 to 5 are assigned for each parameter used. A benefit of the DRASTIC model is its flexibility to adjust the relative weight of the seven primary parameters for a particular application, such as evaluating the effect of pesticide contaminants. For example, earlier GIS-based studies using the DRASTIC model, such as those done by Piscopo (2001) involved adjusting the relative weights (in addition to using additional "custom" parameters) to emphasize a particular contaminant of concern. For pesticide related studies such as that conducted by Remesan and Panda (2008), the relative weights from the generic version of the DRASTIC model were adjusted higher for the soil type and topography parameters; they also lowered the vadose zone impact and hydraulic conductivity parameters. Since weed control measures are the focus of this GIS project, the DRASTIC model will use a modified sets of weights for these same four parameters. Ratings from (generally from 1 to 10) are assigned based on the attribute values of the data used for each of the DRASTIC model parameters. For example, the aquifer media parameter will use Geology data; the value of each rock type will be given a rating based on its "porosity" (shale = 2, sandstone = 5, karst, limestone = 10, etc.). Ratings must be assessed for all attribute data selected for use as inputs for the parameters. The DRASTIC Index (DI) value, which is calculated for each mapping unit, is produced by multiplying the weight and rating values together for each parameter, and then adding up all the sum values. A higher DI value equates to a higher likelihood of groundwater contamination; conversely, a lower DI means a lower likelihood of contaminated (Ramesan & Panda, 2008; Piscopo, 2001). For this GIS, risk assessment maps will be based on the calculated DI values, to be used as indicators of water contamination risk caused by application of weed control herbicides. The DI final values will classified into ranges with descriptors indicating low, medium, high, and very high risk. The DRASTIC model: data sources to be used for the nine input parameters The data sources that will be used for building the geodatabase for the GIS are described below in the context of how specific attributes will be used as inputs for the various parameters used by the DRASTIC model. The data input and rating level assignment approaches for the seven primary parameters are based on techniques used in recent GIS-based studies using the DRASTIC model, including Bazimenyera and Zhonghua (2008), Ramesan and Panda (2008), and Piscopo (2001). One data source should be described here first because it describes the primary mapping unit to be used: it is the MAPS Atlas database, created and maintained by Montana State University. This is an updated version of MAPS that was used by Wilson and Inskeep in their earlier studies. The MAPS Atlas database divides the entire state of Montana into a grid of 18,000 individual rectangular shaped (2-mile by 3-mile) cells. Each cell (or polygon) feature contains over 150 attributes describing aspects of land and climate. Polygons P a g e 8 | Matt Stuemky – Research Paper 2 – GEOG 587 (Summer 2009) based on the MAPS Atlas will be used as the primary mapping unit because they offer a high degree of spatial resolution. For Teton County, DRASTIC index (DI) values will be calculated for each of the 324 individual MAPS Atlas mapping units in order to construct a complete risk assessment map of the entire county. Refer to Figure 1 below for an overview of the architecture of the proposed GIS, including data sources selected, the geodatabase to be created, the nine parameter inputs to be used for the DRASTIC model, and the risk assessment maps that will be produced using ArcGIS software. 1. Depth to Water Table Depth to water table (DTWT) is the distance in which a contaminant must travel from the ground surface in order to reach groundwater. The deeper the water level, the longer the travel time for contaminants. To obtain attribute data for the DTWT parameter rating, field-collected water well data that is maintained by the Groundwater Information Center, Montana Bureau of Mines and Geology will be used. For example, over 2000 water wells are located in Teton County. The Groundwater Information Center database contains spatially-referenced point data indicating DTWT values for the current year as well as previous years. 2. Recharge Using techniques described by Piscopo (2001), estimated recharge rate can be calculated by adding together slope percentage, soil permeability, and precipitation. This calculated value is used for the Recharge parameter rating. In general, the greater the recharge rate the greater the potential for groundwater contamination. Slope percentages will be derived from Digital Elevation Models available from the U.S. Geological Survey. Soil permeability values will be derived from the Montana Soil Survey Geographic (SSURGO) database. Average annual precipitation values will be derived from one of two data sources: actual precipitation data is available from the USDA Natural Resources Conservation Service (NRCS) which maintains a database containing spatially-referenced point data from weather and climate stations. Estimated annual precipitation data is also available from the MAPS Atlas database, which can be used if actual values are not available for a given mapping unit (MAPS cell); in many cases, estimated values will need to be used. 3. Aquifer Media Aquifer media affects the route and path that groundwater flows. It is defined by geology and characteristics such as fracturing and porosity (among others) associated with specific rock formation types including sand/gravel, alluvium, shale, sediments, sandstone, limestone, glacial till, basalt, igneous, metamorphic, etc. Aquifer media type attribute data can derived from polygons that make up the 1:100,000 series geologic formation maps for Montana. This information will be obtained from the U.S. Geological Survey's National Geologic Map Database. P a g e 9 | Matt Stuemky – Research Paper 2 – GEOG 587 (Summer 2009) Figure 1 - DRASTIC model based GIS to assess groundwater contamination risk caused by noxious weed control measures P a g e 10 | Matt Stuemky – Research Paper 2 – GEOG 587 (Summer 2009) 4. Soil Type Soil has a significant impact on the recharge rate of groundwater. Soil texture also contributes to the movement of contaminants downward from the ground surface. Fine-textured materials such as silt and clay make for decreased soil permeability whereas sand and gravel increase permeability and more easily facilitate the movement of contaminants such as pesticides to groundwater. The SSURGO database will be used as the data source to obtain soil type attribute values. The Soil Type parameter rating for the DRASTIC model will be a range from low to high (1 to 10) based on the soil permeability characteristics of the soil type. 5. Topography Flatter topography results in minimal contaminant runoff on the ground surface but it allows contaminants to more easily infiltrate into the groundwater sooner. Steeper slopes obviously promote runoff, which also affects soil transport, allowing contaminants to be moved from one location to another before sinking down into the groundwater. As with the Recharge parameter, slope percentage values can be derived from DEMs obtained from the U.S. Geological Survey. Slope maps can be created using any one of various commercial GIS software packages to extract discrete slope values (either in degrees or percent), which can then be stored in an attribute table to be used for the Topography parameter required by the DRASTIC model. 6. Impact of Vadose Zone The vadose zone is the zone above the water table which is "unsaturated or discontinuously saturated" (Piscopo, 2001, p. 6). The two factors considered for defining the impact to vadose zone are soil permeability and depth to water table (DTWT). Soil permeability is based on the soil type, which will be obtained from the SSURGO data. A soil permeability value will be assessed with a factor ranging from a value of 1 (very low) to 5 (high). DTWT data will be obtained from well water data obtained from the Groundwater Information Center. A DTWT value will be assessed with a factor ranging from a value of 1 (greater than 20 meters) to 5 (less than 5 meters). The Impact of Vadose Zone parameter's rating for the DRASTIC model will range from 1 to 10. It is calculated simply by adding the soil permeability and DTWT factors together. 7. Hydraulic Conductivity Hydraulic conductivity is defined as the ability of materials in the aquifer to transmit water, which in turn controls the rate at which groundwater flows and thus the rate at which contaminants enter the aquifer (Piscopo, 2001). Ramsen and Panda (2008) developed hydraulic conductivity ratings for the Kapgari watershed in India based on field data where head permeability testing was conducted on soil samples collected from their small study area. Other studies using the DRASTIC model have relied on hydraulic conductivity data produced by local water resource agencies (Bazimenyera & Zhonghu, 2008). Based on some P a g e 11 | Matt Stuemky – Research Paper 2 – GEOG 587 (Summer 2009) preliminary research, soil type and soil depth (soil horizon) attribute data from the SSURGO database for Montana should be able to be used to calculate hydraulic conductivity values. Parameter ratings from 1 (very low conductivity) to 10 (very high conductivity) will be used. Additional research is needed to determine how best to use SSURGO data to derive values for this parameter. 8. Additional parameter: Impact of Roads One of the factors mentioned but not considered in Wilson's original study was the influence that roads might have on herbicide intrusion into the ground. The inclusion of roads data into this GIS is to try and account for the likelihood of higher concentrations of herbicides both directly on and adjacent to roads where agricultural vehicles and equipment regularly dispense weed control products during the growing season. The Impact of Roads parameter is derived from land use and roads data obtained from the State of Montana's Natural Resource Information System (NRIS). Land Use (LU) coverage (polygons) data will be assigned a factor from 1 to 3 (urban, rural/non-agricultural, and agricultural), three general categories that are produced by generalizing multiple LU types. Roads coverage (lines) data will be assigned a factor from 1 to 4 to indicate the permeability the road surface type (paved, improved/gravel, improved/other, unimproved). An Impact of Roads parameter rating is assessed by a two step process. First, each road line segment that is "contained within" a single mapping unit (a polygon derived from the MAPS Atlas) is analyzed by adding the LU factor (1-3) and the road surface type factor (1-4) together in order to indicate the level of impact the single road has. Second, an average value for the entire mapping unit is calculated by adding up all individual ratings, then dividing by the total count of road line segments analyzed - this becomes the Impact of Roads rating for the mapping unit. As a manual process, this would be extremely time consuming, so this task will need to be automated using tools in ArcGIS or by some other means to derive Impact of Roads parameter values. A DRASTIC weight value of 5 is used for the Impact of Roads parameter. 9. Additional parameter: Noxious Weeds Zone The Noxious Weeds Zone parameter is also derived from data obtained from NRIS. The same land use coverage (polygons), reclassification, and factoring methods described above for the Impact on Roads parameter will be used. Weed coverage (polygon) data will be obtained from NRIS's new Noxious Weed Survey spatial database. Wilson's earlier studies focused on the leafy spurge and knotted knapweed infestations. Factors from 1 to 5 (few to high noxious weed concentrations) can be assessed based on the density of all or selected weed species that are regularly treated with systemic herbicides. A rating value for the Noxious Weeds Zone parameter is calculated by combining both LU type and noxious weed type factors together. This parameter attempts to reduce some uncertainty that existed in Wilson's studies regarding the P a g e 12 | Matt Stuemky – Research Paper 2 – GEOG 587 (Summer 2009) locations where noxious weed grow. Reality demonstrates that weed growth patterns do not uniformly cover the ground in the same density over a given soil type. Since weed coverage data shows artificial boundaries demarcating zones of growth, adding the LU component provides a means for the DRASTIC model parameter to indicate a broader "zone of influence", based on the type of land where noxious weeds are likely to grow and where weed control measures are also likely to be applied (e.g. assessing a high factor value of 3 for agricultural land use areas). As with the Impact of Roads parameter, a DRASTIC weight of 5 is used for the Noxious Weeds Zone parameter. Risk assessment maps Risk assessments maps produced by the GIS will eventually be capable of showing the entire state, region, county, or a small study area in large scale (1:24,000). For each map produced, base map data can be either remotely-sensed satellite imagery or land cover data. Selected weed map polygons can also be superimposed over the base maps. The calculated DRASTIC Index values for each 2-mile by 3-mile mapping unit will be classified using thresholds based on four "vulnerability indices". The actual threshold range values to be used will be determined once the project reaches the stage where maps can finally be created from data that has been collected and integrated into the geodatabase. However, the descriptors will be low, medium, high, and very high to indicate the risk of groundwater contamination by weed control measures. For visual display, different colors will be used for the four pixel ranges: blue colors for areas of low risk, green to yellow colors for areas of medium risk, orange colors for areas of high risk, and red colors for areas of very high risk. A transparency effect will be added to allow the underlying layers to be visible. Project plan: Phase 1 - six month duration (begin January 2010) Since the geodatabase will be designed to eventually house project data for the entire state of Montana, considerable time must be spent to properly architect the feature datasets, feature classes, and non-spatial attribute tables. Note that some of the tasks occur concurrently over the six month timeframe. PROJECT TASKS TIME FRAME Data collection Jan - Feb (2 months) 2 Geodatabase design using ArcCatalog Jan - Mar (3 months) 3 Data integration using ArcCatalog Mar - Apr (2 months) 2 Data cleanup / editing using ArcCatalog and ArcMap Apr (1 month) 1 Analysis and modeling with DRASTIC using ArcMap Apr - May (6 weeks) 1 Risk assessment maps using ArcMap May (1 month) 1 Poster and paper showcasing results from selected study area (Teton County) May-Jun (6 weeks) 2 P a g e 13 | Matt Stuemky – Research Paper 2 – GEOG 587 (Summer 2009) STAFFING Project resources and costs: staffing, equipment, GIS software, miscellaneous Initial funding for the project is $350,000. This project phase must be completed in six months. Work will be conducted at an existing facility, with offices containing phones, desks, chairs and other standard office equipment; gigabit Ethernet network, wireless network, and high speed Internet access are already in place. Staffing (first six months) JOB TITLE COUNT SALARY Lead researcher 1 $75,000 GIS database specialist / programmer 1 $50,000 GIS analyst 2 $40,000 GIS technician 1 $30,000 Research assistant 1 $25,000 TOTAL 6 $260,000 Equipment 1. Workstation computer: HP Z400 Workstation Quantity: 4 Unit Cost: $2,900 http://h10010.www1.hp.com/wwpc/us/en/sm/WF06b/12454-12454-296719-307907-296721-3718668-3718669-3937735.html 2. Laptop computer: HP EliteBook 8730w Quantity: 2 Unit Cost: $2,700 http://h10010.www1.hp.com/wwpc/us/en/sm/WF06b/321957-321957-64295-3740645-3955549-3784202-3999723-3960438.html 3. Large format printer: HP Designjet T1120 HD Quantity: 1 Unit Cost: $21,000 http://h10010.www1.hp.com/wwpc/us/en/sm/WF05a/18972-18972-3328061-12600-3328078-3878813.html TOTAL: $38,000 GIS software 1. ArcGIS Desktop Suite - ArcInfo Lab Kit 2. ArcGIS Extensions http://spatialscience.usc.edu/ESRI.html Quantity: 1 Quantity: 1 Unit Cost: $3,750 (per year) Unit Cost: $1,500 (per year) TOTAL: $ 5,250 Miscellaneous Data acquisition costs Miscellaneous software (GIS or other), software utilities Coffee: Starbucks / Pete's Other unanticipated costs TOTAL: $25,000 TOTAL PROJECT COST: $328,250 P a g e 14 | Matt Stuemky – Research Paper 2 – GEOG 587 (Summer 2009) IV. Conclusions Most current groundwater contamination studies make use of an ever increasing variety of spatial and non-spatial data. Some of these were utilized in the DRASTIC model based GIS project proposal described above. Some data sources that should be considered for GIS-based groundwater contamination studies include: hydrogeological as well as hydrological data, soil and water sample data collected from field investigations, GPS acquired data with high quality survey-grade instruments, slope, aspect and elevation data from DEMs, fine scale (1:50,000-scale or better) soil survey and topographic maps, and high-resolution digital terrain model (DTM) data based on digitized topographic contour lines. Remotely sensed imagery is another important data source worth consideration for a GIS-based groundwater risk assessment project. For example, Werz and Hötzl (2007) made use of multispectral imagery from the LANDSAT Enhanced Thematic Mapper (ETM) in their groundwater vulnerability mapping project. A resource such as LANDSAT ETM offers the advantages of extensive worldwide geographical coverage, wide spectral range (six bands of data in the visible, near-infrared and mid-infrared), up to 15-meter resolution (in one of the panchromatic bands), good vegetation discrimination, and a relatively cost-effective resource compared to conventional data collection approaches (Werz & Hötzl, 2007). Another remotely sensed data source worth consideration would be high resolution color aerial photography which, at extremely fine 1:10,000 scale, can potentially offer sub-meter resolution of land cover information. When considering aspects of the earlier studies such as those conducted by Wilson and more recent studies gleaned from the literature, it is important to remember that no matter what data sources and methods are used, varying amounts of uncertainty will always remain when attempting to assess complex phenomenon such as groundwater contamination risk factors. The advantages offered by selecting one groundwater vulnerability approach over another, whether overlay and index methods, process-based simulations, or statistical methods, only seem to focus on minimizing uncertainty related to one or more key factors while leaving other elements of uncertainty intact. P a g e 15 | Matt Stuemky – Research Paper 2 – GEOG 587 (Summer 2009) References Bazimenyera, J. & Zhonghua, T. (2008). A GIS Based DRASTIC Model for Assessing Groundwater Vulnerability in Shallow Aquifer in Hangzhou-Jiaxing-Huzhou Plain, China. Research Journal of Applied Sciences, Vol. 3, No. 8, pp. 550-559. Retrieved August 9, 2009 from http://medwelljournals.com/fulltext/rjas/2008/550-559.pdf Inskeep, W.P., et al. (1996). Input Parameter and Model Resolution Effects on Predictions of Solute Transport. Journal of Environmental Quality, Vol. 25, No. 3, pp. 453-462. Montana Groundwater Information Center. (2009). Website: http://mbmggwic.mtech.edu/ Montana State University - MAPS Atlas. (2009). Website: http://www.montana.edu/places/maps/index.html National Research Council. (1993). Ground Water Vulnerability Assessment: Predicting Relative Contamination Potential Under Conditions of Uncertainty. Washington, D.C.: National Academy Press. Retrieved August 11, 2009 from http://books.nap.edu/openbook.php?record_id=2050&page=R1 Oklahoma State University, Department of Plant and Soil Sciences, Soil Physics. (2009). Retrieved August 7, 2009 from http://soilphysics.okstate.edu/software/cmls/cmls94a.htm Piscopo, G. (2001). Groundwater vulnerability map explanatory notes – Castlereagh Catchment. Center for Natural Resources, NSW, Australia. Retrieved August 12, 2009 from http://www.dnr.nsw.gov.au/water/pdf/castlereagh_map_notes.pdf Remesan, R. & Panda, R.K. (2008). Groundwater Vulnerability Assessment, Risk Mapping, and Nitrate Evaluation in a Small Agricultural Watershed: Using the DRASTIC Model and GIS. Environmental Quality Management, Vol. 17, No. 4, pp. 53-75. Retrieved August 8, 2009 from http://www3.interscience.wiley.com.libproxy.usc.edu/cgi-bin/fulltext/119880975/PDFSTART Sinkevich, M.G., et al. (2005). A GIS-Based Ground Water Contamination Risk Assessment Tool for Pesticides. Ground Water Monitoring & Remediation, Vol. 25, No. 4, pp. 82-91. Retrieved August 8, 2009 from http://www3.interscience.wiley.com.libproxy.usc.edu/cgi-bin/fulltext/118664333/PDFSTART Werz, H. & Hötzl, H. (2007). Groundwater risk intensity mapping in semi-arid regions using optical remote sensing data as an additional tool. Hydrogeology Journal, Vol. 15, No. 6, pp. 1031-1049. Retrieved August 9, 2009 from http://www.springerlink.com/content/k63q7315433q7427/fulltext.pdf Wilson, J.P., et al. (1993). Coupling Geographic Information Systems and Models for Weed Control and Groundwater Protection. Weed Technology, Vol. 7, No. 1, pp. 255-264. Wilson, J.P., et al. (1996). GIS-Based Solute Transport Modeling Applications: Scale Effects of Soil and Climate Data Input. Journal of Environmental Quality, Vol. 25, No. 3, pp. 445-453. Wilson, J.P. (2009). GEOG 587 - GPS/GIS Field Techniques Course Notes, Sections 7-9. Los Angeles, California. University of Southern California: Department of Geography. P a g e 16 | Matt Stuemky – Research Paper 2 – GEOG 587 (Summer 2009)