Surface Water Modeling for the Marcus Hook Refinery

• Abstract
• Introduction
• Annual Runoff
• Regional Terrain and Runoff
• Facility Terrain and Runoff
• Flow Calculations
• Conclusions
• References
• Data Dictionary

Abstract

This report documents the development of a surface water runoff model for the former Marcus Hook Refinery and the area surrounding the refinery. The process of developing these models involves obtaining an elevation grid, and then calculating the direction and accumulation of flow on this grid. This is a steady state calculation where water simply flows from a higher elevation head to a lower elevation head. Streams are delineated based on a user defined threshold of how many cells are upstream of the current cell. All of these tasks are performed using the Watershed Delineator extension with ArcView. The facility runoff characterization is unique in that the scale of the model is much smaller than what is typically used in a flow representation process. The facility elevation model has a cell size of 3 feet. While this size should yield more accurate representations of runoff, a few problems were encountered such as flow through structures and limited flow accumulation in the on-site creeks. Most of these problems are a result of the process used in the filling sinks step. ArcTools and a procedure referred to as "burning" were used to solve these problems. In order to get a sense of the flow magnitudes, an annual runoff value was determined using a USGS runoff coverage. The resulting models can be used to calculate loads of chemicals to each surface water body at the facility and to calculate the amount of runoff that accumulates upstream of the two creeks that flow through the facility. This report assumes a working knowledge of ArcView and the Geographic Information System (GIS) application. More information on ArcView and its capabilities can be found on the GIS Hydro '97 page.

Introduction

For my research, I am working on an environmental assessment of an operating oil refinery located in Marcus Hook, Pennsylvania. The refinery lies along the Delaware River and covers approximately 340 acres. It has been in operation since the early 1900s, and there were many years of unregulated spills and releases before the current era of EPA regulations. The refinery was sold by British Petroleum Exploration & Oil (BP) three years ago, and BP is responsible for the remediation of any environmental problems that existed at the time of the sale.

• Facility layout

The focus of this research is to investigate the use of a Geographic Information System (GIS) application and a Decision Analysis Framework for making risk-based decisions. Using an already existing database of information characterizing the site, the goal is to develop a program for corrective action. Some of the primary tasks include source area characterization, receptor identification, fate and transport modeling, exposure quantification, and communication of information and results of the risk-based decision process.

This report documents the development of a facility terrain model and a regional terrain model in order to characterize surface water runoff and drainage. Movement of water across the ground is an important transport mechanism that provides insight into how chemicals of concern (COC) will move from a surficial source area to a receptor. Surface water modeling provides the insight into this process. For example, assume that there is a known release of a COC from a stack, and a certain amount of this COC deposits from the air onto the ground. Knowing the amount deposited and the annual runoff, loads can be calculated for the COC at outlet points, such as drainage to a river or pond. A load is simply the product of flow and concentration, and it represents the amount of mass input to a system on a time basis (e.g., a year for annual runoff). Once the load has been calculated, one can estimate the downstream concentration where a person could come in contact with or ingest the water. In order to be protective of human health at this receptor location, the calculated concentration must be below a specified value, such as an EPA action level. Therefore, by knowing how water flows across a site, one can make better decisions about potential pathways and receptors. For this particular site, the COC loads to the creeks and river will also be important for characterizing impacts to ecological receptors.

Before modeling how water accumulates and drains from an area, it would be useful to have an idea of how much runoff exists in the region. In order to calculate an appropriate value for the site and surrounding area, an annual runoff coverage was obtained from the USGS website. Because of the size of the refinery and the limited detail in available data, only one runoff value was obtained for the site. This value is used to prepare some initial calculations of runoff flow at the refinery and in the surrounding area.

The analysis of regional terrain involves establishing watershed boundaries for Marcus Hook Creek and Stony Run Creek. These creeks flow through the site, and it would be beneficial to know how much flow accumulation occurs upstream of the refinery. The regional runoff is determined using a 30 meter digital elevation model (dem) and the Watershed Delineator extension for ArcView, which the University of Texas has developed in conjunction with the Texas Natural Resources Conservation Commission (TNRCC). This extension is also called HEC PREPRO, and the methods used for this project are very similar to those used in exercise 2 of the GIS class.

The analysis of facility terrain involves characterizing surface water runoff on site and drainage to the creeks and the Delaware River. This task is also performed using a dem and the Watershed Delineator extension. Air Survey Corporation (ASC), who was contracted to prepare orthophotographs and coverages of the refinery, has supplied a digital terrain model (dtm) in arctin format. This tin was converted into a dem, or flat grid, so that the runoff calculations could be completed.

Annual Runoff

The first step for this project was to calculate a representative runoff value for the region. To accomplish this task, an annual runoff coverage was obtained from the USGS spatial data website. This coverage has been constructed using available data for the United States from 1951-1980 and is a set of contour lines:

The only problem with this coverage is that none of the contour lines run through the facility. Even though the coverage is fairly detailed, this result is not surprising because the size of the refinery is so small compared to the scale seen here. One way to solve this problem is to interpolate a surface (e.g., a grid) from this set of contour lines. However, the features of Spatial Analyst only allow for a surface to be interpolated from a point coverage.

A couple of different methods were investigated for creating a runoff surface. The first idea was to create a grid and then intersect this grid with the runoff contours. The resulting coverage would also be a line coverage, but there would be a node at every location where the grid intersected the contour lines. The nodes could then be converted to a point coverage where each point would have a runoff value equal to the contour line from which it was generated. A runoff surface could then be interpolated from the point coverage. While this method was not pursued, these tasks can be performed using ArcInfo.

An alternative method is simply to create a point coverage from the contour lines. The method is not as elegant because a point will be created wherever there is a bend or change in the contour line. However, the procedure is simple, effective, and can be completed completely within ArcView. I have created a generic script that anyone can use to interpolate a surface from a set of contour lines. The script is available here, and the figure below shows the methodology.

• Interpol.ave

This script was used to generate a runoff grid for the entire United States with a cell size of 1 km and a runoff grid for Pennsylvania with a cell size of 500 m.

The orange dot in the southeast corner of Pennsylvania is the Marcus Hook Refinery. Looking at this grid, the annual runoff for the refinery is 20 inches. Only one value is obtained because the detail of the contour lines does not allow for a more precise interpretation. Creating a much smaller grid would be reading too much from the data, and even with a smaller grid, the value changes across the site would be very small (probably less than tenths of an inch). A more detailed approach is often used to obtain a runoff grid that incorporates precipitation, infiltration, and evaporation. However, the precipitation data that is currently available over the internet has cell sizes that are bigger than the entire area of the refinery. Thus, the use of only one value is appropriate, and this value can be used for calculating runoff volumes. This topic is discussed in the Flow Calculations section.

Regional Terrain and Runoff

The development of a regional terrain model is important for a couple of reasons. First, it provides a look at how much water drains into Marcus Hook Creek and Stony Run Creek upstream of the refinery. It also identifies runoff from other facilities that could affect either of these creeks. Utilizing the capabilities of ArcView and free spatial data available on the web, one can delineate watersheds and then determine which facilities fall in each watershed. This area of Pennsylvania is highly industrialized and the Marcus Hook Refinery is not the only facility contributing discharges and drainage to the two creeks.

In order to characterize runoff in the region, a 30 meter dem for Marcus Hook and a coverage of streams in Delaware County was obtained from the Pennsylvania State Spatial Data Center (PASDA). A coverage of regulated facilities was obtained from the EPA's Envirofacts page.

After collecting the needed data, the dem was cut according the Pennsylvania state boundary line, which proceeds down the middle of the Delaware River. The elevations on the other side of the river in New Jersey and southwest of the refinery in Delaware are not needed for any analysis at this point in time. The resulting dem is shown here:

The stream coverage obtained from PASDA was "burned" into this dem in order to ensure proper flow accumulation in the streams and rivers of the region. The process of burning in streams requires that the elevations of each stream remain constant while the rest of the dem is raised by some arbitrary value, such as 2000 feet. This raised value must be greater than the highest point in the dem. Using the Map Calculator available with Spatial Analyst, the following steps are performed to burn in the streams: convert the stream coverage to a grid; divide the stream grid by itself; multiply the new stream grid by the dem (demstream); add 2000 to the dem (demplus); and then merge demstream and demplus with an Avenue script. A more detailed explanation of this process can be found in the GIS exercise Spatial Hydrology of the Urubamba River System in Peru. Even though the elevations in the burned dem are not accurate, the differences in elevation between each cell remain constant (except at the stream edges). Therefore, the flow direction model only change around the streams where water should be draining anyway.

After burning in the streams, the Watershed Delineator extension was used to model surface water runoff. The steps in the process are fill sinks, flow direction, flow accumulation, stream definition, stream segmentation, and watershed delineation. During the fill sinks step, the program searches for all the elevation depressions where water could accumulate. The program then fills these depressions, ensuring that all water which falls on the terrain drains to an outlet off of the grid. In the flow direction step, the program determines how water flows from each cell to another. Water can drain from a given cell into any one (but only one) of the surrounding eight cells. In the flow accumulation step, the program determines the number of cells upstream of the given cell based on the flow direction model. Streams can then be defined based on a threshold number of upstream cells using the flow accumulation grid. Once the streams are defined, they can be broken down into separate segments or links. In the final step, the links grid and flow direction grid are used to delineate watersheds. Here are the results for the regional terrain:

The streams have been defined in this figure using a threshold of 500 upstream cells, which models the actual stream coverage very well. Marcus Hook Creek runs through the left portion of the refinery, and Stony Run Creek runs along the eastern boundary of the refinery. The red dots represent EPA regulated facilities, and a lot of these facilities lie near the two creeks. The attributes table for the regulated facilities includes the facility name, id number, and permit information. See the data dictionary for a complete description.

Facility Terrain and Runoff

The modeling of surface water runoff at the facility is particularly important because this model can be used in making remedial action decisions. Surface water runoff can carry a COC from a source area with high concentrations to a receptor where this COC could affect a human, an animal, or an aquatic organism. For instance, during construction activities, soil is often dredged and placed in piles. These piles can have high concentrations of COC's, and any precipitation runoff from these piles could carry these COC's across the refinery. This runoff could come in contact with a construction worker or drain into one of the creeks where it could affect fish and other organisms. Another example of surface water transport is from standard water bottom release practices of storage tanks. A water bottom release refers to drawing water from a product storage tank through the use of a valve on the side of the tank. Eventually, all the water is drawn, and the tank will begin to release free product (e.g., gasoline or cruel oil). If a refinery worker is not there to shut off the valve, large amounts of this product will be discharged from the tank. A lot of this product will infiltrate into the soil, but some will be carried off with surface water runoff. Again, this runoff could come in contact with a worker or drain into one of the creeks. By knowing how water moves across the surface, one can make better decisions about exposures and exposure pathways.

The process of modeling surface water runoff for the facility is identical to that for the regional model. A dtm was obtained from ASC in tin format. This tin can be viewed in ArcView using the 3D-Analyst extension or in ArcInfo using the Arctin extension. Shown below is a small portion of the dtm as seen in the 3D-scene viewer of ArcView. The elevations have been exaggerated so as to enhance detail.

Before beginning the modeling process, the dtm had to be converted to a dem because the watershed modeling tools cannot be used with a tin. The dtm for the refinery has a total of 421,472 triangles. The planimetric area is 22,645,058 square feet, and the surface area is 22,954,375 square feet. The average area of a triangle (surface area/# of triangles) is approximately 54 square feet. This area would correspond to a grid cell size of about 7.5 feet. However, when trying to maintain accuracy in the conversion of a tin to a grid, the average area of a triangle should not be used because this area includes some large triangles. The area of the smallest triangle should not be used either because only so much detail can be obtained from the data. Choosing a grid cell size equal to the smallest area would lead to wasted disk space and energy. Instead, some middle ground value should be used. For this project, a few different values were tested, and a 3 foot cell size was chosen as the final value. A tin can be converted to a grid using either ArcView or ArcInfo.

After converting the dtm, the watershed delineator tools were used with the facility dem to model runoff. However, a few problems were encountered in the initial development attempts. Most notably, flow was protruding through existing buildings and tanks. There was also limited flow accumulation in the creeks. The errors were due in large part to the processing methods of the filling sinks step. At the refinery, dike areas exist around each tank, and these dike areas were being filled to a level terrain in order to prevent flow from accumulating in them. Even though the tank foundations are incorporated in the dem, the dike areas were being filled to the top of the berm, which is usually a few feet higher than the foundation of a storage tank. In the actual system, drainage pipes cut through the berm so that water drains from each dike area to another.

Even if this dike area problem could be fixed, there would still be the issue of flow through buildings. It turns out that even though the building foundations are incorporated in the dem, they are not significant enough to prevent flow from running across them, especially after the filling sinks process. Besides this building problem, the surface water ponds were being filled, and as already mentioned, flow was not accumulating in the creeks properly.

Some unique methods were used to correct these problems utilizing ArcTools and the burning procedure. ArcTools is a component of ArcInfo with which grid cell values can be edited. To account for the drainage pipes between dike areas, trenches were created in the dem with ArcTools. The actual locations of these pipes are not presently known, so for now they have been set according to elevation changes between the dikes. The trenches are set at an elevation of 1 foot, but during the fill process, they are filled until the elevation reaches that of the one on the lower end of the dike area. This figure shows the methodology:

ArcTools was also used to correct the surface pond problem. A no data cell was created in each of these ponds to prevent them from filling. The program setup allows water to drain into no data cells. Thus, any runoff to a surface pond will flow into the assigned no data cell. To solve the building problem, an ASC building coverage and an ASC tank coverage were burned into the dem. The procedure is similar to burning in the streams, except this time the original dem elevation is held constant, and anywhere there is a building or tank, the elevation is raised 50 feet. Again, 50 feet is an arbitrarily set value that is above the highest point in the original dem. The creek problem was solved by burning an ASC stream coverage into the revised dem. Therefore, the original dem was modified to include trenches, no data cells, buildings, and tanks before proceeding with the watershed calculations.

Once the dem manipulation was complete, developing the surface water model was fairly simple. The procedure is identical to the one described in the regional runoff section except that the stream segmentation and watershed delineation steps are eliminated. Here is a zoomed in portion of the results:

This figure shows the water runoff for the southwest corner of the facility. There is no flow through any of the buildings or tanks, and most of the water drains to Marcus Hook Creek, and then eventually to the Delaware River. This figure also shows how some of the drainage routes between dike areas were set. Both of the tank farms shown above outlet to the creek. This result is pretty cool. Just looking at this example, one notices that it would not take long for any release in the West Tank Farm (left of the creek) to runoff into the creek and the river.

Flow Calculations

 In order to obtain a more complete model, some preliminary calculations of flow magnitudes were developed. These magnitudes are computed using a weighted flow accumulation process, in which a value such as runoff is assigned for each grid cell, and the values upstream of a given cell are summed. Instead of accumulating a number of cells, a total amount of runoff is accumulated. However, in this project, the runoff value is the same in every cell: 20 inches. Thus, this value can be multiplied by the accumulated number of cells at the point of interest and by the grid cell area to determine a volume of runoff. Here are some results:

The accumulation was determined by using the identify tool () with the flow accumulation grids. Of particular interest for this project are the amount of runoff upstream of the refinery and the runoff from the refinery. Therefore, the number of upstream cells was identified where the two creeks cross the northern boundary of the site and where the creeks outlet to the Delaware River. The facility grid has a much smaller mesh than the regional grid, so the accumulation of cells will be larger with the facility grid, but the cell area is also much smaller.

Notice that Marcus Hook Creek receives a lot more runoff flow than Stony Run Creek upstream of the refinery. This result was expected, as MH Creek is larger than SR Creek. SR Creek is fairly small when it enters the refinery, and receives a considerable amount of flow as it passes through because of runoff and outfalls. This result is also evidenced in the above table.

Conclusions

The capabilities of ArcView and other GIS applications are progressing at a rapid pace. Only a few years ago, a significant amount of time was needed just to develop a map. Now detailed models can be developed such as the one for surface water runoff presented here. This report primarily discusses the methodology and procedures used to develop the models for regional and facility runoff. However, these results will be used to determine how much of a chemical of concern could move from a source to a receptor. By knowing the exposure concentration at a receptor, one can make decisions about what sort of remedial action will be necessary at the source.

Future work with these surface water models will include determining the actual location of the drainage pipes between tank dike areas. The location of these pipes will allow for the development of a more precise model. Also, as just mentioned, some calculations will be made to see how COC's move with runoff. Because of the two creeks and the Delaware River, it will be important to know how these COC's could affect ecological receptors in these surface water bodies. The chemical loads to each body can be calculated, and then stream and river models can be used to track concentrations downstream of an input. It will also be useful to know the chemical concentrations contributed by facilities upstream of Marcus Hook. These contributions can be determined using the coverage of regulated facilities. Using the facility name listed in the attributes table, more information on NPDES permits and discharages can be obtained from EPA's Query Mapper.

References

I did not use any books for reference. However, I receive a lot of help from the members of Dr. Maidment's research team. I would particularly like to thank Seann Reed, Francisco Olivera, and Lesley Hay Wilson for their help on this project.

Data Dictionary

This data dictionary contains a listing of all the coverages I used in the development of the models along with the attributes of each coverage. The dem, flow direction grid, flow accumulation grid, and stream grids are only listed once although many were created in the project. No metadata is available for the coverage of streams in Delaware County.