One of the most important hazards facing the United States and the world is the disposal and long-term management of waste sites in the 21st century. The wastes range in character from municipal waste to hazardous waste. There are thousands of hazardous waste sites in the United States (EPA, 2002). The wastes are stored primarily in buried, near-surface containment systems with varying levels of effectiveness. The primary objective of current containment system designs is to protect human population, flora, and fauna from exposure to the materials via release to the surface, the atmosphere, or into the subsurface groundwater.
The development of an improved remote sensing-assisted decision support system for monitoring hazardous waste sites is extremely important for two reasons:
- The continued monitoring of these hazardous waste sites will become increasingly significant as the containment systems (e.g., soil closure caps) put in place during the late 20th century age and deteriorate.
- Since September 11, 2001 there is now the possibility that the hazardous waste sites might be subjected to terrorism unleashing toxic materials into the atmosphere, soil, or groundwater. The physical isolation and low frequency of monitoring of some of these hazardous waste containment systems could result in releases going undetected for months or years under current practices of visual inspection. Thus, this hazardous waste site application has serious homeland security implications.
Federal and State Laws Requiring Monitoring of Hazardous Waste Sites
Multiple federal and state agencies are responsible for overseeing the construction and operation of hazardous waste sites through several federal regulations. In particular,
- RCRA addresses the construction and management of hazardous and municipal landfills;
- CERCLA addresses disposal sites where releases have already occurred; and
- UMTRCA addresses the management of residual materials from the mining of uranium.
These regulations are administered by the U.S. Environmental Protection Agency (EPA), the Department of Energy (DOE), the Nuclear Regulatory Commission (NRC), and have varying levels of participation by relevant state agencies where a federal agency has delegated authority.
Existing Hazardous Waste Site Monitoring Decision Support System(s)
To develop an improved hazardous waste site monitoring DSS that incorporates geospatial information it is necessary to focus on certain types of waste sites. This technological hazard research will document the nature of the biophysical (e.g., temperature, vegetation stress, vegetation moisture content, soil moisture content, and terrain subsidence) and demographic variables (e.g., proximity to the nearest population or potable water aquifer) necessary for the development of a hazardous waste site monitoring decision support system.
Dr. John Jensen will work closely with the Dr. John Gladden of the Westinghouse Savannah River Company (WSRC). WSRC manages the Department of Energy hazardous waste sites at the Savannah River Site near Aiken, SC. The creation of the SDSS will also be overseen by Dr. Terrence Slonecker of the EPA’s Environmental Photographic Interpretation Center who is responsible for conducting hyper- and ultraspectral remote sensing research and technical support of environmental media statutes such as the CERCLA and RCRA. Letters of support from the Westinghouse Savannah River Company, DOE, and EPA concerning the development of the remote sensing-assisted hazardous waste site SDSS are found in Section 14.
One of the first tasks will be a thorough examination of all existing hazardous waste site monitoring decision support systems used by WSRC, DOE, and/or EPA. Right now we are not aware of any. If they do exist, the degree to which they incorporate spatial information will be assessed (e.g., Jensen et al., 1998). Based on what is discovered, the logic for a new or improved hazardous waste site monitoring SDSS will be developed that utilizes spatial information. This research will focus first on hazardous waste sites on the Savannah River Site that contain a diversity of wastes covered by a variety of materials. The initial site for evaluation will be the SRS Test Pad Facility, a series of four experimental closure caps that can be manipulated to demonstrate failure modes. A portion of the site is shown in Figure 2. The hazardous waste site SDSS will then be evaluated at other government facilities identified during the initial research period and after interviews with agency personnel.
Figure 1. GER 1750 hyperspectral image of the clay-capped Mixed Waste Management Facility at the Savannah River Site obtained on July 22, 2002 at a spatial resolution of 2 x 2 m. The color composite hyperspectral bands are RGB = (NIR 20, Red 10, green 5).
Technology Development
There are several tasks areas that must be systematically addressed to develop a hazardous waste site monitoring SDSS. These research areas will be investigated over a period of four years. Important considerations include a) clear enunciation of the questions to be addressed by the SDSS, b) its architecture, c) biophysical inputs, c) demographic-socioeconomic inputs, d) creation of the SDSS, and e) methods of testing its effectiveness (i.e., metrics).
Task 1. Encoding the Questions Addressed by the Hazardous Waste Site SDSS
Experts at WSRC, DOE facilities, and EPA responsible for hazardous waste site management will be interviewed at numerous locations throughout the U.S. The goal is to learn what types of logic, information, and decision rules are typically used when monitoring and making decisions about hazardous waste sites. Substantial time will be spent obtaining this critical information, as the SDSS can only be effective if it addresses and answers the most important hazardous waste site management questions. This will allow us to create an accurate knowledge domain.
Task 2. Architecture of the Hazardous Waste Site SDSS
It is important to select the most appropriate architecture for the SDSS. It is likely that the architecture will be based on an expert system shell, a back-propagation artificial neural network (ANN), or a hybrid of the two (e.g., Huang and Jensen, 1998; Jensen et al., 2000). The nature of the information and the questions that must be considered to address specific management decisions will dictate the final architecture of the hazardous waste site SDSS. Well-conceived management questions and quality information are only of value when the SDSS can identify the optimum monitoring and management strategy. Significant effort will go into the design and testing of the hazardous waste site SDSS architecture.
Figure 2. Cyclical work flow diagram of the remote sensing-assisted hazardous waste site monitoring decision support system for the Savannah River Site.
Task 3. Biophysical Information Inputs to the Hazardous Waste Site SDSS
An effective hazardous waste site SDSS must incorporate numerous biophysical variables to facilitate early identification of any breach in the site, movement of water or other materials through the site, and/or translocation of materials from the site to other areas.
Terrain Elevation and Subsidence: Hazardous waste sites that have been capped (e.g., using clay) must be monitored on a repetitive basis to determine if subsidence is taking place. Subsidence can be caused by a number of factors including the collapse of interior containers (e.g., drums) and/or percolation of water through the system with subsequent translocation of materials. Identification of subsidence on the order of just a few centimeters as soon as it occurs allows hazardous waste site managers to rapidly investigate the problem area in situ and correct the problem prior to serious waste site compromise.
The proposed research will investigate the use of several elevation (subsidence) data collection techniques at the Mixed Waste Management Facility (MWMF) at the Savannah River Site (Figure 2), including:
·
- a network of in situ sites that will be periodically systematically surveyed using traditional kinematic global positioning system (GPS) techniques,
- soft-copy photogrammetric techniques based on large-scale stereoscopic aerial photography to create a raster DEM (e.g., Jensen , 1996a), and
- Light Detection and Ranging (LIDAR) data will be acquired and processed to create a raster DEM (e.g., Raber et al., 2003).
Statistical analyses will be conducted to determine the accuracy and sensitivity of the remote sensing-assisted methods to provide a quality digital elevation model as input to the subsidence information required by the hazardous waste site SDSS.
Soil and Vegetation on Hazardous Waste Sites: Hazardous waste materials (e.g., drums, concrete encased liquids) often are covered in varying combinations by an impermeable barrier of low permeability soil, geosynthetic fabric, or plastic. These layers are then covered with topsoil and vegetation. It is imperative that the topsoil and vegetation cover be monitored on a systematic basis over time to determine if there are any changes in the condition of the soil and/or vegetation. This research will involve a number of in situ and remote sensing investigations focused on assessing the integrity of these components.
Topsoil is unconsolidated material at the surface of the Earth that serves as a natural medium for growing plants. Agronomists refer to this as the solum. We no longer identify a ‘soil type’ (Petersen, 1999). Rather, soil scientists determine the soil taxonomy based on soil color, texture-class, moisture content, bulk density, porosity, and chemistry (Loynachan et al., 1999). Ideally, soil-capped hazardous waste sites contain a single soil taxa (e.g., an oxisol). Unfortunately, this is rarely the case. Rather soil from the surrounding countryside from a variety of locations is often used to construct the topsoil layer of the cap. This variability in the solum introduces differences in permeability and soil moisture content throughout the cap. This in turn can influence the density of vegetation found on the cap. Therefore, it is important to determine the spectral characteristics of the various topsoil components on a hazardous waste site.
The vegetation placed on a soil-capped hazardous waste site is selected to perform specific functions, from stabilizing the soil to managing moisture in more advanced designs. This vegetation should exhibit a uniform leaf-area-index (LAI) and biomass as it grows on the theoretically uniform soil taxa. Two of the most important and diagnostic ways of identifying the early compromise of a hazardous waste site are (Jensen et al., 2001): 1) the identification of unusual vegetation cover type combinations on the clay cap (e.g., trees or deep rooted shrubs and herbs), and 2) the diminished or enhanced biomass or leaf-area-index (LAI) of the vegetation that may be associated with differences in soil moisture across the cap. Such soil moisture differences can be caused by subsidence induced ponding, or excessive drainage due to loss of integrity of the below ground impermeable layers. In either case, inspection and potential maintenance of the cap is required to ensure continued integrity.
A spectral library of the major soil taxa and vegetation types that are often used to cover hazardous waste sites must be created. WSRC will first collect and analyze soil and vegetation types at waste sites at the SRS. WSRC staff will also prepare a location where various types of soil and vegetation are arranged in a randomized block design to facilitate examination of treatments of soil and vegetation commonly used in hazardous waste site restoration. In subsequent years, similar sites will be investigated in other geographic regions. The goal is to prepare a soil and vegetation spectral library that will have broad applicability for soil and vegetation covered hazardous waste sites throughout the U.S. Information from available spectral libraries (e.g., NASA, USGS, NASA) will also be incorporated.
In situ measurement will involve the collection and extraction of soil and vegetation spectra over the region 0.4 – 2.5 um (visible through middle-infrared) and 8 – 14 um (thermal infrared) using a hand-held spectroradiometer and precision radiation thermometer. The in situ optical measurements will be taken both in the field and in a lab where illumination can be controlled.
Remote sensing measurement will involve the collection of sub-orbital hyperspectral data obtained at 1 x 1 m spatial resolution over the spectral region from 0.4 – 2.5 and 8 – 14 um. HyMap, CASI, and/or GER sensors will be used to collect the hyperspectral data. If possible, high spatial resolution AVIRIS data will be acquired. GER and/or Terra ASTER sensors will be used to collect thermal infrared temperature data. Funds are allocated each year for the collection of the remotely sensed data.
Soil taxa and vegetation spectral “endmembers” are extremely important when conducting multiple-date hyperspectral investigations of soil capped hazardous waste sites to identify compromise. When they are not available, it is necessary to use “image-derived endmembers” that are not as reliable as those based on a quality library of soil and vegetation spectral “endmembers” (Filippi et al., 2001).
First and second year research will concentrate on the development of the fundamental soil taxa and vegetation in situ and optical remote sensing-derived endmembers. Subsequent years will investigate the spectral characteristics of the most important soil taxa as they are subjected to various compaction and soil moisture treatments. Subsequent years will also investigate the spectral characteristics of the biomass and LAI of the most important vegetation types as they are subjected to: a) variations in moisture content (i.e., relative turgidity), b) the introduction of invasive vegetation types, and c) stress due to the introduction of pathogens or other agents. The in situ and remote sensing-derived endmember spectra of important invasive species (e.g., scrub-shrub, pine seedlings, undesirable grasses) will be documented and placed in the spectral library. Knowing how the various soils and vegetation types should appear in a baseline condition and their characteristics when they are subjected to various treatments should provide significant valuable information about what is taking place on a hazardous waste site cap.
Hydrologic Information: A detailed hydrologic map is indispensable for modeling the flow of water onto and off of a hazardous waste site. Digital elevation models derived from in situ surveying, soft-copy photogrammetry, and LIDAR will be analyzed to determine which is best suited for the development of a detailed hydrologic model (including stream network) of a hazardous waste site. Unfortunately, the aforementioned methods only provide for surficial drainage information. Sometimes it is critical to know what is taking place underground. Therefore, ground-penetrating radar technology may be investigated at the MWMF on the Savannah River Site in a controlled environment.
Task 4. Demographic-socioeconomic Inputs to the Hazardous Waste Site SDSS
Human populations are at risk when a hazardous waste site is compromised due to natural causes or human intervention, including terrorism. It is imperative that the hazardous waste site SDSS include detailed information about the human population in the immediate, local vicinity that are at greatest risk. The SDSS will then synthesize this information and suggest robust, practical steps that will protect the human population while mitigating the effects of the release. This research will investigate the following types of demographic-socioeconomic information required, suggest how the data might be collected, and then place it in the SDSS to be modeled. A sensitivity analysis will be undertaken to see which of the demographic-socioeconomic variables are likely to account for the greatest amount of variance in the operation of the SDSS.
Therefore, it is essential that the hazardous waste site SDSS contain information on:
the spatial distribution (location) of the population within a certain distance (buffer or hydrologic distance) of the hazardous waste site (e.g., < 5 km buffer),
the socioeconomic characteristics of the spatially distributed population (e.g., number of children < 10 or aged > 65 within a 5 km radius),
where the population will be at various times of the day (e.g., a 6:00 am, 12:00 noon, or 5:00 pm), and
on a specific day of the year (e.g., a summer day in June or winter day in January).
As discussed in Task 1 of Section 1.1.2, research will be conducted to determine how to update and then distribute the decennial census of population through the incorporation of ancillary and remote-sensing derived housing information (derived from land use type or actual building counts) as described in Jensen et al. (2002). Extremely accurate local population distribution and demographic information is necessary for the hazardous waste site SDSS.
Task 5. Fauna Distribution
The fauna surrounding a hazardous waste site must not be forgotten. We will investigate the optimum types of fauna information that can be obtained on a local, regional, and national basis. It is likely that the primary information will be obtained from various state GAP and biodiversity analysis files that provide information on prevalent species.
Task 6. Infrastructure (Roads, Railroads, Airports, Utility)
Assuming that we know where the human population is residing during a critical time period after a hazardous waste site is compromised, there are still significant decisions to be made to mitigate the effects of the disaster. This is when it becomes important for the SDSS to contain detailed information about various types of infrastructure, including, roads, highways, interstates, railroads, airports, shipping lanes, and utility lines (water, sewer, electricity, gas, potable water reservoirs). The research will identify the optimum source of infrastructure information on a local, regional and national scale (e.g., Cowen and Jensen, 1999) that can be input to the hazardous waste site SDSS.
Task 7. Test the Hazardous Waste Site SDSS
The hazardous waste site SDSS will be tested when:
Scenarios will be run that simulate the compromise of various types of hazardous waste sites. These scenarios will be investigated, first on Savannah River Sites and then at other locations to be determined in conjunction with the EPA and DOE.
all the elevation, soil, vegetation, hydrologic, fauna, demographic and infrastructure information are placed in the SDSS database,
the domain knowledge consisting of important rules and/or training data have been assimilated by the expert system and/or neural network, and finally
the most important management questions have been input to the system.
Metrics
It is likely that substantial relearning/retraining of the expert system/neural network will have to take place before the appropriate combination of datasets, domain knowledge, and appropriate questions are decided upon. After this takes place, the hazardous waste site SDSS will be tested on several types of waste sites first at SRS and then at other locations. Terrance Slonecker (EPA) and Dr. John Gladden (WSRC) will be involved in the selection of the test (metric) sites and supervise the analysis of the results. They will also select personnel not involved in the project from EPA and DOE to evaluate the utility of the hazardous waste site SDSS. The most important metric will be whether private firms such as WSRC and/or public agencies such as DOE and EPA utilize the SDSS for monitoring hazardous waste sites. Inquiries and comments about the SDSS will be logged to assess community interest.
ReferencesCowen, J. and J. R. Jensen, 1998, “Extraction and Modeling of Urban Attributes Using Remote Sensing Technology,” in People and Pixels: Linking Remote Sensing and Social Science, Washington: National Research Council, 164-188.
EPA, 2002, EPA EnviroMapper National Priority List (NPC) Sites, Washington D.C.: Environmental Protection Agency, www.epa.gov.
Filippi, A. M., Jensen, J. R., Schill, S. R., Kelch, D. J. and M. M. Pendergast, 2001, “Hyperspectral-neural Investigation of Vegetation Conditions at the DOE Savannah River Site Using Low-altitude AVIRIS Data,” Proceedings, American Society for Photogrammetry and Remote Sensing, St. Louis, April 26, CD.
Jensen, J. R., 1996a, “Creation of Digital Elevation Models and Terrain Corrected Orthoimagery Using Soft-Copy Photogrammetry,” Manual of Photogrammetry, Ed. Cliff Greve, 2nd Ed., Bethesda, MD: American Society for Photogrammetry & Remote Sensing, 167-179.
Jensen, J. R. et al., 2002, Down To Earth: Geographic Information for Sustainable Development, Washington: National Academy of Sciences Press, 155 p.
Jensen, J. R., Halls, J. N. and J. Michel, 1998, “A Systems Approach to Environmental Sensitivity Index (ESI) Mapping for Oil Spill Contingency Planning and Response,” Photogrammetric Engineering & Remote Sensing, 64(10):1103-1014.
Jensen, J. R., Qiu, F. and M. Ji, 2000, “Predictive Modelling of Coniferous Forest Age Using Statistical and Artificial Neural Network Approaches Applied to Remote Sensor Data,” International Journal of Remote Sensing, 20(14):2805-2822.
Jensen, J. R., Schill, S., and D. Kelch, 2001, “The Use of Hyperspectral Imagery to Monitor and Model Characteristics of Clay-capped Hazardous Waste Sites on the Savannah River Site,” Earth Observation Commercial Application Program (EOCAP) Conference, SSC, MS: NASA Stennis Space Center, Jan. 16.
Loynachan, T. E., Brown, K. W., Cooper, T. H. and M. H. Milford, 1999, Sustaining Our Soils and Society, Alexandria: America Geological Institute, Soil Science Society of America, and USDA, 66 p.
Petersen, G., 1999, correspondence, President of the Soil Science Society of America, Pennsylvania State University.
Huang, X. and J. R. Jensen, 1998, “A Machine Learning Approach to Automated Construction of Knowledge Bases for Image Analysis Expert Systems that Incorporate GIS Data,” Photogrammetric Engineering & Remote Sensing, 63(10):1185-1194.
