Application Challenges for GIScience: Implications for Research, Education, and Policy for Risk Assessment, Emergency Preparedness and Response (RAEPR)

Abstract: Understanding geographic information is critical if we are to build and maintain livable communities. Since computing has become almost ubiquitous in planning and managing our communities, it is probable that advances in geographic information science will play a founding role in smarter decision making, available to all. We examine the challenges that occur between humans and their environment under conditions thought to be hazardous to life and habitat. Risk assessment, emergency preparedness and response are reviewed, and results from focus groups at the UCGIS Summer Assembly (1999) which identified and recommended priorities for research, educational and policy contributions to RAEPR are documented.

THE APPLICATION CHALLENGE - RISK ASSESSMENT, EMERGENCY PREPAREDNESS AND RESPONSE (RAEPR)

This application challenge is mainly concerned with the interaction between humans and their environment under conditions thought to be hazardous to life and habitat. This application challenge is not only multi-faceted as its title implies, it covers a wide range of disasters, many with fundamentally different underlying processes (such as earthquakes, hurricanes and war). Even though the processes that generate the disaster might be fundamentally different, techniques to assess risk, evaluate preparedness and assist response appear to have much in common and can share and benefit from advances in Geographic Information (GI) Science (such as data acquisition and integration; data ownership, access and liability issues; interoperability; and more).

Natural hazards and most often human generated hazards do not recognize political boundaries, yet in order to effectively mitigate against disaster, manage rescue and response operations, or to organize and deliver relief, policy must be generated and usually administered within politically defined boundaries. Geographic Information and the systems within which they are collected and managed, have particular utility in modeling and analysis which transcend political boundaries, while providing the necessary structure for assisting the implementation of policy within spatially unique administrative areas.

In a similar vein, while hazards do not often recognize land use differentiation, the recovery, the cost and impact on society is often greatly affected by this land use differentiation. In some circumstances, the hazard itself is modified and often magnified by heterogeneous landscapes and land use, such as those found where humans meet nature. These conditions are difficult to map and virtually impossible to model without the use of concepts, tools and technologies which are evolving within GI Science. In order to assess and mitigate risk to human life and property, and in order to respond effectively, we must develop predictive and operational models that are embedded within GI Systems (GIS).

A post-disaster statement might conclude that if we knew then what we know now, we could prevent or at least reduce the risk, damage and loss, and shorten the recovery period. Since GIS and related technologies provide an operational forum for realizing this statement, the effort here begins the process of answering the question: what are the challenges for GI Science?

The Contribution of GIScience to RAEPR might best be navigated within a paradigm which at the very least might be represented as a three dimensional grid but more likely depicted as a graph with three axis as illustrated in Figure 1.0. One axis represents the hazards as we commonly refer to them: i) natural hazards such as earthquakes, volcanoes, tsunami, landslides, fires, floods, tornadoes, hurricanes, drought, freeze; and ii) human induced hazards such as health related epidemics, social unrest, war, infrastructure failure and collapse, toxic spills, explosions and fires (accidental or otherwise). Along a second axis we represent time which can characterize actions taken such as: pre event (proactive – risk assessed), during the event (reactive - response), and post event (reactive - recover). The third axis encodes action taken or response sought by the application of GI Science to RAEPR, such as: discovery, planning, mitigation, management, settlement and policy.

GI Science and related technologies have already contributed in many areas encoded within the paradigm. There are numerous examples of ongoing projects to predict hazards, assess the risk to human life and property, assist response during an emergency, discover and recover from damage, manage ongoing hazardous conditions, plan and mitigate for future hazards, and impact policy and decision making. To navigate a small sample of these will not only serve to point out where GI Science is already contributing to RAEPR, it will also help us understand the future GI challenges for this application area.

Although the following list of hazards is not all inclusive and some content may appear to overlap with other UCGIS application challenges for GI Science, it is an appropriate list to begin the discussion for RAEPR. In the interests of brevity and breadth of coverage, each hazard is not addressed in great detail, but rather is reviewed for the role GI has played which include predicting, responding to, managing and recovering from these disasters.

As with any hazard, in order to reduce the loss of life and damage to property, public safety officials, policy decision makers, and the general public must be aware of potentially hazardous conditions well in advance. In many past disasters the general public would have been able to themselves respond in a crisis if they had knowledge of existing conditions. GI Science in its research and education initiatives appears to be able to offer concrete support here.

In most of the cases researched it was discovered that the focus on mapping was a major part of the effort. This is not surprising as most solutions involving GIS are data poor until they become part of an accepted set of procedures. However the mapping procedures, and how information is being displayed, appears to have been impacted by the advancing of technologies within GI Science. Simply encoding where some single variable existed is being replaced by maps depicting combinations of variables and their contribution to and potential for failure in hazardous conditions.

These data and information are being made more available to the general public due to advances in and the acceptance of the World Wide Web (Web) technology. It is likely that advances in Web technology that have greatly impacted RAEPR mirror the rate and potential impact of advances in GI Science to RAEPR. Rather than look too far into the future, we choose to respond to existing conditions as we discuss RAEPR challenges.

Earthquakes can destroy human infrastructure and habitat, killing and impacting large populations, especially in urban areas [47]. Although the1989 Loma Prieta earthquake was considered by some to be a wake-up call, it certainly reminded others that proactive mitigation efforts pay off as damage and loss of life was minimal for such a large quake in such a populated area.

Major earthquakes of the recent past include Adana-Ceyhan - Turkey (1998), Izmit - western Turkey (1999), Taiwan (1999) and Hector Mines - Joshua Tree, CA (1999). Earthquakes can effect any area within a broad zone and pose great risk to human life and infrastructure. Unlike hurricanes and often volcanoes, predicting when an earthquake will happen still eludes us, however where they will occur is well mapped by existing fault lines.

Of the many seismic digital mapping projects that have been undertaken, one of the most notable state projects stems from the 1990 California Seismic Hazards Mapping Act which required the California Department of Conservation, Division of Mines and Geology (DOC/DMG) to map seismic hazard zones and identify areas of risk that are subject to potential ground failure. The purpose of these maps is to help cities and counties regulate development in hazardous areas, to indicate areas requiring mitigation and to assist in making disclosures for the California Natural Hazard Disclosure Act (AB 1995). These maps show amplified shaking hazards zones, which are defined as areas where historic amplified ground shaking has occurred or local geological and geotechnical conditions indicate a potential for ground shaking to be amplified to a level such that mitigation would be required. They also depict areas of past or potential liquifaction (ground displacements) and past or potential earthquake induced landslides. Urbanized areas have the highest priority for mapping and to date DOC/DMG has mapped most parts of Alameda, Los Angeles, Orange, San Francisco, Santa Clara and Ventura Counties at a 1:24,000 scale [48]. There are plans to release and distribute these maps to the public on the Web in a variety of data formats to likely include GIF, PDF and various other formats which would be compatible with the most popular GIS software.

At the federal level, the USGS has produced the National Seismic Hazard Maps, which were made available on the Web in 1996. These maps, which cover the conterminous US, depict probabilistic ground motion and spectral response with return times of approximately 500, 1,000 and 2,500 years. The nation is divided into two regions (central/east and west) which use separate calculations for attenuation relations. For the western portion, the maps use a grid spacing of .1° (for the east it is .2° ). For grid cells with historic seismic events, seismic hazard is determined based on the number of events greater than the minimum magnitude. For areas with little historic seismicity, "background zones" were created based on discussions at regional workshops [16].

At the federal level FEMA's predictions (using GIS to assist) brought unprecedented efficiency to the process of speeding relief to victims of natural disaster. After the Northridge earthquake in northern California on January 17, 1994, FEMA said that 560,000 households would be affected; the agency received about 600,000 applications for help.

A notable local or regional organization in the area of seismic hazard mapping is the Association of Bay Area Government (ABAG) [51]. With the help of the USGS and the National Science Foundation, ABAG has been using GIS technology since 1975 to produce seismic hazard maps for the Bay Area. The maps, which include designations of fault study zones, ground shaking intensity fault traces and tsunami inundation zones, are easily combined with other data sources, such as the US Census Bureau Tiger street and boundary files, to help local planners in land use decisions and mitigation planning [50].

Volcanic phenomena can destroy vast areas of productive land and human structures destroying and killing the population of entire cities [63]. Major eruptions of the recent past include Mount St. Helens (1980), El Chichon (1982), Nevado del Ruiz (1985), Unzen (1991) and Pinatubo (1994). Mount Rainier currently has a potential to threaten Tacoma and Seattle, and Popocatepetl [65] menaces an area near Puebla and Mexico City home to about 10 million people. Hazardous volcanic phenomena range from passive gas emission and slow effusion of lava, to volcanic explosions accompanied by development of a stratospheric plume with associated dense descending currents of incandescent volcanic ash and rocks that race at high speeds along the surface away from the volcano. Mass movement of surficial materials take the form of rock falls and avalanches or even the sudden collapse of large sectors of the volcanic edifice. These phenomena and their associated water-saturated debris flows are extremely dangerous geologic events and have caused tens of thousands of deaths during the past two decades [66]. In many cases the loss of life could have been reduced if public safety officials and the general population were aware of the potential effect of the phenomena on their local environment.

Hazards related to volcanoes in the USA are administered by the US Geological Survey. The eruption of Mount St. Helens that began in 1980 killed 79 people and disrupted the areas for several years is the worst case in the USA [64]. Long Valley, California, experienced several crises in the past three decades and is potentially dangerous. Mount Rainier (and other Cascade volcanoes) present a risk to a very large population and infrastructure. Hazard maps at various scales exist for most of the potentially active volcanoes of the United Statesnited. In contrast, most dangerous volcanoes in developing countries lack adequate hazard assessment and map coverage.

The use of GIS in volcanic hazards studies is very modest. The first papers appeared in the late 1980s and about two papers per year have been published during this decade. About half of the topics addressed have been mass movements (landslides & debris flows), and the remainder treated general topics. Sophisticated themes like distributed computing, visualization, use of large data sets, and interactive modeling and analysis are lacking.

Volcanoes usually present a known source area of threat, in contrast to earthquakes which could effect any area within a broad zone. This makes them particularly appropriate for GIS analysis. In the United States the geologic histories of most volcanoes are sufficiently understood to forecast the types of phenomena to be anticipated. The relative magnitude and frequency of future events are harder to predict. A complicating factor for volcanoes is that the repose time since the previous event may be very long. The inhabitants surrounding the volcano may have a belief that even if there were eruptions in the past, nothing will happen in their lifetime. At any rate, they are willing to take a chance that they will be safe.

The last decade of the 20th century witnessed the development of several forms of computerized models for volcanic eruptions and their associated hazards [63]. Unfortunately, these have not been linked to interactive GIS systems. Computation, communication, and information technologies during this period advanced at a faster rate than the development, testing, and utilization of controlled scientific models. In general, posters, still images, or video scenes of events at other volcanoes were the main methods used to explain the phenomena to the public safety officials and to illustrate potential events at a volcano in crisis to the local inhabitants. Only in a few cases were advanced technologies or computer models used in the development of volcanic hazard maps.

Tsunami, like earthquakes, are difficult to predict, but their impact zone along the coastline can be mapped and early warning can result. A National Tsunami Hazard Mitigation Program was initiated in July 1994 when the Senate Appropriations Committee directed NOAA to formulate a plan for reducing the tsunami risks to coastal residents [44]. The program is designed to reduce the impact of tsunamis through hazard assessment, warning guidance and mitigation. The first step of producing Tsunami Inundation Maps is essential to assess the tsunami hazard. The center for Tsunami Inundation Mapping Efforts (TIME) within the Pacific Marine Environmental Laboratory of NOAA (NOAA R/PMEL) [5] was created for the purpose of development, maintenance, and upgrading of maps which identify areas of potential tsunami flooding.

The Pacific Marine Environmental Laboratory maintains large data bases related to the research and exploration of hydrothermal vent processes and applies GIS to integrating multidisciplinary data sets to create both a map gallery and an internet live map [45]. The states involved in the PMEL Tsunami Program are: Hawaii, California, Oregon, Washington and Alaska, and they seek to mitigate tsunami hazards by focusing development on improved tsunami inundation maps, hazard assessment tools, and advanced technology to increase the speed and accuracy of tsunami forecasts and warnings [30].

Although landslides can destroy human infrastructure and potentially be deadly, except for a few famous incidents, their impact is generally localized and predictable. The USGS has been extremely active in mapping landslide hazards and in developing new methods and models for assessing and analyzing these hazards. In anticipation of the heavy El Niño rains in 1997-98, scientists from the United States Geological Survey’s (USGS) San Francisco Bay Landslide Team (SFBLT) created landslide hazard maps of the Bay region [52]. Following the rains, the San Francisco Bay Area Region Project and their Landslide Hazards Program, both of the USGS, conducted inventories of landslides in the Bay Area which were then used to develop digital landslide distribution databases, computer landslide models and landslide hazard maps [51]. The SFBLT created digital maps that depict areas of potential slides (slumps and translational slides), earth flows (flows of clayey earth) and debris flows (rapidly moving slides). The map layers include topography in shaded relief, road networks, hydrography, mapped distributions of slides and earth flows, rainfall thresholds for debris flows and likely debris-flow areas. Most of these data are mapped at the 1:125,000 scale (for local emergency planning) and at the 1:275,000 scale (for regional planning). These maps are part of an overall strategy to help planners mitigate and respond to disasters [23].

Additionally, the California Department of Conservation, Division of Mines and Geology (DOC/DMG) has been active in mapping landslide hazards in the state. They produce six types of maps that depict landslide hazards. Among them are: Landslide Hazard Identification Maps, 1:24,000 maps showing landslide features, landslide susceptibility, and debris-flow susceptibility. They were produced from 1986 to 1995 under the now-repealed Landslide Hazard Mapping Act; and Watershed Maps, 1:24,000 maps which include landslide features to assist in timber harvest planning and water quality protection. They were produced in concert with the California Department of Forestry. Four categories of active and dormant landslides are depicted, including debris flows, debris slides, translation slides, earthflows, debris flows and torrent tracks. These maps cover parts of Mendocino, Humboldt and Del Norte Counties [53].

Fire is a natural phenomenon in the landscape and in the long run is often considered more beneficial than hazardous. However, even Philadelphia's most famous citizen, Benjamin Franklin, understood the hazards of fire when it intrudes upon human habitat and wrote on the need to regulate urban growth in order to decrease fire and environmental hazards. Although fire is often considered a natural hazard, the extent of the hazard can be mitigated with sound land use practices and management. Today, the practice of fire suppression in both rural and urban environments mostly does not follow sound vegetation management plans and has created potential catastrophic conditions for fires. No where in the United States is the hazard greater than in California. The Mediterranean climate, the rugged topography, a shifting urban-wildland interface, and the practice of fire suppression in recent history all collaborate to create catastrophic conditions. In the hills east of the San Francisco Bay alone 5,298 structures were lost in dozens of fires since 1920 with the majority of them occurring in the last decade [24].

GIS plays a critical role in mapping and documenting fire. The California Department of Forestry (CDF) began an intensive program of mapping fire in response to legislation in the early 1980s. This legislation required CDF to map different classifications of fire hazards with State Responsibility Areas (SRAs), or areas of state fire prevention responsibility (i.e. outside of large, incorporated cities). As a result of the catastrophic Oakland Hills fire, the Bates Bill (AB337) was passed in the California legislature in 1992. This bill required CDF to work with local fire authorities to map Very High Fire Hazard Severity Zones (VHFHSZs) within Local Responsibility Areas (LRAs), generally referring to areas subject to wildfire hazard that are within incorporated city boundaries. These maps are intended for purposes of enforcing roofing and vegetative clearance requirements, in addition to serving as the basis for disclosure statements in real estate transactions under AB 1195 [59]. For both types of maps, fire hazard is determined on the basis of fuel loading, fire weather and slope, among other criteria. Vectorized fire hazard zones were overlaid on USGS topographic maps at 1:24,000, 1:62,500 and 1:100,000 scales [19, 55, 56].

Nationwide, the US Forest Service has implemented the Wildland Fire Assessment System (WFAS), based out of the Forest Service's Rocky Mountain Research Station. Unlike CDF's mapping efforts, this is not designed for long-term hazard assessment as much as for short term fire danger warning. This system constantly generates maps of fire weather and fire danger components of the National Fire Danger Rating System (NFDRS), based on daily observations from 1,500 weather stations throughout the US. Because each station is merely a sampling point, values between stations are estimated with an inverse distance squared technique using 10 km grids. The Fire Danger Rating Maps that result are based on current and antecedent weather, fuel types and the state of live and dead fuel moisture. Fuel models to be used generally are decided upon by local managers. Weather forecasts are based on data from the National Weather Service. Live fuel moisture is generated from greenness maps, derived on a weekly basis from Normalized Difference Vegetative Index data from satellite imagery. Dead fuel moisture is available on digital maps showing 10-hour, 100-hour and 1,000-hour fuels. Additionally, drought maps and lower atmosphere stability index maps are used [57].

At the local or neighborhood scale, mapping and modeling topography and fuel load based on vegetation and structures is gaining in popularity due to advances in GI technology. The 1991 Oakland Hills fire resulted in a local study integrating fire models and data inputs within a GIS to map potential firestorm risk [24]. Although much of this input data was encoded by hand, many more GIS encoded data bases have since become available with Web delivery. This simple advance has not only lead to more modeling, it has also stimulated the development and use of new fire models embedded within GIS. FARESITE, a stand-alone fire growth simulation model, is a good example of such a model. It runs within several GIS softwares (Arc Info, Arc Veiw, or GRASS) and is used to simulate wildland fire growth and behavior under complex conditions of terrain, fuels, and weather [67].

Flood zones can be mapped and floods can be predicted with some degree of accuracy. The widest-scale and most systematic mapping of flood hazard comes from the Federal Emergency Management Agency (FEMA). FEMA produces flood insurance rate maps (FIRMs) for the purposes of determining if properties lie within the "floodway" of a river system or the 100 year flood plain [58]. These maps form the basis of FEMA's policy under the 1969 National Flood Insurance Act (and later amendments). These policies call for restriction of development in the floodway and require purchase of flood insurance and/or flood proofing for structures within the 100 year flood zone. FEMA has worked in recent years to make these maps digitally available.

As part of a Map Modernization Program, Digital Q3 Flood data were developed by scanning FIRM hardcopies and vectorizing flood zones as a thematic overlay, including the 100 and 500 years floodplain (i.e. 1% and .2% annual probability of flooding). Q3 data do not contain all information from the FIRM and are not as accurate. Rather, Q3 data are intended to support regional-scale uses, such as planning activities, insurance marketing and mortgage portfolio reviews. For more precise parcel based queries, or for engineering analysis, the more detailed Digital FIRMs (DFIRMs), or paper FIRMs are more appropriate. DFIRMS include all of the information required to create a hard copy FIRM in digital form. This includes base map information, graphics, text, shading, and all other geographic data necessary to meet the standards and specifications set for FIRMs. These data provide the basis for the digital line graph of flood risks, known as DFIRM-DLG [58].

Another very different application of flood mapping technology was used to help emergency managers in North Carolina to evacuate flood-prone areas prior to Hurricane Fran in 1996. Before this hurricane, the North Carolina Center for Geographic Information and Analysis had used the Sea, Lake and Overland Surges from Hurricane (SLOSH) model to prepare several Hurricane Storm Surge Inundation Area maps for coastal areas of the state, showing the historic extent of hurricane storm surge inundation. The model was used to produce maps showing flood extent under conditions of slow and fast velocity hurricanes. These flood extents were then overlaid on 1;24,000 USGS topographic quads. Based on the SLOSH model, Hurricane Evacuation Restudy Maps were prepared which were used to guide the evacuation of residents from low lying and coastal areas. These maps were also used by other agencies, such as the Division of Forest Resources, which performed overlays of these maps with forest cover layers to predict the amount of forest damage [15].

The Office of Emergency Services of California (OES) has produced digital flood maps depicting areas at risk from dam failure. These maps are intended to be used by local and state officials in devising emergency procedures under the Emergency Services Act (Section 8589.5 of CA Government Code) and in making natural hazards disclosure statements under the California Natural Hazard Disclosure Act (AB1195) [59]. The inundation maps produced by OES represent the best estimate of where water would flow if a dam failed suddenly and completely under full capacity conditions, recognizing that later downstream land use changes may effect the extent and intensity of inundation. These digital maps were produced by scanning paper blue line copies of the original maps and are organized by county and are available from the OES Web site as PDF files [60].

Tornadoes are one of nature's most violent storms. In an average year, 800 tornadoes are reported across the United States, resulting in 80 deaths and over 1,500 injuries which is the most severe of any country in the world [37]. These violently rotating columns of air extend from a thunderstorm to the ground and are capable of tremendous destruction with wind speeds of 250 mph or more [38]. Damage paths can be in excess of one mile wide and 50 miles long.

Although GIS is employed to map and summarize the events of tornadoes, in a growing number of communities it is used in real time on the front line. On the evening of May 3, 1999, The National Weather Service (NWS) issued a tornado warning for southeastern Sedgwick County, Kansas [14]. Though not officially part of the emergency response personnel, the Sedgwick County GIS Department (SCGIS) produced more than 300 maps for the Emergency Operations Center (EOC) to get initial locations of damage reports and identify the actual properties. SCGIS provided the EOC with a probable path and damage map and estimated values based on the damage reports received up to that time.

Hurricanes can destroy human infrastructure and habitat, killing and impacting large populations across vast territory [46]. We only have to refer to a few: Agnes (1972), Hugo (1989), Andrew (1992), and now Floyd (1999) to illustrate the damage and loss to society.

After the devastation of Hurricane Andrew, FEMA upgraded its pre- and post disaster planning and response capabilities. The GIS-based system, called the Consequences Assessment Tool Set (CATS), developed by Science Applications International Corporation (SAIC), enables FEMA to predict the effect of impending disasters, like hurricanes, and quickly mobilize a well-coordinated and directed response [13, 41]. This allows FEMA to pinpoint critical evacuation areas as well as make accurate damage predictions for phenomena such as storm surge and wind damage that facilitates a quick recovery [30].

As disaster strikes, CATS using combined government, business, and demographic databases produces reports and graphics that provide emergency managers and the national media with timely information. Known damage is reported along with mapped estimates of the extent of damage and affected population. Suitable mobilization sites are identified along with nearby airstrips, empty warehouse space, and information about federal and local sources for disaster relief. When hurricane Eduardo (1996) was threatening to endanger the U.S. coastline, FEMA identified areas of potential water contamination and quickly moved freshwater supplies to those sites ahead of the storm.

In the aftermath of Hurricane Mitch, the U.S. Geological Survey's Center for Integration of Natural Disaster Information (CINDI) created a digital atlas containing more than 60 different types of geospatial information [40]. These new maps showed the locations of landslides and floods, damage to roads, bridges, other infrastructure, precipitation information, and impacts on agricultural lands. The information used to create these maps came from remote sensors as well as existing ancillary data bases such as geologic maps, airphotos, and dozens of other digital and paper sources. The maps, which are available at the CINDI web site (http://cindi.usgs.gov/), serve as a critical resource for allocating resources in short-term relief efforts, for understanding the disaster's long-term impact on ecosystems, and for planning the region's economic recovery and reconstruction.

Human Induced Hazards (Human induced disasters that impact humans and environs)

Unlike many natural hazards most human induced hazards could be prevented, reducing or even eliminating loss of life and damage to property. With a better understanding of the underlying forces which induce disasters, we can work toward mitigation and possibly elimination of some of them.

Epidemiologists use maps to log location, encode association and study the spread of disease [12, 39]. Add to the map the ability to undertake spatial analysis through advances in GI tools and a technology results which is well suited to tracking disease. Studies that quantify lead hazards [29], model exposure to electromagnetic fields [32], and monitor air and water borne diseases all benefit from the development of technologies in GI Science.

GIS was used to identify and locate environmental risk factors associated with Lyme Disease in Baltimore County, Maryland [17]. Watershed, land use, soil type, geology, and forest distribution data were collected at the residences of Lyme Disease patients and combined with data collected at randomly selected addresses to fuel a model detecting most probable locations where Lyme Disease might occur. With GIS it is much easier to combine epidemiology data and ecological data to model and predict disease spread and transmission. This data integration is essential if we hope to mitigate in this hazardous area through better health policy planning,

At a national level GIS has been used to help design a surveillance system for the monitoring and control of malaria in Israel [34]. The GIS-based surveillance system located breeding sites of Anopheles mosquitoes, imported malaria cases, and population centers in an effort to better respond in the cases of outbreaks.

On a global scale the National Aeronautics and Space Administration (NASA) established the Global Monitoring and Disease Prediction Program at Ames Research Center to identify environmental factors that affect the patterns of disease risk and transmission [1]. The program developed predictive models of vector population dynamics and disease transmission risk using remotely sensed data and GIS technologies and applied them to malaria surveillance and control [3].

Although one could argue that war is a good candidate for a health related epidemic via germ warfare, the use of GI technologies by the military has been more proactive than simply monitoring and surveillance.

The National Imagery and Mapping Agency (NIMA)[43] a major combat support agency of the Department of Defense and a member of the intelligence community, was established in 1996 to provide accurate imagery, imagery intelligence and geospatial information in support of the nation. For example, during the 1995 Bosnia peace accord, the Defense Mapping Agency (now NIMA) employed technology called Powerscene, developed by Cambridge Research Associates, to recalculate the territorial balance between rival factions as the borders were modified and adjusted based on landscape and political conditions. This interactive process was undertaken at Wright-Patterson Air Force base in Dayton, Ohio where orthorectified imagery and digital terrain elevation data was integrated to produce a Terrain Visualization Manuevering Support system [42]. This system enabled NATO commanders and peace negotiators to tour the 650-mile cease fire border and any disputed territory without endangering lives on the ground.

GIS is also used as a tracking tool for troops in training and combat, and as planning and negotiation tool for peacemakers. A prototype terrain visualization system was installed at the National Military Command Center in the Pentagon in 1994 to help support national command level missions such as locating downed aircraft (Scott O'Grady's F-16), and troop withdrawals (from Somalia) [43]. To better fuel such technology the Shuttle Radar Topography Mission (SRTM) will collect important data during an 11-day space shuttle mission planned for the year 2000. The mission, a partnership between NASA and NIMA, will use the SRTM data to generate digital elevation models and three dimensional pictures of the Earth's surface. Besides scientists using the data to study flooding, erosion, land-slide hazards, earthquakes, ecological zones, weather forecasts, and climate change, the military will use it to plan and rehearse missions, and for modeling, and simulation purposes.

GI technology is also being used for environmental monitoring and cleanup at several Navy installations as part of the Navy’s comprehensive long-term environmental action (CLEAN) program, and at the Rocky Flats Nuclear Weapons Complex [7].

There are numerous other human induced disasters which could be lessened and even prevented by integrating GI technologies. Although many successful initiatives are already underway at both the local and national level, they could greatly benefit from advances in GI Science.

The City of Winston-Salem, North Carolina, built an Integrated Network Fire Operations (I.N.F.O.) system which is designed to reduce the time it takes for firefighters to respond to emergency (911) calls and to provide information about the address of an incident to aid fire fighters in making better informed decisions and plan the fire fighting effort while en route [69]. I.N.F.O. automatically uses the address of an incident to search for any prefire and HazMat planning information that might be available (e.g., building floor plans, hazardous waste information, occupants, etc.).

The Federal Emergency Management Agency (FEMA) developed HAZUS a natural hazard loss estimation methodology software program which is useful for earthquake-related mitigation, emergency preparedness, response and recovery planning, and disaster response operations [68]. HAZUS is implemented within PC-based GIS software.

The search of hazards to determine UCGIS application challenges for RAEPR not only illustrated how Geographic Information is being integrated into solutions, it also illustrated the important role the Web now plays in communication and disseminating information to the public for mitigation, management and recovery from a disaster. Although much of the information on the Web might be represented as a document, the results and often their graphic output is clearly the result of applying GI technologies to the problem. In some instances it is clear that GI technology has advanced the information from simply data display to output from an advanced modeling effort.

UCGIS research priorities (http://www.ncgia.ucsb.edu/other/ucgis/CAGIS.html) are all applicable to RAEPR. While we leave recommendations for RAEPR specific research contributions to the last section in this paper, it is valuable to report on activity and needs for specific UCGIS research priorities here. To best illustrate RAEPR needs within these UCGIS Research Priorities, some scenario building is undertaken here.

Data acquisition and integration may be the single largest contribution area needed for RAEPR. Although models can be developed for handling disasters, making them operational on a day to day basis means huge investments in data acquisition and integration.

There are essentially three parties that have spatial information needs in an emergency management arena. These include public sector personnel like emergency managers and government agencies, private citizens, and researchers. Also, as discussed above, the disaster cycle can be divided into the temporal stages of before, during, and after a disaster. Using these two dimensions, we can define a matrix where each cell represents a given party’s spatial information requirements at each stage in the disaster cycle.

This matrix can be used to examine the information needs of various parties at various stages in the disaster cycle. For example, if we focus on the cells in the diagonal of the matrix, the cell in the upper left corner of the matrix would represent the public sector’s spatial information needs before a disaster. This would include risk mapping, emergency simulation, and any other activities that involve spatial information in emergency planning or analysis. The cell in the center of the matrix represents the spatial information needs of private citizens during a disaster. This would include evacuation orders and routing, information about the spatial extent of the hazard, and any other information that citizens might require during a disaster. Finally, the cell in the lower right hand corner would represent the spatial information needs of researchers after a disaster. This might include data on the processes that led to the disaster, the routes taken by evacuees, or any other spatial information that researchers might want to know about a particular event. It should be noted that there is a significant amount of overlap in the information needs of these parties during the various cycles of a disaster. However, the matrix supports the notion that the information needs of these parties are not identical.

A significant challenge in emergency management is delivering the appropriate information to the proper party at the appropriate place and time in a useful form. ‘Useful form’ in this context refers to the scale, accuracy, and detail of the delivered information. The UCGIS research agenda for spatial data acquisition and integration should focus on research associated with questions and problems related to acquiring and integrating spatial information to meet the various needs of the parties listed in the table above for the given time periods.

As an example, assume that an engine company from county A that has been instructed to respond to an emergency call in county B. The problem is to provide the company with the nature of the incident and the needs of the parties in distress, the location of the incident, and, ideally, and the best route to the site. This information must be delivered in a timely manner in an appropriate form, where errors in the information may have serious consequences. The fact that the information must also be delivered in a timely manner puts unique demands on any system designed to deliver this information. It implies that there is a time window within which the information must be delivered to have value.

The overall research challenge in spatial data acquisition and integration for emergency management can be viewed as one of delivering accurate, appropriate information to all the parties involved in a disaster at the proper stages of the disaster in a timely manner. There are a number of research questions that can be generated from this overarching research challenge. Namely:

Modern computer simulations of complex natural phenomena, such as rapid forest fire growth or development of a volcanic plume, require supercomputer facilities with distributed simultaneous computing on many processors. Linked to GI Systems, these models for pre-disaster planning, crisis management, and post-disaster recovery could become extremely valuable mitigation and response tools.

Although this level of analysis is not possible today, during a crisis, such a system could be highly interactive allowing real-time communication between parties and aiding in the execution of models that could be viewed remotely. This would allow scientists and civil protection agencies to apply results immediately to the current extremely dangerous conditions. Here the data must be output in various levels of format complexity allowing images and animation of various scenarios to be viewed by scientists, decision makers, and the general public (with the approval of the appropriate public safety and government officials).

It is important that any new data systems be developed on a platform that is widely compatible with those of existing data users. It is also important that these systems be designed to run on thin clients as in an emergency it is likely portable, wireless computers will be the communication tool in the field.

A key area to pursue is the dynamic representation of physical and human processes in RAEPR. GIS have not traditionally been designed to represent dynamic phenomena, but this is critical in assessing and responding to emergencies. Very little research has been conducted in this area, despite the obvious consequences of making critical decisions with inaccurate or incomplete information.

There is a need to improve our representation of risk and human vulnerability. The computational representation of human vulnerability has lagged behind the theoretical advancements in this area. As such, GIS is not representing the depth and richness of the theoretical frameworks and empirical research on human vulnerability to environmental hazards remains incomplete. Risk and human vulnerability are much more dynamic than the representations that are now being used in GIS. There is a need to be able to rapidly model and summarize alternative scenarios especially when the future is uncertain (e.g. tornado, hurricane, fire).

Before proceeding to extending representations, we must make sure that the representations we have are up to date. There are many cases where the data that emergency managers are relying on are simply not up to date. As an example, in Oklahoma during the tornadoes, the 1997 TIGER files did not have many of the schools included in the database. We need to progress to hazard, risk, and vulnerability classification systems that include multiple hazards. Most research in this area has focused on single hazard scenarios. In other words, classifying based on just one hazard. Finally, there is a need to develop representations of past disasters and events both static and dynamic: what factors led to a particular disaster? Where did the event occur? What development has taken place since the last disaster? How many hazardous events have occurred at a particular location?

The scientific domain of the cognition of GI includes humans, computers, and the earth. Research in this area centers on questions related to human conceptualizations of geographic spaces, computer representations of geographic space, and human perceptions of computer representations of geographic space. An obvious area for research in this area is risk perception and the affect of GI on risk perception. How do GI representations affect people’s perceptions of risk? How do people perceive risk, hazard, and vulnerability? Does GI amplify the perception of risk to certain hazards?

The technical problem of interoperability arises from the need to share data, algorithms, and models (DAM). What DAM need to be shared in emergency management arenas? Institutions must know about DAM that exists elsewhere before the need to share data arises. International attempts are in progress in the area of sharing geographic information for emergency management purposes. The Global Disaster Information Network (GDIN) is a prominent example [70].

GDIN is an interagency program undertaken at the initiative of Vice President Gore to assist fire and emergency management personnel. GDIN has two proposed components: 1) a national disaster information network; and 2) a global system. GDIN will operate on the Internet and possibly Guardnet (National Guard Network) during disasters to broadcast and integrate disaster management information from all sources and provide it rapidly and readily. GDIN will also promote training and communication in the areas of emergency preparedness and mitigation. It is expected to produce many benefits to include: saving lives and minimizing costs, enhancing coordination and sharing of compatible capabilities, facilitating the leveraging of existing resources; and assisting in validating and verifying information. To date, completed projects include the State of Florida Hurricane Simulation Exercise, and the State of Alaska "Information Process Flow Report". In addition, a regional, theme-based disaster information network is being developed to promote collaborative activities between the U.S. and Canada. The Red River Basin Disaster Information Network was established in response to the 1997 flood affected North Dakota, Minnesota, and Manitoba.

Experience has shown that a top-down approach to data sharing in disaster management is not entirely effective. The problem of interoperability in RAEPR must be approached by first assessing user’s needs: what GI needs to be shared? What GI needs to be acquired? What GI exists in other agencies, institutions, companies? Data sharing and interoperability of GI must occur under tremendous time constraints in RAEPR. There are incompatibilities between physically-based forecast models and the data stored in geographic databases. This is the data-integration, or coupling, problem in all its forms.

Digital Elevation Model (DEM) resolution is not adequate for many applications in risk mitigation. Present applications use simplified flow models that display results in 2-D as data files or images. Such small data sets are easy to use on PCs or workstations and computational nodes for models are widely spaced. For example, standard DEM data sets have 90 meter spacing of nodes and working files are on the order of megabytes. Although these data make very large files for ordinary computers, the detail is insufficient for realistic prediction for many natural phenomena for which small differences in topography or other parameters could have a large effect on the spatial pattern of the result. A more suitable grid spacing could be meter-scale and the data sets could be as large as one or two gigabytes. Development of such large data sets on a supercomputer could provide a valuable source that could be distributed for use at various remote GIS sites.

Current risk simulation codes work on small areas with large grids and are slow. Future codes should operate on fine grids of data sets that include the entire area of risk surrounding, for example, a volcano. For optimal use the data will be a high-density grid of topographic points (x,y,z data) at a horizontal spacing of 10 to 30 meters and a vertical increment of 1 to 10 meters. In some circumstances meter-scale might be appropriate. The areas encompassed by a single network may be as large as 50 x 50 grids. Such a large computational grid is too large for a single processor computer and hence supercomputers are necessary for the computation and visualization.

Computer simulations linked to GIS systems could permit analysis of loss of life and disruption of infrastructure that is not possible today with the current set of available tools. Sophisticated visualization systems allow public safety officials, scientists, and the general population to understand the effect of the various phenomena in their areas of interest and to design appropriate mitigation plans. A 3-D visualization system could provide an animated illustration of the areas threatened by volcanic phenomena at several scales.

Perspective views of the phenomena could be interactively manipulated to include a spectrum of possibilities ranging from individual rivers, streets, and buildings to entire disaster scenes. Overlays of images on topographic grids would create a realistic 3D appearance of the phenomena that will move in real time with data moving in and out of the system dynamically. The GIS interface could allow query and manipulation at various levels and between multiple viewers at different sites.

It is important that scientists involved in the GIS analysis can interact in several different ways. Multiple windows on their computer desktops could allow the interaction via the Internet. Such an interaction allows them to send explicit equations or mathematical expressions at the time of crisis. This facility could permit scientists to continuously update parameters and expressions that represent the current scientific state of the lava flow, for example. These changes could then filter through all others areas of the system infrastructure, providing scientists the ability to see the effects immediately. Another window displays a representation of the volcano or area around it, whether an actual picture or graphic or some other visual representation. Another window could control some general functionality of the other windows. For example, automatic refreshing of all information, what files are viewed, etc. At any time a scientist could change this set-up. If the scientist wants multiple graphic windows and no equation interface, this can be easily done. This type of interface should focus on communicating the nature of the disaster, including the magnitude, extent, uncertainty of the event. A very important element is to ascertain the risk of making a wrong decision.

Emergency managers rely on a system for managing emergencies called incident control system (ICS). ICS specifies exactly which party (e.g. police, fire, highway patrol, mayor) is to do what during an emergency and precisely how communication, authority, and many other critical facets of the emergency management process are to take place. It is nationwide at local and state levels. Is there an equivalent institutional protocol, procedure, or approach for agencies to determine exactly who will collect and share GI before, during, and after an emergency? The UCGIS could be central in developing and dissemination model data sharing procedures that address the institutional and technical issues associated with GI data sharing in RAEPR; an "ICS for GI", if you will. This might exist in the form of information sources, flows, and ultimately application. There is also the need to develop foundation data models for sharing GI that is multidimensional, multi-scale and multi-source in nature.

As methods and models of GIS analysis become more sophisticated the quality of data increases in importance. Many data sets undergo temporal adjustments which add an uncertainty to the analysis. For example, using one or two day old data in volcano forecasting at the time of the crisis would lead to a faulty conclusion. The same is true of other disasters where geopolitical or natural conditions change from moment to moment. We must be able to analyze and incorporate such temporal uncertainty in the analysis and forecast that we make.

We must be able to quantify the uncertainty in the data (and the analysis) and express this in a satisfactory mode. A major problem exists in how we report uncertainty in GIS. For example, what significance do we place on the lines on your various maps and diagrams? How do we address this issue? A case in point would be designing a hazard map for flooding on an alluvial fan where there is no defined channel and the flood has different probabilities of spreading in various directions. Another major problem is the propagation of uncertainty through the data set as we combine several sets of data of different levels of confidence and even potentially different types. Research on this topic should help to resolve this problem.

There are many interesting areas where issues in GIS and society arise in RAEPR. The media is a interesting topic for research in this area. What is their role, responsibilities, ethics, motivations in disseminating GI for warning, preparedness, response, etc.? How are new technologies like pagers, hand-held devices, and other electronic innovations affecting equity, vulnerability, and the perception of risk? How are issues like socio-economics, insurance, race, and other issues related to the application of GI in RAEPR? Public participation GIS and group decision making are other obvious areas for research in RAEPR?

The search of hazards to determine UCGIS application challenges for RAEPR illustrated the important role the Web now plays in communication and disseminating information to the public. It appears that an informed population is more prone to accept and even embrace mitigation, respond and participate in the management of a hazard or emergency, and be better equipped to assist and appreciate recovery from a disaster. Much of the information on the Web is commonly represented as a document, yet images, maps and graphics illustrating the results of some analysis are slowly finding their way onto RAEPR related Web sites.

UCGIS education priorities (http://www.ncgia.ucsb.edu/other/ucgis/ed_priorities/contents.html) are applicable to RAEPR. The most important educational needs or components that surface when one looks at individual hazards, either natural or human induced, focus on issues of certification of specialists to undertake response and settlement, public education and awareness of response during a disaster, the development of a model curriculum to develop GI Science experts for RAEPR, and the development of simulators to train rescue workers and settlement specialists. It is clear that only a few people that work within this area will require in-depth education in GI Science while most others will benefit by training on installed GIS related technology. However, training needs to be presented within the context of the profession with appropriate amounts of spatial literacy and integrated with other technologies common to the profession.

To best illustrate RAEPR needs within these UCGIS Education Priorities, some discussion is undertaken here.

Technology is playing a central role in education at the college and university levels. In some instances it serves to lower the cost of education, while in others it enhances and even makes possible some opportunities never before imagined. Distance learning taught by domain experts, Web based programs, and simulators to create better and cheaper technology, are all served by these emerging technologies.

Emerging technologies make it possible to educate more people and are even more effective in training. It is now relatively common to find Web based training courses where one can enroll and conveniently become well versed in GIS. However, unlike the rigors of the college classroom, quality, accreditation and assurance are not clearly defined and regulated. Assessing liability and assuring accountability in a disaster may call for the regulation of emerging technologies as they are applied to education.

As training and modeling in GIS become more the status quo for personnel in RAEPR, demands on technology classrooms and Internet portals will rapidly increase. Who will bare the cost of such infrastructure in the short run and will this persist and set a trend? RAEPR support is recognized during a disaster when many groups step forward to lend a helping hand, but what is being done over the long haul to help mitigate and be prepared if a disaster strikes?

The GI Science community must ensure access to the technologies and data to disadvantaged groups and impaired individuals so that they may also be effective in RAEPR in their communities. The first goal of this education priority is to ensure access and determine what is the necessary "spatial literacy" to effectively use GIS. However, in an emergency, access and equity issues quickly shift from how effective is the trained emergency worker to under what circumstances should rescue workers have access to private information? Under what circumstances may a community breach ownership rights in order to acquire and access data? In the heat of a disaster is it impossible to even address some of these issues let alone come up with solutions?

Although the basic GI Science concepts might be the same, as you cross domains, specific concepts vary. Likewise, in RAEPR, the level of GI knowledge necessary for emergency workers to carry out their jobs varies. At one end of the spectrum a worker may need to be able to read a map while at the other end a sophisticated understanding of spatial statistics might be in order. Delivering GI education under such extreme needs calls for a scaleable curriculum to increase the likelihood that GIS will be deployed properly and effectively in the RAEPR area.

Adopting these technologies and employing them in the field to save lives and property does not come easy. The emergency worker must not only have faith in the technology, they must have confidence in the data. Building this confidence starts with a sound education where the student participates in data collection so that ownership and a stake in the data buys into its use in RAFPR.

The majority of RAEPR workers need training on how to use the technology to extract information about infrastructure, follow guidelines in assessing risk, navigate and follow procedures during a crisis, and assess damage after a disaster. It is likely the majority of workers in this area will not have been widely exposed to GI technologies and professional training will play a key role in filling this gap.

To advance the state of any science, researchers must be educated so that they may lead on the frontiers of research and then train and collaborate with those emerging researchers to push those frontiers forward. Unlike professional GIS education programs where a student is trained to use the state of the art systems, research-based graduate GIS education designs and creates new technology that will eventually define the path that new technology will take. RAEPR researchers will benefit from this high level focused approach and be able to build better sensors, predictors, models and data management environments.

Learning with GIS in RAEPR employs a curriculum that emphasizes RAEPR specific topics and uses GI to study them. Since most disasters can be mapped, a GIS can provide a very effective navigation tool for dissecting problems and learning the steps necessary to deliver an effective response.

Proactive approaches to disasters lead to practice sessions where GI fueled simulators play out a variety of disaster scenarios. These simulators will play a critical role in educating the RAEPR community and its response strategies during an emergency.

Although accreditation and certification may carry with it problems associated with licensing, for RAEPR workers liability is a serious issue and accreditation and certification are most often embraced. Just as emergency response workers must be certified on their search and rescue techniques and technologies, certification on how to properly use GI data is a necessary component if quality control and assurance is to be taken seriously.

RECOMMENDATIONS FOR RAEPR SPECIFIC RESEARCH, EDUCATIONAL AND POLICY CONTRIBUTIONS

At the UCGIS Summer Assembly (1999), focus groups were formed to identify and recommend priorities for research, educational and policy contributions for each application topic. The following priorities summarize the result of that process for RAEPR.

Both natural and human generated hazards usually transcend political boundaries that are effective for defining regions used to effectively mitigate against disaster, manage rescue and response operations, or to organize and deliver relief. Since policy is most often generated and administered within politically defined boundaries we must develop new policies which emulate hazards rather than human administrative structures.

Policy and regulation are commonly applied on the landscape as a function of form. For example, brush must be cleared to create a specific size protective buffer zone around homes in a urban- wildland intermix region. Although the specific size buffer zone, represented here as a form, can easily be complied to and administered, it is naive and unrealistic to assume the impact of this specific buffer zone will be uniform over space. Advances in GI Science will bring about a shift where policy and regulation can become a function of the underlying process rather than relying on an easily administered but limited form based policy. The greater our confidence in data and models, the more likely policy will be process rather than form based.

Research and education in RAEPR is crucial as we search for conditions thought to be hazardous to life and habitat, undertake mitigation efforts, respond during emergencies to reduce loss of life and property, and settle and restore a damaged environment. In some instances, we found that early warning systems need to be built while in other instances we need to change more fundamental elements such as land use and life style. In almost all instances large data bases that contain information on humans, their activity and their habitat are necessary. We need to insure that these data are accessible to assess risk, prepare to engage disaster and aid in effective response and settlement. Although these data sets must be engineered to effectively assist emergency workers, we must also insure privacy of the individual so that exploitation cannot occur.

We set out to discover whether advances in UCGIS research and education priorities might contribute to needs within the application RAEPR. By identifying and recommending priorities for research, educational and policy contributions to RAEPR and cross referencing them with the UCGIS priorities, we identify a focus for GI Science for this application challenge. We discovered that interaction between humans and their environment under conditions thought to be hazardous to life and habitat can be facilitated through advances in GI Science.

Only after successes in UCGIS priorities are accomplished and applied in RAEPR will we be able to realize shifts where policy can be more directly linked to underlying process rather than simply the form that appears during and as a result of a disaster.

Emergency: An emergency is a deviation from planned or expected behavior or course of events that endangers or adversely affects people, property or the environment.

Disaster: Disasters are characterized by the scope an emergency. An emergency becomes a disaster when it exceeds the capability of the local resources to manage it. Disasters often result in great damage, loss, or destruction.

Risk: Risk is the potential or likelihood of an emergency to occur. For example, the risk of damage to a structure from an earthquake is high if it is built upon, or adjacent to, an active earthquake fault. The risk of damage to a structure where no earthquake faults exist is low.

Hazard: Hazard refers generally to physical characteristics that may cause an emergency (for example earthquake faults, active volcanoes, flood zones, highly flammable brush fields, are all hazards).