This chapter discusses two kinds of defensive strategies for protecting U.S. air transportation spaces from chemical/ biological attacks: one type is based on detecting an attack through a technological sensor-based detection system and then reacting in an appropriate way; the other type involves taking actions that can ameliorate or help prevent an attack without reliance on a detection event. One example of the latter approach would be continuously cleaning the air to remove contaminants. These two strategies are not mutually exclusive; indeed, they can be complementary for certain attack scenarios.
In a defensive strategy that is based on the detection of a chemical/biological agent in order to initiate a response, the time required for authorities to respond to an attack has three components: the inherent response time of the detection system, the time required to verify the validity of a detector alarm, and the time required to decide on what action to take in response to the alarm. These three elements are discussed in more detail below.
Response Time of the Detection System
The detection of harmful substances in air requires time. Three kinds of detection schemes are discussed here: (1) continuous chemical sensors (Figure 3-1) and (2) discontinuous chemical-detection systems, both of which might be suitable for the detection of both fast- and delayed-acting chemical agents, and (3) biosensing systems (see Figure 3-2), which would typically be used to detect and/or identify biological (slow-acting) agents.
The signal from continuous chemical sensors is continuous in time. It follows changes in the concentration of the analyte up and down. The signal often originates from the interaction of the analyte with a chemically selective layer on the sensor.1 This interaction can occur either in the bulk of the layer or at its surface. The signal is then amplified by the transducer using one of four basic types of amplification mechanisms—thermal, mass, electrochemical, or optical—listed in Figure 3-1. The continuous signal typically enables the chemical sensor to respond quickly to large changes in the concentration of the target analyte, which would occur, for example, with a nearby release of chemical agents.
Discontinuous chemical-detection systems do not provide a signal that is continuous in time, but rather cycle rapidly through a series of phases such as sample collection, preconcentration, separation, and detection in such a way that the overall system is capable of providing a detection report every minute or so. Examples of such systems include ion mobility spectrometers, mass spectrometers, and chromatography-based systems. Many technologies are possible candidates for each of the different phases.2
The distinction between “detection” and “identification” is important, since it may affect the overall response time and options. A detection occurs when a chosen parameter exceeds its threshold value. The detection may be nonspecific—that is, it registers the occurrence of an anomaly but does not necessarily indicate the presence of a particular threat substance. By contrast, identification establishes the identity of the threat substances in a given set. Nonspecific detection systems may have a relatively rapid response time compared with that of specific identification systems, but the former typically provide a lower confidence level that a threat substance is in fact present. In some cases, an alarm from a rapid but nonspecific detection system may be used
Sensors that operate on the basis of interactive chemical surfaces are chosen for discussion here because they are likely candidates for the detection of chemical agents. However, many other sensor types are possible.
Figure 4-2 in the next chapter offers a partial list of technologies being investigated for various stages of chemical/biological agent detection systems.
to initiate “low-regret” responses (those that could provide some protection to potentially exposed people without causing too much disruption—e.g., shutting down a heating, ventilation, and air conditioning [HVAC] system) or to trigger an analysis by a more specific, but slower, identification system.
The response curves of all analytical methods, including continuous sensors and discrete assays, have important common characteristics, shown in Figure 3-3. The range of concentrations in which the sensor responds is called the dynamic range. It is bounded on its upper end by the saturation limit and on its lower end by the detection limit. Increasing concentrations of interfering compounds shift the detection limit to higher concentrations, thus reducing the dynamic range. Therefore, the detection limit always depends on the chemical complexity of the sample. The sensitivity is the slope of the response curve in the dynamic range.
As with discontinuous chemical-detection systems, the information derived from a biosensing system or assay is discontinuous in time. As noted above, the discontinuous chemical detector may be capable of cycling through its stages and providing a report every minute or so, whereas an assay typically takes some hours to complete. The assay consists of two or more discrete steps, such as sample introduction, addition of reagents, incubation with reagents, measurement of signal, and so on, as shown in Figure 3-2.3 Considerable time is generally required before the steps shown in Figure 3-2 to acquire the sample and prepare it for the assay (e.g., concentration, separation from background contaminants, lysing of cells to release the target moiety, and so on). The acquisition of the sample always determines time t = 0 for the sequence. In other words, even if a biological agent release occurs in the vicinity of the detection system at time t, the analysis sequence does not begin until the next sample acquisition time.
Typical times required to complete the detection steps for chemical and biological agents are shown on the left-hand side of Figure 3-4. In the case of chemicals, the total time required for sampling, measurement, and evaluation is typically less than 1 minute. In the case of biological agents, the assay typically takes longer than 1 hour to perform. For example, in a typical assay involving the binding of target DNA strands to complementary strands on a sensor surface, the incubation time to achieve a measurable signal is about 1 hour. In the case of both chemicals and biological agents, the quality of the analysis is a function of agent type and concentration, as well as the complexity of the environment in which it is detected.
In general, there is a trade-off between the time required to obtain analytical results and the specificity of the information desired: for example, it may take somewhat less time to determine that an unusually high level of biological particles is present in the air (nonspecific detection) and considerably longer to identify the specific biological organisms associated with those particles.4 Detection systems that provide continuous monitoring of airborne organic-based particles in the respirable 1 to 10 µm size range are currently available via laser-based technologies employing Mie scattering and ultraviolet (UV) fluorescence techniques.5 These methods are significantly less sensitive and specific6 than the biological assays discussed above and would have to be validated by rigorous field tests to ensure that they have an acceptably low false-alarm rate; however, in the future they could potentially detect very large attacks with particle counts much higher than the background, or serve as a real-time “trigger” for the initiation of a more specific detection technology. Given the potentially lengthy period that may be
required to fully identify the specific organisms that may have caused a detector to alarm, it may be desirable to initiate a response before the final identification is completed.
Time to Verify Alarm Validity
All detection systems feature a trade-off between the probability of detection (POD) of the target substance and the probability of false (positive) alarms (PFA). POD refers to the probability that the instrument will detect a threat material that is present; PFA refers to the probability that the instrument will alarm when a threat material is not present at a given threshold level. The overall concentration of the target substance affects this trade-off: higher concentrations are easier to detect, resulting in performance closer to the optimum operating point (perfect detection with zero false alarms). In addition, where data are accumulated over time, one can increase POD and decrease PFA by increasing the accumulation time.
Depending on the performance of the detection system technology (that is, on the POD and PFA), one might have greater or less confidence that an alarm represented a real chemical/biological attack, and if PFA were significant it would be necessary to have an independent means (e.g., a separate detection technology) of verifying the validity of the alarm. This process would take additional time—the alarm resolution time. In some cases, it might be possible to initiate certain low-regret responses while an alarm was being resolved.
The trade-off between POD and PFA is affected by the alarm threshold setting of the instrument; a low alarm threshold results in a higher POD but also more frequent false alarms, whereas a higher threshold results in a lower POD as well as a lower PFA. A higher threshold increases the rate of false negatives—that is, the instrument fails to alarm when the threat material is present. By combining two orthogonal detection technologies, one can lower the detection threshold and the false-alarm rate achievable with a single technology, although at a higher cost.
One reason for false alarms from detection systems is the presence of interfering molecules or organisms from the background environment that may be present in the sample. These may be chemically or biologically similar to the target threat agents, causing a similar response in the detector. Thus, the specificity of the detection system (i.e., its ability to distinguish the target analyte from the background) is a critical factor determining its performance. Although chemical threat agents tend to be quite distinct from most background molecules likely to be encountered in the airport environment, the same may not be true of biological agents. The conditions in an airport terminal, in which large numbers of people are coming and going, removing coats, coughing, and so on, are likely to produce a significant and fluctuating background of biological aerosol particles that could mimic the fluctuations likely to be seen in an attack with biological agents and therefore create false alarms. In the case of some threat agents (e.g., anthrax or plague), there may also be small natural concentrations in the environment that could trigger false alarms. In order to set detection thresholds appropriately and to minimize false alarms, it will be important to characterize the background levels and fluctuations of various kinds of relevant bioaerosol particles over time and in various airport locations.
Time Required for Deciding How to Respond
Once a chemical/biological detector has alarmed, authorities must make decisions about how to respond. One immediate action is to try to determine the validity of the alarm, as discussed above. In general, the level of response should be commensurate with the level of confidence that the alarm reflects a real terrorist attack. In the case of an attack with slow-acting agents, the alarm may be the only indication that an attack has taken place. Thus, if the detection system has a relatively high PFA, or if the identification of the organisms causing the alarm will require an extended period, an appropriate initial response may be to initiate low-regret actions such as shutting down the HVAC system or increasing the pressure in adjacent air spaces, rather than immediately notifying the potentially exposed individuals that they need to seek medical attention. However, if the detection system has a very low PFA, more extreme responses may be appropriate. The POD/PFA trade-off can be used to evaluate the effects of the various alarm settings and response policies, noting again that there is a trade-off between decision quality (high POD, low PFA) and response time.
To minimize the time needed to make a decision on the response to choose in a detection-based defensive strategy, it is necessary to have contingency plans in place for responding appropriately to the alarm situations likely to be encountered. These plans should include an array of options of graduated intensity keyed to the quality of information available. They should include emergency changes to the operation of the HVAC system, evacuation of potentially exposed individuals, isolation of affected terminal areas and passengers in order to limit additional exposures, notification of passengers and workers potentially exposed who may have left the area, provision of supplies for early medical treatment, and decontamination of affected areas to facilitate timely reopening of the facility and to minimize the economic impact of an attack. Response time will be reduced if the roles and responsibilities of all the various decision-making authorities (those in charge of airport operations, security, fire prevention and control, health and safety, environmental protection, and so on) are clearly defined and coordinated, and appropriate plan documents, required approvals (e.g., for decontamination plans), and training are in place ahead of time.7
Of equal importance is that decision makers have a well-designed visual display of the information needed to make the appropriate decision and that they are well practiced in making decisions under time urgency. If the political and economic consequences of a course of action are great, the decision time for response is likely to be longer. There has been little research on these important aspects of human decision making in the security context, although much relevant research is available from comparable domains, such as operating rooms and fire-control centers.
Improved Visual Surveillance
As discussed above, in an attack involving the use of a fast-acting chemical/biological agent, it is likely that released agent would begin producing symptoms in the exposed population before or at about the same time that the agent reached any deployed detector. Therefore, the fastest “detector” of this type of attack may be the visual observation that individuals in the terminal or aircraft are collapsing or behaving in an unusual manner. One option for improving the capacity to recognize such attacks rapidly would be to deploy enough surveillance cameras to observe the spaces where large numbers of people gather and to feed the output back to a central monitoring point. Unlike a technological detector, which is only capable of responding to a limited number of toxic chemicals that have been anticipated and whose signatures have been placed in a reference library, the surveillance camera is a “functional detector” that can be expected to “see” the effects of all fast-acting toxic chemicals that are present in sufficient concentration to cause symptoms.
Given the large amount of data generated by surveillance cameras and the infrequency of attacks, human monitors would require a technological alert system, and the overall response time to a possible attack might well be reduced if
human monitors were provided with technological backup. Software programs could be developed that could be trained to recognize crowd behavior patterns that would be characteristic of an attack with a fast-acting chemical/biological agent, and these could provide a rapid, continuous method of monitoring the surveillance data.8 The issue will be the optimum allocation of function between humans and algorithms to maximize POD while minimizing PFA. Neither humans nor algorithms alone are completely effective for these purposes; hybrid systems typically produce better performance.
Critique of Detection-Based Defensive Strategies
As outlined above, a critical parameter determining the value of a detection-based strategy for defense against chemical/biological attacks is response time. In an attack involving a fast-acting agent, the agent would reach the detector and produce a response in about the same amount of time that it took to begin producing symptoms in the exposed population. If a chemical-detector system were reliable, were sensitive to most possible attack agents, and had an acceptably low false-alarm rate, such an alarm could have value in confirming the fact of an attack, helping to pinpoint its location, or alerting authorities to a possible release of agent in unoccupied areas of a facility. Furthermore, if the detector could identify the agent used unambiguously, this information could help in formulating an appropriate response and speeding up the administration of antidotes and subsequent medical treatments. Such an early-warning detector might be particularly useful in some scenarios, such as the release of agent in an unoccupied area that could migrate to occupied areas, or a release into air intakes.
The value of a detection-based strategy in responding to an attack involving a slow-acting agent depends on the response time. If the detection can be made within about a minute of the initiation of such an attack, the spread of the agent may be limited to the area in the vicinity of the release, and it may be possible to warn and evacuate individuals who are farther away, thus preventing their exposure. If the detection can be made within approximately 1 hour (a typical passenger residence time in an air terminal), it may be possible to notify the potentially exposed population before people leave the area and isolate the affected area so as to limit new exposures. Provision needs to be made for safe-handling areas and procedures for any sources of hazard. If the detection of slow-acting agent can be made within several hours, it may be possible to alert potentially exposed populations through news broadcasts and other mass-media outlets to seek medical treatment or to divert, land, and quarantine affected airplanes. Even if a detection is not made for several days, an identification of the specific organisms involved and analysis of any factors that make them unique can aid the medical treatment of victims, the restoration of affected areas, and the forensic investigation to find the perpetrators.9
It is likely that with continued investment in research and development, it will be possible in the future to reduce the detection time for biological agent attacks below 1 hour, while maintaining or increasing the POD.10 A faster response time with high POD would increase the benefits of detection systems and the effectiveness of detection-based strategies.
The feasibility of detection-based defensive strategies depends on the performance and cost of detection systems. Some of the factors that determine performance have been discussed above (e.g., POD and PFA, sampling frequency, assay analysis time). Others factors include the detection limit (the smallest amount detectable, always a function of the chemical or biological specificity); the detector sensitivity (the ability to detect small changes in the target concentration); the chemical specificity (the ability to separate analyte from background); and the range of agents that can be simultaneously detected. Further, a well-designed operator interface and training are needed to help the operator establish an appropriate level of trust in the system. Cost factors include the initial cost of the instrumentation as well as the recurring operating costs for maintenance, calibration, replacement of spent reagents, regeneration of the sensor surfaces, and other assay costs, including training and labor costs.
A successful detection system would have to be capable of detecting a large palette of chemical/biological agents simultaneously. This would likely require an array of sensor elements that would respond in a unique way to different agents. Nevertheless, the system would only be capable of recognizing agent signatures that had been anticipated and included in a reference “library.” This problem, together with the fact that there are thousands of toxic chemicals that could potentially be used in a terrorist attack, represents a fundamental limitation on the technological detection-based strategy. The situation is somewhat better in the case of a biological attack, since the range of agents that could be used in practice is significantly smaller. Unanticipated agents, those whose signatures had been modified by an attack perpetrator, or those deliberately embedded in complex mixtures of interfering compounds might well go undetected.
See, for example, information available at http://www.vistascape.com. Accessed October 11, 2005.
An additional option might be the collection of historical air samples for later analysis if an attack is suspected. However, there is a significant cost to collecting and storing samples, and these costs may well outweigh the expected benefits.
National Research Council, Sensor Systems for Biological Agent Attacks: Protecting Buildings and Military Bases, Washington, D.C.: The National Academies Press, 2005.
The feasibility of detection-based defensive strategies also depends on how the detectors are deployed and how they are actually used. Deployment considerations include the number and placement of detectors, whether in open spaces or in HVAC ductwork. In this respect, airport terminals are likely to be more difficult to protect by this strategy than are aircraft, owing to the vastly greater air volume and necessarily greater physical spacing between detectors in terminals. To the extent that more than one type of independent detection or verification system is needed to achieve acceptable POD and PFA, the system costs are multiplied.
NON-DETECTION-BASED DEFENSIVE STRATEGIES
Defensive strategies that do not depend on the technological detection of a chemical/biological attack include steps that can be taken prior to any attack to reduce the probability of the attack or to reduce its severity, and the use of continuous air treatment to remove all contaminants. For attacks involving slow-acting agents, these strategies obviate the lengthy delays needed for sample collection, preparation, assay, and reporting (see dashed lines of Figure 3-4 and the discussion below).
Steps for preventing or deterring a chemical/biological attack might include improved security—for example, ensuring that only authorized personnel have access to the air intakes of terminal HVAC systems and the ventilation systems connected to aircraft on the ground. As noted in Chapter 2, the air intakes in terminals could be used to rapidly spread threat agents throughout a large area in any attack. More visible security personnel might also have a deterrent effect on would-be perpetrators.
Proactive steps that could help to mitigate the impact of any such attack include balancing the air-handling systems in different regions of the terminal to reduce drafts that could spread threat agents (although with thousands of people coming and going, the effectiveness and practicality of this step would have to be verified), and enabling the rapid shutdown of HVAC systems in the event of a suspected attack in order to reduce the spread of agents. Empirical studies involving the transport of released agent simulants in various transportation spaces could help to illustrate how threat agents having various physical, chemical, and biological properties would spread through the spaces; this process could help create suggestions of other proactive steps that might be taken.11 It might also be prudent to ensure that critical spaces within the air transportation system (emergency-control rooms, control towers, and so on) have access to an independent supply of clean air and are maintained at a positive pressure with respect to their surroundings, in order to reduce the likelihood that chemical/biological agents released in surrounding publicly accessible areas might disrupt these critical nodes.
Continuous Air Treatment
The strategy of continuous air treatment would be aimed at providing “clean air” to all users of transportation spaces, both in airport terminals and aircraft. This approach would be analogous to that of municipal water-treatment programs that continuously treat water supplies so as to provide clean water to city residents. The current “gold standard” for air cleaning in hospitals and industrial clean rooms is a combination of high-efficiency particulate air (HEPA) filters to remove particles along with carbon adsorbent filters to remove organic compounds. This approach can be expensive and bulky; in addition, carbon filters have a limited capacity and so must be changed at regular intervals. Other approaches being explored include catalytic oxidation of the air stream and use of hybrid systems such as a plasma oxidation pretreatment followed by carbon filtration, in which the plasma oxidation helps to offset the limited capacity of the carbon.
Given the much smaller air volume inside an aircraft as compared with that in an airport terminal, it could be technically easier and cheaper to implement the continuous air-treatment strategy in an aircraft.12 As discussed in Chapter 2, the air filtration system in modern airliners efficiently removes particulate matter, including aerosolized pathogens, that passes through them, although current filter systems would not provide protection against fast-acting chemical agents.
In terminal areas, the air volumes are much greater, and typical HVAC filtration systems do not remove aerosol particles from the air as efficiently as do aircraft Environmental Control Systems (see Chapter 2). Thus, the costs and benefits of various enhanced filtration and air-cleaning strategies would have to be carefully assessed. An ancillary benefit to be considered would be the reduction of the transmission of common ills such as cold and flu viruses (or more serious viruses, such as the severe acute respiratory syndrome [SARS] virus) among airport patrons.
Critique of the Non-Detection-Based Strategy
A defensive strategy against chemical/biological attacks that does not rely on a detection event to initiate a response
has many appealing aspects. It would avoid many problems: the likely high cost of purchasing and operating a network of detector systems; the coverage problems that may be inherent in the network owing to the large spaces involved (especially in terminals) and the concerns associated with the number and placement of the detectors; issues associated with the resolution of false alarms; and the time delay between a detection event and the initiation of a response, which may extend to hours or longer, especially for the identification of slow-acting biological threat agents.
A non-detection-based strategy could be implemented immediately with existing technologies and would not require the years of research, development, and testing that would be needed to qualify detection technologies for deployment in the air transportation environment. Although this approach would not provide protection against all threats and there would be increased costs—for example, for improved filtration systems and increased security—implementation could be stepwise, and costs would likely be much lower than for detection-based strategies.13 Costs might also be offset to some extent by the ancillary economic benefits of cleaner air; in fact, it is conceivable that airports and/or airlines that advertise the provision of “clean air” might enjoy a competitive advantage.
In an attack involving a fast-acting agent, the occurrence of the attack per se would not be in doubt, and a strategy using videocamera surveillance coupled with pattern-recognition software could allow for a more rapid response than could a technological detection-based strategy. To identify the specific agent used in an attack, specialized equipment could be brought to the attack site as part of the initial response. In other words, a technological sensor-based strategy has no apparent advantage over a visual observation-based strategy for an attack involving a fast-acting agent; in fact, a surveillance camera will “detect” a much broader range of toxic substance releases. Its specificity is defined in terms of acute toxicity rather than in terms of a predetermined list of anticipated chemical agents.
In the case of an attack involving a slow-acting agent, the non-detection-based strategy would begin to mitigate the impact of the attack immediately by continuously removing biological aerosol particles from the air, as illustrated in Figure 3-4. This approach would bypass the response time delay of a detection-based approach associated with a biological assay, alarm resolution, and response decision making.
Neither strategy would prevent the exposure of individuals in the immediate vicinity of the agent release site;14 however, in the non-detection-based strategy, there may be no recognition that an attack has occurred until days later when victims begin to exhibit symptoms. By then, exposed individuals would be geographically dispersed, and it would take considerable time to connect the cases forensically to an initial exposure point.15 With no early indication that an attack had occurred, there would be no means of locating and isolating the release point so as to prevent further exposures as passenger traffic through the site continued.
In the scenario involving slow-acting agent, the detection-based strategy offers significant benefits: even if the detection of an attack and the identification of the agent used took an hour or more, action could still be taken to notify and pretreat exposed individuals while they were still in the vicinity, prevent new exposures from occurring, remediate the site, and initiate a forensic investigation. Thus, in terms of mitigating the impact of an attack with a slow-acting agent, the early detection and identification of agent and continuous air-treatment approaches could be complementary.
Each of the defensive strategies discussed in this chapter—the detection-based strategy and the non-detection-based strategy—has strengths and weaknesses in protecting U.S. air transportation spaces from terrorist attack with chemical/biological agents. The non-detection-based strategy has several undeniable advantages—it is cheaper, has a faster initial response, is more robust with respect to a wide variety of potential threat agents, and can be implemented now. Thus, combined with improved video surveillance, it represents a reasonable baseline defensive strategy for protecting U.S. transportation spaces against chemical/biological attacks.
In the scenario in which an attack is perpetrated with a slow-acting biological agent, the detection-based strategy has some added benefits, particularly if the agent detection and identification time can be reduced to less than 1 hour. For this reason, developments in detector-system technology bear watching, and, if systems become reliable and cheap enough, consideration should be given to deploying them along with the baseline air-treatment systems to counter this threat. Owing to the smaller, more controlled air volumes involved, it is likely that both detection and air-treat-
ment systems will be technically feasible for the protection of aircraft before they become feasible for the protection of airport terminals.
The costs and benefits of both strategies could be evaluated more easily if models could be developed to simulate the spread of chemical/biological agents released under various scenarios in various transportation spaces, as well as their removal or neutralization via continuous treatment technologies. The committee’s recommendations for the roles and responsibilities of government agencies in developing these models and its recommendations for exploring these defensive strategies are discussed in the next chapter.