Gerard P. O Reilly, Raymond Hohl, Gus DeGennaro, Thomas Kopko LGS Innovations LLC.

Transcription

1 Presidential Alert Modeling Gerard P. O Reilly, Raymond Hohl, Gus DeGennaro, Thomas Kopko LGS Innovations LLC. Abstract The USA is developing an Integrated Public Alerting and Warning System (IPAWS) for its citizens led by the Federal Emergency Management Agency (FEMA) which is part of the Department of Homeland Security (DHS). FEMA/ DHS has established a goal for the system to deliver presidential alerts to 85% of the population within 10 minutes. That goal is extremely challenging, especially during the difficult period from 2 AM to 5 AM. In this paper, we show how it can be achieved for all hours of the day by a combination of TV, Radio, Internet, and most importantly, by the Commercial Mobile Alert Service (CMAS) cell phone delivery mechanisms. Index Terms Alerting, IPAWS, Emergency Management.

2 TABLE OF CONTENTS I. Introduction 01 A. Coordinate Systems 01 B. The IPAWS System 02 II. SEPARATE MODELS OF ALERTING PENETRATION 03 A. Radio 03 B. TV 04 C. Internet 05 D. Cell Phone 06 III. COMBINED MODELS OF ALERTING PENETRATION 08 IV. CONTAGION MODEL 09 A. Simple Contagion Model 09 V. EXAMPLE CMAS ARCHITECTURE 10 VI. POTENTIAL MODEL EXTENSIONS AND VALIDATIONS 11 VII. CONCLUSION 12 VIII. ACKNOWLEDGMENT 13 IX. references 14 ii

3 I. introduction Presidential Executive Order [1], in combination with the Warning, Alert, and Response Network (WARN) Act [2] have mandated the creation of an integrated public alert and warning system (IPAWS) to inform the public during periods of national emergency. This will ensure that under all conditions the President can rapidly and effectively address and warn the public over a broad range of communications devices and under any emergency condition. The FEMA/DHS has set a goal to alert 85% of the population within a 10 minute time frame [3]. This will be achieved by leveraging a combination of existing systems such as the Emergency Alert System (EAS) and HazCollect (part of NOAA) and emerging technologies such as CMAS, to form the Integrated Public Alert and Warning Service (IPAWS). These services, managed by FEMA through the IPAWS Project Management Office will enable Emergency Operations Centers (EOCs) across federal, state, county, city, territories and tribal lines of government to access the federal level alerting system using their existing alerting tools. This paper models how the 85% goal can be achieved by a combination of TV, Radio, Internet, and cell phone (CMAS) alerting for all hours of the day. It is important to note that CMAS cellular alerting is distinct from today s common SMS text messaging on cellphones. This study shows that CMAS cell phone alerting is expected to be the critical contributor to this goal since cell phone alerting can wake up a sleeping person and will be widely available. A. The 85% goal The 85% goal means that a presidential alert message has to be received by 85% of the population (directly or indirectly) at any time of the day. By directly, we mean that an alerting message is received from DHS/FEMA by an individual person. By indirectly, we mean that a direct alert received by one person is relayed to another person, e.g., in the same household, a co-worker, or call to a neighbor or relative. The authors are with LGS Innovations LLC, in Florham Park, NJ, Washington, DC, and Colorado Springs, CO. 1

4 B. The IPAWS System The IPAWS System will be the backbone that allows all levels of government to disseminate alert and warning information to the public through a common set of methods, as shown on Figure 1 from FEMA [4]. It integrates a number of existing systems into a single coherent whole: Figure 1 IPAWS Vision Architecture The message disseminators at the center of Figure 1 show the methods for getting information to the public. They include: Emergency Alert System (EAS) and Digital EAS - Used for national and local weather alerts on TV and radio. Most familiar alerting system; the crawl across TV screens. Commercial Mobile Alert System (CMAS) - Planned system for issuing alerts via cell phones (described below). Internet Services , alerting services, etc. National Oceanic & Atmospheric Administration (NOAA) Includes the Weather Service (NWS) and National Weather Radio (NWR) and the HazCollect system which provides an automated capability to streamline the creation, authentication, collection, and dissemination of non-weather emergency messages. State/Local Unique Alerting systems these include emergency telephone networks, sirens, and signage systems on roads. Standards Based Alert Message data exchange format, alert message aggregation, shared, trusted access & distribution networks, alerts delivered to more public interface devices. Emergency Manager Message Disseminators Public Emergency Alert System National EAS State EAS Emergency AM FM Satellite Radio; Digital, Analog, Cable, and Satellite TV Federal State DM-Framework DM- OPEN Cellular Broadcast Internet Services Cellular Carriers Networks Cellular Phones Web Browsers, widgets, web sites Local NOAA IPAWS compliant CAP Alert Origination Tools State/Local Unique Alerting Systems (e.g. ETN, Siren, Signage systems) 2

5 II. SEPARATE MODELS OF ALERTING PENETRATION We first define some terms. Let TV = Television Media R = Radio Media I = Internet Media C = Cell phone Note that our model is focused on the US Government s implementation of the IPAWS as opposed to a broader discussion inclusive of state and local unique alerting systems. Since an alert could come at any time t of the day, let Alert%(t) = percent of USA population alerted at ANY time t, where t runs from 0 to 24 hours of the day and is given for a specific time zone, say Central time zone. Then the other time zones are Pacific: t-2; Mountain: t-1; Eastern: t+1. Hence the DHS goal is: Alert%(t) 85% for all t ε [0,24) The US population is million people [5]. Now, we exclude young children, say less than 7 and a half years old (about 10.2% of the population[11]), which gives us million people as the target population for alerting. For simplification, we ignore adjusting these numbers further to exclude people who are also under the guardianship of another person such as, patients in hospitals or nursing homes and prisoners, who arguably do not need direct alerting because they likely would receive indirect alerts through their guardians. A. Radio We start with a discussion of radio, and the alerting possibilities with radio. Let Radio(z,t) = probability that a person in time zone z is listening to Radio at time t (Central time). See Figure 2 which shows the percentage of the population listing to the radio as a function of the time of the day [6]. During the daytime hours, it reaches a maximum of around 23%. Figure 2 - Radio listening as a function of time of day Population Penetration 25% If an alert were to go out during the daytime, about 20% of the population are listening to Radio. 20% 15% 10% At 3 AM, only a few percent are listening to the radio. % Radio Pacific % Radio Mountain % Radio Central 5% 0% % Radio Eastern % Radio average listening weighted by population Hour of day (Central Standard Time) 3

6 We also compute the average, %Radio Average(t), across all these 4 time zones (Pacific, Mountain, Central, and Eastern) by weighting the individual time zone listening percentage with the population of those states within the time zone. We then use the computed average in subsequent calculation for the radio media rather than dealing with separate time zones. Also, Hawaii and Alaska with smaller populations (total of ~2 million) were excluded here for simplicity. If desired, they could be broken out separately with their respective time zones. Let %Radio Alert(t) = average probability that a person listening to Radio at time t (central time) is alerted by the Radio media as shown in Figure 3. = Radio Effectiveness Parameter * %Radio Average(t) Figure 3 - Radio Alerting by Radio media Population Penetration 25% 20% % Radio average listening weighted by population % Radio average alerted 15% 10% 5% 0% Hour of day (Central Standard Time) The Radio Effectiveness Parameter is assumed to be 0.85 as represented by the dotted line on Figure 3. This is an important parameter which could be as high as 1.0 if everyone listening to the radio would actually be alerted by a radio alert message. It depends on the attention of the listener. Here we assumed a rather conservative number that only 85% (.85) of those listening would actually be alerted. Similar conservative estimates were applied to TV and Internet below. B. TV We now show a similar result for people watching TV over the course of a day [7]. Figure 4 TV watching as a function of time of day If an alert were to go out at 9 PM at night central time, we could get about 43% of the population 50% would be alerted with just TV. 45% % TV Pacific % TV Mountain 40% % TV Central 35% % TV Eastern 30% % TV average weighted by population 25% For much of the day, the percentages are much lower. For 20% example, at 5 AM, only a few 15% percent are watching TV. 10% 5% 0% Hour of day (Central Standard Time) %TV Alert(t) = average probability that a person watching TV at time t (central time) is alerted by the TV media as shown in Figure 4. = TV Effectiveness Parameter * %TV Average(t) 4

7 The TV Effectiveness Parameter is assumed to be 0.80 as represented by the dotted line on Figure 5. Again, it depends on the attention of the person watching TV. Figure 5 - TV Alerting by TV media Population Penetration 25% 20% 15% % TV average watching from previous chart % TV average alerted 10% 5% 0% Hour of day (Central Standard Time) C. Internet Similarly, we show a similar result for people using the internet over the course of a day [7]. Figure 6 Internet usage as a function of time of day 40% 35% 30% 25% % Internet Pacific % Internet Mountain % Internet Central % Internet Eastern % Internet average 20% 15% 10% 5% 0% Hour of day (Central Standard Time) %Internet Alert(t) = average probability that a person using the Internet at time t (central time) is alerted by Internet media as shown in Figure 6. = Internet Effectiveness Parameter * %Internet Average(t) The Internet Effectiveness Parameter is assumed to be just 0.45 as represented by the dotted line on Figure 6. The alerting could be done by , but then there is no guarantee that the might be read. There are also no national requirements for alerting over the internet, as there are with Radio and TV with the Emergency Broadcast methods. Clearly, the patterns for radio listening, TV watching, and Internet usage are very different. Radio listening peaks in midday, TV watching peaks during the night hours, and Internet usage has two peaks. However, all of these three media exhibit the dilemma of alerting, namely, how we alert people during the middle of the night, from 2 AM to 5 AM, when most people are sleeping. See Figure 11 (discussed later) which shows that only 10-15% of the population would be alerted by these methods. 5

8 D. Cell Phone We now discuss what we believe is the solution to this alerting dilemma, that is, using CMAS cell phone alerting. In August 2009, the FCC gave its 3rd order on CMAS (Commercial Mobile Alerting System). The major mobile carriers have committed to transmitting the three classes of alerts - Presidential, Imminent Threat, and AMBER alerts which would consist of 90 character messages broadcast over their cell phone networks. Cell phone users may not disable Presidential alerts and we assume that a negligible number of people will optionally disable the valuable Imminent Threat and AMBER alerts. The carriers are required to target alerts down to the county-level. Most importantly, the carriers must include an audio attention signal (special ringing) and a special vibration cadence for alerts on CMAS-capable handsets. Hence, cell phone alerting can wake up and alert a sleeping person, which solves the key problem of public alerting, as we will see. The alerted person would then turn on their radio or TV or internet access to find out the details of the alert. We mentioned previously that our target was million people to be alerted. According to the CTIA [8], there were million wireless subscribers in the USA. Hence the cell phone penetration is almost 100% of the targeted alerting population. Furthermore, Figure 8 shows the number of cellphone subscribers by carrier. The top 5 carriers had 92% of all subscribers in 2008 [9]. (note that Alltel has since become part of Verizon Wireless). In subsequent filings, the major carriers have committed to deploy CMAS on their networks. Another allowance from the FCC was that the carriers could support CMAS on new phones, and that legacy cell phones do not need to offer CMAS service. This suggests that the population will only get access to CMAS as they get the new phones with the ability to carry the alerts. Figure 7 Number of subscribers by carrier [9] (000s of subscribers) 80,000 70,000 60,000 50,000 40,000 30,000 20,000 10,000 0 AT&T VerizonWireless Sprint/Nextel T-Mobile Major carriers have said they will deploy CMAS capabilities on new phones Alltel USCellular MetroPCS Leap SunCom Centennial Other top 20 beyond top 20 CMAS deployment is twofold: Cell infrastructure by carriers choosing to deploy this new technology CMAS capable mobile phones (more limiting) This is a function of how fast people upgrade their phones over time. If we assume a 3 year refresh cycle, then 2.77% of phones will be upgraded each month over a three year period (Figure 9). Starting in June of 2010 for example, the cell phone alerting capability could be fully deployed by June of 2013 with 25%, 50%, and 75% deployments along the way. 6

9 Figure 8 Possible penetration of CMAS equipped phones CMAS % phone capable CMAS 50% deployed 75% % 100% 0 Jun-10 Sep-10 Dec-10 Mar-11 Jul-11 Oct-11 Jan-12 May-12 Aug-12 Nov-12 Jun-13 Feb-13 The dotted line on Figure 10 gives the percent of the population that is awake over a 24 hour day [10]. We assume that an awake person with a cell phone will be alerted with 100% effectiveness, that is, most people have them on and ready to use when they are awake. The Cell Phone Effectiveness Parameter is assumed to be just 0.60 for sleeping persons. This number was estimated using an internal focus group (mostly engineers) where some people didn t have the cell phone in the room where they were sleeping and others did. This parameter is the most important one which needs continued research validation. Note also that there are a growing number of households (20.2% according to CTIA [8]) that do not have wireline service in their homes; they only have cell phone service. As such, that growing population will likely have their cell phones available (in the same room) while they are sleeping. Cell Alert (CMAS%, t) = average probability that a person with a cell phone will be alerted by CMAS at time t (central time) of the day averaged across time zones with the given CMAS percent deployment. Figure 9 - Alerting by cell phone access by time of day 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% CMAS can wake up a sleeping person % Awake average CMAS 100% deployed CMAS 75%deployed CMAS 50% deployed CMAS 25% deployed Hour of day (Central Standard Time) When CMAS is fully deployed, almost 58% of the population could be alerted by this method alone during the critical period, and over 85% during wakeful periods. 7

10 III. COMBINED MODELS OF ALERTING PENETRATION Now we combine the multiple methods of alerting together, by taking the union of the alerting sets over the 24 hour period: R TV I C Figure 10 Union of sets broken into intersections R only TV only R+C TV+C R+I+C TV+I+C I+C I only C only Note that TV and Radio don t intersect as a person is usually listening to one or the other and not both. But a person can be watching TV or listening to the radio and simultaneously using the internet [12]. And with all of these, cell phone alerting is available. The result of combining these sets, assuming independence of the sets except for Radio and TV which do not intersect, is shown on Figure 11 which shows the combinations as a function of the CMAS% deployment. The bottom three curves show the separate alerting models for Radio, TV, and Internet. The fourth curve from the bottom, shows the union of these three alerting sets with NO cell phone (CMAS) deployment. During the critical period from 2 AM 5 AM, alerting potential remains under 30%, far from the 85% goal. Figure 11 - Contagion with Radio, TV, Internet and Cell Alerting Population Penetration 100% 90 TV or Radio or Internet Cell Alerts (CMAS 100%) % % Cell Alert (CMAS 100%) 70 Radio or %TV or Cell Alerts (CMAS 50%) % % Cell Alert (CMAS 50%) TV or Radio or Internet and No Cell Alerts (CMAS 0%) 20 % TV Alert 10 % Radio alert 0% % Internet Alert Hour of day (Central Standard Time) The additional 4 curves on Figure 11 show the results for 50% CMAS deployment and 100% CMAS deployment. At 100% CMAS deployment, the 4 methods combined yield only about 60% alerting potential during the critical period, still short of the 85% goal; however, during the daytime hours the goal is met. The direct alerting percents do not achieve the DHS goal at all times of the day, but contagion can spread these alerts to a much higher percent, meeting the DHS goal. 8

11 IV. CONTAGION MODEL Lastly, we present a contagion model where a person can be: Alerted directly from IPAWS alert Alerted indirectly from another person that received a direct alert, typically a family member in the same household. This allows us to estimate the percent of people alerted by others who were directly alerted. For example, the direct methods might yield an average 81% alerting penetration (for 100% CMAS deployment) at a given time t of the day (say at 8 AM), but then those 81% of alerted people contact another 16% within the 10 minutes target, i.e., family, friends, relatives, co-workers, etc. Hence, the total alerted is 81% + 16% = 97% which is well above the DHS goal of 85%. The most difficult time for alerting will be at night when most people are sleeping during the critical period, but CMAS can be used to wake them. A. Simple Contagion Model Radio, TV, Internet or CMAS alerts one family member and they wake their remaining family members and perhaps even nearby friends. Households are either single person or multi-person. For multi-person households, on average, there are three members in multi-person households [11]. Hence, if one member of a family is alerted, they alert the other two members of the household. This simple contagion model applies just to households. More elaborate contagion models could be applied resulting in higher penetrations. But even this simple model gets us where we want to go during the critical period. The contagion results are shown on Figure 12 assuming that one household member alerted by Radio, TV, Internet, or cellphone alerts other household members. We see that with 100% deployment of CMAS, the alerting penetration is 94% or greater for ALL hours of the day, including the critical period. With a 75% deployment of CMAS, estimated to be accomplished by June 2012, we also achieve the DHS goal of 85% for ALL hours of day. Figure 12 - Example CMAS Architecture Contagion CMAS 100% Contagion CMAS 75% Contagion CMAS 50% Contagion CMAS 0% Radio or TV or Internet or Cell (CMAS 100%) Radio or TV or Internet or Cell (CMAS 50%) Radio or TV or Internet or No Cell (CMAS 0%) 0 o

12 V. EXAMPLE CMAS ARCHITECTURE Since the critical path for meeting the FEMA/DHS alerting goal is CMAS infrastructure deployment, Figure 13 presents an example architecture. It shows alerts originating at the top and flowing through alert gateways to Commercial Mobile Service Provider (CMSP) gateways which then broadcast the messages through each of their mobile network infrastructures to their subscribers. Figure 13 Example CMAS Architecture Federal, State, and Local Agencies CAP CAP Alert GW Alert GW Alert GW XML XML CMA (Alerts) generation Authentication and validation of alerts Maintans CMSP Profiles Delivery of CMA to CMSPs Multiple alert gateways in goverment domain Government Domain Carrier Domain CMSP Gateway BMC Data Replication CMSP Gateway BMC Mated pair BMC supporting external CMSP gateway function. Single point of entry for CMAs in CMSP network MC function collocated with the CMSP Gateway for message delivery to CDMA network MC CBC MC CBC CBC function collocated with the CMSP Gateway for message delivery to GSMU/UMTS network IS-824, IS-637 SMDPP MSC Cell tite CDMA Network CDMA Handsets 3GPP CBS-BSC BSC BTS GSM Network GSM Handsets 3GPP lu-bc NNG Node B UMTS Network UMTS Handsets 10

13 VI. POTENTIAL MODEL EXTENSIONS AND VALIDATIONS Rather than dealing with the total USA population as a homogeneous body, we could break down the population into sub-populations for in-depth analysis, which would also provide individual sub-population analysis: 2.7 M in prisons ~1 M Hospital beds in 5800 Hospitals More millions in nursing homes/assisted care Non-English speaking people These model extensions are not expected to change the fundamental results shown previously. More importantly, we should target alerting for people with disabilities: In 1994, 1.3 million people were legally blind In 2004, 800 thousand people were unable to hear normal conversation. Congressional Committee testimony stated about 49 million Americans are disabled in some way [13]. These extensions of the models for people with disabilities would require further in depth modeling. However, it is expected that the alerting methods for these populations could also achieve the targeted alerting goal with suitable aids. Lastly, our models contain a set of assumptions on the effectiveness parameters for Radio, TV, Internet, and most importantly, the ability to wake a sleeping person who has a cell phone. These assumptions need validation and is a possible target of future market research. 11

14 VII. CONCLUSION Alerting Penetration Models allow estimation of the impacts from deployment of various alerting systems: Radio, TV, Internet, and Cell Phone. We believe that FEMA could potentially use these models with modifications as needed to predict the effectiveness of the IPAWS platforms as they are deployed. To achieve the DHS goal of 85% alerting of population within 10 minutes, significant deployment of cell phones capable of CMAS alerting is needed. When deployment of cellphones capable of CMAS alerting hits 75% (expected by August 2012), when combined with a conservative contagion model (one person alerted by cell phone wakes the rest of the family) it is estimated that the DHS goal will be achieved for all hours of the day. The Commercial Mobile Alert Service (CMAS) should be the major player in the overall IPAWS deployment and will be critical to the alerting success required by the U.S. Government. ABOUT LGS INNOVATIONS LGS Innovations delivers next generation solutions that solve the most complex networking and communications challenges facing the U.S. Federal Government, state and local governments, foreign governments, and commercial organizations worldwide. LGS offers groundbreaking research and development and builds advanced wireless, optical, and wired products and applications customized for specific mission environments. These solutions provide unique information and security advantages that lead to the operational success of its customers. LGS offerings include: Campus and building networking solutions for military bases, hospitals, and corporate centers Maritime applications for in-port and at sea communications Global networks (long-haul communications, including undersea cable) Enterprise voice, video, and data networking 4G wireless deployable communications for public safety, battlefield, and emergency and first responder communities Network engineering, integration, and installation Cloud and data center infrastructure Video teleconferencing and IPTV suites Research and development in advanced multimedia/rf communications, cybersecurity, sensing technologies, and photonics LGS Innovations is a U.S.-owned company headquartered in Herndon, Virginia, with offices in Colorado, Florida, Illinois, Maryland, New Jersey, New Mexico, and North Carolina. Formerly a subsidiary of Alcatel-Lucent, LGS is the exclusive reseller of Alcatel-Lucent products and services to the U.S. Federal Government and any other entity when the end customer is the U.S. Federal Government. LGS maintains strong ties to Bell Labs and its technologies, employing more than 450 scientists and engineers and over 720 employees worldwide. To learn more about LGS Innovations, visit 12

15 VIII. ACKNOWLEDGMENT The authors would like to thank our colleagues at LGS Innovations for their support. 13

16 IX. references 1 Presidential Executive Order 13407, June 26, 2006, Public Alert and Warning System. 2 Warning, Alert, and Response Network Act, Title VI of the Security and Accountability for Every Port Act of 2006, Public Law , Titles I through III of the Communications Act of 1934, as amended. 3 R. David Paulison, Administrator,The Federal Emergency Management Agency, U. S. Department of Homeland Security, Committee on Homeland Security and Governmental Affairs, United States Senate, April 3, FEMA handout, Technologies for Critical Incident Preparedness 2010 Conference, Philadelphia, PA, February 2, Fema regions, % of population. 6 Arbitron. Radio Today. How America listens to radio, 2008 edition. 7 TV and Internet Usage in Brazil, 8 CTIA, the Wireless Association, th Annual Report and Analysis of Competitive Market Conditions With Respect to Commercial Mobile Services, WT Docket No , January 16, Bureau of Labor Statistics, American Time of use study US Census: population by age, people in households, 12 Nielsen, A2/M2 Three Screen Report, Television, Internet and Mobile Usage in the US, Volume 5, 2nd quarter 2009, updated December 18, LGS Innovations LLC - All Rights Reserved LGS, LGS Innovations, and the LGS Innovations logo are trademarks of LGS Innovations LLC. 14