 |
  |
MISSOURI DIVISION |
|
 |
 |
PROCEEDINGS OF THE SEPTEMBER 2000 POST EARTHQUAKE HIGHWAY RESPONSE AND RECOVERY SEMINAR HELD IN ST. LOUIS MISSOURI
ESTIMATING LOSSES TO HIGHWAY SYSTEMS BY STUART WERNER
It is a pleasure to be here to talk to you about some of the work that we've been doing on seismic risk analysis on highway systems and estimation of losses due to the seismic damage to the highway system. This was funded through a contract by Federal Highway Administration in Washington through the Multi-Disciplinary Center of Earthquake Engineering Research in Buffalo. This is a highway project that has been going on for about six years and it's included a lot of different tasks related to the seismic performance of existing highway construction and work that we did was one of those tasks.
I'd like to acknowledge the people on our team that worked on this. Craig Taylor, John Walton, Jim Moore, and Sun Bin Cho are from the University of California. In addition to our project team members, there were a number of other people who provided us with very helpful information and data and these include Howard Hwang and John Jernigan as well as John Ander and Les Yo.
I'll start out with some introductory comments about why we started to look at seismic risk analysis of a highway system. Then I'll talk about procedure and briefly summarize it for you, go through a demonstration application of the procedure to an actual highway system. Then I'll end with some closing comments.
A highway roadway system; what is it? It's really a network of interconnected roadways and bridges, embankments, tunnels, and a whole range of different types of components that all work together as a system to transport people and goods by land based roadway systems.
The idea is that the system should maintain efficient, effective traffic flows and it should also work safely. What happens when an earthquake hits and there's damage to the highway system? This can disrupt the traffic flows and in turn can hamper and impact the economic recovery of the region. It can also impact short-term emergency response and recovery operations.
Now, what are some of the factors that affect an overall network, the highway system response to an earthquake? One set of factors relates to the components: the bridges and roadways and other types of components such as geometry, the materials, construction, seismic design details, and soil conditions at the site. There are other factors that enter into how well the system will perform during an earthquake. These are the system characteristics, location of the roadways, and the overall configuration of the roadway network. How redundant are these various networks? What are the capacities for the various roadways? How well are they able to accommodate the demands of traffic on the system?
In earthquake engineering practice for highway system components and emergency planning, these system´s issues are not considered in concurrent, decision-making regarding seismic risk reduction applications to actual highway systems. This is the reason we are trying to develop a new procedure for analyzing seismic risk to highway systems. The procedure that we have developed does evaluate the effects of earthquake damage on traffic flows. It considers the characteristics of the bridges and individual components and includes the system characteristics that I mentioned earlier as well.
Because it brings in issues of traffic flows and system and network characteristics, it really provides a much more rational basis for making decisions regarding what's the most appropriate seismic risk reduction options that ought to be implemented for a given system.
This is a tool that can eventually be incorporated into real-time traffic flow estimates immediately after an earthquake. If models are developed of traffic flows and we know approximately where the earthquake is, there can be an initial quick estimate of where traffic flows will be affected. This could help in rapid response of DOT and other emergency response people as well. That's really the motivation and why we've addressed this issue of seismic risk analysis and highway systems.
Next, I'd like to briefly summarize the procedure itself then describe it in more detail as we go through a demonstration. This procedure basically consists of four parts. It involves what's called a Monte Carlo analysis of system response for a whole range of different scenario earthquakes and simulations.
It's really a four-step process. The first step we develop all the input data, incorporate all the models we're going to use, and so on. Then in the second and third step, we carry out an evaluation of the seismic performance of the highway system for each scenario earthquake simulation we're looking at. We repeat this for all the simulations and scenarios that have been identified in that first step. Then in the final step, we aggregate the results and provide loss estimates for the system, either deterministically or probabilistically.
The heart of this process is the GIS database. This database contains four modules. These modules contain input data, the models, and the analysis procedures necessary to characterize the earthquakes, the seismic and geologic hazards to which the system will be subjected. The vulnerability of the bridges and the components and how the traffic flows in the system will be affected by earthquake damage. That's the system module. The losses due to any highway system damage are what we call the transportation cost module.
This procedure has a number of important features. First of all, it's GIS based. This greatly helps in our organization of data, our analysis, and the display of the results. It's also modular. We deliberately designed this to be modular. When there are future improvements from engineering research developments, they can be readily incorporated into this procedure. Third, it is a multidisciplinary procedure. The application of this process really is a team effort involving seismologists, geologists, geo-technical and structural earthquake engineers, as well as transportation engineers and emergency planners working together to get the results that are needed for the user to make an appropriate seismic risk reduction and decision.
Finally, the procedure as we see in the demonstration application, there's a number of different forms of results that can come out of this depending on the needs of the user. The results can be provided either probabilistically or deterministically depending on user needs. That is a brief introduction to the procedure itself.
Let's go into a demonstration application to show you how the procedure works and information that can be developed. We've applied this to the highway roadway system in Shelby County Tennessee. These show the borders of Shelby County by the dash line. It's in the southwestern corner of Tennessee, just north of the state of Mississippi and just alongside the Mississippi River.
The main hub of Shelby County is the City of Memphis. It is surrounded by highways and freeways. There are two major bridges that cross the Mississippi River and Interstate 40 and Interstate 55 and a number of major arterioles that emanate from the center of the city in all directions.
I'm going to be using the term component damage state. This means the type of damage that occurs to a bridge structure or component, the location of the damage, and extent of the damage under a given level of ground shaking. This is determined from earthquake engineering analysis, geo-technical and structural analysis. This is what I mean when I talk about damage state.
Once the bridge is damaged, the next step is to evaluate how that damage will be repaired, how long it will take, how much it will cost and how will traffic along that bridge be affected during the repair process. When I talk about traffic state, that's what I mean. This traffic state will vary with time after the earthquake because as the repairs to the damage proceed, it may be possible to gradually reopen the bridge to at least partial traffic flows.
Finally, I'm going to use the term system state. A system state is if we envision a picture of the whole highway roadway network before the earthquake. After the earthquake, we start to modify that network and put in the traffic states at the various components that were damaged. This means that some of the links in the system will not have as much capacity to carry traffic as it did before the earthquake. The overall system state will also vary with time.
I'll go into the application of our seismic risk analysis procedure to the Shelby County highway system. Some of the input data that is needed for the first step include the various highways and roadways that have been modeled in our seismic risk analysis. This is all information that is available throughout the Metropolitan Planning Organization for the region and they handle all this data. For each roadway, we have to put in information like number of lanes, traffic volume information, congestion information and so on. There's a whole range of different types of roadways ranging from freeways down to ramps that have been modeled.
We also need to provide information on the bridges. In this particular application, we've modeled bridges and approaches as the main components. There are some 431 bridges in Shelby County. We need to input various attributes of these bridges that will enable us to evaluate their seismic vulnerability.
For most of the bridges, we use information from the NBI database and deduce seismic structural response of parameters for our vulnerability analysis from these NBI parameters.
This is now a plot of the highway system and bridges. We overlaid it with the local soil conditions throughout Shelby County. This is information that was very kindly provided to us by Howard Hwang of the University of Memphis through some excellent work that he's done in this area. We see here that Shelby County is primarily composed of two types of soils. There's either soil Type D shown in green and soil Type E that is shown in brown. Soil type E follows the various creeks and rivers and waterways in the area. These are soils that will amplify ground motion because they're softer soils.
We also need to provide input on the traffic demands on the system. To do this, we use what we call series of origin-destination zones. This is also information that's provided by the MPO for the area. The MPO for each of these origin- destination zones there is a trip table that's provided that defines the number of trips emanating from each zone in the region to all other zones in the region. This is based on traffic models, on demographics, and on traffic counts that are carried out by the MPO. These are continuously carried out.
The yellow areas in this slide are locations of what we call key zones. Part of the results that we can obtain from our seismic risk analysis procedure is travel times to and from any zone in the system.
We need to model the earthquakes. We've carried out the probabilistic analysis of the seismic risk but these results can also be used in a deterministic way.
For doing a probabilistic analysis, we need to develop a whole range of different scenario earthquakes for use in the model that represent the geologic and seismologic characteristics of the region. We did this by adopting an earthquake model for the Central United States that was developed by Art Frankel and Associates at U.S. Geological Survey under their National Hazards Mapping Program. From this, we carried out this evaluation looking at a 50,000 year time period. We developed over 2,300 earthquakes occurring over this time period. We use this long time period in order to get an adequate statistical sampling of the occurrence over the whole range of different possible earthquakes, ranging from relatively frequent earthquakes with moderate magnitude to rarely occurring earthquakes with much higher magnitude. We didn't only look at the most extreme earthquake events. We looked at the whole range of possible earthquake events that could possibly occur in the region.
This is a GIS plot of the region showing the spatial distributions of the earthquakes with Shelby County, the state of Tennessee and the Mississippi River. This shows the whole area we considered in developing our scenario earthquakes. That's basically the earthquake modeling.
As part of this initialization step, we also need to illustrate our models for the seismic hazards. For this particular example, for ground shaking, one needs to use a ground-shaking model that is applicable to the particular region of the U.S. where his or her system is located. For this particular area, we used a bedrock attenuation model and the soil amplification model that was developed by Howard Hwang and his co-workers from the University of Memphis.
We've also modeled liquefaction hazards as well. This is a relatively new development in our procedure. For each soil zone in the system, we divide the whole region into a number of different soil zones. There's geological screening that is done before starting analysis in order to evaluate which zones could have potential or liquefaction. Then for each scenario of earthquakes, we evaluate whether there is a potential for liquefaction. For those areas where there is a potential for liquefaction, we then compute permanent ground displacement that occurs during that scenario. We compute lateral-spread displacement using procedures developed by Leo Yo and his associates. We also included liquefaction settlement using a Tokimas Seed model.
We've modeled bridge structures and modeled approaches. As Erol mentioned earlier, this is a pretty widespread damage mode, at least for bridges in California during the past earthquakes. We thought this was important to include.
So we have a range of different models that we use for modeling the site performance of bridges under both ground shaking and liquefaction hazards, including approach settlements. We model damage states and traffic states of the individual bridges as a function of the level ground shaking and permanent ground displacement at the bridge site.
What a model looks like for use in a probabilistic seismic risk analysis that calls for a series of fragility curves that plots probability versus the level of ground shaking. Each of these different lines represents a different traffic state for the bridge.
In this case, it ranges from the onset of damage with no traffic closure all the way up to extended closure for several months and a full range of different traffic states in between.
Once we know we've our components and hazards, we have to also incorporate into the initialization process how we analyze the traffic flows in the system after an earthquake. After each earthquake event, we assembled these different system states.
For each earthquake, we know the level of ground shaking that has occurred throughout the system. We then use our vulnerability model to evaluate whether the various bridges in the system have been damaged. We then go from that to the overall system state for the whole system for various times after the earthquake. We have all these modified highway system networks.
We apply these transportation network analysis procedures to each of these post-earthquake system states in order to estimate post-earthquake travel times, travel paths and traffic volumes throughout the system. We use different methods. If we're just doing a deterministic evaluation, just looking at a couple of different earthquakes. We use a mathematical procedure called user equilibrium method.
If we're looking at many different system states for many different scenario earthquakes and simulations. There's a new method developed by Jim Moore and his co-workers at UMC called associate of memory method. It works very efficiently when analyzing large numbers of system states.
That's basically the initialization, the input data and types of analysis procedures used. Now, let's show how we apply this information. I've deliberately not chosen a Magnitude 8 earthquake on the New Madrid seismic zone in order to illustrate that it doesn't take a Magnitude 8 earthquake to cause damage. There can be more frequently occurring earthquakes of lower magnitude that can also affect the highway roadway system. So here we have a Magnitude 6.9 that's located some 80 kilometers to the north/northwest of Shelby County. That's the earthquake we're going to be going through. We know its location and we know its distance to all the bridge sites. We know the soil conditions throughout Shelby County so we can estimate the levels of ground shaking at each bridge site throughout the system.
Here on this plot of the highway network, we also show the different soil conditions. The dark brown represents the Type E, the locations of the loose soil deposits. We see how the most severe levels of ground shaking do correspond pretty readily to where these loose soil deposits are. We also estimate permanent ground displacement from liquefaction at those bridge sites that are judged to be potentially liquefiable soil sites. Here are some of our liquefaction results for this particular scenario earthquake. Now, we know the ground shaking. We know the liquefaction hazard. We then use our vulnerability models for the bridge to estimate the damage states of the various bridges.
We see there are some bridges that have collapsed up in the northern part of Shelby County in the area closest to the epicenter of the earthquake. We also see that the most extensive damage follows the dark zones: the zones where soft soil deposits are located.
Once we get the damage states, now we have to assemble the traffic states for the highway system at different times after the earthquake. This is your highway system model. We now plug in the traffic capacities, the traffic states, for the various links where the bridges are located. Some of these links have reduced capacity to carry traffic. This isn't all Shelby County. This is just a zoom of the area around Memphis. This is just the beltway surrounding the city of Memphis with the two crossings of the Mississippi River.
We see that within Memphis, there are a few links that shut down seven days after the earthquake and varying degrees of damage to some of the other links in the system. Now, this is at seven days after the earthquake and before any repairs have been able to begin. We also look at systems states at various times after the earthquake and after the repair process has started. We see a lot of the links that were closed immediately after the earthquake have now begun to be reopened 60 days after the earthquake. Then about five months after the earthquake, we see further improvement in the traffic capacities of the various links. These are the various system states that can be developed.
The next step in the analysis is the transportation network analysis. For each post-earthquake system state that we've developed in the first step, we want to evaluate the traffic flows. We apply this network analysis procedure to the system state to each scenario earthquake
We can't get travel times: travel times throughout the system or travel times to and from selected locations in the system for the various earthquake events. These are post-earthquake travel times that can be compared to pre-earthquake travel times in order to see what the effects of earthquake damage is on travel throughout the highway network.
In addition to travel times, we can also get information on travel paths, minimum time paths from any selected location in the system to any other location. A number of the speakers yesterday spoke about emergency response routes and the need to keep certain routes open, and these routes will be related to certain emergency response facilities such as medical centers or airports. This will give a planner an idea of which routes will be most affected, will represent the minimum travel time to or from a given location. This can help in designating those emergency routes. We can also examine the various roadways in the system in order to see what level of traffic is being carried by each of the roadways at various times after the earthquake. This is what we call a traffic volume bandwidth plot in which the width of the bands represents different levels of traffic volume that are carried after the earthquake.
That's the result we can get from our network analysis. Next, we now gather and aggregate all these results and come up with estimation of earthquake-induced losses. One form of results is the probabilistic estimates of economic losses for various exposure times. We've developed these economic losses by using very simple models that were used at CALTRANS after the Northridge earthquake. Losses were estimated due to increased commute times or increased travel times and considering such factors as the price of fuel, the time value of the truck, the estimated time value of the time of persons and automobiles, and how many trucks and automobiles might be on the roadways at a given time and so on.
These are very simple economic loss models. They do give a picture of the effects that could happen. One type of criteria that may be useful in the future in helping to establish what seismic risk reduction options to look at, might be the terms of some kind of economic loss model, where a decision maker may want to say for a probability to exceed 10 percent in 50 years, I want my economic losses to be no more than, let's say, $200 million. So if we then look at that plot, here's 10 percent, go to the 50-year curve, we see that the losses are on the order of $300 and $400 million. This is considering the full range of different earthquake events and simulations that we've considered. This would be an economic loss picture, let's say, with no seismic retrofit. We can redo the process looking at various levels of seismic retrofit or seismic risk reduction and see how these losses are decreased as a result of this seismic retrofit.
In addition to looking at economic losses probabilistically, we can also look at them deterministically for individual earthquake events. We show some results for various magnitude earthquakes at different locations relative to downtown Memphis. We could compute the economic losses for each of these earthquakes. If one wants to look at individual earthquake event, they can do that as well.
In addition to looking at economic losses, we can also look at access and egress times to certain key locations. We can see the damage to the highway network due to a given earthquake event, the effects on access times or egress times to and from a key location in the region. This can also be very helpful for emergency planning.
I've shown the results here for four different locations: the Government Center in Downtown Memphis, the Memphis Airport, which is south of the City of Memphis, University of Memphis, and Germantown, which is a suburb of Memphis located to the east of the beltway. We see the spatial distribution of the effects of earthquake damage to the highway system from a table like this.
I've been talking about is a seismic risk analysis procedure whose function is to evaluate the ability of the highway system to carry traffic after an earthquake and then estimate the losses due to earthquake damage to a highway system.
We can define losses here very generally. These are not necessarily only economic losses. These can also be losses in increases to travel times to key locations in the area. They can be losses of access along emergency routes and so on. So there's a whole range of losses that can be examined here.
I want to just briefly go into how this kind of process can be used to help guide making a decision regarding seismic risk reduction for a given highway system. One would carry this procedure out multiple times in which you'd start out with no seismic risk reduction. That's your starting point. You estimate the losses for that case then you start looking at different candidate seismic risk reduction strategies that might be used. You rerun the seismic risk analysis, including each option that you're considering. You then compare the cost of the seismic risk reduction option. Let's say the option is to include seismic retrofit or not to include seismic retrofit. So if we say we want to include seismic retrofit to some level, we have some costs associated with that decision because there's going to be a cost needed to construct and to upgrade the various bridges to the appropriate selected level of seismic upgrade. So we want to compare those costs to the benefits that are achieved in terms of how much this level of seismic retrofit of the bridges reduces the losses due to earthquake damage to the highway system.
This brings in issues of traffic flows, as well as the performance of the components itself that's system wide and component response characteristics. Again, it can give a more complete picture of the consequences of a particular decision.
Where are we going with this process? One goal of this is to eventually develop a software package that will be available to the public to use through the internet. We've started work on this software now. It's called REDARS, which stands for Risk from Earthquake Damage to Roadway Systems.
We have just developed a version of this program. There's a lot of validation and there's going to be a lot more programming to be done. It's a few years away from being available but we are working on it. This is a goal we hope to accomplish in the next few years.
What kinds of information would you need to help you make decisions regarding seismic-risk reduction? We hope to, through conferences like this, through separate meetings, meet with you and see and work with you on how we can help structure methodology that will meet your needs.
I've shown you models for Central U.S. We want to incorporate other ground motion models for other parts of country as well. We want to expand our component vulnerability models as well, improve our bridge models, develop and incorporate models for other highway components as well.
 |
Send mail to scott.holder@fhwa.dot.gov with questions or comments about this web site. |
