DEVELOPING AN AGENT BASED SIMULATION MODEL FOR EARTHQUAKES IN THE CONTEXT OF SDI Mahdi Hashemi1, Ali Asghar Alesheikh2 Department of Geospatial Information Systems, K.N. Toosi University of Technology, Tehran, Iran 1
[email protected] 2
[email protected] Abstract Natural disasters such as Earthquakes are devastating hazards that affect the environment, and lead to financial, environmental and/or human losses. The resulting loss depends on the capacity of the population to support or resist the disaster, and their resilience. Simulation of earthquakes can greatly influence the quality of decisions in all steps of crisis management. Numerous technologies such as agents are developed for proper spatial simulation. Spatial agents are entities within an environment that can emulate mental processes or simulate rational behavior. This paper attempts to simulate the Earthquake by means of NetLogo software. In order to construct and assess the simulation, the information about Bam earthquake is used. The methodology used consists of three major steps. In the first step, all spatial information including satellite images (before and after the earthquake) and topographic/cadastral maps of the area were mosaicked and georeferenced. The parts of the city that contain various levels of destructions are selected. Three types of features namely buildings, roads and recreational areas were classified and extracted from the satellite images. In the second step, the governing factors of destructions were identified; a mathematical model that integrates the factors was constructed, and the resistance values for each pixel were computed. The classes were imported into raster space of NetLogo software and the resistance values were assigned to corresponding features. In the final step, the simulation was constructed for various parameter values (different earthquake strength, time elapses, etc.). The results were then presented to users. It is suggested that the software be used by various crisis management organization to have an in depth understanding of the destructions and the decisions to be made in various phases of crisis management. Keywords: Multiagent modeling, Earthquake simulation, crisis management, SDI 1.
INTRODUCTION
A number of natural disasters, such as earthquake, storm surge, Tsunami wave and the flood, occur annually in various parts of the world. On average, one very severe earthquake Richter 8, 100 relatively severe earthquakes, thousands of mediocre earthquakes and about 3 million poor earthquakes occur on earth annually (earthquake.usgs.gov/earthquakes/eqarchives). One of the most calamitous countries on the planet, Iran is imposed by nearly 2 Billion USD of damage each year. Earthquake in Iran is the most natural disasters where 97 percent of villages and cities are exposed to earthquakes. Earthquakes are not like other disasters (such as typhoons or epidemics) for which one can receive early warnings. Thus, in this study, we seek to provide a practical mechanism for earthquake disaster damage assessment in an urban environment. In order to determine the extent of a disaster quantitatively, it is necessary to estimate the behavior of natural phenomena which causes such disaster. There have been presented a number of numerical methods to evaluate the damage by the natural disaster. Numerical simulation is one of the safe and cost effective ways to
investigate the damages. Agents can conveniently be used for such spatial simulations. Russell (2002) defined Artificial Intelligence (AI) as the study of agents that receive percepts from the environment and perform actions. Multiagent systems consist of multiple interacting computing elements, known as agents. An agent is a computer system that is placed in some environment, and that is capable of autonomous action in such environment to fulfill its design objectives (Wooldridge, 2002). As an intelligent entity, an agent operates logically in different environmental circumstances, if its perceptual and effectual equipment is provided. Agents exist physically in the form of programs that run on computing devices (Weiss, 1999). There are multiple methods for multiagent based programming (Shoham, 2008). This paper attempts to provide a logical construct for simulating an earthquake by means of NetLogo software. In order to develop and assess the simulation, the information about Bam earthquake is used. A powerful earthquake (6.5 Richter) struck southeastern Iran on December 26, 2003, killing forty thousand and destroying much of the city of Bam (Fig. 1). About 60 percent of the buildings in Bam were destroyed. Figure 1: The study area, the amount of destruction is showed by three color: blue shows 90 percent of damage, red shows 65 percent of damage and green shows 35 percent of damage
Fig. 2 illustrates the general workflow of the proposed intelligent earthquake simulation system. The methodology used consists of three major steps. In the first step (showed by color brown in Fig. 2), all spatial information including satellite images (before the earthquake,1998) of the area were mosaicked and georeferenced. The destruction map after the earthquake were taken from National Cartographic Center (NCC) and overlaid on satellite images. The parts of the city that contain various levels of destructions were selected (as showed in Fig. 1). Buildings, trees and roads were visually extracted from the satellite images and imported into raster space of NetLogo software. In the second step (showed by color blue in Fig. 2), the governing factors of destructions were identified and classified. They were: the amount of seismic energy released by earthquake in Richter scale; the distance to earthquake source in Meters; and the earthquake lasting time in Seconds. A mathematical model that integrates the factors was constructed, and the resistance values for each simulated feature were computed. In the final step (showed in Fig. 2 by color red), the simulation was performed for various parameter values (different earthquake strength, time elapses, etc.).
In Section 2 we briefly present the most relevant researches related to our work. In Section 3 we argued about georeferencing layers and extracting features from images and importing them into raster space of NetLogo software. In section 4 we described the mathematical model. In section 5 we developed the earthquake simulation environment so that users can try the model with different earthquake source-parameters. Lastly, we concluded our research in Section 6. Figure 2: The overall workflow of the intelligent simulation system for earthquake disaster assessment Identifying the governing factors
Earthquake magnitude (Richter )
Distance to earthquake source (Meters)
Earthquake lasting time (Seconds )
Destruction map of Bam city after the earthquake
Satellite images of Bam city before the earthquake
Georeferencing
Mosaicking and georeferencing
Overlaying and selection case study
Model construction
Extracting buildings , trees and roads
Bam earthquake simulation
Importing features into raster space of NetLogo software
User interface construction
2.
RESEARCH BACKGROUND
The application of agents in disaster management is studied by numerous researchers (Petit, 2006; Liu, 2004). Keisuke and Kazuo (2008) presented a simulation system for the disaster evacuation based on multiagent model considering geographical information. The model can estimate not only the damages to the structures but also the damages of human being. Furthermore, it was possible to investigate the appropriate evacuation route by the simulation. Gervais and his colleques (2007) designed Intelligent Map Agents (IMA) architecture which was aimed at replacing the monolithic approach to geographic information systems with a new dynamic, lean, and customizable system supporting spatially-oriented applications. The IMA was designed as an open architecture that can accommodate a large number of mobile users and services possibly distributed across a wide geographical area. Tang and Ren (2008) developed an agent-based simulation model which incorporates the fire scene and the building geometry using a fire dynamics simulator (FDS) based on the computational fluid dynamics and GIS data to model the occupant response. Evans and Kelley (2008) employed an agent-based model to analyze the process of forest regrowth in south-central Indiana from 1939 to 1993. Benenson and his colleques (2008) presented PARKAGENT, an agent-based, spatially explicit model for parking in the city. PARKAGENT simulates the behavior of each driver in a spatially explicit environment and is able to capture the complex selforganizing dynamics of a large collective of parking agents within a nonhomogeneous (road) space. The model generates distributions of key values like search time, walking distance, and parking costs over different driver groups. Lin and Liu (2007) used multicriteria evaluation techniques to determine some of the parameters for the agent-based model of urban simulation. Empirical data from GIS are used to define agent’s properties. Sensitivity analysis is also carried out to assess the influences of parameters on simulation outcomes. Nute and his colleques
(2004) developed an agent-based decision support system for forest ecosystem management. A graphical user interface written in Visual C++ provided inventory analysis tools, dialogs for selecting timber, water, ecological, wildlife, and visual goals, and dialogs for defining treatments and building prescriptive management plans. Tang and Wen (2009) presented an intelligent simulation system for an earthquake disaster assessment system based on development platforms of a GIS and Artificial Intelligence (AI). The system is designed to identify the weakness of the structure and infrastructure system in pre-earthquake conditions, quickly assess earthquake damage and make an intelligent emergency response for the public and government during the earthquake and post-earthquake. Feng and his colleques (2008) developed an integrated urban earthquake simulation system that uses GIS as the model source, Computer Aided Drafting (CAD) as the model generating tools, Finite Element Analysis (FEA) as damage prediction, and Virtual Reality (VR) as the post-process platform. Tsai and Chen (2009) developed an earthquake disaster assessment model to apply to risk management in the tourism industry. Cinicioglu and his colleques (2007) presented a methodology that incorporates a literature-derived probabilistic assessment of damage-causation, and is interpreted and presented as single numbers deemed “Damage Grades.” These damage grades integrate the initial probabilistic evaluation with professional experience and judgment to determine potential damage to a particular structure at a particular location. Ansal and his colleques (2009) developed a GIS-based loss estimation model to evaluate different loss scenarios depending on the ground shaking input, as well as to consider the implications of mitigation actions. 3.
METHODOLOGY USED
3.1. Mosaicking And Georeferencing Layers Satellite images of Bam city (before earthquake) are gathered and mosaicked. The image is then georeferenced using 12 control points. The points are evenly distributed to have an accurate referencing (Fig. 3). Destruction map of the earthquake has been obtained from NCC of Iran. Pixels in the destruction map are categorized in three classes based on the amount of their destruction. Figure 3: Control points used for georeferencing of images
In the last step, RMSEs of control points were estimated while using Spline function in ArcGIS environment. The destruction map were overlayed on satellite images to determine the magnitude of destructions for all spatial feature such as buildings, trees and roads. A part of this correspondence is shown in Fig. 4. Figure 4: The overlay of satellite image and destruction map of study area
3.2. Extracting Features And Importing Them Into NetLogo A portion of study area that has various levels of destructions and landuse categories - buildings, trees and roads - were selected. Then, a square grid imposed on the area and a lable was assigned to each pixel that identifies its feature type. The grid is shown in Fig. 5 where buildings are black, trees are green and roads are blue. The grid was then reconstructed in NetLogo software. Each type of features (buildings, trees and roads) are an agent class and have their attributes like resistance against earthquake, distance to earthquake source, size and etc.. Figure 5: Squre grid for feature extraction
4. CONSTRUCTING MATHEMATICAL MODEL FOR EARTHQUAKE ENERGY OF EACH AGENT No matter where an earthquake occurs, in the city or the countryside, it causes damages, injuries and death as well as economic loss in varying degrees to the area (Ergonul, 2005). Bam earthquake with magnitude of 6.5 Richter took 12 seconds. The earthquake source was 3590 meter far from the center of study area. Because of smallness of this area, the area divided by four sections and the distance from earthquake source assumed to be the same for each section. A mathematical model is developed to estimate the damages. The function consists of the following parameters: • The energy release of an earthquake. It closely correlates to its destructive power. (A difference in magnitude of 1.0 is equivalent to a factor of 31.6 ( = (101.0)(3 / 2)) in the energy released; a difference in magnitude of 2.0 is equivalent to a factor of 1000 ( = (102.0)(3 / 2) ) in the energy released (earthquake.usgs.gov/learn/topics/richter.php)). • The distance to earthquake source. The waves are largest where they are formed and gradually deflate. Seismic waves also become attenuated as they move away from the earthquake source (earthquake.usgs.gov/learn/animations). • The lasting time of the quake. The more time the earthquake takes, the more damages it causes. Therefore the energy of earthquake is computed from the following formulae. In this model, “Richter” is the magnitude of earthquake, “Time” is lasting time for earthquake and k1, k2 are unknowns. For computing unknowns three sections of study area used and results were certificated another section. The unknowns are computed based on least squares adjustment. Final mathematical model is then formed by: The above model is constructed based on the data availability constraints. It should be mentioned that the above model is a crude estimation of earthquake energy as in reality factors such as topography, velocity, acceleration and the way of movement of crust can also affect the earthquake energy (Wald, 2007). The earthquake energy arrived in each feature is compared with its resistance and that feature react to this effect. If the resistance of feature is less than earthquake energy, it will be destroyed completely, otherwise it bears damages. We determined the resistance of features so that the result is similar to the real destruction map. 5.
BAM EARTHQUAKE SIMULATION IN NETLOGO SOFTWARE
In this step, the features in the software act independently so that they can simulate real damages for Bam earthquake. The simulated area is shown in Fig. 6 in three steps of earthquake: first second, fourth second and last (12th) second. Ruined and damaged features are shown graphically and numerically. Sections at the left of Fig. 6 show the ruined and damaged percent of each type of features.
Figure 6: Simulated Bam earthquake and its damages for three different times during earthquake; first second (left), forth second (middle) and last second (right)
Various parameters can be set by users to alter settings such as the magnitude of earthquake, tolerances of resistances of features and the earthquake lasting time to see how features react to those changes. Users can determine the earthquake source by clicking on graphical interface and distances to features will be reviewed. For example we tried the model with these settings: • • •
Magnitude of earthquake=5.5 Richter Resistance tolerance=1 Earthquake lasting time=30 seconds
And the earthquake source has shown by red circles in Fig. 7. Resulting damages are shown in the same picture. Figure 7: An experimental earthquake and its results
6.
CONCLUSIONS AND FUTURE DIRECTIONS
In this paper, we presented a simulation of a real earthquake in Iran. After obtaining and preparing appropriate data, we imported them into NetLogo software. Besides we developed a mathematical model for earthquake energy that arrives in a point far from earthquake source and found unknowns of this formula in the study area. Finally, we simulated the earthquake in NetLogo software and added some options for altering parameters values and running the model. We observed some limitations with raster space of the software. Raster space is convenient for programming but needs unrealistic assumptions about features shapes. It is aimed to provide similar simulations in vector space to provide more flexibility to objects' shapes. REFERENCES Ansal, A., Akinci, A., Cultrera, G., Erdik, M., Pessina, V., Tonuk, G., et al. (2009). Loss estimationinIstanbulbasedondeterministicearthquakescenarios of theMarmaraSearegion(Turkey). Soil DynamicsandEarthquakeEngineering , 29, 699–709. Benenson, I., Martens, K., & Birfir, S. (2008). PARKAGENT: An agent-based model of parking in the city. Computers, Environment and Urban Systems , 32, 431439. Cinicioglu, S. F., Bozbey, I., Oztoprak, S., & Kelesoglu, M. K. (2007). An integrated earthquake damage assessment methodology and its application for two districts in Istanbul, Turkey. Engineering Geology , 94, 145–165. Ergonul, S. (2005). A probabilistic approach for earthquake loss estimation. Structural Safety , 27, 309–321. Evans, T. P., & Kelley, H. (2008). Assessing the transition from deforestation to forest regrowth with an agent-based model of land cover change for south-central Indiana (USA). Geoforum , 39, 819-832. Feng, X., Xuping, C., Aizhu, R., & Xinzheng, L. (2008). Earthquake Disaster Simulation for an Urban Area, with GIS, CAD, FEA, and VR Integration. Tsinghua Science And Technology , 13, 311-316. Gervais, E., Liu, H., Nussbaum, D., Roh, Y.-S., Sack, J.-R., & Yi, J. (2007). Intelligent map agents — An ubiquitous personalized GIS. ISPRS Journal of Photogrammetry & Remote Sensing , 62, 347-365. Keisuke, U., & Kazuo, K. (2008). Development of simulation system for the disaster evacuation based on multi-agent model using GIS. Tsinghua Science and Technology , 13, 348-353. Li, X., & Liu, X. (2007). Defining agents’ behaviors to simulate complex residential development using multicriteria evaluation. Journal of Environmental Management , 85, 1063-1075. Liu, K. (2004). Agent-based resource discovery architecture for environmental emergency management. Expert Systems with Applications , 27, 77-95. Nute, D., Potter, W. D., Maier, F., Wang, J., Twery, M., Rauscher, H. M., et al. (2004). NED-2: an agent-based decision support system for forest ecosystem management. Environmental Modelling & Software , 19, 831-843. Petit, C., & Magaud, F.-X. (2006). Multiagent meta-model for strategic decision support. Knowledge-Based Systems , 19, 202-211.
Russell, S., & Norvig, P. (2002). Artificial intelligence: a modern approach (2nd ed.). New Jersey: Pearson Education International. Shoham, Y., & Leyton-Brown, K. (2008). Multiagent systems: algorithmic, gametheoretic, and logical foundations. London: Cambridge University Press. Tang, A., & Wen, A. (2009). An intelligent simulation system for earthquake disaster assessment. Computers & Geosciences , 35, 871-879. Tang, F., & Ren, A. (2008). Agent-based evacuation model incorporating fire scene and building geometry. Tsinghua Science and Technology , 13, 708-714. Tsai, C.-H., & Chen, C.-W. (2009). An earthquake disaster management mechanism based on risk assessment information for the tourism industry-a case study from the island of Taiwan. Tourism Management , 31, 470–481. Wald, D. J., & Allen, T. I. (2007). Topographic slope as a proxy for seismic site conditions and amplification. Bulletin of the Seismological Society of America , 97, 1379–1395. Weiss, G. (1999). Multiagent systems: a modern approach to distributed modern approach to artificial intelligence. London: The MIT Press. Wooldridge, M. (2002). An introduction to multiagent systems. London: John Wiley & Sons, LTD. Xu, F., Chen, X., Ren, A., & Lu, X. (2008). Earthquake disaster simulation for an urban area, with GIS, CAD, FEA, and VR integration. Tsinghua Science and Technology , 13, 311-316.