Innovative Decision Support System based Artificial Intelligence & Spatial Planning for Disaster Risk ReductionSession VI: Reduction of Risk in Environmental Disaster Dr. Hussain Aziz SALEH1,2 & Prof. Georges ALLAERT2 1Ministry of Local Administration and Environment, Damascus, Syria Tel:+963 11 211 9955, Fax:+963 11 2119954, E-mail:Hussain.Saleh@UGent.be 2Institute for Sustainable Mobility, Faculty of Engineering,Ghent University, Krijgslaan 281 (IDM, S8), B-9000 Gent, Belgium. Tel:+32 9 264 47 17, Fax:+32 9 264 49 86, E-mail:Georges.Allaert@UGent.be
Contents of the Paper • Introduction to Processing Architecture of the Disaster Management Information System. • Information Flow & Early WarningSystem Architecture in Flood Management. • Geomatic & Information Communication Technology. • Multiobjective Combinatorial Optimisation Problems (MCOPs) • Metaheuristic Techniques • Disaster Warning Network for the Danube Basin. • Optimiasting some real-life applications (Earthquake & Flooding) • Conclusion and future work.
Disaster Management Cycle Disaster • Prevention & Mitigation • Hazard prediction & modelling. • Risk assessment & mapping • Spatial planning • Structural & non-structural measures • Public Awareness & Education Preparendness -Scenarios develpment Emeregency Planning Training Alert -Real time monitoring & forcasting -Early warning -Secure & dependable telecom -Scenario identification -all media alarm • Post Disaster • Lessons learnet • Scenario update • Socio-economic & environment impact assessment • Spatial (re)planning • Response • -Dispatching of resources • Emergency telecom • Situational awareness • Command control coordination • Information dissemination • Emergency healthcare • Recovery (Rehabilitation & Reconstruction) • Early damage assessment • Re-establishing life-lines transport & communication infrastructure • Reinforcement
Information Flow & Early WarningSystem Architecture in Flood Management Data collection: Space, Airborne ground Ancillary Risk Assessment Early Warning Hydrological Stations Provision of Forecasts Monitoring Damage Assessment Relief provision Dynamic Data Processor Desktop GIS Tools • Distribution to the Users Internet IP/http, Dissemination Main Services (File, Spatial Data Engine, Internet Map, Web, etc) Web site, bulletin, e-mail, fax, radio, telephone, etc.
Flood Management & Risk Reduction (FMRR) Disaster reduction as part of sustainable development through: Strengthen institutions (especially in communities) to build resilience Build risk reduction into emergency management and recovery 1. Regional FMRR centre Capacity Building (Training Unit) 2. Structural Measures & Flood Proofing Data Collection & Processing FMRR 3. Trans-boundary mediation Forecasting, Warning & Dissemination 4. Flood Emergency Mngmt 5. Land Use management Annual Flood Forum, Workshops, Communications 6. Spatial Planning
Remote Sensing Technology A Close View of Camorta Island, Tsunami - 2004 Post-Event Pre-Event IRS-P6 AWiFS Image of 21-Dec-04 IRS-P6 AWiFS Image of 26-Dec-04
Sat1 Sat2 Sat3 Z WGS-84 Receiver Sat4 (0, 0, 0) Y Local X Global Navigation Satellite Systems GNSSs GPS GALILILEO GLONASS
Geographic Information ProcessDatabase & Visualization The geographic information process consists of three stages: Data acquisition, data processing, and data dissemination Visualisation “Worth a Thousand words” Database “Not easy to interpret”
Rules of Spatial Planning in Flood Management flood reduction is part of sustainable development and the various action possibilities to improve preventive flood protection based on spatial planning and urban development can be described as follow: • Protection of existing retention areas: e.g., declaration of flood areas, etc. • Extension of retention areas: e.g., creating detention ponds, etc. • Retention in the catchments: e.g., restoration of small streams, etc. • Minimisation of damage potential: e.g., preventive land-use management, etc. • Technical flood protection measures: e.g., dikes, retention ponds, etc. As shown from above, flood risk can only effectively be reduced if, in addition to the technical measures, spatial planning regulates land-use in flood-prone areas.
Meteorological forecast Heavy rain/snowmelt or high inflow Data collection and transmission Minimise Boundary estimation (rainfall, tide, ..) time Forecast calculation Warning dissemination Emergency action maximise Flooding starts Flood Forecasting & Early Warning Provide accurate forecasts with a suitable lead time; and a timely and effective dissemination
Multiobjective Combinatorial Optimisation Problems (MCOPs) Many real-life applications involves two types of problem difficulty: 1) multiple, conflicting objectives, 2) a highly complex search domain. fi(x) is the ith objective function to be optimised xis a set of decision vectors Question: How do I combine all the fieldwork components and find the optimal schedule for receivers. Therefore, we need to answer not only the question “What is the best?”, but also “What is sufficiently robust?” Answer: Metaheuristic Techniques. X is the search domain. Minimize cost
Objective function Local minimum Global minimum Iteration Minimization(Global and local minima) Maximisation(Global and local maxima) Objective function Global maximum Local maximum Iteration In mechanical engineering, an engineer wishes to design a car consisting of composite materials. The engineer, by optimization (maximization in this case), will conceivably design a lighter, stronger, attractive and safer composite car In GPS surveying, a surveyor wants to design a network taking in consideration components of the field-work. The GPS surveyor, by optimisation (minimization in this case) will design a cheaper and more acceptable schedule
Schematic Representation of the Search Progress of Metaheuristic Techniques Select a given solution S I(S) and compute its value C(S). Generate a schedule S’ I(S) and compute its value C(S’). If C(S’) < C(S) then, S’ replaces S as a current solution. Otherwise, retain S and generate other moves until C(S’)<C(S) for all S’ I(S). Terminate the search and return S as the local optimal solution. Values of Si or the objective function Q(S) Q(S) I(S0) I(S12) I(S0) I(S6) S12 S0 S6 S1 S2 S7 S11 S3 S0 S5 S8 S4 S10 I(S4) I(S9) S9 Iterations
Initial Solution Formation Neighbourhood Search by Move Formation (Local Search) Provisional Neighbourhood Formation Search by Solution Formation (Development and guided search) Neighbourhood for the next solution formation Acceptance Criteria Termination of the Search Metaheuristic Techniques • Metaheuristic technique: is an iterative, self-learning procedure for quickly and efficiently identifying a high quality solution for COPs. • Types of Metaheuristics: 1- Solution improvement techniques: Simulated AnnealingSA, Tabu SearchTS. 2- Solution construction techniques: Ant Colony OptimisationACO, Genetic AlgorithmGA, Neural NetworkNN. • Fundamental concepts of the metaheuristics: 1- Selection or construction of the initial solution. 2- Generation of the neighbouring solutions. 3- Acceptance of solutions. 4- Stopping criteria.
Û High free energy (high temperature) Case I Temperature K The material as a system of particles L Start Ti (Cooling process) F Annealing (slow cooling) Quenching (fast cooling) Tf Stop Û Û Iterations Global minimum of the energy Local minimum of the energy Case II Case III The Simulated Annealing (SA) Technique • Based on analogy between combinatorial optimization and annealing process of solids • Improvement of solution for move from S to S’ always accepted • Accepts uphill move, only with given probability (decreases in a number of rounds to zero) • Cooling parameters: the initial starting value of the temperature, the temperature length (Markov Chain), L. the cooling ratio, F Crystalline solid (frozen) state (T=0) Amorphous solid (frozen) state (T=0)
The Tabu Search (TS) technique • Add memory to LS (Prevent Cycling, increase diversity of exploration, encourage escape from local optima) • Avoids cycling (visiting same solution more than once) by use of short term memory for the very recent history and long-term memory for the distant history. • TS parameters: Size of Memry, Contents of memory (solution based, attributes: e.g. changes to solution features), long term memory • Tabu List • Candidate List • Tabu Tenure
The Ant Colony Optimization ACO The modified ACO will have some major differences with a real (natural) one: • Artificial ants will have some memory • Artificial ants will not be completely blind • Artificial ants will live in an environment where time is discrete
ACO We will finish on time. Do not worry OK?? • Construction the Initial Schedule • Local Search • Local Updating Rule • Global Updating Rule
The Genetic Algorithms (GAs)The evolution process for generating a population Gene Initial Generation One- point crossover Chromosome Father Mating Selection Mother Crossover N Child 1 Mutation Child2 Next Generation One-flip mutation in child 2
Neural Networks NNimitate the human brain NN consists of a collection of interconnected simple computational units that work together cooperatively (like synapses in the human brain). These units are input units for receiving information from the problem domain, hidden units as internal weight structure, and output units for broadcasting data.
Different types of the GPS Surveying Networks The Seychelles GPS Network (linear) The Malta GPS Network (triangulation)
Handicapped Person Transportation in the City of Bruxelles using Grouping Genetic Algorithm
Disaster Warning Network for Danube Basin (DWN) • DWN is a system of satellites and ground stations for providing real time early warning of the impact of a disaster. • DWN implements a network of Control Stations (CSs) spread over the whole geographic area of the hazard and transmits their observations to a master station. • DWN provides this reliable information on a continuous basis through the parallel process of coverage accuracy prediction and integrity risk simulation functions.
Disaster Warning Network for the Danube Basin Differential GPS GPS Signal 1 GPS Signal 2 Satellite GPS Signal 4 GPS Signal 3 GPS Signal RR Roving Receiver 1 Error correction message 2 Error correction message 1 Error correction message 3 Error correction message 4 Roving Receiver 2 Roving Receiver 4 Reference Receiver RR Roving Receiver 3
The DWN model for Environmental Pollution Control • Multiobjective Optimisation Problems(MOPs) • f(x) is the objective function to be optimised and X is the search domain • Ps is a multiplier for measuring the size of the user population near the location s (e.g., safety teams) • f(acy, ing) are the evaluations of accuracy coverage and integrity performance respectively for user location s. • f(cost) is the cost function of a given DWN design. • N is the number of user locations. • n is the number of obtained DWN designs.
Disaster Warning Network (DWN) • Dynamic Metahuristics Algorithms (DMA) potentially have these capabilities to produce set of high quality real-time designs that can model more closely and easily the objective functions and tolerate the noisy evaluations given by these functions. • The final design should be robust (i.e., performs well over a wide range of environment conditions), and flexible (i.e., allows easy adaptation after the environment has changed).
Disaster Warning Network (DWN) • DMAs can successfully handle a mix of continuous and discrete parameters as well as selecting individual components from database. Therefore, another objective of this project is to connect DWN to a database which includes all the related information (physical, geographical,and biochemical) colleted by other observation techniques (e.g., Geographic Information System (GIS), Remote Sensing (RS), photogrametery, internet, air (water and soil) survey technology, etc).
Disaster Warning Network (DWN) DWN can effectively optimise this problem by: 1) providing access to a wide range of data types in real-time, 2) combining the observational data with innovative data analysis to improve forecasting and risk assessment, 3) and being a user friendly tool in decision support of environmental management.
Current Applications (Earthquake) (1) • GEO-INFORMATION TECHNOLOGY AND ARTIFICIAL INTELLIGENCE FOR OPTIMIZING THE DISPLACEMENTS OF EARTHQUAKES: A CASE STUDY FOR SYRIAN ACTIVE FAULT ZONES • With Dr. Mohamad Rukieh, Syrian Ministry of Commuincation and Technology. • International Union of Geodesy and Geophysics, IUGG07 • Perugia, Italy, July 3-14 2007 • http://www.iugg2007perugia.it/
Current Applications (Flooding) (2) • Intelligent Decision Support System based Metaheuristics and Spatial Planning for Flood Management in Flanders, Belgium • With: Wouter VANNEUVILLE (Flandes Hydraulics Research), Georges ALLAERT, Philippe DE MAEYER (Ghent University) • The 4th International Conference on Information Systems forCrisis Response and Management, • Special Session on SPATIAL PLANNING (SP) FOR DISASTER RISK REDUCTION (DRR) • Delft, the Netherlands, May 13-16 2007 • http://www.iscram.org
Conclusion & Future Work • The most significant development in these recent years for disaster management lies in the better integration of GNSS, image processing and GIS systems coupled with intelligent algorithms Early Warning Challenges … • Warnings / Risks are NOT Understood • Information is Scattered • Dissemination is Limited • “High Availability” and Redundancy are Essential • Time is LIFE Outlook more robust models
Innovative Decision Support System based Artificial Intelligence & Spatial Planning for Disaster Risk ReductionSession VI: Reduction of Risk in Environmental Disaster Dr. Eng. Hussain Aziz SALEH Ministry of Local Administration & Environment, Damascus, Syria. Tel: 00 963 11 211 9955 Fax: 00 963 11 211 9954 http://iridia.ulb.ac.be/~hsaleh/ Hussain.Saleh@UGent.be • The Joint Regional Conference On: Disaster: Relief and Management : International Cooperation & Role of ICT • 14-17 April 2007 • Alexandria • Egypt Syrian Ministry of Higher Education