Debris is the waste generated by a hazardous event such as a natural disaster or a terrorist attack. Debris operations are categorized into three phases: (i) Clearance which include operations performed during or right after a hazardous event with the goal of clearing the debris from major arteries to give access to critical facilities and to aid in emergency operations; (ii) Collection which include transporting the debris from the disaster area to collection sites; (iii) Disposal which include transporting the debris to the final disposal sites and choosing the disposal method (e.g., landfill, reduce, recycle, or reuse). The research objective of this award is to develop mathematical models that capture the important characteristics of the debris related operations in each stage along with methodologies for solving these mathematical models efficiently. The models will be enhanced by including considerations on fairness (given the public impact nature of the application) and robustness (given various types of uncertainty in these settings). The ultimate goal is the development of an integrated model that considers the interactions between the decision problems in all three phases simultaneously.
The results of this research will facilitate the development of decision support tools for debris clearance operations during the planning, response, and short and long term recovery stages of a hazardous event. The decision support tools built will be used in tactical, operational and comprehensive policy evaluation phases by both local and federal emergency management agencies such as FEMA and USACE. These tools will aid in evaluating the impact of resource availability and prioritization decisions on the costs and service levels of debris operations and ultimately will result in faster and more cost effective debris management operation.
The main goals of this project are to develop quantitative models and decision support tools for post-disaster debris management operations, which are challenging due to the size as well as the costly and complicated nature of these activities. Post-disaster debris operations are categorized into three phases: (1) Debris Clearance: Operations performed during or right after a hazardous event with the goal of clearing the debris from major arteries to provide access to critical facilities and to aid in emergency operations. (2) Debris Collection: Transportation of the debris from the disaster area to collection sites. (3) Debris Disposal: Transportation of the debris to the final disposal sites and the choice of the disposal method (e.g., landfill, reduce, recycle, or reuse) for a given debris type at a given location. The debris clearance stage presents a computationally difficult network problem over multiple periods. We focus on two debris clearance decisions under two different settings. In the first one, all debris amounts are known with certainty. For this setting, we introduce heuristics based on solving periodic problems sequentially to build an overall solution. To further improve the heuristic output, multiple models are solved for each period, each sequentially more complex. We introduce a secondary heuristic that uses multiple relaxations of the problem to generate valid bounds on the optimal solutions. The combination of these heuristics has demonstrated that our methods can find solutions within a few percent of optimality on smaller sized networks. For larger network instances, which have been generated as datasets using the hazard estimation tool Hazus by FEMA, we can significantly improve both solution and bound quality compared to commercial solvers in a fraction of the time. The second setting involves cases where debris amounts are only partially known due to lack of geographical and demographic data on the disaster area. For this case, we develop a stochastic model that finds the optimal road clearance schedule, and enhance the computational efficiency by providing a continuous-time approximation of the model. Using simple network settings, we show that this approach provides a close approximation of the original model, and can be solved in significantly shorter time. Computational experiments on disaster instances reveal the quality of our heuristic solutions, and provide policy-based results to aid the decision makers. The debris collection decisions are modeled using a graph partitioning problem. This problem contains decisions of assigning regions of a city to contractors, deciding which dumping sites should be opened, and the actual routing of vehicles from collecting debris to taking them to the dumping sites. We decompose the problem into three separate models that allows us to quickly obtain solutions for large-scale problems. We have implemented and computationally tested this model on various US-based and non-US based instances. In the last stage, the main decisions consist of which debris processing facilities to open, which processes to apply in each facility, and whether the debris on each road segment should be disposed of or recycled. We propose a mathematical model that can be solved in at most a few minutes. The model has been tested on various potential earthquake scenarios on Trinidad and Tobago, and the results indicate that slower collection (due to on-site separation) is convenient for reducing disposal time when debris separation at temporary sites is a bottleneck. We have developed a prototype for a decision support tool for the last stage. The tool can be run on MS Excel and uses an open source solver to solve the underlying mathematical model. The decision support tool can be used for what-if analysis in the preparedness stage by running a set of potential scenarios. It can also be used as a post-disaster decision making tool by entering actual disaster data and solving accordingly. The current version of the tool and a tutorial are available at http://debrismanagement.gatech.edu, where we have also summarized the results of our work. Throughout the project, we have introduced the ideas of complex decision making to earthquake engineers in the civil engineering community. Those we have worked with are exploring how to recycle debris from an earthquake into raw building material, which can potentially be added into our models as a decision for processing debris. Furthermore, we have been in discussion with USACE for incorporating our decision aid tools in their emergency response training and simulation models that are used by FEMA and USACE for training domestic and international responders. We have also established contact with the UN in Haiti, private contractors, and local emergency management agencies both from the US and abroad to assess and understand their needs.
