Machine Learning Solution for Predicting Flood Damage

Project Overview

During the last few years, the United States (U.S.) has experienced unprecedented rainfall and flash flooding in a number of metropolitan areas, which has resulted in some of the costliest natural disasters in U.S. history. Government agencies have sought to apply data-driven solutions to aid in relief efforts to assess the damage resulting from storms and other natural hazards.

Special flood hazard areas are those areas that have been determined by federal agencies to have special flood, mudflow or flood-related erosion hazards based on historical events and local geology. After a flooding event has taken place, communities participating in the National Flood Insurance Program (NFIP) are required to determine whether structures within these flood areas are considered "substantially damaged" through inspection by local floodplain administrators. However, floodplain administrators assigned to this task are challenged by:

• Limited appraisal or construction backgrounds
• Developing reasonable estimates of structure values and structure-specific damages in accordance with NFIP requirements in a timely manner

Understanding that efficiency, speed and transparency are critical during large-scale disaster recovery, we worked closely with a U.S. government agency to develop a rapid response triage tool to predict damage levels and mitigate risk. Our Data Analytics and Data Management solution helped regional agency officials support timely damage determinations and reduce unnecessary costs associated with the inspection process.

Client Needs

• Lower cost for conducting flood damage inspections and estimates
• A less intrusive experience for communities recovering from flooding disaster
• Faster completion time to expedite contingent recovery activities
• Meet various stakeholder requirements
• Develop an accurate process which can be followed in different locations to support timely damage estimate determinations