Conference Agenda

Flooding is one of the most dangerous and costly natural hazards, and has a large impact on infrastructure, mobility, and safety. Despite the disruptive impacts of flooding, there is very little quantitative data available on the occurrence, frequency, or extent of urban floods.

To address this, our team, FloodNet, has been designing, building, and deploying low-cost, ultrasonic sensors to systematically collect data on street-level floods. FloodNet is a partnership between academic researchers (New York University and City University of New York) and NYC municipal agencies working in consultation with community organizations. The innovative FloodNet sensors are designed to be compact, rugged, low-cost, and deployed in a manner that is independent of existing power and network infrastructure. These requirements were implemented to allow deployment of a hyperlocal, city-wide sensor network, given that urban floods often occur in a distributed manner due to local variations in land development, population density, sewer design, and topography.

Over the last few years, in consultation with community stakeholders, FloodNet deployed a pilot sensor network to gather real-time data on floods related to high tides, stormwater runoff, and extreme precipitation events. Multiple use cases for the data have been identified, including improving hydrologic and hydraulic models, planning infrastructure investments for long-term resiliency, planning and advocacy by community members, and use of real-time alerts by community members, emergency response, and disaster recovery teams. NYC included FloodNet in their extreme weather task force agenda, and we plan to deploy hundreds of sensors across NYC over the next five years.

Private wells supply 35% of North Carolinians with drinking water. These well systems are unregulated by the EPA and are susceptible to contamination following flooding events. Despite this risk, there is limited understanding of the impact of flooding events on private well communities. Four common methods are used to delineate floodwater extents: FEMA disaster designated areas, FEMA 100-year floodplains, Height Above Nearest Drainage (HAND) modeled inundation, and satellite derived products like Dartmouth Flood Observatory (DFO) modeled inundation. In this study, we aimed to analyze these delineation methods in the context of Hurricane Florence to (1) determine uncertainties in estimates of area impacted and number of wells inundated, (2) analyze testing rates and contamination rates, and (3) examine increases in contamination post-Florence compared to NC Department of Health and Human Services baseline testing. Preliminary findings indicate vast discrepancies between methods, which estimate between 79.92 and 25,836 mi² of land were affected by Florence. Affected well estimates also differed substantially with 1,382 private wells flooded within the DFO flood boundary to 311,843 wells within the FEMA counties boundary, of which 35.4-55.5% were contaminated with total coliforms. All methods show increases in contamination post-Florence from baseline conditions and all methods agree a small percentage (0.3-1.3%) of flooded wells were tested. When analyzing baseline conditions, years with increased contamination rates are associated with larger testing sample sizes. Increasing testing is critical but optimizing testing locations based on estimated impacts is necessary to obtain a representative sample for effectively guiding recovery efforts with limited resources.

The risk of wildfires occurring at the wildland-urban interface is increasing and dramatically affecting the lives and infrastructure of much of the American West. Widespread volatile organic compound (VOC) and semi-VOC (SVOC) contaminations have been found in several water distribution systems in the aftermath of wildfires. One potential source of these contaminations is thought to be the thermal degradation of plastic components in drinking water distribution systems.

The goal of this study was to assess the type and variability of VOC/SVOCs generated when water infrastructure plastics were thermally degraded in the presence of water. Thermal degradation occurred by using a reactor design where plastic drinking water pipes were first submerged in water, heated to 100-300°C for 60 min, and then diluted with a continuous flow of clean water to assess the effect of flushing on decontamination. The plastic materials tested included new and exhumed high-density polyethylene (HDPE) and cross-linked polyethylene drinking water pipes. Results have shown benzene leaching at exposure temperatures as low as 150°C in all materials tested. Elevated exposure temperatures increased both the number of leached compounds and their concentrations. Plastic type was found to impact the leaching profile as well. Building on these results, we have compounded HDPE with a variety of antioxidants to understand the underlaying phenomenon causing the formation and leaching of contaminants following thermal degradation. Results from this study will have significant implications for wildfire and structure fire recovery as plastic materials are increasingly being used in drinking water conveyance.

Over 44 million United States residents rely on private wells for their drinking water supply which are not regulated like municipal systems and may be at higher risk of storm-related contamination if they lie within or adjacent to flooded areas. In terms of public health, the prevalence of acute gastrointestinal illness (AGI) following storm-related well contamination is of particular interest. Accordingly, this study examines private well data following Hurricane Harvey in Houston, TX, evaluates multiple techniques used to map the flooding extent of Hurricane Harvey, and examines the uncertainty amongst the flood mapping techniques in the context of their post-flood AGI prevalence correlation and the post-flood perception of well safety among survey respondents. The flood mapping techniques evaluated include output from the Dartmouth Flood Observatory, the Height Above Nearest Drainage method coupled with National Water Model retrospective output, and a Google Earth Engine based imagery mapping technique. Private well data come from in-situ sampling campaigns shortly following Hurricane Harvey, and preliminary results indicate that of the 263 respondents drinking well water without post-flood treatment, 48% tested positive for Total coliform, 13% tested positive for E. coli, and 35% reported their well as “not safe” to drink. By understanding how flood mapping methodologies correlate with AGI prevalence following flooding events, future research can better predict the spatial likelihood of AGI prevalence based on flood extent and can also help inform the need for well stewardship amongst well users, thus aiding public health authorities in their disaster response.

Bridges play a critical role in transportation networks. They are vulnerable to deterioration and aging, especially under the changing climate and increasingly frequent extreme weather events. Ensuring the longevity and avoiding costly failures of a bridge system is a vital yet challenging task that requires a well-planned maintenance strategy. This study introduces a multi-agent Reinforcement Learning (MARL) algorithm aimed at resolving the maintenance optimization challenge for bridge systems, with particular emphasis on factoring in the impacts of long-term deterioration and floods. The algorithm is designed to effectively allocate financial resources to bridges that are susceptible to failure caused by scour and flooding. The MARL framework models a stochastic Markov Decision Process that captures the interactions between agent decisions (i.e., maintenance decisions) and the state changes (e.g., floods with different return periods and bridge condition changes due to deterioration or repair) in a 50-year time span with 1-year time steps. To illustrate the application of the framework, the developed MARL framework was applied to a collection of bridges in Boston, MA. The findings suggest that multi-agent deep reinforcement learning algorithms have great potential in providing cost-effective maintenance strategies for aging infrastructure systems. The significance of this study lies in its valuable insights for cities worldwide on how to effectively optimize their limited resources for the maintenance and rehabilitation of critical infrastructure systems.