Low-cost UAV surveys of hurricane damage in Dominica: automated processing with co-registration of pre-hurricane imagery for change analysis

Abstract

In 2017, hurricane Maria caused unprecedented damage and fatalities on the Caribbean island of Dominica. In order to 'build back better' and to learn from the processes causing the damage, it is important to quickly document, evaluate and map changes, both in Dominica and in other high-risk countries. This paper presents an innovative and relatively low-cost and rapid workflow for accurately quantifying geomorphological changes in the aftermath of a natural disaster. We used unmanned aerial vehicle (UAV) surveys to collect aerial imagery from 44 hurricane-affected key sites on Dominica. We processed the imagery using structure from motion (SfM) as well as a purpose-built Python script for automated processing, enabling rapid data turnaround. We also compared the data to an earlier UAV survey undertaken shortly before hurricane Maria and established ways to coregister the imagery, in order to provide accurate change detection data sets. Consequently, our approach has had to differ considerably from the previous studies that have assessed the accuracy of UAV-derived data in relatively undisturbed settings. This study therefore provides an original contribution to UAV-based research, outlining a robust aerial methodology that is potentially of great value to post-disaster damage surveys and geomorphological change analysis. Our findings can be used (1) to utilise UAV in post-disaster change assessments; (2) to establish ground control points that enable before-and-after change analysis; and (3) to provide baseline data reference points in areas that might undergo future change. We recommend that countries which are at high risk from natural disasters develop capacity for low-cost UAV surveys, building teams that can create pre-disaster baseline surveys, respond within a few hours of a local disaster event and provide aerial photography of use for the damage assessments carried out by local and incoming disaster response teams.

Automated High-Resolution Time Series Mapping of Mangrove Forests Damaged by Hurricane Irma in Southwest Florida

“…This is the only period for which ground-truth data were available, and so the results are assumed to be representative of all maps. Single-date validation is not the most robust means for assessing accuracy when change detection is involved, but is often the only available approach when assessing natural disaster impacts [24,25]. Using field surveys, we attempt to compensate for the single-date validation issue by acquiring the most accurate validation data [26].…”

Quantifying Sediment Deposition Volume in Vegetated Areas with UAV Data

“…Zekkos et al [67] used the GCPs as checkpoints and calculated the root mean square error (RMSE) as vertical accuracy (Table 2). Schaefer et al [68] performed UAV surveys four months after hurricane Maria in different parts of Dominica to investigate the topographical changes. They assessed the uncertainty in their DEMs using the pre-event UAV data from Zekkos et al [67] as reference.…”

“…They assessed the uncertainty in their DEMs using the pre-event UAV data from Zekkos et al [67] as reference. Schaefer et al [68] did not lay out any GCPs, due to safety considerations in accessing survey areas on foot. They matched their data with the data collected by Zekkos et al [67] in three dimensions by comparing the point clouds or the GCPs derived from pre-event data.…”

“…The comparison was made for unchanged features such as roofs and roads. Schaefer et al [68] applied the resulting transformation matrix to the whole cloud. They assessed the DSMs accuracy (Table 2) by creating profiles in ArcGIS using the Stack Profile tool along the airport runway or roads and comparing in a Python script.…”

Automated Satellite-Based Assessment of Hurricane Impacts on Roadways

“…Aerial images are a valuable source of data for hurricane damage assessment [2]. However, flying helicopters and drones over damaged areas are highly prone to weather conditions.…”