Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Mapping Housing Stock Characteristics from Drone Images for Climate Resilience in the Caribbean

Isabelle Tingzon, Nuala Margaret Cowan, Pierre Chrzanowski

TL;DR

The study addresses the need for timely baseline housing stock data in climate-vulnerable Caribbean SIDS to support resilience planning. It proposes an end-to-end workflow that combines very high-resolution drone imagery, the Segment Anything Model for building footprints, and CNNs for roof type and roof material classification, with cross-country evaluation between Dominica and Saint Lucia. Results show that locally trained models often excel for roof material while cross-country generalization varies, highlighting domain shifts and the value of local data and potential domain adaptation. The work also emphasizes capacity building and co-creation with government partners to operationalize EO-AI workflows for sustainable climate resilience.

Abstract

Comprehensive information on housing stock is crucial for climate adaptation initiatives aiming to reduce the adverse impacts of climate-extreme hazards in high-risk regions like the Caribbean. In this study, we propose a workflow for rapidly generating critical baseline housing stock data using very high-resolution drone images and deep learning techniques. Specifically, our work leverages the Segment Anything Model and convolutional neural networks for the automated generation of building footprints and roof classification maps. By strengthening local capacity within government agencies to leverage AI and Earth Observation-based solutions, this work seeks to improve the climate resilience of the housing sector in small island developing states in the Caribbean.

Mapping Housing Stock Characteristics from Drone Images for Climate Resilience in the Caribbean

TL;DR

The study addresses the need for timely baseline housing stock data in climate-vulnerable Caribbean SIDS to support resilience planning. It proposes an end-to-end workflow that combines very high-resolution drone imagery, the Segment Anything Model for building footprints, and CNNs for roof type and roof material classification, with cross-country evaluation between Dominica and Saint Lucia. Results show that locally trained models often excel for roof material while cross-country generalization varies, highlighting domain shifts and the value of local data and potential domain adaptation. The work also emphasizes capacity building and co-creation with government partners to operationalize EO-AI workflows for sustainable climate resilience.

Abstract

Comprehensive information on housing stock is crucial for climate adaptation initiatives aiming to reduce the adverse impacts of climate-extreme hazards in high-risk regions like the Caribbean. In this study, we propose a workflow for rapidly generating critical baseline housing stock data using very high-resolution drone images and deep learning techniques. Specifically, our work leverages the Segment Anything Model and convolutional neural networks for the automated generation of building footprints and roof classification maps. By strengthening local capacity within government agencies to leverage AI and Earth Observation-based solutions, this work seeks to improve the climate resilience of the housing sector in small island developing states in the Caribbean.
Paper Structure (6 sections, 4 figures, 4 tables)

This paper contains 6 sections, 4 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Data
  3. Methods
  4. Results and Discussion
  5. Local Capacity Building
  6. Conclusion

Figures (4)

  • Figure 1: Examples of VHR drone-derived roof image tiles for each of the roof material categories (top row) and roof type categories (bottom row).
  • Figure 2: Building footprints from (a) Microsoft, (b) Google, (c) OpenStreetMap, and (d) SAM overlaid on a drone image taken in Salisbury, Dominica from OpenAerialMap smith2015openaerialmap.
  • Figure 3: Proposed workflow for the automatic generation of housing stock information from drone images using DL models.
  • Figure 4: Drone images (top) and corresponding roof material classification maps (bottom) of Coulibistrie, Dominica taken before (left) and after (right) Hurricane Maria.