mwBTFreddy: A Dataset for Flash Flood Damage Assessment in Urban Malawi

Abstract: This paper describes the mwBTFreddy dataset, a resource developed to support flash flood damage assessment in urban Malawi, specifically focusing on the impacts of Cyclone Freddy in 2023. The dataset comprises paired pre- and post-disaster satellite images sourced from Google Earth Pro, accompanied by JSON files containing labelled building annotations with geographic coordinates and damage levels (no damage, minor, major, or destroyed). Developed by the Kuyesera AI Lab at the Malawi University of Business and Applied Sciences, this dataset is intended to facilitate the development of machine learning models tailored to building detection and damage classification in African urban contexts. It also supports flood damage visualisation and spatial analysis to inform decisions on relocation, infrastructure planning, and emergency response in climate-vulnerable regions.

mwBTFreddy: A Dataset for Flash Flood Damage Assessment in Urban Malawi

The paper "mwBTFreddy: A Dataset for Flash Flood Damage Assessment in Urban Malawi" presents a novel dataset specifically designed for studying flash flood damage in African urban settings, focusing on Blantyre, Malawi, after Cyclone Freddy in 2023. This paper addresses the critical need for localized data resources to enhance machine learning models aimed at automatic damage classification.

The research introduces the mwBTFreddy dataset, which includes pre- and post-disaster satellite imagery accompanied by comprehensive JSON annotations outlining building coordinates and evaluating damage levels. This dataset serves as a tool for developing models that can effectively classify flood damage, aiding in informed decision-making for urban planning, emergency response, and infrastructure development in Malawi's urban areas prone to climatic events.

Key Contributions

The dataset fills a significant gap in geospatial resources tailored to urban Malawi, precisely detailing the impacts of cyclonic events like Cyclone Freddy. It offers substantial ground for:

• Building Detection and Damage Classification: Leveraging machine learning algorithms to automatically detect and classify structural damages, ranging from no damage to complete destruction.
• Environmental Impact Analysis: Supporting flood damage visualization and spatial analysis, which are crucial for understanding the environmental implications and disaster impact on urban settings.

Numerical Insights

The dataset comprises 696 annotated instances out of 1026 images initially collected, with annotations based on a refined process that excludes buildings with significant visibility issues or geographical overlaps. These images represent a focused geographical area, ensuring coverage where the most severe damage from Cyclone Freddy occurred.

Implications and Future Directions

Practically, the dataset allows local authorities to make informed urban planning decisions, particularly in developing infrastructure to mitigate the effects of future climatic disasters. Theoretically, it offers a foundation for examining the efficiency of machine learning models in context-aware disaster assessment.

Future research may explore integrating this data with community-generated insights or simulate predictive models for assessing potential future climate scenarios. This integration could enhance the accuracy and applicability of machine learning approaches to broader African contexts, beyond Malawi.

This paper lays a critical foundation for geospatial data-driven solutions in disaster risk reduction and urban resilience planning in Malawi and presents opportunities for extending insights to other regions with similar ecological vulnerabilities.