Flood Science Communication Through Software Engineering

Abstract:

Massive amounts of earth observation data enable powerful flood science analyses, but flood risk management decision makers like governments and NGOs are often prohibited from taking full advantage of flood science due to its inaccessibility. Cloud to Street’s flood information system has bridged this information gap in 15 countries since 2015. In 2018, for example, our flood information system enabled the relocation of refugees in the Republic of Congo shortly before a flood emergency. The flood information system also facilitated requests for international flood aid from the World Food Programme in November 2019. This paper outlines how we designed and engineered the flood information system as a cloud computing near real-time flood mapping and monitoring system built on Google Cloud Platform. This system produces insights from a wealth of data in near real-time from sources such as MODIS, Landsat-8, Landsat-7, Sentinel-1, Sentinel-2, and GSMaP. Use of the latest standards in geospatial data formats enables us to combine earth observation data from multiple sources to create flood maps and data visualizations. An in-house Python API makes it easy for remote sensing scientists to experiment with the data and a REST API allows other systems to integrate the results. Crowdsourcing tools allow users to input observations from their phones allowing for two-way information exchange through the platform. The data from these systems is combined and presented in a web-based user interface designed alongside non-expert users from governments and NGOs and built with open source software like Angular, OpenLayers, and GeoNode. We conclude with principles and best practices for human-centered design of disaster information systems for places with otherwise limited access to top-tier flood science.