The deployment of effective flood early warning systems (EWS) requires reliable hydrometeorological data, including historical and real-time measurements of river levels and rainfall. This is a major challenge in emerging markets, as they often have limited access to historical and real-time hydrometeorological datasets. While global and regional forecasting services can enhance the efficiency of EWS, these services are typically designed for regional-level forecasts and may not meet the specific needs of national, district, or community-level forecasts.
To address this challenge, we incorporate frontier technologies such as artificial intelligence (AI) and machine learning (ML) into our EWS. These technologies enhance the accuracy and effectiveness of flood warnings, leading to better preparedness and response strategies. Specifically, we employ Long Short-Term Memory (LSTM) networks, which are widely used for weather forecasting, stock market predictions, and other applications requiring time-series predictions based on diverse data inputs. LSTM networks enable us to predict outcomes from continuous weather data using a wide range of inputs, allowing us to generate risk profiles based on real-time and predicted data. Additionally, we can test our trained models against simulated inputs to explore hypothetical conditions and "what if" scenarios that have not yet occurred.
To operationalize these models, we are developing an open-source ML-informed risk assessment software framework. This framework will integrate the ML model with demographic and GIS data to provide not only forecasted alerts of impending floods but also insights into the expected impact of disasters on specific communities.
The open-source nature of this software framework will enable other organizations to enhance their disaster preparedness and response capabilities. Developing this project as open source and "in the open" fosters collaboration with the broader climate risk reduction community, which may have different use cases from our own. These collaborations allow our initial vision to expand rapidly and efficiently, covering scenarios that we had not previously considered or were further along in our development roadmap. The collective knowledge of the wider community helps guide the software development process, making it more useful more quickly within an inclusive and positive community of stakeholders.
Simile is composed of 26 individuals who come from various backgrounds, cultures, and professional experiences. Our team includes experts in technology, design, climate change, hydrology, marketing, finance, and project management. This diversity in experiences and perspectives encourages innovative thinking and fosters a collaborative environment where everyone’s unique strengths contribute to our collective success.