Increasing amounts of data, together with more computing power and better machine learning algorithms to analyse the data, are causing changes in almost every aspect of our lives. This trend is expected to continue as more data keep becoming available, computing power keeps improving and machine learning algorithms keep improving as well. Flood risk and impact assessments are also being influenced by this trend, particularly in areas such as the development of mitigation measures, emergency response preparation and flood recovery planning. Machine learning methods have the potential to improve accuracy as well as reduce calculating time and model development cost. It is expected that in the future more applications will become feasible and many process models and traditional observation methods will be replaced by machine learning. Examples of this include the use of machine learning on remote sensing data to estimate exposure and on social media data to improve flood response. Some improvements may require new data collection efforts, such as for the modelling of flood damages or defence failures. In other components, machine learning may not always be suitable or should be applied complementary to process models, for example in hydrodynamic applications. Overall, machine learning is likely to drastically improve future flood risk and impact assessments, but issues such as applicability, bias and ethics must be considered carefully to avoid misuse. This paper presents some of the current developments on the application of machine learning in this field and highlights some key needs and challenges.

During the aftermath of Typhoon Hagibis, we made flood and damage proxy maps, rapidly derived from synthetic aperture radar (SAR) data using change detection approaches. The maps have large spatial coverage over the Tokyo, Fukushima, Ibaraki, Iwate, and Nagano prefectures of Japan. The maps are also largely in agreement with various validation sources including aerial imagery, optical imagery and news sources. Apart from visual maps, we provide flood and damage extents in various formats compatible with geographic information system (GIS) applications. The data may potentially be used for applications such as typhoon risk modelling, investigating spatial correlations of typhoon impacts, and comparing alternative flood or damage mapping techniques.

In this study we explored the application of synthetic aperture radar (SAR) intensity time series for urban flood detection. Our test case was the flood in Lumberton, North Carolina, USA, caused by the landfall of Hurricane Matthew on 8 October 2016, for which airborne imagery—taken on the same day as the SAR overpass—is available for validation of our technique. To map the flood, we first carried out normalization of the SAR intensity observations, based on the statistics from the time series, and then construct a Bayesian probability function for intensity decrease (due to specular reflection of the signal) and intensity increase (due to double bounce) cases separately. We then formed a flood probability map, which we used to create our preferred flood extent map using a global cutoff probability of 0.5. Our flood map in the urban area showed a complicated mosaicking pattern of pixels showing SAR intensity decrease, pixels showing intensity increase, and pixels without significant intensity changes. Our approach shows improved performance when compared with global thresholding on log intensity ratios, as the time series-based normalization has accounted for a certain level of spatial variation by considering the different history for each pixel. This resulted in improved performance for urban and vegetated regions. We identified smooth surfaces, like asphalt roads, and SAR shadows as the major sources of underprediction, and aquatic plants and soil moisture changes were the major sources of overprediction.