Post event flooding maps are currently extracted from synthetic-aperture radar (SAR) and/or optical satellite images or developing using hydraulic model simulations. Several sources of uncertainties impact the accuracy of such flood maps constructed from each method, especially in urban areas. An integrated approach that combines satellite imagines of flooded areas, hydraulic models, and markers from social media that should reduce these uncertainties and allow a more accurate reconstruction of flooded urban areas, is presented in this paper. The flooding associated with Hurricane Harvey in Houston, TX was chosen as a case study. Model validations demonstrate the effectiveness of our integrated approach in reconstructing an accurate flooding map, as well as the temporal and spatial patterns of flooding. Using the experience from this case study we discuss the possibility to use satellite data, instead of groundbased rainfall gauge measurements as precipitation inputs to the hydraulic model; and possible error sources in simulating flooding in urban areas using the hydraulic model.

<p>Recent disasters stress the demand of fast and reliable tools for flooding forecasting, where the real-time prediction of extreme events becomes essential to avoid potential hazards for the population. In this work, we focus on the flash flooding phenomenon, given by the combination of temporally concentrated rainfall and steep slopes. Such configuration is typical in the St. Lucia island, in the eastern Caribbean Sea, that we exploit as a case study. It is possible to simulate the full evolution of rainfall by numerically solving the Shallow Waters equations (SW) on a computational domain. A preliminary comparison with historical events proved that an accurate solution is achieved only when the Digital Elevation Model (DEM) presents a resolution equal or inferior to 5 meters. With this grid resolution the whole island is discretized in over 60M cells at best, forbidding a real-time application of the SW solvers in flash flooding events.</p><p> In this work we present a machine-learning surrogate model for a SW solver to estimate the level of the flooding danger. It is evaluated through a synthetic parameter, hereafter referred as flag, that takes in account both the water depth and its velocity. Therefore, flooding patterns in the island are represented through high-resolution maps with discrete values of flags, varying from 0 – safe to 4 – extremely dangerous.</p><p>The final aim is to solve a supervised regression, training a Multi-Layer Perceptron Neural Network (MLPNN) to map sequences of time- and spatial-varying rainfall (input features) to the corresponding previsions of flags (output features) shifted ahead of time. To do so, we first generate a rough database by simulating more than 30 flash flooding events, using an in-house validated code, whose input is the temporal and spatial rainfall distribution obtained by radar measurements of events occurred in the past. DEM resolution is set to 5 meters and SW solver solutions is sampled every 6 mins. Given the high dimensionality of the problem, both the inputs and the outputs of the simulations are preprocessed using an Incremental Principal Component Analysis (IPCA) to extract the scores and loadings. The elbow charts indicate the correct number of principal components, set to 8, that explains the 95% of the cumulative explained variance. The scores given by IPCA processing of rainfall are built into sequences of five elements, endowing the algorithm a memory. The min/max regularization are applied to the database. The MLPNN training phase is fastened through batch feeding and monitored to prevent overfitting, relying on Tensorflow library. To test the generalization capability of the synthetic model was verified by forwarding events that were not included in the original database.</p>

<p>A     recent report “The Future is Now: Science for Achieving Sustainable Development” Global Sustainable Development Report 2019 - SDG Summit’      as part of the activity of Agenda 2030 of UN, highlights the opportunity to develop Early warning system for drought, floods and other meteorological events, that by providing timely information can be used by vulnerable countries to build resilience, reduce risks and prepare more effective responses. Following the suggestion,      combining outputs from Global Circulation models, remote sensing, hydraulic models and machine learning tools,       a local scale flooding Early Warning System (EWS) is proposed for the St. Lucia island (     Caribbean). The objective of the EWS is to provide forecasts of potentially dangerous flooding phenomena at different time scale: a) 0-2 hours, nowcasting; b) 24-48 hours, short range; c) 3-10 days, middle to long range. Data used to build the model are: Geopotential Height (GPH) fields at 850 hPa and Integrated Vapor Transport (IVT) fields from European Centre for Medium-range Weather Forecasts (ECMWF) - Reanalysis v5 (ERA5); Tropical Cyclone tracks from NOAA-NHC; 18 weather stations homogeneously distributed in the island; rainfall map data from the weather radar in Saint Lucia. GPH and IVT fields were defined between 110°W - 10°W and 45°N - 10°S. The EWS is constituted by an ensemble of flooding risk forecast subsystems which is potentially applicable to Atlantic tropical and extra-tropical regions. Different approaches are used for each subsystem      to link large scale atmospheric features to local rainfall and flooding: a) Non-homogeneous Hidden Markov and Event Synchronization models to translate IVT and GPH at 850 hPa  fields (from ECMWF-Set II- Atmospheric Model Ensemble) in local      daily rainfall amount and probability of  exceedance of  a prefixed heavy rainfall threshold; b) a physical based cyclone/rainfall  model to convert      Tropical cyclone attributes – position and      maximum wind      velocity       (provided from National Hurricane Center)- in rainfall intensity spatial distribution on the island; c) a surrogate model for a  fast and accurate prediction of flooding events that is obtained from a multi-layer perceptron neural network (MLPNN), which is trained on a high-fidelity dataset relying on solution of the full two-dimensional shallow water equations with direct rainfall application.        Results show an excellent ability of the models to identify the climatic configurations that determine the occurrence of extreme events and the exceeding of threshold values ​​that generate floods. In particular, during the late hurricane season September-October-November, when is highest the probability of flood events, the EWS was able to forecast the occurrence of critical climatic configurations 86% of the times they occurred. The EWS was able to predict the exceeding of the rainfall threshold that generated floods 80% of times.</p>