Floods are recurrent hydrological phenomena characterized by the inundation of land areas due to an excessive accumulation of water. The causes of flood may vary, and their severity can be increased by the conjunction of causes; they may occur due to elevated water levels in rivers, intense precipitation in local and/or upstream areas, soil saturation, suboptimal drainage infrastructure, local topographic features, rapid snowmelt, Human infrastructure break, or glacial lake outburst. The impact of these hydrological phenomena is potentialized by the increase of population density or Human activities in flood-prone areas.

Fluvial floods occur when the water level of a river, stream, or other water body rises and overpasses the bankfull stage, causing it to overflow into the surrounding areas. They occur due to excessive or prolonged rainfall, abnormal inflow from tributaries, snow melt, and other factors.  This type of flood is common along riverine areas, where the overflowing of a nearby river or stream leads to flooding. Fluvial floods typically develop from hours for flash floods (see below) to weeks for the biggest rivers  and can cause more extensive and long-lasting inundation.

Pluvial floods are caused by intense or prolonged rainfall in a specific area, leading to excessive runoff and a rapid increase in water level in flat and low-lying areas, independently of any river stream. They develop quickly, often within hours of heavy rainfall.  The impact can be more severe in areas with poor soil infiltration capacity (e.g., clay soils), impermeabilization of the surface (e.g., urbanization) or steep slopes.

Flash Floods can be caused by various factors but most commonly result from extremely intense rainfall during storms. Flash Floods can also occur due to Dam or Levee Breaks, and/or Mudslides (Debris Flow). The speed and location of flash flooding depend on factors such as rainfall intensity, spatio-temporal distribution, land use and land cover, topography, vegetation type and density, soil type, and soil moisture content. Urban areas are particularly prone to rapid flooding, as rainfall from the same storm can cause more severe and faster flooding in cities than in suburban or rural areas due to impermeable surface.

Flash Flooding occurs rapidly, often catching population off-guard. People may be in dangerous situations if they encounter high, fast-moving water while traveling. If people are at home or work, the water may rise quickly, trapping them or causing property damage before they have a chance to protect it.

Earth Observation (EO) data can help distinguish between different types and sources of floods. By using various satellite technologies, it is possible to analyze and monitor flood events in detail. Some ways in which EO data is utilized:

• Flood Mapping: Satellites equipped with Synthetic Aperture Radar (SAR) or optical sensors can map the extent of floodwaters.
• Water Level: EO data can measure water levels using altimetry data, tracking water levels at virtual stations, aiding in flood assessment.
• Flood Prediction and Monitoring: Integrating EO data with hydrological models improves flood forecasting and monitoring, allowing for better preparedness and response (for more info, see question 11 below).
• Long-term Trends: By analyzing historical satellite data, it is possible to identify patterns and trends in flooding, which can help in understanding the causes and give an indication of the flood hazard in future events.

Although there has been some recent advances in this topic, EO-based flood detection alone cannot differentiate or categorised between flood types; additional data is required for this purpose.  However, due to their persistence and large spatial coverage, fluvial floods are more likely to be detected by Earth Observation data. In contrast, flash floods are less likely to be captured by satellites unless the satellite overpass coincides with the flood event.

Flood hazard models use observed measurements of river flows, rainfall, and/or coastal water levels, combined with topographic data and flow equations, to generate flood hazard information such as extent, depth and velocity.

Fluvial and Pluvial floods are modelled independently as they are governed by different processes and equations. As a result, distinct maps are produced for each flood type. Modelled floods are characterized by specific magnitudes (return periods), which are homogeneous for the entire simulation and area considered.

In the EO4FLOOD project, simulated river flows will be generated using the following hydrological models: MGB (Hydro Matters), DHI-GHM (DHI), HYPE (SMHI), and AI-based model incorporating both pure data-driven and hybrid approaches (DHI). Simulated flood maps will use the following models: Mike+(DHI/DTU), HEC-RAS (RSS-HYDRO), LISFLOOD-FP (SMHI), and DassFlow (Hydro Matters).

Flood hazard maps based on Earth Observation uses satellite imagery data to identify flooded areas at specific moments in time, corresponding to the satellite passes over the area of interest. EO-detected floods can capture flood occurrence and extent during the observation period without distinguishing flood generation mechanisms.  Since this flood detection approach relies on observations of past events, it cannot incorporate information such as flood magnitude (return periods), which may vary across different regions when analysing large areas.

In the EO4FLOOD project, EO-based flood maps will be derived from Copernicus Sentinel-1 or Sentinel-2 sensors and/or NOAA VIIRS sensor. 

Flood forecasting refers to the process of predicting when, where, and how severe a flood might occur. This involves using data from various sources, such as weather forecasts, water levels of the river, and soil moisture, to anticipate flood events and provide early warnings. The goal is to help communities prepare and reduce the impact of floods on lives and property.

Reliable flood forecasts depend on having accurate data and models available early enough to issue timely warnings, typically hours to days before a flood occurs. The time varies depending on the size, shape, and characteristics of the basin – small basins can respond within hours, while larger ones may take days.

Yes, satellites play a crucial role in monitoring key factors like precipitation, soil moisture, and water levels or river discharge. While they may not always provide real-time data, many satellites now offer frequent updates that are near real-time, enabling them to contribute to flood forecasting models. Advances in satellite technology and data processing have significantly improved the speed and accuracy of the information provided.

EO4FLOOD aims to demonstrate the potential of satellite data in advancing flood forecasting capabilities. By integrating satellite observations with hydrological/hydraulic models, EO4FLOOD enhances our ability to monitor rivers, predict floods, and assess their potential impact. This project bridges the gap between cutting-edge satellite technology and practical applications, supporting more reliable and timely flood forecasts to protect lives and livelihoods.

AI, particularly machine learning and deep learning, can analyze vast amounts of meteorological and hydrological data to identify patterns and relationships. This allows AI models to predict variables like water discharge, evapotranspiration, or total water storage, which are critical for anticipating flood events.

AI, particularly hybrid models, enhances flood forecasting by combining the data-processing power of machine learning with the physical laws governing hydrological systems. For example, deep learning models can map satellite-derived meteorological features to hydrological variables such as river discharge. By replacing uncertain parameters in traditional models with AI-driven simulations, hybrid models maintain physical consistency while leveraging the rich information provided by satellite data.

EO4FLOOD will capitalise on the developed products and tools for advancing the scientific understanding of human activities on the distribution, frequency, and intensity of flood, and advancing towards an operational capacity to use EO data for flood forecasting. The main three topics to cover will be:

• Impact of dams in the river regime: devoted to the study on the impact of dam construction and operation on flood distribution, frequency and intensity.
• Land use change: devoted to the study of the changes in land use and their impact on flood exposure.
• Population trends, flood dynamics and impact on flood exposure: devoted to the study of population trends and flood dynamics and their combined impact on flood exposure.

EO4FLOOD will test the impact of EO through calibration, forcing data, initial condition and data assimilation in three rainfall-runoff models (Hype, GHM, MGB) and one AI model. The testing modelling framework will be implemented over selected areas within five specific basins (Torne, Negro, Congo, Niger and Brahmaputra). The dataset of EO will be provided also in the bigger European basins like Po, Danube, Rhine and Ebro. To optimize the use of the available tools for flood forecasting, we will develop a hybrid approach that integrates the strengths of a physics-based approach and advanced AI techniques.

The insights derived from this research will not only refine current methodologies but also serve as a benchmark for future hydrological studies and applications aimed at mitigating flood impacts and enhancing water resource management strategies globally. The key innovations areas are:

• Retrieval of multi-mission water level data, combining nadir, SAR, and SWOT altimeters;
• Utilization of multiple sensors for river discharge estimation, incorporating various water variables like level, width, reflectance indices and slope;
• Regionalization of parameters also for space-based river discharge estimation;
• Hybrid approach merging physical-based hydrological models with AI methods for robust flood forecasting;
• Integration of satellite data into established flood forecasting systems to evaluate their potential impact and benefits;
• Extensive model comparison to optimize application fields based on data and environmental factors;
• Provision of probabilistic flood forecasting results to simulate real-world conditions for stakeholders during early warning phases.