My recent research focuses on the integration of remote sensing, geomorphology, and data-driven methods to improve the detection, mapping, and understanding of flood dynamics.

This work combines:

  • Satellite observations (SAR and multispectral)
  • Hydrogeomorphic descriptors derived from DEMs
  • Probabilistic and machine learning approaches

The goal is to develop robust, scalable, and physically meaningful tools for flood monitoring and risk assessment across heterogeneous environments.

1. Multi-Source Flood Detection and Data Fusion

Flood detection is inherently complex due to the interaction of land cover, topography, and hydrological conditions. To address this, my research has contributed to the development of data fusion approaches integrating multiple sources of information.

A key contribution is the use of Bayesian Networks to combine:

  • Multitemporal SAR imagery
  • Interferometric coherence
  • Geomorphic and ancillary data

This probabilistic framework allows modeling dependencies among variables and improves flood detection accuracy by reducing false alarms and missed detections (D’Addabbo et al., 2016).  

The results demonstrate that:

  • Multi-source integration significantly outperforms single-sensor approaches
  • Probabilistic models can effectively represent uncertainty and complex interactions

2. Multispectral Remote Sensing for Water and Flood Mapping

Multispectral satellite imagery represents a fast and cost-effective tool for monitoring surface water and floods at large scales.

Research in this area has:

  • Reviewed and analyzed spectral indices such as NDWI, MNDWI, and NDVI
  • Assessed their performance across different land cover types
  • Identified strengths and limitations of optical data

Findings show that:

  • Multispectral data enable rapid delineation of flooded areas
  • Performance depends strongly on land cover, turbidity, and atmospheric conditions (Albertini et al., 2022).  

This work provides guidance for selecting appropriate indices depending on environmental conditions and application needs.

3. Integration of Spectral and Hydrogeomorphic Information

A major advancement in this research is the integration of satellite-derived information with hydrogeomorphic descriptors.

In particular:

  • Spectral indices from Sentinel-2 imagery are combined with DEM-based indicators
  • The Geomorphic Flood Index (GFI) is used to identify flood-prone areas

This integrated approach:

  • Improves flood detection accuracy
  • Reduces false positives caused by spectral ambiguities
  • Provides physically consistent flood maps

Applications demonstrate that combining spectral and geomorphic layers significantly enhances mapping performance compared to single-source approaches (Samela et al., 2022).  

4. Machine Learning for Flood Mapping

Recent work explores the use of machine learning algorithms to exploit the growing availability of Earth Observation data.

A particular focus has been on the Random Forest (RF) classifier, applied to multi-source datasets including:

  • Sentinel-1 SAR data
  • Sentinel-2 multispectral imagery
  • DEM-derived geomorphic features

Key contributions include:

  • Evaluation of predictor importance and robustness
  • Analysis of model performance under varying training conditions
  • Identification of stable features for reliable flood mapping

Results show that:

  • RF provides high accuracy and robustness even with complex datasets
  • The integration of satellite and geomorphic variables significantly improves classification performance (Albertini et al., 2024).  

5. Toward Operational Flood Monitoring

The integration of remote sensing and modeling approaches supports the development of operational tools for flood monitoring.

Key advantages of the proposed methodologies:

  • Use of freely available satellite data (e.g., Sentinel missions)
  • Applicability over large spatial scales
  • Capability for near real-time assessment

These approaches are particularly valuable for:

  • Emergency response
  • Damage assessment
  • Risk management and planning

They provide a scalable framework for global flood monitoring under changing climate conditions.

Concluding Remarks

This research advances flood science by integrating:

  • Remote sensing (SAR and optical)
  • Hydrogeomorphic analysis
  • Probabilistic and machine learning models

The resulting framework improves the accuracy, robustness, and interpretability of flood mapping, offering practical tools for hydrology, environmental management, and disaster risk reduction.

References

  • D’Addabbo, A., Refice, A., Pasquariello, G., Lovergine, F. P., Capolongo, D., & Manfreda, S. (2016). A Bayesian Network for Flood Detection Combining SAR Imagery and Ancillary Data. IEEE Transactions on Geoscience and Remote Sensing.  
  • Albertini, C., Gioia, A., Iacobellis, V., & Manfreda, S. (2022). Detection of Surface Water and Floods with Multispectral Satellites. Remote Sensing.  
  • Samela, C., Coluzzi, R., Imbrenda, V., Manfreda, S., & Lanfredi, M. (2022). Satellite flood detection integrating hydrogeomorphic and spectral indices. GIScience & Remote Sensing.  
  • Albertini, C., Gioia, A., Iacobellis, V., Petropoulos, G. P., & Manfreda, S. (2024). Assessing multi-source random forest classification and robustness of predictor variables in flooded areas mapping. Remote Sensing Applications: Society and Environment.  

About the author

He is Full Professor of Hydrology and Hydraulic Constructions at the University of Naples Federico II. He is currently chair of the IAHS MOXXI working group. His research primarily centers on hydrological modeling and monitoring. Recognizing the challenges posed by the complexity and limitations of traditional hydrological observations, he actively explores advanced and alternative monitoring techniques, such as the utilization of Unmanned Aerial Systems (UAS) coupled with image processing.