Geomorphic Flood Index (GFI) version 2.0

Overview

The Geomorphic Flood Index (GFI) 2.0 is a MATLAB-based toolbox designed for the rapid and reliable delineation of flood-prone areas, particularly in data-scarce environments. 

Building upon the original GFI method, version 2.0 introduces a Hierarchical River-First Update (River-to-Confluence mapping). This advancement accounts for downstream backwater effects and confluence influences, improving spatial accuracy and water depth consistency by propagating downstream flood depths to upstream segments.

The example present in the .ZIP is using a DEM of 5m. Due to the high resolution, the time consuming should be considered based on the computer capacity, specially in when PARFOR is used.

QGIS version: gfi_toolbox

Theoretical Background

The model is based on the calculation of the Geomorphic Flood Index:
$$GFI = \ln\left(\frac{h_r}{H}\right)$$

Where:
* $H$ is the elevation difference between the hillslope pixel and its nearest hydrologically connected river pixel.
* $h_r$ is the potential water level in the river, estimated using hydraulic scaling:
    $$h_r = a \cdot A_r^n$$
* $A_r$ is the contributing area (drainage area) of the river.
* $n$ is a predefined scaling exponent (set to $0.354429752$).
* $a$ is a coefficient calibrated against an observed flood map.

To define the flood extent, a binary classification is performed. A threshold ($\tau$) is calibrated against an observed flood map (e.g., satellite imagery or 2D hydraulic models) by performing a Receiver Operating Characteristic (ROC) analysis and finding the optimal $\tau$ that maximizes the classification performance (AUC).

Once the optimal threshold ($\tau$) is identified, the scaling coefficient $a$ is mathematically derived from it:
$$a = e^{-\tau}$$

With $a$ defined, the final potential water level in the river ($h_r$) is established:
$$h_r = a \cdot A_r^n$$

Finally, the Water Depth ($WD$) for each pixel is calculated as the difference between the potential water level in the river ($h_r$) and the relative elevation of the hillslope ($H$). To ensure physical consistency (i.e., avoiding negative water depths for dry areas), it is defined as:
$$WD = \max(0, h_r - H)$$

• If $h_r > H$, the pixel is considered flooded with an estimated depth equal to $h_r - H$.
• If $h_r \le H$, the pixel is outside the flood extent and remains dry ($WD = 0$).

Prerequisites

To run this toolbox, you need:

• MATLAB (Tested on R2024a or newer) 
• [TopoToolbox v3](https://topotoolbox.wordpress.com/): Mandatory for DEM conditioning and flow routing.
• gradient8.m: Available on the MATLAB Central File Exchange.
• Parallel Computing Toolbox (Optional but highly recommended to utilize `parfor` loops for faster execution).

Installation

1. Clone or download this repository.
2. Install TopoToolbox and add it to your MATLAB path.
3. Ensure `gradient8.m`, `traceFlow.m`, and `ROCcurve_maggiore.m` are in the same directory as the main script.
4. Place your input raster files in the working directory.

Usage / Quick Start

1. Open `GFI_v2_main.m`.
2. Update the filenames in Section 1, Section 5.1, and Section 7 to match your data. The default example uses:
   •  `Bradano.tif` (Digital Elevation Model)
   • `Conditioned DEM or Filled DEM, Flow Accumulation (D8, 1-128), Flow Direction and Slope (optional if you already have those files)
   • `Flood T = 200.tif` (Observed flood hazard map for calibration)
   • `Tiranti T = 200.tif` (Observed water depths for validation)
3. Run the script. The toolbox will automatically:   
   • Condition the DEM and define the drai  
   • Map hillslopes to nearest channels and channels to downstream confluences.
   • Calibrate the coefficient $a$ via ROC curve analysis.
   • Apply the v2.0 confluence-aware network up
   • Generate visual performance metrics, ROC curves, and error maps.

Project Structure

• `GFI_v2_main.m` - The main executable script.
• `traceFlow.m` - Custom function to trace flow paths downslope.
• `ROCcurve_maggiore.m` - Custom function for binary classification calibration.
• `Example/` - (Optional) Folder containing example `.tif` datasets.

Authors and Creators

• Jorge Saavedra Navarro (nandres049@gmail.com) - Main Developer of version 2.0
• Cinzia Albertini - Developer of Coastal version (https://doi.org/10.1111/jfr3)
• Caterina Samela - Developer of GFI and Water Depth calculation
• Salvatore Manfreda - Creator and first developer of GFI (https://github.com/HydroLAB-UNIBAS/GFA-Geomorphic-Flood-Area)

Citation

if you use this toolbox in your research, please cite the associated DOI (via Zenodo) and the foundational literature.

License

This project is licensed under the GNU General Public License v3.0 - see the `LICENSE` file for details. This ensures that the toolbox remains free and open-source for the academic and scientific community.