Calibration in flood models.:

Flood models simulate water flow behavior in specific areas, and their accuracy largely depends on key parameters such as:

• Manning’s roughness coefficient (Manning’s n): controls flow velocity based on surface roughness.


• Soil infiltration rate: determines how much water penetrates the soil, reducing runoff.

In the system, global calibration is performed to adjust these parameters based on observed global datasets. This process is essential to ensure the model adapts correctly to different regions and basin characteristics, especially when local data is limited. Below is an overview of how calibration is carried out:

1. Basin Selection

HydroBasins data is used, which provides vectorized layers of sub-basin boundaries at a global scale. Three sub-basin levels are considered:

• Level 12: Small basins


• Level 08: Medium basins


• Level 04: Large basins

A Python script randomly selects basins within these levels to run flood simulations and calibrate model parameters.

2. Design Storm Simulation

The model is run using 10-year return period design storms generated from ERA5 reanalysis data. These storms serve as the basis for simulating flood depth and peak discharge in the selected basins.

3. Observation Database

An internal database of flood extent and observed discharge is used, sourced from:

• GloFAS (Global Flood Awareness System)


• Global Surface Water Explorer

These observed datasets allow the model to be adjusted so that flood simulations match real-world extent and discharge data.

4. Brute Force Optimization Method

A brute force optimization approach is applied to find the best multipliers (50%-150%) for Manning’s n and infiltration, minimizing the error between simulations and observations. This method evaluates the model across a grid of parameter values, exhaustively searching for the configuration that achieves the lowest error.

5. Bias Correction

GloFAS discharge data is bias-corrected using GRDC (Global Runoff Data Centre) datasets to improve accuracy and reduce discrepancies between reanalysis models and actual discharge observations.

Once infiltration and Manning’s n parameters are adjusted for all selected basins, these values are interpolated using an Inverse Distance Weighted (IDW) approach, creating three global raster layers for calibration parameters corresponding to each basin size.

Application of Manning’s n Multipliers

Manning’s n multipliers are used to modify surface roughness and adjust flood extent. These multipliers allow the model to better reflect real terrain conditions:

• Higher multipliers (e.g., 1.5): indicate rougher terrain than default, slowing water flow and increasing water accumulation and flood extent in flatter areas.


• Lower multipliers (e.g., 0.5): indicate smoother terrain, allowing water to flow faster and reducing flood extent.

Go to Wiki