Fathom’s Product Stack

The quality of the terrain data used within a flood model directly affects the overall accuracy of its results.

Our products need a high-resolution, bare-earth digital representation of the Earth’s terrain in order to simulate the flow of water across a 2D grid.

To achieve this, we use a terrain processing methodology to correct and blend digital elevation data, remove surface objects and minimize error or bias.

When simulating water risk globally, how do you characterize the location and properties of every river?

Fathom’s Channel Solver uses computational fluid dynamics to calculate how large a channel needs to be in order to convey a specified flow of water within its banks. The method calculates the appropriate cross sectional area for every point along the river network, solving for depth given that other variables such as width and slope can be observed. This calculation is necessary because the depth of rivers cannot be measured over very large areas using remote sensing techniques.

In order to build physically based simulations of extreme events, we need to understand what the potential magnitude of these events may be.

Traditional statistical techniques for this determine the magnitude and frequency of flood events based on observational records, taken from gauges.

However, for most of the world gauge data does not exist. Fathom’s Flood Frequency Analysis uses data from gauged regions to predict the behavior of extreme events in regions where no gauge data exists.

In order to anticipate how river flows may respond to climate change, to produce flood forecasts, and to understand how individual catchments relate to their neighbors during extreme events such as tropical cyclones, we need dynamical models computing how much water flows in rivers worldwide.

These models are called hydrological models. They simulate water exchanges between the atmosphere and the land surface, how water is stored and released by catchments, for instance through snow accumulation and melt, and how it travels through the river network.

To represent risk today and into the future, Global Climate Model (GCM) Data is needed.

However, in their raw form, current GCMs are too coarse to directly drive high resolution flood models.

In order to understand how risk today is different to the past, and how it may further change in the future, Fathom leverages data from a number of carefully selected GCM experiments.

Using a range of transparent statistical techniques, our team translates climate model outputs into relevant modifiers that can be applied to our fine-scale flood models at various points in the modeling chain.

Flood hazard data alone does not tell you everything about flood risk: for this you need to study the interaction of flooding with people, buildings, and businesses.

Therefore, once flood hazard data has been collated, the next step is to intersect it with data and methods that characterize global exposure and vulnerability. Fathom uses proprietary data and methods to simulate this translation from hazard to risk. These are then used to produce bespoke risk layers and to complete the calculation chain within our high-resolution OASIS-enabled catastrophe models.

The final layer of the Method Stack is the Model Builder. This is where we bring all of the methods together and automate their delivery.

Automation is key to scaling traditional techniques for flood modeling and enabling efficient large-scale modeling.

Bare earth terrain data has applications that reach far beyond flood modeling. Whether it is gaming, mining or flood mapping, terrain models form an integral foundation for the formation of a digital-twin world. 

Working in collaboration with academic partners, Fathom has created the world’s most accurate bare-earth map of the world’s terrain, ranging from ~10 – 30m resolution.

This is achieved by cleansing the COP GLO-30 DEM and enriching it with LiDAR data where it is available. Fathom has now incorporated LiDAR across the planet. The end product is Fathom’s Global Terrain Data – FABDEM+.

Fathom combines MERIT-DEM with additional water body data from sources like OpenStreetMap and Landsat satellite imagery to create a global river network.

The dataset, known as MERIT-Hydro, is considered the best global hydrography dataset for flood modeling. It accurately identifies main rivers and tributaries, and defines the upstream catchment area of every grid cell worldwide.

Estimating river flows for the entire globe is challenging. In many parts of the world, the data simply does not exist. Where there is data, flow records are generally limited with short record durations for most rivers.

To overcome this, Fathom has developed an approach for estimating extreme river flows globally. The method pools and transfers information from data-rich regions to data-poor regions for flood estimation in ungauged catchments.

This is achieved through machine learning techniques that have been trained using over 10,000 global gauging stations.

Intense, short-duration pluvial events pose significant risk to many communities globally. However, in many parts of the world we do not have accurate local observations of extreme precipitation. This poses a challenge when attempting to simulate the behavior of pluvial events within a flood model.

To overcome limitations in globally available data, Fathom has established a machine learning process that links the characteristics of the locally observed sub-daily precipitation data to a suite of globally available parameters.

When scaled to the global extent, the result are global Intensity-Duration-Frequency (IDF) curves, ‘extreme rainfall’, that provide the foundation of our pluvial flood layer.