Global Adaptation Layer - Methodology

How the Global Adaptation Layer is built from raw geospatial features to hazard-specific adaptation scores

Methodology

The Global Adaptation Layer is a globally consistent, high-resolution climate adaptation dataset engineered from a diverse set of geospatial data sources that provide information on the built environment at up to 30-meter resolution and are calibrated to each type of climate hazard.

The overall methodology can be broken down into four steps, as shown in the flowchart below. The process is designed to be iterative, allowing insights generated at each stage to continuously inform and refine decisions across the wider workflow.

The diagram below illustrates in greater technical detail how spatial attributes are collected from raw data sources and mapped to thematic features before being transformed into adaptation scores at each location on Earth.

Define adaptation themes from domain expertise

We have conducted extensive research on best practices in mitigation and adaptation for each hazard. We then define the thematic areas that can be quantified to explain the degree of adaptation.

For example, water treatment facilities — including filtration systems, wastewater treatment plants, and water storage solutions — play a critical role in managing dry spells. During prolonged periods of drought, a location’s proximity to these facilities and its access to their services become key indicators of drought resilience. Thus, we define the thematic feature group ‘Availability & Capacity of Water Works’ to indicate whether a location falls within the service area of water treatment facilities.

The table below shows example thematic feature groups that we define for drought, wildfire, inland flooding, and coastal flooding, along with relevant data sources.

Hazard	Thematic feature groups	Example spatial features
Drought	water_storage, water_amenity, water_well, water_works	reservoir, water tower, water treatment plant, drinking water amenity, water well, etc.
Wildfire	fire_detection, fire_prevention, fire_response	fire hydrant, fire station, fire lookout, emergency landing site, etc.
Inland flooding	storage_and_control, direct_barrier, drainage, nature_based_solution	dam, reservoir, flood wall, drain, wetland, etc.
Coastal flooding	storage_and_control, coastal_defense, drainage, natural_buffer	seawall, retention basin, drain, mangrove, etc.

Collect required spatial attributes

After defining the features to be engineered, we identify suitable data sources to construct each proposed feature. The selected data sources must meet at least one of the three criteria below:

• Peer-reviewed: the data is published in academic journals that underwent a peer-review process. For example, local climate zone data from World Urban Database and Access Portal Tools (WUDAPT).

• Well-documented: the data is published with clear explanations of methodologies and/or code repositories for third parties to reproduce the results. For example, OpenStreetMap.

• Well-established: the data is published by reputable organizations and is widely used by the industry. For example, 10m Global Land Cover by Impact Observatory.

In the case of the features defined above, OpenStreetMap provides a global map of geospatial features containing water works, wastewater treatment plants, water access points, and other related vector features. Additionally, data from the Global Dam Watch Database (GDW) and Global Georeferenced Database of Dams (GOODD) provide the locations, capacity, and outlines of freshwater reservoirs and catchment areas, which complement the data from OpenStreetMap.

The raw spatial attributes are then categorized into different thematic feature groups to reflect the diverse aspects of adaptation strategies.

The maps below show examples of different adaptation feature groups for drought, coastal flooding, inland flooding, and wildfire in Santa Cruz, California, USA.

Construct thematic features using geospatial operations

In most cases, raw source data alone is insufficient to explain the defined adaptation themes. The raw data must be cleaned, processed, and engineered into the desired thematic features for better explainability and compatibility with other data layers. While each data source and adaptation theme requires a unique processing pipeline, the general workflow is as follows:

1. Clean and fix source data anomalies at native resolution.

2. Enhance the resolution or complete missing areas of the source data if necessary.

3. Convert raw data into structured data with spatial indexes.

4. Apply spatial and statistical operations to obtain the targeted thematic feature.

5. Rescale the thematic feature to a scale of 0-100, with 100 indicating the highest resilience in each theme. This can be achieved using min-max normalization or percentile ranking depending on the distribution of the thematic feature.

Using the drought resilience example, vector data from OpenStreetMap is first cleaned and buffered as a proxy for functional service areas and to account for spatial uncertainty, and then aggregated into a unified spatial index. Similarly, data from reservoir datasets is processed by buffering reservoir polygons with a corresponding radius. By overlaying these layers, we can identify the number of services or adaptation features available at each location, which is then converted into an adaptation score that captures the cumulative contribution of diverse infrastructure types at that location. A location that falls within the service radius of all Water Works, Water Storage, Water Amenity, and Water Well categories would have access to diverse water storage, processing, and access infrastructure, benefit from high water capacity, and thus score 100, indicating the highest resilience for drought.

For inland flooding adaptation, reservoir storage density is a useful first-order proxy for how much a catchment can buffer excess rainfall. Higher storage density implies more potential to absorb/attenuate runoff, while lower storage density implies faster translation of rainfall into river flow (higher peaks).

We use the watersheds data from GOODD and reservoir capacity data from GDW to calculate the storage density for each watershed. The result is spatially joined to our H3 spatial index grid to provide additional information or context for each cell's adaptation against flood.

Aggregate adaptation scores for each hazard

Finally, all adaptation features for a single hazard are aggregated into one distribution, which is then used to adjust the intensity of the hazard. A location that scores 100 in all constituent themes enjoys the highest possible offset for the relevant hazard. This process is applied to each climate hazard to obtain the Resilience-adjusted Risk using the methods described in our Resilience-adjusted Risk Framework.