A crowdsourced global data set for validating built-up surface layers.

Area of Science:

• Earth Observation
• Geographic Information Systems (GIS)
• Remote Sensing

Background:

• Recent advancements in satellite technology have led to numerous global high-resolution built-up surface products.
• These products rely on open-source data from satellites like Landsat and Sentinel.
• A critical need exists for validation datasets that are independent of the product creators.

Purpose of the Study:

• To develop an independent validation sample set for global built-up surface products.
• To provide a reliable resource for assessing the accuracy of existing and future urban area mapping products.
• To facilitate the inter-comparison of different built-up area datasets.

Main Methods:

• A stratified sampling approach was used to create a validation sample set of 50,000 locations.
• A crowdsourcing campaign on Geo-Wiki was employed for visual interpretation of built-up surfaces.
• Very high-resolution satellite imagery was used as reference data for labeling, with a minimum of five validations per location.
• Data were collected at 10m sub-pixel resolution within an 80x80m grid to account for geo-registration errors and enable various validation modes.

Main Results:

• A comprehensive validation dataset of 50,000 locations has been generated.
• The dataset is suitable for validating and comparing multiple global built-up surface products.
• The crowdsourced approach ensured a robust and independent validation effort.

Conclusions:

• The developed dataset addresses the need for independent validation of global built-up surface products.
• This resource will enhance the reliability and comparability of urban area mapping initiatives.
• The methodology provides a framework for creating similar validation datasets in the future.