The world's rivers are covered over and fragmented by road infrastructure. Road–river infrastructure result in many socio-environmental questions and documenting where different types occur is challenged by their sheer numbers. Equally, the United Nations has committed the next decade to ecosystem restoration, and decision makers across government, non-government, and private sectors require information about where different types of road–river infrastructure occur to guide management decisions that promote both transport and river system resilience. Field-based efforts alone cannot address data and information needs at relevant scales, such as across river basins, nations, or regions to guide road–river infrastructure remediation. As a first step towards overcoming these data needs in Great Britain, we constructed a georeferenced database of road–river infrastructure, validated a subset of locations, and used a boosted regression tree model-based approach with environmental data to predict which infrastructure are bridges and culverts. We mapped 110 406 possible road–river infrastructure locations and were able to either validate or predict which of 110 194 locations were bridges (n = 60 385) or culverts (n = 49 809). Upstream drainage area had the greatest contribution to determining infrastructure type: when <10 km2 our model correctly predicted culverts 73% of the time but only 60% of the time for bridges. Road type and stream gradient also influenced model results. Our model-based approach is readily applied to other locations and contexts and can be used to inform decisions about management of smaller infrastructure that are frequently overlooked worldwide.
|Number of pages||9|
|Journal||Environmental Research: Infrastructure and Sustainability|
|Early online date||28 May 2021|
|Publication status||Published - 1 Jul 2021|
- boosted regression trees
- great britain