Climate change exposure estimates for the UK at 1 km resolution, 1901-2080

For historical and contemporary climate periods, we considered the early 20th Century (1901–1930), mid-late 20th Century (1961–1990), and the last decade (2010–2020). For each period, HadUK-Grid data on monthly precipitation and maximum, minimum and mean temperatures were downloaded at 1 x 1 km resolution; averages for the 1961–1990 period were available, since this is a widely used climate “normal” period; for the other periods annual data were averaged together to create long-term mean datasets. UKCP18 Local data for the coming decades (2021–2040) and later 21st Century (2061–2080) were downloaded at 5 x 5 km resolution and resampled to 1km resolution using bilinear interpolation to match the HadUK-Grid data.

For each time period, an extended set of 37 bioclimatic variables were produced from the monthly climate data. These variables were scaled and their dimensionality reduced using a Principal Components Analysis. The first two principal components (PCs), generally reflecting gradients from warmer and drier to cooler and wetter conditions (PC1, 58.25% of total variance) and from wetter (especially in winter) to more temperature-stable conditions (PC2, 21.43% of total variance), were used as a summary of the resulting multivariate climatic “space”. The Euclidean distance between pixels’ (i.e. 1 km sites’) positions in PC1–PC2 space (weighted by the PCs’ eigenvalues) in adjacent time points was calculated and used as an index of a locations’ climate change exposure through time; this approach was used for all four of the time period comparisons deposited here.