Metadata for Assessing the speed impact of urban traffic calming devices using functional data analysis

Metadata is for the analyses presented in the article “Assessing the speed impact of urban traffic calming devices using functional data analysis” are based on floating car data (FCD) derived from GPS trajectories of vehicles. These data provide continuous information on vehicle position and speed along road sections and allow the construction of detailed speed profiles as a function of distance. Two case studies were analysed. The first dataset originates from an urban through-road in Sardón de Duero, Spain, where a road marking treatment (“dragon’s teeth”) was implemented at the entrance to the municipality. The analysed road segment was trimmed to 2,200 m. The data covered the period from January 2020 to October 2021, with before–after comparisons based on the January–October periods of 2020 and 2021. After data cleaning, the sample included 7,993 individual drives from 7,249 unique vehicles, specifically 1,374 drives in the before period and 6,619 drives in the after period. By direction, the numbers were 628 vs. 3,243 drives in direction 1 and 746 vs. 3,376 drives in direction 2. GPS-based speed profiles were used to compare driver speeds before and after the implementation of the treatment. The second dataset consists of Czech urban road locations where dynamic speed feedback signs were installed. The original Czech dataset covered 43 locations installed between November 2014 and May 2015, but due to limited FCD penetration and sample size, only 8 locations were suitable for the functional-data-based analysis. These locations were Staňkov, Kolinec, Červené Poříčí, Horšovský Týn, Janovice nad Úhlavou, Mochtín 1, Mochtín 2, Domažlice, and Planá. Across these 8 locations and 16 direction-specific trajectories, the cleaned dataset contained 535 drives in the before period and 326 drives in the after period; one direction in Domažlice had no usable post-measure observations. The Czech data covered approximately October 2014 to May 2015. In the article, two representative Czech examples were presented in detail: Planá (direction 1) and Červené Poříčí (direction 2). In both studies, the floating car data were processed to obtain speed as a function of distance along the road. The preprocessing included assigning unique drive IDs, defining road centrelines and cumulative distance, determining driving direction, localising traffic calming devices, and trimming trajectories to a common analysed segment. The resulting speed profiles were subsequently analysed using functional data analysis, based on cubic B-spline smoothing with 100 basis functions and a smoothing parameter of λ = 10^5. Statistical inference was performed using interval-wise testing at 1-metre resolution with 10,000 permutations to detect spatially varying differences between the pre- and post-intervention periods. Due to licensing and data protection restrictions associated with the original floating car data providers, the underlying raw datasets cannot be publicly shared.