Field-based modelling of segment speed in trail running using GPS and course characteristics

Trail running performance is characterised by large fluctuations in speed due to changes in slope, terrain, and race progression, which are not well captured by models derived from road running. This study aimed to develop a field-based model of segment speed in trail running using GPS data and course characteristics. Data were collected from 517 runners competing in 20 short-distance trail races (10–35.6 km) in Belgium, primarily involving recreational participants and low-to-moderate technical terrain. Race courses were segmented using the Trail Segment Analysis Tool (TSAT), with segment boundaries and technical difficulty determined through on-site reconnaissance. GPS data were used to quantify segment length, mean slope, and participant speed. Exploratory analyses revealed a strong non-linear relationship between segment speed and slope, with weaker associations for technical difficulty and race position. A multiple regression model incorporating these variables and average trail speed explained 81% of the variance in segment speed. Maximal speeds occurred on moderate downhills (~–8%), with marked reductions on steeper gradients. Within this specific race context, these findings show that accessible field-derived data can support robust segment-level performance modelling and provide practical insights for pacing analysis and race planning in trail running.