Improved prediction of hiking speeds using a data driven approach

Area of Science:

• Geospatial analysis
• Human movement science
• Algorithm development

Background:

• Established walking speed models primarily use walking slope.
• Fell-running research suggests hill slope and terrain obstruction are important factors.
• Advancements in GPS tracking enable large-scale analysis of these variables.

Purpose of the Study:

• To develop and validate a more accurate walking speed prediction model.
• To investigate the statistical significance of hill slope and terrain obstruction.
• To compare a new model against existing walking speed algorithms.

Main Methods:

• Utilized extensive public GPS data (88,000 km) from UK walking/hiking tracks.
• Filtered GPS data to ensure accuracy and relevance.
• Developed a new generalised linear model (GLM) incorporating hill slope, terrain type, and obstruction.

Main Results:

• The GLM demonstrated higher accuracy (lower root-mean-square-error) than established models.
• Hill slope, terrain type, and terrain obstruction were found to be highly significant predictors.
• Model error increased with steeper hill slopes, highlighting the importance of this variable.

Conclusions:

• The new GLM provides a more accurate prediction of walking speed.
• Incorporating hill slope and terrain characteristics is crucial for off-road travel models.
• This research offers improved tools for route planning and understanding pedestrian movement in varied terrain.