Elevation & Statistics

Computed statistics

Category	Statistics
Distance	Total 2D distance (Haversine), 3D distance (with slope)
Elevation	D+ / D- (3 algorithms available), max altitude, min altitude
Time	Total duration, moving time, stopped time, start date, end date
Speed	Average speed, average moving speed, max speed
Pace	Average pace (min/km), average moving pace
Stops	Number of stops, total duration, longest stop, average duration
Biometrics	Heart rate (avg/max/min, HR zones with --max-hr), power (avg/max, normalized power), cadence (avg/max), temperature (avg/min/max)
Metadata	Number of points, number of segments, point density per km

Elevation correction

Raw GPS elevations are often very noisy (10 to 50 meter errors are common). This artificially inflates D+ and D-. The tool offers two correction mechanisms that can be combined.

Software smoothing (--smoothing)

Two-pass filter applied to elevation data before any computation:

1. Median filter — removes isolated spikes (an outlier point is replaced by the median value of its neighbors)
2. Moving average — smooths remaining high-frequency noise

Preset	Median window	Average window	Recommended use
none	(disabled)	(disabled)	Already clean data or debugging
light	3 points	3 points	Good quality GPS (recent Garmin)
medium	5 points	5 points	General use (default)
heavy	7 points	11 points	Very noisy GPS (watch, phone)

DEM/SRTM correction

Replaces GPS elevations with those from a digital elevation model (NASA SRTM). This is the most accurate method.

Automatic download (default)

By default, missing SRTM tiles are automatically downloaded from the AWS Elevation Tiles service (SRTM1, 30m resolution when available). Tiles are cached locally:

• Windows: %LOCALAPPDATA%\gpx-utility-analyzer\srtm\
• macOS: ~/Library/Caches/gpx-utility-analyzer/srtm/
• Linux: ~/.cache/gpx-utility-analyzer/srtm/

# Works out of the box, tiles are downloaded on the fly
gpx-analyzer analyze my-hike.gpx

To disable automatic download:

gpx-analyzer analyze my-hike.gpx --dem-auto-download=false

To change the cache directory:

gpx-analyzer analyze my-hike.gpx --dem-cache /path/to/cache

Local tiles (--dem-dir)

To use SRTM1 tiles (30m, more accurate) or work offline:

1. Download SRTM tiles covering your track from NASA Earthdata or CGIAR-CSI
2. Place the .hgt files in a directory (e.g., ./srtm/)
3. Pass --dem-dir ./srtm/

gpx-analyzer analyze pct.gpx --dem-dir ./srtm-tiles/

Files use the standard HGT format (SRTM1 at 30m or SRTM3 at 90m resolution). Naming follows the convention N48W003.hgt (coordinates of the tile's southwest corner).

When --dem-dir is provided with --dem-auto-download (default), local tiles take priority. If a tile is missing locally, it is downloaded to the cache. If the download fails, the GPS elevation is kept with a warning.

Limitations

• Automatic download requires an internet connection
• The AWS service provides SRTM1 tiles (30m) between 60°N and 56°S, and SRTM3 (90m) elsewhere

Example: impact on a 4000+ km track (Karel Sabbe's PCT)

Configuration	D+	Max altitude
--smoothing none	599 323 m	7 583 m
--smoothing medium (default)	226 908 m	5 720 m
--smoothing heavy	155 015 m	5 645 m
DEM + --smoothing medium + 5m threshold	~126 000 m	~4 001 m
DEM + --elevation-algo segments	~104 000 m	~4 001 m

The actual D+ of the PCT is approximately 96 000 m. The segments algorithm combined with DEM gives the closest result.

Elevation gain algorithms (--elevation-algo)

Three algorithms are available for computing D+ and D-. They are applied after elevation smoothing (--smoothing) and DEM correction.

threshold (default)

Accumulates D+/D- only when the elevation change since the last reference point exceeds the threshold (--elevation-threshold). Simple and effective for filtering GPS noise.

gpx-analyzer analyze trace.gpx --elevation-algo threshold --elevation-threshold 3

douglas-peucker

Simplifies the elevation profile (cumulative distance, altitude) using the Douglas-Peucker algorithm, then computes D+/D- on the retained points. The epsilon (--dp-epsilon) controls the maximum tolerated vertical deviation in meters.

gpx-analyzer analyze trace.gpx --elevation-algo douglas-peucker --dp-epsilon 3

Works well on GPS data without DEM. With DEM, the terrain profile retains many legitimate micro-variations, which limits the filter's effectiveness.

segments

Divides the profile into quasi-constant slope segments using greedy linear regression. D+/D- is computed from the fitted elevations at segment endpoints.

gpx-analyzer analyze trace.gpx --elevation-algo segments --seg-min-length 200 --seg-max-deviation 2

Parameter	Description	Default
--seg-min-length	Minimum horizontal segment length (meters)	200.0
--seg-max-deviation	Maximum RMS residual before splitting a segment (meters)	2.0

This is the most effective algorithm with DEM data: it absorbs SRTM grid noise and produces results close to actual terrain.

GPS track smoothing (--track-smoothing)

Applies a moving average to lat/lon coordinates before DEM correction. Reduces horizontal GPS noise that causes artificial altitude oscillations when points oscillate between different DEM cells.

Preset	Window	Use
none	(disabled)	Default, no lat/lon smoothing
light	3 points	Good quality GPS
medium	5 points	Standard GPS
heavy	9 points	Very noisy GPS

gpx-analyzer analyze trace.gpx --track-smoothing medium --elevation-algo douglas-peucker

Note: lat/lon smoothing modifies the coordinates used for distance calculation and stop detection. Total distance will be slightly reduced (horizontal noise is filtered out).

Full pipeline

The processing order is:

Track smoothing (lat/lon) → DEM correction → Elevation smoothing (--smoothing) → Distance calculation → Elevation gain algorithm

Stop detection presets

Preset	Max speed	Min duration	Use
hiking	0.2 m/s (0.7 km/h)	3 min	Hiking
walking	0.2 m/s (0.7 km/h)	3 min	Urban walking, casual walks
trail	0.3 m/s (1.1 km/h)	2 min	Trail running, mountain running
running	0.5 m/s (1.8 km/h)	5 min	Road running, track
cycling	1.0 m/s (3.6 km/h)	30 sec	Cycling, mountain biking
swimming	0.15 m/s (0.5 km/h)	2 min	Open water, pool swimming

A stop is detected when the computed speed (distance between points / elapsed time) remains below the threshold for at least the minimum duration. Thresholds can be customized with --stop-speed and --stop-duration.