Reproduction R2: HJA TV081 LiDAR stem map (2008), trees >= 10 m, benchmarked with NativeGIS
About this report
Every number in the text is followed by a bracketed reference, such as [3]. It points to the tool call, in the step list at the end, that produced the number, with that call's arguments, input files and output files. Nothing here is typed in by hand. Numbers in the validation tables are labelled V and come from the validation documents, not from this run.
Summary
ot_HJ_Andrews_bin_100.laz: 5,493,081[3] points, 7.4 first returns/m²[4], ground from the delivered class[6], canopy to 83.1 m[10], 12,812[15] trees (227.8 per ha[15]).
1. Data set: ot_HJ_Andrews_bin_100.laz
1.1 Input
| File | ot_HJ_Andrews_bin_100.laz, 25.8 MB[3], sha256 bf81fa07ed8a… (the file) |
|---|---|
| Format | laz, LAS 1.0 point format 1[3] |
| Coordinate system | EPSG:26910 NAD83 / UTM zone 10N [m] + EPSG:5703 NAVD88 height [m][3]; horizontal units m, vertical m |
| Points loaded | 5,493,081[3] of 5,493,081[3] scanned; reprojected on open to EPSG:26910[3] |
1.2 Point cloud quality
The cloud holds 5,493,081[4] points over 56.25 ha[4]: 9.77 points/m²[4] in all, 7.41 first returns/m²[4] and 0.35 ground points/m²[4] in the delivered ground class. Cover of 1 m cells is 99.7 %[4] and of 5 m cells 100.0 %[4].
| Class | Name | Points |
|---|---|---|
| 1 | unclassified | 5,296,438[4] |
| 2 | ground | 196,643[4] |
- Delivered ground (class 2) is sparse: 0.35 points/m2. Re-classify ground (gis_lidar_ground mode "reclassify" or "auto").[4]
1.3 Ground
Ground was taken from the ground class delivered with the file. delivered class 2 density 0.35/m2 >= 0.25[6]. 196,643[6] of 5,493,081[6] points are ground, 0.35 per m²[6].
1.4 Terrain model (DTM)
The terrain model is a triangulated (TIN)[8] surface of 196,643[8] ground points on a 1.00 m[8] grid of 750[8] by 750[8] cells. 97[8] cells had no ground nearby and were filled from their neighbours. Elevations run from 751.92 m[8] to 887.65 m[8], mean 815.37 m[8].
1.5 Canopy height model (CHM)
The pit-free canopy height model stacks 18[10] triangulated layers of the 4,170,507[10] first returns, with triangle edges above 1.50 m[10] dropped, on a 0.50 m[10] grid of 1,500[10] by 1,500[10] cells. Canopy height reaches 83.1 m[10]; the mean over all cells is 20.3 m[10]. Method: Khosravipour et al. 2014, PE&RS 80(9):863-872, doi:10.14358/PERS.80.9.863.
1.6 Trees and crowns
This stage ran 2 times in the session (steps 12, 15); the last run is described.
12,812[15] trees of 10 m[15] or more were found, 227.8 per ha[15]. The tallest is 0.0 m[15]. Tops are local maxima of the smoothed canopy height model in a window that grows with height; crowns are grown from the tops by region growing and drawn as convex hulls.
| Height class | Trees |
|---|---|
| 5-20 m | 7,747[15] |
| 20-40 m | 3,399[15] |
| 40-50 m | 617[15] |
| 50-60 m | 377[15] |
| 60+ m | 672[15] |
What this count is. Detection follows the visible canopy: suppressed and intermediate trees under a dominant crown are largely missed (see docs/validation/slice1.md).[15]
2. Files written
| Step | File | Bytes | sha256 (first 12) | Hash of |
|---|---|---|---|---|
| 8 | dtm_2008_1m.tif | 2,239,124[8] | 4ab757315ac2 | file |
| 10 | chm_2008_0p5m.tif | 10,709,057[10] | 857566541e31 | file |
| 12 | tops_2008_ge10m.geojson | 1,793,432[12] | 1d79a0566b03 | file |
| 15 | tops_2008_ge10m_nosmooth.geojson | 2,240,545[15] | 7edd1efa21f0 | file |
| 23 | trees_vs_tv081.svg | 6,896,146[23] | 1890dc444816 | file |
| 25 | trees_vs_tv081.png | 1,990,085[25] | 29d3d2795955 | file |
3. Validation
These tables are copied from the NativeGIS validation documents. They were measured on the HJ Andrews test window against independent software and a denser survey; they describe the method, not the accuracy of this run's data. The tree counts are of dominant and codominant trees, not of every stem.
| DTM | vs 2021 vendor ground: RMSE | median | 95% abs |
|---|---|---|---|
| 2008, delivered ground (auto) | 0.23 m | -0.05 m | 0.41 m |
| 2008, SMRF re-classified | 0.47 m | +0.09 m | 0.90 m |
| 2008, CSF (oracle, slope smoothing off) | 3.14 m | +0.13 m | 3.03 m |
| 2021, SMRF re-classified | 0.24 m | +0.06 m | 0.29 m |
| 2021 thinned, delivered ground | 0.07 m | 0.00 m | 0.15 m |
| 2021 thinned, SMRF | 0.26 m | +0.07 m | 0.42 m |
| Comparison (2,049,335 canopy cells, 10-cell margin) | bias | RMSE | 95% abs | r | within 1 m |
|---|---|---|---|---|---|
| raw (different ground) | -0.27 m | 3.28 m | 3.83 m | 0.980 | 83% |
| same ground (ours + DTM_ours - DTM_csf) | 0.00 m | 1.01 m | 1.15 m | 0.998 | 94% |
| Oracle height class | oracle trees | ours | omission | commission | height RMSE |
|---|---|---|---|---|---|
| 5-20 m | 6,437 | 6,455 | 5.2% | 5.7% | 0.65 m |
| 20-40 m | 2,865 | 2,901 | 3.2% | 2.3% | 1.41 m |
| 40-50 m | 580 | 563 | 1.9% | 2.0% | 3.55 m |
| 50-60 m | 367 | 357 | 1.6% | 0.8% | 4.34 m |
| 60 m + | 651 | 623 | 0.8% | 0.2% | 4.92 m |
| all | 10,900 | 10,899 | 4.1% | 4.1% | 1.90 m (bias +0.06 m) |
| Metric (window mean) | 2008 | 2021 thinned | 2021 full |
|---|---|---|---|
| p95 height | 39.4 m | 41.8 m | 41.8 m |
| mean height | 18.3 m | 21.2 m | 21.2 m |
| cover > 2 m | 0.88 | 0.94 | 0.94 |
| rumple (0.5 m CHM) | 7.03 | 6.93 | 6.80 |
| trees >= 50 m / ha | 18.5 | 17.3 | 17.5 |
| tree height CV | 0.34 | 0.32 | 0.32 |
| old-growth index | 0.53 | 0.52 | 0.51 |
| Oracle height class | oracle | ours | omission | commission | height RMSE |
|---|---|---|---|---|---|
| 5-20 m | 1,085 | 1,098 | 7.8% | 9.0% | 0.74 m |
| 20-40 m | 5,718 | 5,718 | 3.4% | 2.6% | 1.03 m |
| 40-50 m | 772 | 767 | 2.2% | 1.2% | 2.41 m |
| 50-60 m | 359 | 340 | 1.1% | 0.9% | 5.10 m |
| 60 m + | 605 | 577 | 0.8% | 0.9% | 5.00 m |
4. Limits
- Trees are found from the canopy surface. Trees under a taller crown are mostly missed, so stem counts in multi-layered old growth are too low.
- Heights come from the first returns above a terrain model built from the ground points; where ground is sparse under dense canopy, heights carry the terrain error.
- Crowns are convex hulls of grown regions; they are not field-measured crown areas.
- No field plots were used in this report. Compare with plot data before relying on stem counts.
5. Steps
Replay: gis_replay with action “record” writes these steps as a run.json, and action “replay” runs it again in a fresh session and says which outputs are identical.
| Step | Tool | Result | ms | Arguments | Files read and written |
|---|---|---|---|---|---|
| 1 | gis_open | ok | 21,175 | {"path":"stem_map.shp","gpsTime":false,"maxPoints":5000000,"async":true} | stem_map.shp 35.5 MB sha256 82fe0b31a73b… |
| 3 | gis_open | ok | 23,805 | {"path":"ot_HJ_Andrews_bin_100.laz","toEpsg":26910,"gpsTime":false,"maxPoints":10000000,"async":false} | ot_HJ_Andrews_bin_100.laz 25.8 MB sha256 bf81fa07ed8a… |
| 4 | gis_lidar_qa | ok | 455 | {"handle":"ddf2a119_ds_2","cell":1} | |
| 5 | gis_query | ok | 295 | {"handle":"ddf2a119_ds_1","match":"all","bbox":[565350,4900150,566100,4900900],"includeGeometry":false,"limit":0,"offset":0,"saveAs&quo | |
| 6 | gis_lidar_ground | ok | 13 | {"handle":"ddf2a119_ds_2","mode":"auto","minGroundDensity":0.25,"cell":1,"slope":0.15,"window":18,"threshold":0.3,"scalar&quo | |
| 8 | gis_lidar_dtm | ok | 1,590 | {"handle":"ddf2a119_ds_2","resolution":1,"method":"tin","out":"dtm_2008_1m.tif","async":false} | dtm_2008_1m.tif 2,186.6 kB sha256 4ab757315ac2… |
| 10 | gis_lidar_chm | ok | 64,959 | {"handle":"ddf2a119_ds_2","resolution":0.5,"maxHeight":100,"out":"chm_2008_0p5m.tif","async":false} | chm_2008_0p5m.tif 10,458.1 kB sha256 857566541e31… |
| 12 | gis_lidar_trees | ok | 4,876 | {"handle":"ddf2a119_ds_2","minHeight":10,"cwA":2,"cwB":0.18,"wsFactor":0.6,"wsMin":3,"thSeed":0.45,"thCrown":0.55,"smoothSigm | tops_2008_ge10m.geojson 1,751.4 kB sha256 1d79a0566b03… |
| 14 | gis_compare | ok | 382 | {"mode":"trees","a":{"handle":"ddf2a119_ds_2","layer":"trees"},"b":{"handle":"ddf2a119_sel_1"},"params":{" | |
| 15 | gis_lidar_trees | ok | 3,350 | {"handle":"ddf2a119_ds_2","minHeight":10,"cwA":2,"cwB":0.18,"wsFactor":0.6,"wsMin":3,"thSeed":0.45,"thCrown":0.55,"smoothSigm | tops_2008_ge10m_nosmooth.geojson 2,188.0 kB sha256 7edd1efa21f0… |
| 17 | gis_compare | ok | 322 | {"mode":"trees","a":{"handle":"ddf2a119_ds_2","layer":"trees"},"b":{"handle":"ddf2a119_sel_1"},"params":{" | |
| 18 | gis_open | ok | 125 | {"path":"chm_2008_0p5m.tif","gpsTime":false,"maxPoints":5000000,"async":false} | chm_2008_0p5m.tif 10.2 MB sha256 857566541e31… |
| 19 | gis_process | ok | 1,311 | {"op":"translate","inputs":[{"handle":"ddf2a119_ds_9"}],"params":{"dx":0,"dy":0,"band":1},"name":"Canopy height, 20 | |
| 20 | gis_process | ok | 137 | {"op":"translate","inputs":[{"handle":"ddf2a119_ds_6","layer":"matched"}],"params":{"dx":0,"dy":0,"band":1},&q | |
| 21 | gis_process | ok | 58 | {"op":"translate","inputs":[{"handle":"ddf2a119_ds_6","layer":"omitted"}],"params":{"dx":0,"dy":0,"band":1},&q | |
| 22 | gis_process | ok | 2 | {"op":"translate","inputs":[{"handle":"ddf2a119_ds_6","layer":"committed"}],"params":{"dx":0,"dy":0,"band":1}, | |
| 23 | gis_render | ok | 4,629 | {"layers":[{"handle":"ddf2a119_ds_10","style":{"ramp":"greys","min":0,"max":80,"units":"m"}},{"handle":"d | trees_vs_tv081.svg 6,734.5 kB sha256 1890dc444816… |
| 25 | gis_render | ok | 1,791 | {"layers":[{"handle":"ddf2a119_ds_10","style":{"ramp":"greys","min":0,"max":80,"units":"m"}},{"handle":"d | trees_vs_tv081.png 1,943.4 kB sha256 29d3d2795955… |