BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//pretalx//talks.osgeo.org//foss4g-2026//speaker//Q7TQTQ
BEGIN:VTIMEZONE
TZID:JST
BEGIN:STANDARD
DTSTART:20000101T000000
RRULE:FREQ=YEARLY;BYMONTH=1
TZNAME:JST
TZOFFSETFROM:+0900
TZOFFSETTO:+0900
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
UID:pretalx-foss4g-2026-MC8VGT@talks.osgeo.org
DTSTART;TZID=JST:20260903T143000
DTEND;TZID=JST:20260903T150000
DESCRIPTION:The rapid development of high-resolution topographic informatio
 n obtained through unmanned aerial vehicle (UAV) LiDAR systems has signifi
 cantly improved the capability to analyse terrain morphology\, detect geom
 orphologic processes\, and support spatial hazard analysis at centimetre-s
 cale resolutions. At the same time\, the maturity of open-source geospatia
 l software frameworks\, including QGIS\, GRASS GIS\, SAGA GIS\, PDAL\, ESA
  SNAP\, and the Python scientific stack\, has enabled the development of t
 ransparent and reproducible analytical methods outside closed software env
 ironments. Despite these developments\, a considerable number of terrain h
 azard modelling studies still rely on closed or only partially documented 
 processing chains\, which limits transparency\, reproducibility\, and scie
 ntific reuse of methods. In addition\, while machine learning-based landsl
 ide and terrain instability analysis methods are well established\, only a
  small number of studies present fully reproducible end-to-end workflows b
 ased on free and open-source geospatial software\, including UAV LiDAR-der
 ived geomorphometry.\nIn this study\, we present an open\, reproducible\, 
 and fully documented machine learning-based framework for high-resolution 
 terrain classification and hazard susceptibility mapping using UAV LiDAR-d
 erived geomorphometry. The main purpose of this study is to demonstrate th
 at robust scientific analysis of terrain morphology\, including terrain ha
 zard modelling\, can be achieved using an open geospatial software stack\,
  while at the same time ensuring that the predictive accuracy of the propo
 sed framework is comparable to closed software environments. A secondary p
 urpose of this study is to quantify the relative importance of key geomorp
 hometric factors controlling terrain instability at very high spatial reso
 lutions\, while at the same time exploring the reproducibility of the prop
 osed framework within the broader context of open geospatial science and F
 AIR data principles.\nThe research is based on a dense UAV LiDAR survey th
 at results in a centimetre-scale digital terrain model after classificatio
 n of ground points\, removal of outliers\, and interpolation via an openly
  scripted PDAL processing pipeline. All the data were acquired in Croatia 
 (Europe)\, on different geomorphology terrain characteristics. UAV LiDAR d
 ata were collected using DJI Matrice 350 RTK with L2 payload. To ensure ce
 ntimetre level accuracy\, UAV was paired with GNSS receiver Emlid Reach RS
 + with the connection on Croatian Positioning System (CROPOS). From the no
 rmalized terrain surface\, a full range of geomorphometric derivatives was
  computed via GRASS GIS and SAGA GIS plugins integrated into QGIS. The com
 puted derivatives include slope gradient\, aspect\, plan and profile curva
 ture\, terrain position index\, surface roughness\, flow accumulation\, an
 d the LS erosion factor. These parameters are recognized as key indicators
  of morpho dynamic processes linked to slope instability and erosion. The 
 raster layers were normalized to a consistent spatial resolution and exten
 t and a consistent coordinate reference system to obtain a multivariate pr
 edictor stack for machine learning-based research.\nThe reference data for
  the supervised models were derived from detailed geomorphological interpr
 etation of the LiDAR terrain surface and existing engineering-geological i
 nformation in the study area. The stable and unstable terrain units were i
 dentified to develop the training and validation datasets. To guarantee me
 thodological reproducibility and computational reproducibility\, all the s
 teps involved in the research were implemented via openly accessible Pytho
 n scripts and QGIS models.\nThe machine learning classification was perfor
 med via open-source Python libraries. The most dominant libraries utilized
  were scikit-learn. The classification results were evaluated via spatiall
 y independent validation. The models were tested via confusion matrix\, ac
 curacy\, F1 score\, and receiver operating characteristic score. Feature i
 mportance was also computed to evaluate the relative contribution of indiv
 idual geomorphometric variables to terrain instability hazard. The feature
  importance results provided physically interpretable results on terrain i
 nstability via LiDAR terrain morphology.\n\nThe results show that fully op
 en-source machine learning workflows\, when applied to UAV LiDAR-derived g
 eomorphometric parameters\, can attain high predictive reliability in diff
 erentiating stable from potentially unstable terrain conditions. Ensemble-
 based methods\, particularly Random Forest\, show the most balanced perfor
 mance and robustness against predictor multicollinearity. For all models\,
  slope gradient\, curvature\, and flow accumulation are consistently ident
 ified as dominant predictors\, which is consistent with geomorphological t
 heory and further underscores the analytical value of centimetre-scale ter
 rain representation. Beyond predictive accuracy\, the principal scientific
  contribution of this work is its operational reproducibility within the o
 pen geospatial community. The entire workflow\, from point cloud pre-proce
 ssing to final susceptibility map generation\, can be executed using exclu
 sively free and open-source tools\, along with fully disclosed parameters 
 and computational steps. Moreover\, all processing scripts\, derived data 
 products permitted under data sharing restrictions\, and workflow document
 ation will be made openly available under an open-source license via an op
 en repository. The proposed framework is inherently transferable and appli
 cable to the rapidly developing and expanding UAV LiDAR acquisition initia
 tives across the globe. By demonstrating the capability of high-end terrai
 n hazard modelling without recourse to proprietary tools\, this work makes
  a significant contribution to a community-based methodological blueprint 
 applicable to environmental monitoring\, infrastructure planning\, and ris
 k management under resource-constrained or open science-based scenarios. B
 y integrating UAV LiDAR geomorphometry\, interpretable machine learning\, 
 and fully reproducible open-source processing\, this work advances the sci
 entific and operational role of free and open geospatial technologies in h
 igh-resolution Earth surface science.
DTSTAMP:20260718T211451Z
LOCATION:Cosmos2
SUMMARY:Integrating Machine Learning with Open-Source GIS for Reproducible 
 High-Resolution Terrain Classification and Hazard Mapping from UAV LiDAR D
 ata - Nikola Kranjčić
URL:https://talks.osgeo.org/foss4g-2026/talk/MC8VGT/
END:VEVENT
END:VCALENDAR