Volume IV-3/W1
ISPRS Ann. Photogramm. Remote Sens. Spatial Inf. Sci., IV-3/W1, 33-40, 2019
https://doi.org/10.5194/isprs-annals-IV-3-W1-33-2019
© Author(s) 2019. This work is distributed under
the Creative Commons Attribution 4.0 License.
ISPRS Ann. Photogramm. Remote Sens. Spatial Inf. Sci., IV-3/W1, 33-40, 2019
https://doi.org/10.5194/isprs-annals-IV-3-W1-33-2019
© Author(s) 2019. This work is distributed under
the Creative Commons Attribution 4.0 License.

  01 Mar 2019

01 Mar 2019

PLANNING HARVESTING OPERATIONS IN FOREST ENVIRONMENT: REMOTE SENSING FOR DECISION SUPPORT

M. Piragnolo1,2, S. Grigolato1, and F. Pirotti1,2 M. Piragnolo et al.
  • 1TESAF Department , University of Padova, Via dell’Università, 16, 35020 Legnaro (PD), Italy
  • 2CIRGEO – Interdepartmental Research Center in Geomatics - University of Padova, Via dell’Università, 16, 35020 Legnaro (PD), Italy

Keywords: Machine learning, forest harvesting, skyline vs. forwarder, digital terrain model, morphology, classification

Abstract. The goal of this work is to assess a method for supporting decisions regarding identification of most suitable areas for two types of harvesting approaches in forestry: skyline vs. forwarder. The innovative aspect consists in simulating the choices done during the planning in forestry operations. To do so, remote sensing data from an aerial laser scanner were used to create a digital terrain model (DTM) of ground surface under vegetation cover. Features extracted from the DTM are used as input for several machine learning predictors. Features are slope, distance from nearest roadside, relative height from nearest roadside and roughness index. Training and validation is done using areas defined by experts in the study area. Results show a K value of almost 0.92 for the classifier with best results, random forest. Sensibility of each feature is assessed, showing that both distance and height difference from nearest road-side are more significant than overall DTM value.