Hierarchical Object Based Image Analysis of Unmanned Aerial System (UAS) imagery for species level mapping of Himalayan treeline vegetation

Abstract : Situated between montane forest and alpine vegetation, treeline ecotone represents the upper limit of forest on a mountain. The Himalayan region in south Asia has the highest and biologically one of the most diverse treeline ecotone in the world. The impact of climate change is projected to be disproportionately high along Himalayan treeline ecotone due to their high sensitivity to changing temperature and precipitation patterns, leading to significant ecological changes. This study analyzed multi-spectral imagery acquired using Unpiloted Aerial Systems (UAS) and plot level field collected data at treeline ecotone in Langtang National park, Nepal to examine the suitability of the ecological proxies derived from UAS imagery for mapping of treeline vegetation at genus level in a spatially explicit manner. The UAS images were pre-processed using Structure from Motion (SfM) algorithms and resulting reflectance maps were analyzed following semi-automated Object Based Image Analysis Approach to develop homogeneous objects at multiple segmentation scales and derive spectral and textural features. Following a hierarchical classification approach, decision tree algorithm was utilized first, to semi-automatically derive vegetated and non-vegetated parts in the imagery and at the second hierarchical level, four woody vegetation genus types were mapped. Results showed that mapping accuracy varied not only between hierarchical levels of classification but also among different species types. The results highlight potential and limitations of UAS acquired imagery and object based image analysis for detailed mapping of vegetation properties in a data scarce Himalayan region and enable understanding of ecological patterns and the underlying processes along treeline ecotone.
Complete list of metadatas

http://hal.univ-reunion.fr/hal-01960456
Contributor : Réunion Univ <>
Submitted on : Wednesday, December 19, 2018 - 2:05:59 PM
Last modification on : Tuesday, April 16, 2019 - 12:58:08 PM

Identifiers

  • HAL Id : hal-01960456, version 1

Collections

Citation

Niti Mishra, Bharat Shrestha. Hierarchical Object Based Image Analysis of Unmanned Aerial System (UAS) imagery for species level mapping of Himalayan treeline vegetation. GEOBIA 2018 - From pixels to ecosystems and global sustainability ​, Jun 2018, Montpellier, France. ⟨https://geobia2018.sciencesconf.org/⟩. ⟨hal-01960456⟩

Share

Metrics

Record views

34