Hydrological feature extraction from LiDAR. Grant Pearse 28 March 2017

Transcription

1 Hydrological feature extraction from LiDAR Grant Pearse 28 March 2017

2 Overview Hydrology and Forestry Background & Motivation Extract more value from LiDAR Case study Geraldine Practical Overview

3 Hydrology and Forestry National Policy Statement for Freshwater Management (NPS-FM) National Environmental Standard for Plantation Forests (NES-PF) Emphasis: Management of potential impacts Monitoring and compliance Land managers will need better information: Identify hydrological features Integrate into planning (esp. roading & harvesting) Identify and manage riparian areas Manage sediment entry

4 Current Sources of Information Field surveys (best but partial coverage) Topo50 (vector) NIWA River Environments Classification (REC) Recognised data source Topologically correct centre lines & junctions Free: ArcGIS Geodatabase Updated and refined

5 REC - limitations Not intended for forestry use cases 1 st and 2 nd order information missing

6 Alternative Sources - LiDAR LiDAR is well suited to hydrological feature extraction Improvements in methods and algorithms Many implemented as software tools Study Method Software Implementation Clubb et al. (2014) Pelletier (2013) Passalacqua et al. (2010a) Drainage Extraction by identifying Channel Heads method (DrEICH) Method using an optimal Wiener filter and a user-defined contourcurvature threshold for channelisation. GeoNet combines local non-linear diffusion filtering with a global, geomorphologically informed geodesic cost function to automatically identify channel initiation points and extract channel paths from LiDAR DTMs. Open source - LSDTopoTools Open source - LSDTopoTools MATLAB (license required) or Python (free) Sofia et al. (2011) Statistical approach based on normalised topographic attributes, such as openness and minimum curvature as a weight for the upslope area. Contact: G. Sofia (giulia.sofia@unipd.it) - ESRI ArcGIS (license required) Tarolli and Dalla Fontana (2009) Uses curvature to assess the capability of high resolution topography to recognise the convergent hollow morphology of surveyed channel heads. Contact: P. Tarolli (paolo.tarolli@unipd.it) - ESRI ArcGIS (license required)

7 Cast Study: Geraldine Forest LINZ funded high-resolution LiDAR capture Range of topics investigated including hydrology Hydrology: range of methods LSDTopoTools: Four algorithms for hydrological features Steep learning curve GeoNet Significant literature supporting the method Already used by Forestry Corp. NSW

8 Case Study: Geraldine Geraldine DTMs at: (A) 0.4 m, (B) 1 m and the best currently available national elevation DTM at 25 m resolution (C).

9 Case Study: Geraldine Results GeoNet channel networks extracted from 1 m and 0.4 m resolution DTMs

10 Case Study: Geraldine Results Contrasting GeoNet and NIWA REC river lines (A) and GeoNet vs ArcGIS channel networks (B)

11 GeoNet: Geomorphic Feature Extraction MATLAB Very easy to get working Fast AUD$9000 license cost Base MATLAB + Toolboxes: Image Processing + Mapping, Statistics. Python: Free and Open Source Linux Virtual Machine Simplified approach Memory bound

12 GeoNet: Getting Started 1. Virtual Machine Host 2. Download the VM 3. Prepare your data: Raster DTM (1 m resolution / pixel size) Projected CRS e.g. UTM 60S or 59S NoData: Split into sub-catchments (resource dependent)

13 GeoNet: Getting Started 4. Transfer your data to the Linux environment 5. Setup GeoNet parameters in Python Covered in instructions + video Defaults generally perform well Fine-tune on small areas

14 GeoNet: Getting Started Results: Intermediate results for fine-tuning Shapefiles: drainagenetwork.shp, channelheads.shp Must be projected to input CRS

15 GeoNet: Limitations Extremely resource intensive Sub-catchment level Manual delineation Stitching results Fine tuning can be time consuming No validation against NZ ground survey data Tutorial and Virtual Machine:

16 Grant Pearse Geomatics Scientist Date: 28 March 2017