[email protected]
Fusing Lidar, SAR and Optical Data to Monitor Wet Area Extent of Prairie Pothole Wetlands in Southern Alberta Joshua Montgomery1, Christopher Hopkinson1, Brian Brisco2 & Laura Chasmer1
1Department of
Geography, University of Lethbridge, Lethbridge, AB T1K 6T5, Canada 2Natural Resources Canada, Canada Centre for Mapping and Earth Observation, Ottawa, Canada
BACKGROUND
Sensitive Prairie Pothole wetlands are under increasing pressure from urban and industrial development as well as agricultural drainage modification. In additional, recent warming trends may have an increasingly adverse effect on wetland wet area extent, a determining factor of wetland riparian health and productivity. Synthetic Aperture Radar (SAR) with HH polarization is best able to separate land from water due to the contrast in backscatter responses between land and water in HH polarisation [1].
KEYWORDS SAR; Lidar; data fusion; wetlands; temporal analysis
CONTACTS Joshua Montgomery
[email protected]
Chris Hopkinson
[email protected]
Integration & QC By selecting a range of dB values instead of a minimum value and applying an additional speckle filter post image extraction, much of the saturated fields and road networks are removed from the water mask, while preserving open water extent of waterbodies. (Blue vs yellow mask).
A methodological framework has been developed focused on data fusion of high resolution SAR, optical and Light Detecting and Ranging (Lidar) data to model changing wet area extents of open water in prairie marshlands over a 300km2 region at the Shepard Slough wetland complex, east of Calgary, Alberta. The framework provides a method of data fusion with user extracted threshold (dB) values and automated k-mean classification to describe changing wetland attributes over large areas at high resolution [2].
SAR Threshold Processing Step 1: To determine the range of thresholds (dB) that represent surface water at a given time of year. Areas of known surface water are sampled with a SAR polarimetry tool over a waterbody with known permanency as an intensity image.
SAR derived water mask.
Lidar DEM
June 6, 2013
July 24, 2013
[email protected]
Step 2: A histogram of the decibel values sampled from the waterbody provides the user with a decibel range which is used as the threshold value.
Step 3: Process the RADARSAT-2 (RS2) data. a) Scale the RS2 data from linear to decibel. b) Apply filters to reduce noise and preserve image edges. c) Apply the threshold range derived in Step 2. d) Apply additional noise and speckle reduction filters on the binary water mask. August 11, 2013
WEBSITE http://artemislab.strikingly.com/
Step 4: Output binary water mask.
REFERENCES [1]Remote
Sensing of Wetlands: Applications and Advances. CRC Press 2015.Chapter 6: Mapping and Monitoring Surface Water and Wetlands with Synthetic Aperture Radar. Brisco, Brian. 119-136.
[2]White
L., Brisco B., Dabbor M., Schmitt A., Pratt A. A collection of SAR methodologies for monitoring wetlands. Remote Sensing, 7(6), 2015: 7615-7645.
WD-CAG
Western Division-Canadian Association of Geographers
Prince George, BC | 11-12 March 2016
October 4, 2013
Temporally coincident SPOT and RapidEye Optical water masks derived using automated k-means classification are used as a quality control measure for SAR derived water masks. Kappa Coefficient and overall accuracy values are reported between 0.85-0.96%. Chestermere Lake functions as a control for water extent in each image and throughout the year artificially controlled as an irrigation reservoir and recreational area.
TERN MODIS GF
Laura Chasmer
September 10, 2013
Shepard Slough SAR and LIDAR Time Series for June-October 2013
Brian Brisco
[email protected]
August 17, 2013
August 10, 2014
August 14, 2015
0
5
10 Kilometers
Shepard Slough SAR and LIDAR Yearly Comparison in August of 2013, 2014, 2015. Note the drastic change in open water (yellow box) of a large ephemeral wetland.
Date
Data Source & Water Mask Area (km2) SAR SPOT RapidEye Greatest % Change
August 2013
2.386 2.387 2.386