post-fire hazard detection using alos-2 radar and landsat-8 … · 2021. 2. 1. · 30 november 2020...
TRANSCRIPT
@CCDM_2020 #GI4DM #URAP2 #SSSI_AU
Three conferences, one challengeGeographic Information for Disaster Management (GI4DM) 1 December – 3 December 2020Urban Resilience Asia Pacific 2 (URAP2) 3 December – 4 December 2020NSW & WA Surveying and Spatial Sciences Institute (SSSI) Conference 30 Nov 2020 & Tutorial Tuesday 1 Dec
Climate Change and Disaster Management Technology and Resilience in a Troubled World
30 November 2020 – 4 December 2020
Post-fire hazard detection using ALOS-2 radar and Landsat-8 optical imagery
Stella Chelangat*,1, Ling Chang2
[email protected] , [email protected]*,1 International Fund for Agricultural Development, Kenya
2 Dept. of Earth Observation Science, ITC, University of Twente, The Netherlands
@CCDM_2020 #GI4DM #URAP2 #SSSI_AU
Outline
• Background
• Objective
• Methods
• Results
• Discussion
@CCDM_2020 #GI4DM #URAP2 #SSSI_AU
• Background
• The effect of wildfires in forests has attracted recognition both locally and globally.
• The main focus being post-fire damage assessment of wildfires in Victoria (Australia).
• There exists a gap in analysing the effect of variation in geographical aspect of area and influence in fire severity estimation.
FFDI (Forest Fire Danger Index), Spring 2019 (wikipedia)
(World weather attribution)
@CCDM_2020 #GI4DM #URAP2 #SSSI_AU
• Objective
• The main objective is to analyze the use of satellite SAR data and its comparison to optical imagery for identification and classification of burnt and unburnt patches after a forest fire.
Specific Objectives• To develop a forest fire burnt severity map that compares the size and extent of
change on pre and post-fire instances.
• To explore the sensitivity of polarimetry decomposition and backscatter intensity in the identification of burnt and unburnt areas.
• To determine the degree of spectral contrast between burnt and unburnt areas.
@CCDM_2020 #GI4DM #URAP2 #SSSI_AU
• Earth Observations/Models/Methods:• Pre-processing steps
• Burnt sensitive spectral composite • Support vector machine (SVM)
• Training data processing
• Accuracy assessment and validation
@CCDM_2020 #GI4DM #URAP2 #SSSI_AU
MethodsOptical-based spectral indices (dNBR) calculation
• The threshold-based classification of dNBR was used as a methodological reference to obtain burnt severity maps.
• 𝑑𝑁𝐵𝑅 = 𝑝𝑟𝑒𝑁𝐵𝑅 − 𝑝𝑜𝑠𝑡𝑁𝐵𝑅• Burn severity levels obtained calculating dNBR, proposed by United States
Geological Survey (USGS).dNBR Values(not scaled)
Burn Severity Classified value
-0.100 to +0.99 Unburnt Unburnt
+0.100 to +0.269 Low severity burn Burnt
+0.270 to +0.439 Moderate to low severity burn Burnt
+0.440 to +0.659 Moderate to high severity burn Burnt
+0.660 to +1.300 High severity burn Burnt
@CCDM_2020 #GI4DM #URAP2 #SSSI_AU
MethodsSupport Vector Machine
• It finds an optimal hyperplane and maximizes the margin between two defined classes using fewer training samples.
Optimal hyperplane
Maximized margin
Support vector
@CCDM_2020 #GI4DM #URAP2 #SSSI_AU
MethodsValidation
• Sampling sites for training are obtained through visual interpretation of images both before and after the fire and drawing of vector training sets using ArcMap.
• The training and test sets are separated using a random sampling procedure.
• Training set contains two-thirds of the total area and test set one third
Data Number of training samples
Number of test samples
Burnt 40 30
Unburnt 30 25
@CCDM_2020 #GI4DM #URAP2 #SSSI_AU
Study area at Victoria, Australia
@CCDM_2020 #GI4DM #URAP2 #SSSI_AU
Data description
Data Landsat 8 (Optical) ALOS-2 (SAR)
Acquisition date (pre/post)
Pre_fire 26/07/2019 post_fire 13/10/2019
Pre_fire 30/07/2019 post_fire 08/10/2019
Product type Level 1.1 Level 1.1
resolution 30 × 30 m 10 × 10 m
Wavelength \ L =24 cm
Polarization HH
Orbit Ascending
@CCDM_2020 #GI4DM #URAP2 #SSSI_AU
• Result• ALOS-2 pre-fire (a) ([-0.47, 63.01] dB in HH) and post-fire (b) ([-0.44, 47.13] dB in HH)
intensity images covering the area of bushfire.
@CCDM_2020 #GI4DM #URAP2 #SSSI_AU
• Result• Landsat-8 pre-fire (a) and post-fire (b) images covering the area of bushfires,
respectively. Band combination is R:G:B=7:5:2. Foreshowing changes before and after fire.
@CCDM_2020 #GI4DM #URAP2 #SSSI_AU
• Results• A forest fire burnt severity map that compares the size and extent of change on pre
and post-fire instances.
• Burn severity levels obtained calculating dNBR, proposed by United States Geological Survey (USGS).
@CCDM_2020 #GI4DM #URAP2 #SSSI_AU
Results• Classified map of Landsat-8 optical imagery and ALOS-2 imagery of study area
respectively. Red colour representing burnt areas and green representing unburnt areas.
@CCDM_2020 #GI4DM #URAP2 #SSSI_AU
Accuracy assessment
• It showed the capability of both the electromagnetic spectrum and SAR backscatter intensity in corresponding to changes in vegetation structure after a fire.
• Similarly, we obtained quite a good classification percentage accuracy and separation of burnt and unburnt areas within fire perimeter zones
Data Landsat 8 ALOS-2
Kappa coefficient
0.80 0.89
Producer accuracy Burnt=84.35 % Unburnt=82.26%
Burnt=80.69% Unburnt=78.65%
User accuracy Burnt=92.43% Unburnt=84.65%
Burnt=85.65% Unburnt=80.25%
@CCDM_2020 #GI4DM #URAP2 #SSSI_AU
• Conclusions• Successful results obtained from identification of burnt and unburnt scars using both SAR sensor
and optical.
• The use of L-band in forest fire analysis varies in identification of burnt and unburnt regions.
• Sensitivity of SAR depends on vegetation structure, geographical aspect of the area of study and fire intensity.
• Critical factors in use of SAR local incidence angle, acquisition geometry and environmental conditions.
• Post fire analysis the timings of datasets and digitization of fire perimeter zones are key.
• Use of optical dataset as reference dataset can be substituted by use of SAR under certain circumstances.
@CCDM_2020 #GI4DM #URAP2 #SSSI_AU
• Immediate Next Steps
• The use of enhanced data fusion of SAR and optical imagery from multiple satellites in order to improve the timeliness and accuracy of burnt area mapping
• Explore other target decomposition theorems and compare its results to the Cloud and pottier decomposition theorem and see its effectiveness in burnt severity estimation compared to vegetation indices with radar experts and forest fire analysts.
• Lastly further method improvement of terrain correction upon SNAP software for ALOS-2 is required, so that one can perform polarimetry decomposition or use alternative tool(s) to improve the quality of the results
•