Using SAR Intensity and Coherence to Detect A Moorland Wildfire Scar

Editor's Notes

• #4: Fires cause an increase in greenhouse gas emissions e.g. Carbon dioxide, methane and nitrous oxide.UK social impacts occur during the Spring Bank Holidays – discarded cigarette ends and disposable BBQ + hotter drier temperatures happening at that time of year e.g. Spring of 2003, 2008 and 2011.
• #5: Video Clip of WainstallsWildfires are unwanted vegetation fires. UK causes Arson Accidental ignition cigarettes and disposable BBQ Lack of rainfall in the spring in combination with winter drying effects on the vegetation decreasing the FMC will increase the potential for a fire Major impact on moorland ecosystems, especially peatlands. The impact varies with the amount of area burnt and severity of the burn.Wildfire EffectsNegativeDOC concentrations increase in drinking waterDeep seated blanket peat fires release CO2into the atmosphereErode landscapePositiveChange ecological composition of moorland environmentDestroy habitat for grouseIncrease in graminoids and decrease in ericoid sub-shrubs
• #6: European Forest Fire Information SystemEFFIS Burnt Area Locator managed to identify and produce a burnt area outline for the 1017 hectares Anglezarke Fire in Lancashire 29/04/11 Did not locate the Wainstalls fire which began on 30/04/11 and burnt approximately 300 hectares of moorland Burnt area product for EFFIS is produced using either 32m DMC data or Advanced Wide Field Sensor on board IRS with spatial ground resolution of 56mThreshold for size of burnt areas detectable is 5 to 10 ha or largerOvenden MoorTherefore use of Optical Data is an issue for monitoring burnt areas of UK fires due to cloud cover. Other approaches need to be explored i.e. Radar which can see through cloud and smoke Fire scar monitoring is important for assessing the recovery of the moorland landscape as some fires such as Wainstalls are deep seated and burn into the peat destroying the roots of heather and impeding recovery
• #7: This research will inform the next steps in my PhD
• #8: Essential requirement in the UK, due to microwaves having a longer wavelength compared to optical dataSAR sensors emit their own illumination source in the form of microwaves For this research C-band data will be used Future research using more recent case studies will also analyse L-band data which can penetrate deeper into the ground due to the longer wavelength.
• #9: There have been many studies in the literature for using SAR for forest fires in the tropics, Mediterranean and boreal ecozones but there is little research on the use of SAR for detecting fire scars in moorland environments. This is a feasibility study.Radar is a distance measuring device There is a Transmitter, a Receiver, an Antenna, and an Electronic system to process and record the data. Transmitter generates pulses of microwaves at regular intervals which are focused by the antenna into a beam The radar beam illuminates the surface obliquely at a right angle to the motion of the platform. The antenna receives a portion of the transmitted energy reflected known as ‘backscatter’ from various objects on the ground in this case a tree
• #10: PDNP would be very vulnerable to temperature increases predicted b the UK Climate Impacts Programme (UKCIP) as its one of the most southern moorland landscapes One of the most visited national parks especially around Bank Holidays18th AprilBleaklow experienced intense fire which burnt deep into the peat, covering 7Km2, 700 hectares Previous fires have occurred in this area logged by the PDNP rangers Vegetation consists of heather, cottongrass and mosses
• #12: Exposed peat bog inside the fire scar had the highest pre-fire intensity signal at 0.16 dB JD 39 Can see relative brightness on the east side of the fire scar Fig a-d Peat bog values stay high post-fire (0.78 dB JD 144 and -0.57 dB JD 179) as shown in fig e & f. Very dry during JD 72 – JD 90 with only one notable rain event of 15.2mm on JD 91, this could explain the downward trend in backscatter intensity then peak in intensity for the ASAR AP image acquired on JD 93 (d). After the fire event rainfall frequently occurred with Fig e and f exhibiting strong backscatter.
• #13: Explain Axis Average intensity values in dB inside and outside the Bleaklow fire scar for CORINE land cover classes. ERS-2 image acquired on JD 74 and ASAR Image Mode image acquired JD 81 show a downward trend in backscatter intensity for all land cover classes except natural grassland intact peat bog outside the fire scar
• #14: Baseline should not be greater than 500m to avoid temporaldecorrelation. Coherence images measure the degree of correlation between two SAR images acquired at different times. Produced during Interferometric SAR (InSAR) pre-processing, using the phase portion of the radar signal and the amplitude (Rykhus and Zhong, 2011) Step 1 Interferogram Generation: this measures the phase difference between two SLC coregistered SAR images. One image is multiplied by the other image producing an interferogram of phase difference. Step 2 Interferogram Flattening: the constant phase (due to the acquisition geometry) and the phase expected for the topography (25m DEM of the site used) usually known as the „low frequency phase‟ is separated out from the residual differential phase known as the „high frequency phase‟ which relates to the temporal phase variation between the master and slave image. Step 3 Interferogram Adaptive Filter and Coherence Generation: The filtering of the flattened interferogram produces a product with reduced phase noise. As a byproduct the coherence is generated as an indicator of phase quality and the intensity filtered images. Step 4 Phase UnwrappingStep 5 Generate Ground Control Points: The _fint and _cc images were opened and a New Vector Layer was generated in ENVI 4.7. GCP‟s were selected off the _fint image using the _cc image as a guide to select points where there is high coherence values (white areas), avoiding fringes and black areas. Step 6 Phase to Displacement Conversion and Geocoding: This step was run to produce a geocoded version of the coherence image.
• #15: 1stInSAR pair there is low coherence ranging from 0.14 – 0.24 depending on the CORINE defined land cover class. 2ndInSAR pair shows a slight increase of coherence for all land cover classes except natural grassland inside the fire scar which remains constant at 0.19 3rdInSAR pair acquired after the fire s(19/04/03 – 24/05/03) show all 3 land cover classes inside the fire scar exhibit an increase in coherence. Greatest increase is moors and heathlands class inside the fire scar value of 0.46 compared to 2nd pair at 0.29Can see this increase visually on the west side of the fire scar Coherence for moors and heathlands outside the fire scar decreased from 0.29 to 0.23 this could be due to phenological change of the heathlands. 4thInSAR pair shows an overall decrease in coherence for all classes, I think this result is due to temporal decorrelation as the initial baseline was highe at 654. Reseeding also occurred on the east side of the fire scar during this time.
• #20: Data selected from ESA Small incidence angle as Huang and Siegert (2006) found backscatter decreased with an increase in incidence angle from the fire scarSARScape 4.2 pre-processing. Focusing and multilooking to produce intensity image Frost, Lee and Degrandi filtering algorithms tested with 2 ERS-2 data – Degrandi smoothed speckle more effectively (amplitude coregistration must be done using this filter as it is a multi temporal filter)Geocoding and radiometic calibration was applied to produce geocoded greyscale images at 25m