Volcano hotspot monitoring

1
Volcano hot-spot monitoring
Klemen Zakek
2
(No Transcript)
3
Physical backgroundDifficulties of hot spots
RSStrategy

• Physical laws important for RS
• Temperature sensitivity
• Subpixel response
• Hot spot detection
• Number of endmembers
• Dual band method
• Radiant flux

4
Physical laws importantfor remote sensing

• Plancks law
• Stefan Boltzmann's law
• Wiens displacement law

5
Temperature sensitivity
MIR
TIR
6
Subpixel response

• The radiance of a pixel is a linear combination
  of the separate radiances.
• Difference TMIR TTIR is an a hot spot indicator.

300 K
500 K
TMIR 450 K TTIR 390 K
300 K
300 K
7
Hot spot detection

• Fixed threshold algorithms
• Contextual algorithms
• Multi-temporal algorithms
• Normalized thermal index (Wright et al., 2002)

8
Number of endmembers

• Homogeneous pixel
• Crust background
• Crust lava
• Crust lava background
• Crust lava background clouds...

9
Dual band method

• Crust background (2 equations, 2 variables if
  TB is assumed Dozier, 1981)
• rad1 p R(?1, TL) (1 p) R(?1, TB)
• rad2 p R(?2, TL) (1 p) R(?2, TB)
• Crust lava (3 equations, 3 variables)
• rad1 p R(?1, TL) (1 p) R(?1, TC)
• rad2 p R(?2, TL) (1 p) R(?2, TC)
• rad3 p R(?3, TL) (1 p) R(?2, TC)
• nmeasurements 2 nendmembers nassumptions
  1

10
Radiant Flux

• The problem is the emissivity.
• With some arbitrariness, radiant flux (W) can be
  converted into effusion rate (m3/s).

11
Physical backgroundDifficulties of hot spots
RSStrategy

• Desired sensor characteristics
• Spatial resolution comparison
• Appropriate spectra
• Detection / saturation problems
• Point spread function

12
Desired sensors characteristicsfor volcano
hot-spot monitoring

• good spatial resolution
• good temporal resolution
• good radiometric resolution
• good spectral resolution
• We have to make a compromise!!!

13
MSG SEVIRI 3000 3000 m
Terra ASTER (SWIR) 30 30 m
Aqua / Terra MODIS 1000 1000 m
Lava flow
14
Appropriate spectra

• 2.02.5 mm
• 3.54.0 mm
• 1011 mm

15
Detection / saturation problems

• Detection limit depends on SNR 0.010.1 of the
  pixel area has to be hotter than the background.
• Sensors usually saturate at temperatures above
  350 K.

MODIS channel 21 AVHRR channel 3B Tbackground
300 K
16
Point spread function

• PSF can be usually approximated by a Gaussian
  function it does not have a shape of a box.
• A single hot spot can contaminate neighbour
  pixels.
• The estimated temperature depends on the hot-spot
  position within the pixel.

17
Physical backgroundDifficulties of hot spots
RSStrategy

• Appropriate sensors
• Minimising error
• Expected problems

18
Appropriate sensors

• Geostationary satellites
• Imager aboard GOESS covering East Asia, Pacific
  and both Americas
• SEVIRI aboard MSG covering Africa and Europe
• no appropriate platform covering Central Asia
• Polar orbiting satellites
• AVHRR aboard NOAA and MetOP
• MODIS aboard Terra and Aqua
• AATSR aboard Envisat
• ASTER aboard Terra
• Landsat TM / ETM was also often used but it is
  easily saturated

T 560 K A  0.4 ha
T 400 K A  1.4 ha
19
Minimising error

• Characterise clusters and not single pixels.
• The most reliable parameter is radiant flux.
• AVHRR 10 images per day
• MODIS 4 images per day
• Kalman filter

combine data
20
Expected problems

• merely a few validation data available
• clouds prevent detection of underlaying surface
• a possible sensor failure (MODIS lifetime has
  already finished)