Discrepancies Between Satellite Detection and Forecast Model Results of Ash Cloud Transport: Case St

1
Discrepancies Between Satellite Detection and
Forecast Model Results of Ash Cloud Transport
Case Study of the 2001 Eruption of Mt.
Cleveland Volcano, Alaska

• David Schneider, USGS-Alaska Volcano Observatory
• Rene Servranckx, Environment Canada, Montreal
  VAAC
• Jeff Osiensky, National Weather Service,
  Anchorage VAAC

2
Motivation/Background

• The understanding that total avoidance of ash
  clouds is required but confusion about what is
  meant by (or how to achieve) zero tolerance.
• A realization that operational decisions
  typically involve resolving conflicts between
  data sources.
• How should warnings and info releases utilize
  model results, satellite data, and observer
  reports? What is the proper weight to give each
  when they are in conflict?
• The question of how long to keep a warning going
• A look at the 1991 eruption of Cleveland volcano
  will illustrate these issues, but not answer any
  of these questions.

3
Location Map
Anchorage
Cleveland Volcano
R. Wessels

• About 900 miles from Anchorage 5675 ft high.
• Eruption in 2001 is the largest from a
  seismically unmonitored volcano since the
  formation of AVO in 1988.

4
Summary of February 19 Eruption

• Eruption detected in AVHHR satellite image as
  part of routine monitoring by an AVO remote
  sensing analyst, about 3 hours after the eruption
  start.
• Ash production for about 6 hours, and detected in
  GOES satellite images for 48 hours.
• No Color Code issued by AVO for Cleveland because
  of the lack of a seismic monitoring network
  (policy since changed).
• Combined response of 3 VAACS (Anchorage,
  Washington, and Montreal), Anchorage Center
  Weather, and the AVO. Pointed out a need for
  additional tools to facilitate collaboration
  (VACT).

5
VAAC Map
6
Ash Cloud Forecast Models

• A number of models used by responding groups
• PUFF Anchorage VAAC and AVO
• Canerm Montreal VAAC
• Vaftad Washington VAAC
• Although there are differences between the
  models, they are typically in general agreement.
• Ash particles are essentially tracers of flow in
  the atmosphere, and output is influenced by a
  number of factors. The ash is predicted, not
  detected.

7
Satellite-based Ash Cloud Detection
Split-window method is a common technique
Brightness temperature difference (BTD) between
2 thermal infrared channels Semitransparent
volcanic clouds generally have negative BTDs
while meteorological clouds generally have
positive BTDS
8
Satellite-based Ash Detection

• The magnitude of the BTD signal depends upon many
  factors
• Cloud opacity (amounts of ash and water in the
  cloud)
• Size and size distribution of the cloud particles
• Temperature contrast between the cloud and the
  surface beneath it
• Satellite viewing angle
• Atmospheric conditions
• Thus, the detection limit varies (/-) between
  eruptions and during cloud transport.

9
AVHRR Band 3 2/19/01 1645 UTC
10
AVHRR Band 3 2/19/01 1645 UTC
11
AVHRR Band 4m5 2/19/01 1645 UTC
12
AVHRR Band 4 2/19/01 1645 UTC
13
1600 UTC 2/19/01 E 1 hour
14
1800 UTC 2/19/01 E 3 hours
15
2000 UTC 2/19/01 E 5 hours
16
2200 UTC 2/19/01 E 7 hours
17
0000 UTC 2/20/01 E 9 hours
18
0200 UTC 2/20/01 E 11 hours
19
AVHRR Band 4m5 2/19/01 1645 UTC
Upper Level Cloud gtFL300
Lower Level Cloud ltFL200
20
1800 UTC 2/19/01 E 3 hours
21
1800 UTC 2/19/01E 3 hours
22
2000 UTC 2/19/01 E 5 hours
Upper Level Cloud Appears in BTD Images
SIGMET Covers this area at 2 levels Supported by
PIREP
23
2200 UTC 2/19/01 E 7 hours
Lower Level Cloud Starts to Fade in BTD Images
24
0000 UTC 2/20/01 E 9 hours
25
0000 UTC 2/20/01E 9 hours
26
0200 UTC 2/20/01 E 11 hours
FL360 Cinders and sulfur odor in cockpit
FL360 Ash and sulfur odor in cockpit
27
0200 UTC 2/20/01 E 11 hours
SIGMET extended to cover cloud ltFL400
28
0400 UTC 2/20/01 E 13 hours
29
0600 UTC 2/20/01 E 15 hours
SIGMET uses more model guidance
30
0600 UTC 2/20/01E 15 hours
FL S-200
31
0800 UTC 2/20/01 E 17 hours
32
1000 UTC 2/20/01 E 19 hours
33
1200 UTC 2/20/01 E 21 hours
34
1200 UTC 2/20/01E 21 hours
FL S-200
35
1400 UTC 2/20/01 E 23 hours
36
1600 UTC 2/20/01 E 25 hours
37
1800 UTC 2/20/01 E 27 hours
As signal starts to fade, the SIGMETs give
satellite images more weight.
38
1800 UTC 2/20/01E 27 hours
39
2000 UTC 2/20/01 E 29 hours
40
2200 UTC 2/20/01 E 31 hours
41
0000 UTC 2/21/01 E 33 hours
42
0000 UTC 2/21/01 E 33 hours
43
0200 UTC 2/21/01 E 35 hours
44
0400 UTC 2/21/01 E 37 hours
45
0600 UTC 2/21/01 E 39 hours
46
0600 UTC 2/21/01 E 39 hours
47
0800 UTC 2/21/01 E 41 hours
48
1000 UTC 2/21/01 E 43 hours
49
1200 UTC 2/21/01 E 45 hours
50
1200 UTC 2/21/01 E 45 hours
51
1400 UTC 2/21/01 E 47 hours
52
1600 UTC 2/21/01 E 49 hours
SIGMET HOTEL 12 cancelled at 1715 UTC on 2/21/01
53
1800 UTC 2/21/01E 51 hours
54
0000 UTC 2/22/01E 57 hours
55
0600 UTC 2/22/01E 63 hours
56
1200 UTC 2/22/01E 69 hours
2/22/01 at 1408 UTC FL360 Particles and strong
odor in cockpit
57
1800 UTC 2/22/01E 75 hours
58
Final Thoughts

• With satellite images, does an absence of
  detectable ash mean that ash is absent?
• Does the prediction of ash in a dispersion model
  mean that ash is present?
• The eruption response demonstrates how the
  balance between data sets can evolve.
• Cloud height is crucial but hard to determine.
• No reports of damage to aircraft. Does this mean
  that no damage occurred?
• Was the decision to end warnings after 48 hours
  prudent given a report of odor 24 hours later?
  (Zero tolerance?)

59
Thank You
60
Composite Ash Movement
Image by K. Papp
61
1800 UTC 2/19/01E 3 hours
62
0000 UTC 2/20/01E 9 hours
63
0600 UTC 2/20/01E 15 hours
64
1200 UTC 2/20/01E 21 hours
FL S-200
65
1800 UTC 2/20/01E 27 hours
66
0000 UTC 2/21/01 E 33 hours
67
0600 UTC 2/21/01 E 39 hours
68
1200 UTC 2/21/01 E 45 hours
Last GOES detection was at 1600 UTC on 2/21/01.
69
1600 UTC 2/19/01 E 1 hour