Transcript vaacTSOP_ejh_final-2..

Real-Time Estimation of Volcanic
Ash/SO2 Cloud Height from Combined
UV/IR Satellite Observations and
Numerical Modeling
Eric Hughes
University of Maryland - College Park
Cooperative Institute for Climate and Satellites (CICS)
Gilberto A. Vicente
NOAA National Environmental Satellite, Data, and Information Service (NESDIS)
Office of Satellite Data Processing and Distribution (OSDPD)
Wilfrid Schroeder
University of Maryland - College Park
Cooperative Institute for Climate and Satellites (CICS)
Overview
•
Project Outline:
 Volcano Monitoring (Review)
 Volcanic Cloud Height Estimation
•
Height Estimation Online System
•
Applications: Eruption of Eyjafjallajokull
• Future Directions:
 Additional Satellite Data
 Forecast Generation
Overall Goal:
Create a set of tools that assist users in the
near-real time (NRT) monitoring, modeling,
and forecasting of volcanic emissions
Project Overview
Volcano Monitoring
Create a platform that allows users to view near real-time volcanic
data products.
Volcanic Cloud Height Estimation
Construct a system which compares near real-time data with
model simulation data.
Project Overview
Volcano Monitoring
* Processed at NASA GSFC
Retrieve satellite data*
SO2/AI Retrieval*
Volcanic Cloud Height
Estimation
Post-Processing
Place SO2/AI maps and
data files on our web server
Model Initialization
Users
Run model simulations
Compare to satellite
observations
These project parts were developed independently, but
work together as a set of tools for users.
Volcano Monitoring
The NOAA/NESDIS OMISO2 product
delivery and visualization user interface
http://satepsanone.nesdis.noaa.gov/pub/OMI/OMISO2/
Satellite orbit
Digital images
Global composites
Volcano sectors
Volcano Monitoring
Volcanic Sector Imagery
SO2
Cloud (Reflectivity)
AI
Volcanic Cloud Height Estimation
The Approach:
Run various dispersion model simulations and
see which initial height conditions reconstruct
satellite observations.
The final product is an operational system where a user provides
minimal input and receives ash cloud height information (based
on model results and satellite observations).
Volcanic Cloud Height Estimation
Implementation:
Step 1: Run the dispersion model (PUFF) using
various initial height conditions
Step 2: Compare the results from the various
simulations to satellite observations.
– Generate Images for a visual analysis.
– Compute a statistical image comparison.
Volcanic Cloud Height Estimation
Implementation: Run the Dispersion Model
• Define initialization parameters (User):
– Location or name of eruptive volcano
– Estimated time of the eruption
– Estimated duration of the eruption
• Run various simulations, iterating through various initial
height assumptions
Run the model assuming a 2 km injection height
–
–
–
–
Adjust the initial injection height by +D1 km
Re-run the model …
Adjust the initial injection height by +D1 km …
Re-run the model …
Continue adjusting and re-running the model until the injection height
has reached 20 km
Volcanic Cloud Height Estimation
Basic Concept
Implementation: Compare the Results
Input data
Gridded data
Compare overlap
AIRS (Ash)
PUFF (2km Simulation)

Overlapping region:
Compare the results from the various simulations to satellite observations.
Volcanic Cloud Height Estimation
Implementation: Statistical Comparisons
Statistical comparison:
A = Number of Coincident Satellite
and Model points
A=
B = Number of Satellite points NOT
coincident with model data
B=
Compute two statistics:
- Probability of Detection (PoD):
PoD = A / (A+B)
- False Alarm Rate (FAR):
FAR = C / (A+C)
Currently, only implementing the PoD statistics
Simulation Height (km)
C = Number of Model points NOT
coincident with satellite data
C=
Probability of Detection
Putting it all together
Constructing a system to perform the height
analysis in the Near-real Time (NRT)
System Construction
Abstract Workflow Diagram
Inputs
Observation
Data
Grid Observation
data
Observation
Conditions
Grid model
output
Compare the
model output to
the observed
data
PUFF
model
Eruption
Parameters
Change eruption parameters (height)
Observation Conditions:
Observed Time, Threshold value
Eruption Parameters:
Height, Start Time, Duration
Save/Output comparison
images and statistics
System Construction
Online Model Setup
Web Server
Firewall
(satepsanone)
Management Server
Model Server
Check to see if a request
has been submitted
…
Submit request
(User)
Retrieve the request,
then submit the request to
PUFF
… show the status of
the analysis …
Generate output images
w/ IDL
Display the
results
Retrieve output images
and data files. Submit
them to the web
Run the PUFF
simulations and
perform the height
analysis.
System Demonstration
Application:
An analysis of the April 2010 Eruption of Eyjafjallajokull
The Eyjafjallajokull eruption
Observations from OMI and AIRS - (April 15th, 2010)
April 14th
April 15th
13:30
12:00
18:00
Time (GMT)
Beginning of Eruption*
AI (Ash)
Ash
SO2
AIRS
*Estimated from
satellite observations
OMI
User server interface
Input parameters
Eruption Parameters:
 Starting time of the eruption
 Duration
 Location (or volcano name)
Satellite/Observation Parameters:




Satellite name
Orbit number
Time of the observation
Threshold value to distinguish between signal/noise
Meteorological Model Parameters:
 Region of the globe where the event occurred
 Meteorological model data input
The Eyjafjallajokull eruption
Input parameters about the eruptions:
Volcano? Eruption time, duration?
Input parameters about the satellite
observations from the NOAA OMI NRT
Volcanic Emissions site
Input Data
April 15th 12:00 UTC
OMI-AI
http://satepsanone.nesdis.noaa.gov/pub/OMI/OMISO2/
Online Interface for volcanic cloud height estimation
Online Interface for volcanic cloud height estimation
Online Interface for volcanic cloud height estimation
Online Interface for volcanic cloud height estimation
Online Interface for volcanic cloud height estimation
Online Interface for volcanic cloud height estimation
Online Interface for volcanic cloud height estimation
Online Interface for volcanic cloud height estimation
PUFF Simulation
OMI AI
The Eyjafjallajokull eruption
Analysis Summary
Observations from April 15th 2010
OMI – AI/SO2
AIRS-Ash
All profiles show two distinct peaks in height: 8-10 km and 5-4 km.
The Eyjafjallajokull eruption
Limitations
April 15th 12:00 UTC
Vertical profile
Visual Analysis
2 km
OMI-AI
8 km
 The statistical and visual analysis do not match exactly
 The statistics predicts the 10km and 4km simulations heights
 A visual analysis suggests the 8-7km and 2-3km heights
 False Alarm Rate (FAR) analysis should improve the statistics
Future Directions
 Incorporating data from other satellites
 Forecast Generation
Future Directions
Additional Satellites
We are currently only working with OMI data.
Plans to add data from other satellites, particularly from:
AIRS and AVHRR
Can we use measurements from different satellites, at
different times, to build a stronger height profile?
Combine the height profiles from various different
simulations to build stronger height profile statistics
Future Directions
Forecast Generation
Re-initialize trajectories at the
locations where the PUFF model
matched the satellite observations.
Overlapping region:
Vertical profile
• A forecast will be run for every simulated height
• Combine all these forecasts into one overall
forecast.
• Those simulation heights which most
accurately match the observations will contribute
more towards the forecast.
Summary and Conclusions
Description of an automated system to compare
dispersion model outputs with Near-Real-Time
(NRT) satellite observations of volcanic emission

Generate a series of maps overlaying various model
simulations atop of satellite observations

Perform a statistical analysis on the simulation/satellite
data to determine which simulation injection heights
produce the best match to satellite observations
 Perform these tasks quickly, requiring little input from the
analyst
Acknowledgements
Arlin Krueger, Simon Carn, and Keith Evans: JCET/UMBC
George Serafino: NOAA/NESDIS
Nick Krotkov and Kai Yang: GEST/UMBC
Jerry Guo: Perot Systems Government Services
Pieternel Levelt: KNMI
- And thank you for your time!