A Comparison of Two Volcanic Ash Height - PowerPoint Presentation

Slide1

A Comparison of Two Volcanic Ash Height Estimation Methods and Their Affects on the HYSPLIT Volcanic Ash Model Output

Kyle Wodzicki (SUNY Oswego), Humberto Barbosa (Advisor at UFAL)

Conclusion

References

Acknowledgements

Introduction

Theory and Methods

Objective

Results

Volcanic ash plumes can cause many hazards, one of which being to aircraft. Not only can ash particles damage aircrafts, more specifically their engines, through abrasion, but the ash can cause engine failure

(U.S. Geological Survey, 2010)

.

To prevent aircraft from flying into ash plumes, the accurate forecast of the plume’s dispersion is necessary. One model commonly used to forecast ash plume dispersion is the

Hybrid Single Particle

Lagrangian

Integrated Trajectory Model (

HYSPLIT) . However, to run this model, the top of the ash plume must be accurately determined.

Using two different cloud top estimation techniques, as well as two different volcanic ash detection techniques to view the plume, the ash plume height of an erupting volcano was determined. The heights from the two methods used were compared and then entered into the HYSPLIT Volcanic Ash model. Output from the various model runs was then compared to the actual ash plumes recorded by satellite imagery to determine which height estimation was the most accurate.

EUMETSAT

, cited 2012: Best Practices for RGB Composting of Multi-Spectral Imagery

.

[Available online at http://oiswww.eumetsat.int/idds/html/doc/best_practices.pdf]Pavolonis, M. J., W. F. Feltz, A. K. Heidinger, 2006: Improved Satellite-Based Volcanic Ash Detection and Height Estimates, Extended Abstracts, 12th Conf. on Aviation Range and Aerospace Meteorology, Atlanta, GA, American Meteorological Society, P8.5. [Available online at https://ams.confex.com/ams/pdfpapers/100829.pdf]Pergola, N., V. Tramutoli, F. Marchese, I. Scaffidi, and T. Lacava, 2004: Improving volcanic ash cloud detection by a robust satellite technique. Remote Sens. Environ., 90, 1–22.NASA LAADS Web [http://ladsweb.nascom.nasa.gov/]NOAA National Operational Model Archive & Distribution System [http://nomads.ncdc.noaa.gov/data.php#hires_weather_datasets]Universtiy of Wyoming Atmospheric Science [http://weather.uwyo.edu/upperair/sounding.html]

Ash Detection

The ash detection schemes used in this research were the split window technique and the the use of RGB composite imagery. The split window technique utilizes the difference in absorptive properties between ash and meteorological clouds

(fig. 1)

. Because ash absorbs more radiation at 11

μm

than at 12

μm and vice versa for meteorological clouds, when subtracting the 12 μm channel from the 11 μm channel, ash appears as negative differences and clouds as positive (Pergola et al., 2004; Pavolonis et al.). RGB imagery assigns a specific channel differences to red, green and blue color planes. In RGB imagery, ash will appear red-orange. The differences used can be found in table 1 (EUMETSAT).

RGBChannel (μm)MinimumMaximumGammaRed12.0 – 10.8-4 K2 K1.0Green10.8 – 8.70 K15 K2.5Blue10.8261 K289 K1.0

DataAt 2015 UTC on 04 June 2011, a volcano in the Puyehue - Cordon Caulle volcano complex in Chile erupted. Imagery from the Terra MODIS satellite inside a domain covering South America and most of the Atlantic Ocean to the east was retrieved from NASA’s LAADS Web server. The images retrieved spanned from 04-06 June. The 04 June 2011 12Z GFS model run was retrieved from the NCDC’s archived server. The 12Z sounding launched from Puerto Montt was retrieved from the Wyoming Sounding Website.

Height EstimationTwo methods for ash plume height estimation were used: a comparison of the cloud top temperature from 11 μm imagery to a near by sounding and a comparison of plume drift to winds forecast by a global model. These two methods are considered inaccurate, however, the comparison of heights determined from each method will show if there is a large amount of variability between the two methods (Pavolonis et al.).

Data ManipulationUsing the McIDAS-V software, linear combinations were performed to create the split window images as well as the differences for the RGB color planes (fig 2). Due to the Terra MODIS being a polar orbiting satellite, images taken at the same longitude are up to an 1 ½ hours apart. However, images are assumed to be taken at the same time due to upper level winds changing slowly with time (fig 3).

Figure 3 - Times

of imagery are 1255 (top right), 1300 (bottom right), 1435 (top left) and 1440 (bottom left). Images taken on 04 June

2011. Red line is the location of the volcano.

Figure 2 – Sample of linear combinations performed on all

.

Table 1 – Channel differences for each color plane for RGB imagery. Scale ranges for each plane are also shown

05 June 2011

06 June 2011MOD021KM.A2011156.0235MOD021KM.A2011157.0140MOD021KM.A2011156.0240MOD021KM.A2011157.0145MOD021KM.A2011156.0415MOD021KM.A2011157.0320MOD021KM.A2011156.1205MOD021KM.A2011157.1245MOD021KM.A2011156.1340MOD021KM.A2011157.1250MOD021KM.A2011156.1345MOD021KM.A2011157.1425

04 June 2011MOD021KM.A2011155.1255MOD021KM.A2011155.1300MOD021KM.A2011155.1435MOD021KM.A2011155.1440

Table 2 – List of images retrieved from NASA’s LAADS Web server

Figure 1 – Index of refraction for water (solid), ice (dotted) and ash (dashed). Vertical lines represent the 11 and 12

μm

wavelengths. (Pavolonis et al.)

Cloud Top and GFS EstimationsUsing the 11 μm imagery from 04 UTC on 05 June 2011, the cloud top temperature was determined to be 218 K or -55° C (fig 4). When comparing the plume drift to the GFS forecast winds at 03 UTC, it was determined that the plume was at the 200 hPa surface (fig 5).

Figure 4 – 11

μm

imagery from 04 UTC on 04 June 2011. Cross hair is pixel where temperature was taken.

Figure 5 – RGB imagery from 04 UTC on 04 June 2011 with 200

hPa

GFS winds overlaid.

Heights HYSPLIT Model RunsUpon comparison with the near by sounding, the atmospheric temperature was -55° C at three different heights (fig 6). This yielded three different possible plume heights. The height of the 200 hPa was also determined using the sounding (table 3). The HYSPLIT model was then run using each of the three heights with two different ash reduction schemes for each; no ash reduction and medium ash reduction.

Cloud Top Temp.GFS Comparison10,200 m11,620 m11,600 m14,000 m

Table 3 – Plume top heights determined from the two different methods

Figure 6 – Sounding launched from Puerto

Montt

. Arrows indicated where the atmosphere was -55° C

HYSPLIT Model OutputModel output 6 hour forecast with the 06, 18, 30 and 42 hour forecast correlating with the satellite imagery. These times from all six model runs were overlaid in McIDAS for comparison with the MODIS imagery (fig 7).

Upon comparison it became apparent that the lowest height estimate (10,200 m) was not close to the actual height due to the first and second HYSPLIT runs not being very close to the actual plume drift. The third and fourth model runs (11,620 m) were very similar to the actual drift with the 06 and 42 hour forecasts. However, the 18 and 30 hour forecasts were slightly off. For the last two runs, the first three forecast hours matched the actual drift very closely, with the medium ash reduction bringing the forecasts even closer. However, the 42 hour forecast over slightly over did the drift. With runs three and four being the closest overall to the actual dispersion of the ash plume, this height was the most accurate estimate. This height was also the height that both the plume top estimation methods agreed upon. From this it can be decided that although the two estimation methods are inaccurate by them selves, when used together they can produce and accurate estimate.

Figure 7 – Run 1 (top left), run 2 (top right), run 3 (middle left), etc.