using gps signal strength data to detect volcanic … et al.pdf · using gps signal strength data...

UsingGPSSignalStrengthDatatoDetectVolcanicPlumesKris7neM.Larson,SiddeshNaik,Sco?Palo,DavidSchneider,YusakuOhta,MarioMaCa,MassimoRossi,Valen7naBruno,MauroColtelli

Department of Aerospace Engineering Sciences, University of Colorado, Boulder USGS Alaska Volcano Observatory, Anchorage

Tohoku University, Japan Istituto Nazionale Geofisica e Vulcanologia, Catania, Italy

References Agnew, D. C. and K. M. Larson, Finding the Repeat Times of the GPS Constellation, GPS Solutions, Vol 11(1), doi:10.1007/

s10291-006-0038-4, 2007. Aranzulla, M., et al., Volcanic ash detection by GPS signal, GPS Solutions, Vol. 17, 485-497, 10.1007/s10291-012-0294-4, 2013. Fee, D., S. R. McNutt, T.M. Lopez, K.M. Arnoult, C.A.L. Szuberla, and J. V. Olson, Combining local and remote infrasound recordings from

the 2009 Redoubt Volcano eruption, J. Volcanol. Geotherm. Res., Vol. 259, 100-114, doi:10.1016/j.jvolgeores.2011.09.012, 2013. Fournier, N. and A.D. Jolly, Detecting complex eruption sequence and directionality from high-rate geodetic observations: The August 6,

2012 Te Maari eruption, Tongariro, New Zealand, Journal of Volcanology and Geothermal Research, Vol. 286, 387-396, 2014. Grapenthin, R., J.T. Freymueller, and A.M. Kaufman, Geodetic observations during the 2009 eruption of Redoubt Volcano, Alaska, J.

Volcan. Geotherm. Res., Vol. 259, 115-132, 2013. Houlie, N., P. Briole, A. Nercessian, and M. Murakami, Sounding the plume of the 18 August 2000 eruption of Miyakejima volcano (Japan)

using GPS, Geophys. Res. Lett., Vol. 32, L05302, doi:10.1029/2004GL021728, 2005a. Houlie, N., P. Briole, A. Nercessian, and M. Murakami, Volcanic plume above Mount St. Helens detected with GPS, EOS Trans. AGU, Vol.

30, No. 30, 277-281, doi: 10.1029/2005EO300001, July 2005b. Larson K.M., E.E. Small, E. Gutmann, A. Bilich, P. Axelrad, and J. Braun, Use of GPS receivers as a soil moisture network for water cycle

studies, Geophys. Res. Lett., Vol. 35, L24405, doi:10.1029/2008GL036013, 2008. Larson K.M., E. Gutmann, V. Zavorotny, J. Braun, M. Williams, and F. Nievinski, Can we measure snow depth with GPS receivers?,

Geophys. Res. Lett., Vol. 36, p. L17502, doi:10.1029/2009GL039430, 2009. Larson, K.M., A New Way to Detect Volcanic Plumes, Geophys. Res. Lett, Vol. 40(11), 2657-2660, doi:10.1002/grl.50556, 2013. McNutt, S.R., G. Thompson, M.E. West, D. Fee, S. Stihler, and E. Clare, Local seismic and infrasound observations of the 2009 explosive

eruptions of Redoubt Volcano, Alaska, J. Volc. Geotherm. Res., Vol 259, 63-76, 2013. Ohta Y and M. Iguchi, Advective diffusion of volcanic plume captured by dense GNSS network around Sakurajima volcano: a case study of

the vulcanian eruption on July 24, 2012, Earth, Planets and Space, 67:157 DOI 10.1186/s40623-015-0324-x, 2015. Schneider, D.J. and R.P. Hoblitt, Doppler weather radar observations of the 2009 eruption of Redoubt Volcano, Alaska, Journ. Volc.

Geotherm. Res., Vol. 259, 133-144, 2013. Small, E.E., K.M. Larson and J.J. Braun, Sensing vegetation growth with reflected GPS signals, Geophys. Res. Lett., Vol 37, L12401,

2010. Solheim, F., J. Vivekanandan, R. Ware, and C. Rocken, Propagation delays induced in GPS signals by dry air, water vapor, hydrometeors,

and other particulates, J. Geophys. Res., Vol 104(D8), 9663-9670, 1999

Acknowledgements This research is supported by NSF EAR 1360810 and NASA NNX14A114G grants to the University of Colorado. GPS data from the Mt. Redoubt and Okmok eruptions were collected by Jeff Freymueller and Ronni Grapenthin (University of Alaska, Fairbanks), and are freely available from UNAVCO. GPS data from Mt. Etna were provided by Mario Mattei and Massimo Rossi (Istitute Nazionale di Geofisica and Vulcanologia) and data from Mount Shindake were provided by Yusaku Ohta of Tohoku University and Geographic Survey Institute of Japan.

Introduction GPS is a L-band navigation system. GPS satellites are operated by the U.S. Department of Defense, and jointly managed with the U.S Department of Transportation. They transmit coded signals at two primary frequencies: 1.5575 GHz (L1) and 1.2276 GHz (L2). As part of a modernization program, a third frequency (L5) was recently added, but this signal can only be tracked with newer instrumentation. GPS has been used by geoscientists for decades to measure crustal deformation. It is also routinely used to monitor water vapor in the atmosphere and timing. GPS units of this type cost from $5,000-$10,000, depending on one’s access to volume purchase prices. GPS has been used to monitor volcanoes for over twenty five years. Generally geodetic-quality GPS units are deployed on the volcano, recording dual-frequency ranging observations at sampling rates of 30 seconds. Fundamentally, GPS receivers measure the one-way travel time (ranges) between multiple satellites and a single GPS antenna. Modeling of satellite orbits, satellite and receiver clocks, relativistic effects and atmospheric delays allows geodesists to estimate position (latitude, longitude, and height) with least squares techniques. With appropriate software, mm-cm level measurements of ground displacements can be easily measured. About 10 years ago, volcano geodesists began to evaluate the effects on volcanic plumes on GPS signals (Houlie et al., 2005a,b). As with most geodetic studies, these focused on using the measured ranges (the distances between the satellites and the GPS antenna) because ranging measurements are longer if there is water vapor in it. More recently volcanic plumes in Alaska, New Zealand, Japan, and Italy have been studied with GPS ranging measurements (Grapenthin et al., 2013; Ohta et al., 2015; Aranzulla et al., 2013; Fournier and Jolly, 2014). Typically the volcanic plume is detected by evaluating least squares residuals used when estimating positions. This has the unfortunate effect of combining the volcanic plume effect (which generally only affects a single satellite) with all the satellites that were in view at the same time. Larson (2013) showed that GPS signal strength data (SNR) could also be used to sense plumes. Currently many groups of volcano geodesists are looking closely at detecting plumes with GPS, which build on these common principles: 1.  GPS is a L-band, global, all-weather system. From 6-12 satellites are visible at any time. 2.  GPS ranging data are significantly affected by water vapor and precipitation. 3.  GPS signal strength data are not affected by water vapor or precipitation, but correlate well with other plume

observations of ash. 4.  The GPS constellation consists of 30+ satellites operating in 6 orbital planes at an inclination of 55 degrees with

respect to the equator. The GPS orbital period is nearly half a sidereal day, and thus the satellites are in the same place in the sky ~4 minutes earlier each day (Agnew and Larson, 2007).

5.  Transmission power of the GPS signals is tightly controlled, and thus signal strength levels (and SNR) are very repeatable from day to day.

6.  In principle any GPS-like signal (GLONASS, GALILEO, and BEIDOU) could be used for volcanic plume sensing; all navigation constellations operate in the L-band.

7.  Because of the satellite inclination, different geometries will be observable at different latitudes. For example, in most of the northern hemisphere, there are no useable satellites at northern azimuths (-30 to 30 degrees). This means that GPS receivers should not be placed to the south of a volcano if desiring to observe a plume.

GPS Signal Strength Data (SNR) Although ranging measurements are the primary observables produced by a GPS instrument, all receivers will calculate an engineering measurement (C/No) which is a proxy for signal strength. These data are frequently called the GPS SNR (signal to noise ratio) data. The figure below on the left shows a typical SNR dataset. The direct signal – the one that travels along the straight line between the satellite and the antenna – has a very simple SNR profile which can be approximated by a polynomial. The lower SNR values as the satellite rises and sets is primarily due to the gain pattern of the antenna. The oscillations seen at low elevation angles in this figure (circled in green) are due to signals reflected from the ground. The amplitude, frequency, and phase of these oscillations can be used to measure vegetation water content, snow depth, and soil moisture (Small et al., 2010; Larson et al., 2009; Larson et al., 2008).

Contact information: [email protected] http://spot.colorado.edu/~kristine

GPS Plume Detections in Italy and Japan Below we show GPS detections of volcanic plumes from Mt. Etna and Mt. Shindake. On the left we show a detection for a summit site in Mt. Etna. There is clear attenuation in SNR (shown in red) when compared to SNR data from the previous day (shown in black). We also point out the oscillations in the SNR data that are representative of ground reflections. The frequency of these oscillations can be used to determine variations in snow depth (right). Plume effects can also be seen at a station (ESLN) at a lower elevation, where it is compared to L-band radar and seismic tremor data.

Ongoing Work 1.  Working with geodetic colleagues around the world, we are compiling datasets of GPS SNR detections of volcanic

plumes. These data will allow us to characterize the performance of the method and provide insight as to the physical mechanisms in play.

2.  With radar remote sensing colleagues, we are developing a forward model for SNR changes due to constituents of volcanic plumes.

3.  Where available (e.g. Mt. Redoubt), we are comparing GPS SNR detections with radar backscatter observations. For past volcanic eruptions, our datasets come from geodetic deployments of GPS instruments. These are generally not

placed in a way that provides optimal resolution of the plume, nor do they record data at high sampling rates. In many cases, as with Mt. Shindake, only a single GPS satellite can see the plume. This significantly limits the value of the observations. In order to improve the temporal and spatial sampling of this method, we need to increase the number of satellites viewed and the number of receivers in the field. Geodetic receivers are expensive, and they are not typically deployed to provide viewing of the plume.

NASA has funded our group to explore ways to use inexpensive GNSS instrumentation for plume sensing. At a cost of ~$50/

unit, it becomes feasible to deploy a large number of receivers at high sampling rates in an optimal geometry. We should be able to track both GPS and GLONASS using this receiver/antenna system, providing a total of 55 satellite sources. When the European and Chinese constellations are completed, 110 satellites will be available. We are currently collaborating with the Istituto Nazionale di Geofisica e Vulcanologia and the Mt Etna Volcano Observatory to develop and deploy this low-cost GNSS plume sensor array.

At the University of Colorado we are developing the software and hardware needed for this low-cost sensor array. Instead of

using existing geodetic GPS sites, we will evaluate via simulation which receiver deployments optimize the greatest number of plume detections at multiple heights. We will test both GPS and GPS+GLONASS scenarios. Two notional low-cost GPS arrays for Mt. Etna are shown below. In the network on the left, the receivers are set at a given distance from the center of the volcano, whereas in the right, the receivers are set along spokes. The simulation here was set to have a plume of radius 3 km, centered on the red triangle. Detections are shown as blue circles, which represents the midpoint of the GPS signal intersection with the cylindrical “plume.” In this scenario, the array on the right finds more plume detections than on the left. The GPS units are being built to include telemetry so that individual sites can transmit their data to nearby sites, ending at two hubs, where a detection algorithm will operate in a PC. We are testing the equipment in Boulder this winter, and hope to deploy 3-4 sites in Italy next summer, with a full deployment the following year. In parallel we are working to improve our models.

Conclusions We have shown that there are strong correlations GPS signal power (also called SNR data) and volcanic plumes. SNR data are insensitive to water/water vapor, and thus have potential to contribute to ash detection activities at volcano observatories. SNR data are computed by a GPS receiver to evaluate the health of its tracking algorithms, and thus are readily available. We are evaluating both the precision and accuracy of SNR data so that we can place uncertainty bounds on our plume detections. We are also working to develop forward models for ash and other plume constituents so that we can convert SNR detections into parameters that would be of more value to this community. We encourage any country that operates a GPS network with open data policies to contact us if they are interested in testing their datasets, as we would be happy to share software with you.

Larson (2013) first detected the presence of a volcanic plume using SNR data. Shown above are SNR data collected by a geodetic-quality GPS receiver operated near Mt. Redoubt at the time of eruption 8 (Fee et al., 2013; McNutt et al., 2013; Schneider and Hoblitt, 2013). This instrument was deployed for the purpose of measuring ground deformation, and thus sampling was set to 30 seconds (Grapenthin et al., 2013). The particular station shown is located ~5 km to the west of the main vent of Mt. Redoubt. A total of five days of SNR data are shown – demonstrating that SNR behaves similarly each day except during event 8. The repeatability of the SNR levels are used to model the “direct signal.” The remaining SNR signal is shown below for events 8 and 19. Seismic duration is taken from Schneider and Hoblitt (2013).

30 35 40 45 50 55 60 6525

30

35

40

45Mar24Mar25Mar26Mar27Mar28

Eve

nt B

egin

s

Eve

nt E

ndsdB−H

z

SNR Data, Station RVBM, Satellite 21, Event 8

Elevation Angle (degrees)

Okmok An eruption began at Okmok Volcano on July 12, 2008. The University of Alaska, Fairbanks and the Alaska Volcano Observatory were operating two GPS receivers near the summit. The GPS receiver OKSO, directly to the south, was not observing any satellites that crossed the plume. The GPS receiver at OKFG, to the east, did not observe the eruption either, but ~hour later, a large disruption in SNR was observed (see below). SNR values remained low until July 17. We hypothesize that the flat GPS antenna was covered by ash, attenuating the signal. This ash was not removed until a rain storm on July 16. One can see in the orange trace (for July 17) that oscillations are now clearly present, indicating a planar reflecting surface below the antenna. These reflections were not observable before the eruption, presumably because the ground was not as smooth as it was after ash was deposited. The USGS reports ~10 cm of ash fell near OFKG.

Left: raw SNR data for a single satellite pass for a typical geodetic GPS site, plotted as a function of time; right: SNR data collected for five consecutive days near Mt. Redoubt, Alaska, plotted as a function of elevation angle. The day of the eruption is plotted in black. Elevation angle is the angle from the GPS antenna to the satellite with respect to the horizon.

17:00 17:10 17:20 17:30 17:40 17:50 18:00−10

−8

−6

−4

−2

0

2

Even

t Beg

ins

Even

t End

s

L2 SNR Data for Station RVBM

Event 8A.

Hours(UTC), March 26, 2009

dB−H

z

13:50 14:00 14:10 14:20 14:30 14:40 14:50−10

−8

−6

−4

−2

0

2Event 19C.

Even

t Beg

ins

Even

t End

s

Hours(UTC), April 4, 2009

dB−H

z

−10

−5

0

5

10

15

20

25

North

(km

)

GPS site2 dB−Hz (SNR change)confidence regiontrack formal error

Event 8

B.

−10 −5 0 5 10 15 20 25−10

−5

0

5

10

15

20

25

East (km)

North

(km

)

RVBM

RBED

DUMM

AC17

Event 19

D.

Mt. Redoubt SatelliteAzimuths

17:00 17:10 17:20 17:30 17:40 17:50 18:00−10

−8

−6

−4

−2

0

2

Even

t Beg

ins

Even

t End

s

L2 SNR Data for Station RVBM

Event 8A.

Hours(UTC), March 26, 2009

dB−H

z

13:50 14:00 14:10 14:20 14:30 14:40 14:50−10

−8

−6

−4

−2

0

2Event 19C.

Even

t Beg

ins

Even

t End

s

Hours(UTC), April 4, 2009

dB−H

z

−10

−5

0

5

10

15

20

25

North

(km

)

GPS site2 dB−Hz (SNR change)confidence regiontrack formal error

Event 8

B.

−10 −5 0 5 10 15 20 25−10

−5

0

5

10

15

20

25

East (km)

North

(km

)

RVBM

RBED

DUMM

AC17

Event 19

D.

Mt. Redoubt SatelliteAzimuths

13 23 33 43 53 63 7318 28 38 48 58 68 78dBZ

15km

10km

5km

15km

10km

5km

15km

10km

5km

15km

10km

5km

15km

10km

5km

20km 30km10kmNW SE 20km 30km10kmNW SE 20km 30km10kmNW SE 20km 30km10kmNW SE 20km 30km10kmNW SE

A

B

14:01:37 14:03:5714:02:47 14:05:07 14:06:17

14:13:17 14:15:3714:14:27 14:16:47 14:17:57

14:07:28 14:09:4814:08:38 14:10:58 14:12:07

14:19:08 14:21:2814:20:18 14:22:38 14:23:48

14:24:57 14:27:1714:26:07 14:28:28 14:29:38

Fig. 8. A) Reflectivity cross sections through eruption cloud from event 19 on April 4, 2009. Times in UTC for start of volume scan, which take 70 s to complete. Location of crosssection is shown in Fig. 9. B) Seismic record of explosive event 5 from temporary broadband seismometer RD03, located 5 km southwest of the vent.

7D.J. Schneider, R.P. Hoblitt / Journal of Volcanology and Geothermal Research xxx (2012) xxx–xxx

Please cite this article as: Schneider, D.J., Hoblitt, R.P., Doppler weather radar observations of the 2009 eruption of Redoubt Volcano, Alaska, Jour-nal of Volcanology and Geothermal Research (2012), http://dx.doi.org/10.1016/j.jvolgeores.2012.11.004

beam height Hwas calculated utilizing a standard atmospheric refrac-tion model as outlined by Rinehart (2004), using the following equa-tion:

H ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffir2 þ R2 þ 2rR sin∅

p−Rþ Ho ð1Þ

where r is the range from the radar to the point of interest, R is 4/3 the ra-dius of the earth,∅ is the elevation angle, andHo is the height of the radarabove sea level (30 m). The down-range distance from the radar to Re-doubt was 82 km, producing central beam heights from 2.9 to 17.7 kmabove sea level for the scan elevation angles shown in Fig. 3. The summitelevation of Redoubt is 3108 m, but the eruptive vent is at approximately2400 m and topography prevented direct observation of the vent by theradar beam. Note that the radar-derived eruption cloud height valuesgiven in this paper are for the central beam height (above sea level) andthe true valuesmay be asmuch as 2 kmhigher. This is due to uncertaintyin howmuch of the radar beam(2 kmhigh at the volcano) isfilled by vol-canic ash. If the beam is completelyfilled, the cloudheightwould be about1 kmgreater, however asmuch as 1/2 of the next higher beamcould con-tain volcanic ash without it being observed as a radar return.

2.2. Estimate of ash detection capability

Like many meteorological radars, the MM-250C transmits a pulseof energy, and then passively listens for the return of energy scatteredby particles. In a simplified form without signal loss, the intensity ofreceived power Pr can be calculated using the Probert–Jones equation(Probert-Jones, 1962):

Pr ¼Rc Kj j2z

r2ð2Þ

where Rc is the radar constant (includes the transmit power, antennagain, angular beamwidth, pulse length, and transmitted energy wave-length), K is the particle dielectric factor, z is the radar reflectivity fac-tor, and r is the range from the radar to the point of interest. ForRayleigh scattering conditions (particle diameter less than 5.35 mmfor the MM-250C) the radar reflectivity factor z is defined as:

z ¼ ∑NiD6i ð3Þ

where Ni is the number of particles of diameter Di per unit volume.Thus, the received power is highly dependent upon particle size.

Radar reflectivity factor (z) spans many orders of magnitude, so it istypically converted to logarithmic reflectivity factor (z) and expressedin units of dBZ by:

Z ¼ 10 log zð Þ: ð4Þ

The relationship between ash particle size, ash mass concentrationand logarithmic reflectivity factor (Z) was calculated and the results areshown in Fig. 4. In these calculations, the simplified target volume is as-sumed to be composed entirely of spherical ash particles with a particledensity of 2.0 g/cm3, and a dielectric factor of 0.39 (Adams et al., 1996)for volcanic ash. Note that the results are for a simplified target volumethat does not account for a variable particle size distribution within thevolume. The MM-250C has minimum detectable reflectivity of approxi-mately 20 dBZ at the range of 82 km (the distance from the radar siteto Redoubt), and this field is indicated by the shaded portion of theplot shown in Fig. 4. For volcanic cloud mass concentrations greaterthan 1 g/m3, only particles with a diameter greater than 0.6 mmwould be detectable. Very fine-grained ash (diameter of 0.01 mm)which can stay airborne for many days is likely not detectable by theMM-250C.

3. Results and discussion

3.1. Eruption column height

There were nineteen explosive events between March 23 and April4, 2009 that are summarized in Table 2. Of these, sixteen had radar-derived maximum eruption column heights in excess of 9 km abovesea level, the approximate altitude of the boundary between the tropo-sphere and the stratosphere at the time of the events. The MM-250Cwas able to detect the eruption column of the seventeen largest andhighest events, and the maximum column heights were similar tothose detected by the WSR-88D radar (Fig. 5). The WSR-88D was ableto detect two smaller explosive events, including the onset of explosiveactivity on March 23, 2009, due to its much higher power and greatersensitivity to low reflectivity eruption clouds. Eruption-cloud heightsgenerally increased through Event 8 at 17:26 UTC on March 26, afterwhich time the maximum cloud height was lower and relatively

0

5

10

15

20

25

020406080100

Alt

itu

de (

km

)

Range (km)

3.5

2.0

6.5

8.0

9.5

11.0

12.5

5.0

Fig. 3. Typical vertical scanning strategy used to produce sector volumes showing the re-lationship between scan angles and central beam height. Topographic profile shows thelocation of Redoubt Volcano at range of 82 km.

0.01

0.1

1

10

100

1000

0.01 0.1 1 10

Co

ncen

trati

on

(g

/m3)

Particle Diameter (mm)

-20 dBZ 0 dBZ 20 dBZ 40 dBZ

60 dBZ

Fig. 4. Relationship between particle diameter (mm), ash mass concentration (g/m3),and logrithmic radar reflectivity factor (dBZ). The shaded portion of the plot indicatesthe variation in particle diameter and mass concentration values that can be detectedby the MM-250C at a range of 82 km.



B.# Satellite#4#

Satellite#29##

C.#Radar#cross#sec4ons#for#event#5##

Lines#are#the#heights#of#the#approximate#GPS#SNR#detec4on#levels.# Schneider)and)Hobli/,#2013#

dispersing ash cloud. The cloud height decreased rapidly during thistime period, at a rate of about 10 m/s.

The seismic record from station RB03 shows an impulsive start tothis explosive event (Fig. 6B), followed by relatively uniform ampli-tude through about 12:40:30, when the amplitude increases. This in-crease in amplitude does not correspond to any noticeable increase indetected cloud height, and occurs as the reflectivity values near thevent begin to decline.

The PPI images at an altitude of 7.9 km (Fig. 7) show a roughlycircular eruption cloud characterized by a high reflectivity core thatdecreases towards the cloud edge. Movement of the cloud at this

altitude is primarily towards the north with lesser expansion to theeast and west. The reflectivity values at this altitude decrease rapidlyafter 12:43 UTC and are barely detectable by 12:53 UTC. The rapid de-crease in reflectivity at this altitude is likely due to the fallout of lapillitephra and a decrease in the cloud ash concentration.

All of the detected explosive events, with the exception of event 19,shared broadly similar reflectivity patterns and evolution. This includesrapid cloud rise, high central core reflectivity values (50–60 dBZ), andrapid decrease in cloud reflectivity and height once the eruptive eventended. The drifting clouds were typically only observable in theMM-250C data for tens of minutes following the eruption end, with

13 23 33 43 53 63 7318 28 38 48 58 68 78dBZ

15km

10km

5km

15km

10km

5km

15km

10km

5km

20km10kmS N 20km10kmS N 20km10kmS N 20km10kmS N 20km10kmS N

A 12:32:17 12:35:1212:33:46 12:36:43 12:38:12

12:39:42 12:42:4312:41:13 12:44:11 12:45:37

12:47:08 12:50:0812:48:38 12:51:37 12:53:07

B

Fig. 6. A) Reflectivity cross sections through eruption cloud from event 5 on March 23, 2009. Times in UTC for start of volume scan, which take 90 s to complete. Location ofcross-section is shown in Fig. 7. B) Seismic record of explosive event 5 from temporary broadband seismometer RD03, located 5 km southwest of the vent.



from local or distant stations and typically provide a range of values.Radar can provide a direct observation of reflecting particles in theeruption column and may be used to help constrain duration. Howev-er, radar reflectivity images show a combination of ash emissions inthe column as well as ash fallout which complicates the interpreta-tion. Eruption durations were estimated from the MM-250C data byobserving the timing of reflectivity values in excess of 40 dBZ at aheight of 6 km above the vent. The reflectivity threshold was chosento identify periods of high concentrations of large particles, and theheight and location was chosen to reduce the influence of ash fallout.Note that radar-derived eruption durations are dependent upon theinherent sensitivity of the radar system, and thus the chosen thresh-old is not universal. These estimates were compared to seismic(Power et al., this issue) and pressure-derived (S. McNutt personalcommunication, May 11, 2011) durations and the results are shownin Table 2 and Fig. 11.

In general, the duration estimates from the seismic and atmosphericpressure approaches are within a factor of 2 of the radar-derived esti-mates. Fig. 11A shows the comparison between the radar-derived dura-tion with the duration derived from seismic station SPU, located 85 kmnorth of Redoubt. At this distance, the effects of surface processes suchas pyroclastic flows and lahars on the seismic amplitude are negligible.A linear regression of these data yield anR2=0.64. Fig. 11B shows the re-lationship between pressure sensor-derived duration estimates fromstation DFR (located 15 km NE of Redoubt) and the radar-derived esti-mates. The agreement between these techniques is worse with anR2=0.36. The better fit with the seismic data suggests a slightly bettercoupling between the seismic energy recorded at a distant station andash emission than for pressure waves at a proximal station, but this isnot definitive. The variability in the various estimates of eruption dura-tion illustrates the uncertainty and difficulty in estimating eruptionsource parameters for ash dispersion, transport and fallout models.

3.5. Comparisons to satellite data

One of the surprising aspects of the 2009 eruption of Redoubt wasthe height of the eruption clouds as determined by the MM-250C andNEXRAD radars. These estimates are generally consistent with eachother, and document that as many as sixteen events produced volca-nic clouds that entered into the stratosphere. Satellite data provideadditional evidence for the cloud height as estimated by radar obser-vations. As volcanic clouds rise in the atmosphere, they equilibrate tothe ambient temperature. The standard technique for estimatingcloud height is to compare the cloud top temperature to a tempera-ture profile of the atmosphere. This works well for clouds emplacedin the troposphere, but it is more problematic for those volcanicclouds that have enough upward momentum to enter the strato-sphere. In those cases, the temperature inversion that exists at thetroposphere-stratosphere boundary results in multiple solutions forthe temperature comparison method.

Fig. 12A shows an AVHRR thermal infrared satellite image of thevolcanic cloud from explosive event 5, imaged at 13:36 UTC on 23March, 2009 (about one hour after eruption onset). At the time ofthis image, the volcanic cloud has become elongated along a NW-SEline in a sector to the north and east of the volcano. The central regionof the cloud is still opaque in the thermal infrared, as suggested by thefringe of warmer temperature values along the cloud edge; thus thetemperature values derived from this portion of the cloud are

0

2

4

6

8

10

12

14

16

18

20

0 4 8 12 16

Alti

tude

abo

ve s

ea le

vel (

km)

Time after eruption start (min)

519

60 m/s 25 m/s

Fig. 10. MM-250C radar-derived cloud heights versus time after eruption start for six-teen explosive events at Redoubt Volcano between March 23 and April 4, 2009. Datafrom explosive events 5 and 19 are labeled and indicated by the thicker lines. Velocitiesof 25 m/s and 60 m/s are shown by the dashed lines for reference.

0

10

20

30

40

50

60

70

80

10

20

30

40

50

60

70

80

0 5 10 15 20

Sei

smic

Dur

atio

n S

PU

(min

utes

)

Radar Duration (minutes)0 5 10 15 20

Radar Duration (minutes)

Pre

ssur

e D

urat

ion

DFR

(min

utes

)

A B

Fig. 11. Eruption duration estimates for sixteen explosive events at Redoubt Volcano between March 23 and April 4, 2009. A) MM-250C radar-derived duration versus seismic durationfromdistant station SPU located 85 kmnorth of Redoubt. B)MM-250C radar-derived duration versus pressure sensor duration from local station DFR located 15 kmnortheast of Redoubt.The solid line is for a 1:1 ratio between duration estimates.



Mt. Redoubt: Comparisons with Radar Shown below are comparisons between GPS SNR detections of plumes and radar observations for Mt Redoubt Events 5, 8, and 19. Note that each detection is based on a satellite that is at a different elevation angle, and thus a different plume height. For event 5, detections for two satellites are available; both are consistent with a velocity of ~25 m/s. More precise detections are somewhat limited because of the GPS sampling rate (one point every 30 seconds).

from local or distant stations and typically provide a range of values.Radar can provide a direct observation of reflecting particles in theeruption column and may be used to help constrain duration. Howev-er, radar reflectivity images show a combination of ash emissions inthe column as well as ash fallout which complicates the interpreta-tion. Eruption durations were estimated from the MM-250C data byobserving the timing of reflectivity values in excess of 40 dBZ at aheight of 6 km above the vent. The reflectivity threshold was chosento identify periods of high concentrations of large particles, and theheight and location was chosen to reduce the influence of ash fallout.Note that radar-derived eruption durations are dependent upon theinherent sensitivity of the radar system, and thus the chosen thresh-old is not universal. These estimates were compared to seismic(Power et al., this issue) and pressure-derived (S. McNutt personalcommunication, May 11, 2011) durations and the results are shownin Table 2 and Fig. 11.

In general, the duration estimates from the seismic and atmosphericpressure approaches are within a factor of 2 of the radar-derived esti-mates. Fig. 11A shows the comparison between the radar-derived dura-tion with the duration derived from seismic station SPU, located 85 kmnorth of Redoubt. At this distance, the effects of surface processes suchas pyroclastic flows and lahars on the seismic amplitude are negligible.A linear regression of these data yield anR2=0.64. Fig. 11B shows the re-lationship between pressure sensor-derived duration estimates fromstation DFR (located 15 km NE of Redoubt) and the radar-derived esti-mates. The agreement between these techniques is worse with anR2=0.36. The better fit with the seismic data suggests a slightly bettercoupling between the seismic energy recorded at a distant station andash emission than for pressure waves at a proximal station, but this isnot definitive. The variability in the various estimates of eruption dura-tion illustrates the uncertainty and difficulty in estimating eruptionsource parameters for ash dispersion, transport and fallout models.

3.5. Comparisons to satellite data

One of the surprising aspects of the 2009 eruption of Redoubt wasthe height of the eruption clouds as determined by the MM-250C andNEXRAD radars. These estimates are generally consistent with eachother, and document that as many as sixteen events produced volca-nic clouds that entered into the stratosphere. Satellite data provideadditional evidence for the cloud height as estimated by radar obser-vations. As volcanic clouds rise in the atmosphere, they equilibrate tothe ambient temperature. The standard technique for estimatingcloud height is to compare the cloud top temperature to a tempera-ture profile of the atmosphere. This works well for clouds emplacedin the troposphere, but it is more problematic for those volcanicclouds that have enough upward momentum to enter the strato-sphere. In those cases, the temperature inversion that exists at thetroposphere-stratosphere boundary results in multiple solutions forthe temperature comparison method.

Fig. 12A shows an AVHRR thermal infrared satellite image of thevolcanic cloud from explosive event 5, imaged at 13:36 UTC on 23March, 2009 (about one hour after eruption onset). At the time ofthis image, the volcanic cloud has become elongated along a NW-SEline in a sector to the north and east of the volcano. The central regionof the cloud is still opaque in the thermal infrared, as suggested by thefringe of warmer temperature values along the cloud edge; thus thetemperature values derived from this portion of the cloud are

0

2

4

6

8

10

12

14

16

18

20

0 4 8 12 16

Alti

tude

abo

ve s

ea le

vel (

km)

Time after eruption start (min)

519

60 m/s 25 m/s

Fig. 10. MM-250C radar-derived cloud heights versus time after eruption start for six-teen explosive events at Redoubt Volcano between March 23 and April 4, 2009. Datafrom explosive events 5 and 19 are labeled and indicated by the thicker lines. Velocitiesof 25 m/s and 60 m/s are shown by the dashed lines for reference.

0

10

20

30

40

50

60

70

80

10

20

30

40

50

60

70

80

0 5 10 15 20

Sei

smic

Dur

atio

n S

PU

(min

utes

)

Radar Duration (minutes)0 5 10 15 20

Radar Duration (minutes)

Pre

ssur

e D

urat

ion

DFR

(min

utes

)

A B

Fig. 11. Eruption duration estimates for sixteen explosive events at Redoubt Volcano between March 23 and April 4, 2009. A) MM-250C radar-derived duration versus seismic durationfromdistant station SPU located 85 kmnorth of Redoubt. B)MM-250C radar-derived duration versus pressure sensor duration from local station DFR located 15 kmnortheast of Redoubt.The solid line is for a 1:1 ratio between duration estimates.



A.#GPS,#event#8#

beam height Hwas calculated utilizing a standard atmospheric refrac-tion model as outlined by Rinehart (2004), using the following equa-tion:

H ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffir2 þ R2 þ 2rR sin∅

p−Rþ Ho ð1Þ

where r is the range from the radar to the point of interest, R is 4/3 the ra-dius of the earth,∅ is the elevation angle, andHo is the height of the radarabove sea level (30 m). The down-range distance from the radar to Re-doubt was 82 km, producing central beam heights from 2.9 to 17.7 kmabove sea level for the scan elevation angles shown in Fig. 3. The summitelevation of Redoubt is 3108 m, but the eruptive vent is at approximately2400 m and topography prevented direct observation of the vent by theradar beam. Note that the radar-derived eruption cloud height valuesgiven in this paper are for the central beam height (above sea level) andthe true valuesmay be asmuch as 2 kmhigher. This is due to uncertaintyin howmuch of the radar beam(2 kmhigh at the volcano) isfilled by vol-canic ash. If the beam is completelyfilled, the cloudheightwould be about1 kmgreater, however asmuch as 1/2 of the next higher beamcould con-tain volcanic ash without it being observed as a radar return.

2.2. Estimate of ash detection capability

Like many meteorological radars, the MM-250C transmits a pulseof energy, and then passively listens for the return of energy scatteredby particles. In a simplified form without signal loss, the intensity ofreceived power Pr can be calculated using the Probert–Jones equation(Probert-Jones, 1962):

Pr ¼Rc Kj j2z

r2ð2Þ

where Rc is the radar constant (includes the transmit power, antennagain, angular beamwidth, pulse length, and transmitted energy wave-length), K is the particle dielectric factor, z is the radar reflectivity fac-tor, and r is the range from the radar to the point of interest. ForRayleigh scattering conditions (particle diameter less than 5.35 mmfor the MM-250C) the radar reflectivity factor z is defined as:

z ¼ ∑NiD6i ð3Þ

where Ni is the number of particles of diameter Di per unit volume.Thus, the received power is highly dependent upon particle size.

Radar reflectivity factor (z) spans many orders of magnitude, so it istypically converted to logarithmic reflectivity factor (z) and expressedin units of dBZ by:

Z ¼ 10 log zð Þ: ð4Þ

The relationship between ash particle size, ash mass concentrationand logarithmic reflectivity factor (Z) was calculated and the results areshown in Fig. 4. In these calculations, the simplified target volume is as-sumed to be composed entirely of spherical ash particles with a particledensity of 2.0 g/cm3, and a dielectric factor of 0.39 (Adams et al., 1996)for volcanic ash. Note that the results are for a simplified target volumethat does not account for a variable particle size distribution within thevolume. The MM-250C has minimum detectable reflectivity of approxi-mately 20 dBZ at the range of 82 km (the distance from the radar siteto Redoubt), and this field is indicated by the shaded portion of theplot shown in Fig. 4. For volcanic cloud mass concentrations greaterthan 1 g/m3, only particles with a diameter greater than 0.6 mmwould be detectable. Very fine-grained ash (diameter of 0.01 mm)which can stay airborne for many days is likely not detectable by theMM-250C.

3. Results and discussion

3.1. Eruption column height

There were nineteen explosive events between March 23 and April4, 2009 that are summarized in Table 2. Of these, sixteen had radar-derived maximum eruption column heights in excess of 9 km abovesea level, the approximate altitude of the boundary between the tropo-sphere and the stratosphere at the time of the events. The MM-250Cwas able to detect the eruption column of the seventeen largest andhighest events, and the maximum column heights were similar tothose detected by the WSR-88D radar (Fig. 5). The WSR-88D was ableto detect two smaller explosive events, including the onset of explosiveactivity on March 23, 2009, due to its much higher power and greatersensitivity to low reflectivity eruption clouds. Eruption-cloud heightsgenerally increased through Event 8 at 17:26 UTC on March 26, afterwhich time the maximum cloud height was lower and relatively

0

5

10

15

20

25

020406080100

Alt

itu

de (

km

)

Range (km)

3.5

2.0

6.5

8.0

9.5

11.0

12.5

5.0

Fig. 3. Typical vertical scanning strategy used to produce sector volumes showing the re-lationship between scan angles and central beam height. Topographic profile shows thelocation of Redoubt Volcano at range of 82 km.

0.01

0.1

1

10

100

1000

0.01 0.1 1 10

Co

nc

en

tra

tio

n (

g/m

3)

Particle Diameter (mm)

-20 dBZ 0 dBZ 20 dBZ 40 dBZ

60 dBZ

Fig. 4. Relationship between particle diameter (mm), ash mass concentration (g/m3),and logrithmic radar reflectivity factor (dBZ). The shaded portion of the plot indicatesthe variation in particle diameter and mass concentration values that can be detectedby the MM-250C at a range of 82 km.



B.# GPS#event#8#

13 23 33 43 53 63 7318 28 38 48 58 68 78dBZ

15km

10km

5km

15km

10km

5km

15km

10km

5km

15km

10km

5km

15km

10km

5km


A

B

14:01:37 14:03:5714:02:47 14:05:07 14:06:17

14:13:17 14:15:3714:14:27 14:16:47 14:17:57

14:07:28 14:09:4814:08:38 14:10:58 14:12:07

14:19:08 14:21:2814:20:18 14:22:38 14:23:48

14:24:57 14:27:1714:26:07 14:28:28 14:29:38




13 23 33 43 53 63 7318 28 38 48 58 68 78dBZ

15km

10km

5km

15km

10km

5km

15km

10km

5km

15km

10km

5km

15km

10km

5km


A

B

14:01:37 14:03:5714:02:47 14:05:07 14:06:17

14:13:17 14:15:3714:14:27 14:16:47 14:17:57

14:07:28 14:09:4814:08:38 14:10:58 14:12:07

14:19:08 14:21:2814:20:18 14:22:38 14:23:48

14:24:57 14:27:1714:26:07 14:28:28 14:29:38




GPS#event#19#

C.#Radar#cross#sec8ons#for#event#19##

Lines#are#the#heights#of#the#approximate#GPS#SNR#detec8on#levels# Schneider)and)Hobli/,#2013#

Events 8 and 19 Event 5

20 30 40 50 60

20

30

40

elevation angle(deg)

OKFG

dB−H

z

20

30

4008Jul11

L2 SNR Data, Okmok Volcano

08Jul1208Jul1308Jul1408Jul17

OKSO

dB−H

z

SNR Data

Station ESLN

GPS SNR Data, Mt Redoubt

Top panel: SNR data for station EPDN; bottom panel: a periodogram of the SNR data. The frequency of the SNR data is converted to meters, the distance between the antenna and the dominant reflector. In principle this method can be used to measure ash deposits on the ground.

using gps signal strength data to detect volcanic … et al.pdf · using gps signal strength data...

Documents