During mid January to early February 2006, a series of explosive eruptions occurred at the Augustine volcanic island off the southern coast of Alaska. By early February a plume of volcanic ash was transported northward into the interior of Alaska. Satellite imagery and Puff volcanic ash transport model predictions confirm that the aerosol plume passed over a polarization lidar (0.694 μm wavelength) site at the Arctic Facility for Atmospheric Remote Sensing at the University of Alaska Fairbanks. For the first time, lidar linear depolarization ratios of 0.10–0.15 were measured in a fresh tropospheric volcanic plume, demonstrating that the nonspherical glass and mineral particles typical of volcanic eruptions generate strong laser depolarization. Thus, polarization lidars can identify the volcanic ash plumes that pose a threat to jet air traffic from the ground, aircraft, or potentially from Earth orbit.
 Volcanic eruptions emit steam, ash, and various gases into the atmosphere. Water vapor and sulfur dioxide injected into the stratosphere during particularly explosive events can over time produce global veils of sulfuric acid droplets that affect the Earth's climate [Bluth et al., 1997; Robock, 2000]. Volcanic ash, on the other hand, is composed of nonspherical mineral particles over a large range of sizes, the largest of which are likely to settle quickly, leading to fallout near the eruption site. This ash component in the troposphere poses a variety of hazards to humans and machinery on the ground, as well as damage to aircraft that fly inadvertently through ash plumes [Miller and Casadevall, 2000]. To mitigate such hazards, a combination of surface observations, satellite remote sensing, and aerosol dispersion model predictions are currently relied on. However, atmospheric clouds can obscure eruptions, and after initial transport and dispersion the volcanic plume will become too diffuse to be detectable using available satellite techniques [Dean et al., 2004].
 Beginning on 11 January 2006, the Augustine volcano (59.37°N, 153.42°W) began a series of episodic explosive eruptions that lasted until early February. During this period, the Alaska Volcano Observatory (AVO) was operationally monitoring the volcanic activity. AVO is a joint program of the United States Geological Survey (USGS), the Geophysical Institute (GI) of the University of Alaska Fairbanks, and the State of Alaska Division of Geological and Geophysical Surveys. Augustine Volcano is an active stratovolcano on an uninhabited island located in the Lower Cook Inlet, 275 km southwest of Anchorage, Alaska. With a summit height of 1.26 km above mean sea level (MSL), eruption clouds reached a maximum altitude of about 12 km MSL, sporadically disrupting aircraft traffic into the Anchorage International Airport. Figure 1 shows an aerial view obtained on 30 January 2006 of the volcanic island with a typical eruption column and spreading plume. The town of Homer, Alaska, 120 km to the northeast of the volcano, experienced volcano ash fallout as shown in Figure 2. This scanning electron microscope image depicts the nature of this basic type of volcanic aerosol: highly irregular particles with complex-surfaces composed of silica glass and crystals of feldspar.
 Serendipitously, on the afternoon of 2 February 2006, the Augustine ash plume was advected by southerly winds over the Arctic Facility for Atmospheric Remote Sensing (AFARS) at the University of Alaska Fairbanks (64.86°N, 212.16°W). There, an AFARS polarization lidar was activated in response to Puff dispersion model predictions to sample the aerosol. We show below that polarization lidar observations, with a known potential for studying aerosol types [Sassen, 2000, 2005], are valid for identifying volcanic ash plumes even after the dispersion/dilution of the ash cloud accompanying downwind transport.
2. Modeled Volcanic Plume Transport to AFARS
 The Puff model is designed to predict the extent and movement of airborne ash particles in an operational mode accurately and quickly [Searcy et al., 1998]. Model simulations place hypothetical particle size distributions above a selected volcano (i.e., Augustine, marked by the star symbol in southern Alaska in Figure 3) and use a modeled, gridded wind field (forecast or re-analysis) to predict particle transport. The initial distribution of the ash particles in the eruption column is assumed using a column shape option. A linear plume shape was used in this case. The particles are then tracked with a 4-D gridded wind field over time according to advection, Lagrangian diffusion, and Stoke's Law of settling [Papp et al., 2005].
 The Puff model was being used operationally by AVO over the period of the Augustine eruption. The forecast wind fields were taken from NAM216, which has a polar stereographic projection centered on Alaska on a 45 km × 45 km grid with 29 levels from 1000 to 100 mb. To make the most of the forecast data, the model was run for an initial 24-h period (2320 UTC 28 to 29 January 2006) using the forecast data, and then restarted at the last time period from the previous run for another 24-h, continuing until 2 February. The model forecasts predicted that from 3l January onwards, airborne ash would be traveling across interior Alaska towards the Fairbanks region at heights below 6.0 km MSL. The Puff model run valid for 2140 UTC coincident with the AFARS lidar measurements on 2 February is depicted in Figure 3. Because the eruption column height is unknown for the period approximately 24-h before the lidar observations were made (i.e., the time needed for aerosol transport to AFARS according to trajectory analysis), the model predictions in Figure 3 are shown for heights of 0 to 6.0 km MSL. Fairbanks (the red ‘F’ symbol) clearly lies under the edge of the main ash plume predicted by Puff at altitudes from about 2.0 to 4.0 km.
3. Satellite Imagery
Figure 4 presents a MODIS (Moderate resolution Imaging Spectrometer) true-color image from the Aqua polar-orbiting satellite obtained at 2210 UTC on 2 February 2006 as it passed over south-central Alaska. The Augustine volcanic island is situated at lower left in the Cook Inlet, and can be seen to emit a widening eruption cloud that curves first to the east and then to the north over the Gulf of Alaska (bottom center). Further north toward Fairbanks (designated by the yellow dot), a veil of clouds or aerosols, which blurs the surface image, is visible crossing the snow-covered Alaska Range and extending to at least the Fairbanks area. The trajectory of this volcanic plume appears to follow closely the initial path predicted by the Puff model (Figure 3).
 Examination of the MODIS standard data products from a number of pixels over the Fairbanks area reveals the difficulties involved in applying satellite multispectral radiance algorithms under conditions influenced by the effects of snow cover, varied terrain, and low sun elevation angle, as in this case. Although aerosol detection algorithms are not even applied under these conditions over land, MODIS cloud data products reveal mostly probably clear pixels.
4. AFARS Lidar Data
 The primary, turnkey remote sensor used at AFARS to sample the Augustine ash layer was the Cloud Polarization Lidar (CPL) [Sassen et al., 2001]. This system uses a high power (1.5 J), 0.1-Hz pulse repetition rate, ruby (0.694 μm wavelength) laser transmitter, and a 25-cm diameter telescope receiver with dual channels. Vertically polarized laser light is transmitted, and the backscattered light is divided into the orthogonal and parallel planes of polarization, from which the linear depolarization ratio (δ) is derived. In Figure 5 we show height versus time displays of CPL backscattering (based on a logarithmic gray scale) and linear depolarization ratios (note δ color scale) of the aerosol over AFARS on the afternoon of 2 February 2006, which brackets the Aqua satellite overpass at 2210 UTC image in Figure 4. (Note that the irregularities in the returned power display at top were caused by variable attenuation from ice fog plumes present just above the lidar in the region of little or no receiver/transmitter beam overlap.) In agreement with the Puff model predictions, the volcanic aerosol is present in the troposphere between about 2.0–4.0 km MSL (Figure 3).
 Maximum δ values in the elevated aerosol layer between 1.7 and 3.8 km MSL are between 0.15 and 0.20: the δ values averaged over the entire period range between 0.09 and 0.14. This magnitude of laser depolarization signifies lidar backscattering from nonspherical particles of a size at least comparable to the wavelength of the incident light [Mishchenko and Sassen, 1998]. This is in compliance with the shapes of the vesicular ash particles shown in Figure 2, although it should be recognized that the largest of these would have suffered fallout during the ∼24-h advection time to AFARS. Nonetheless, even the smallest of these volcanic particles are highly irregular in shape.
 The aerosol layer on this occasion was not readily apparent visually during the measurements, although a reddish glow around sunset was evident. Thus, because the aerosol was rather tenuous, the totalδ values (from the sum of molecular and aerosol backscattering) measured by the CPL have been influenced by the essentially non-depolarizing contributions from air molecules (with δ ≈ 0.025). In a first-order attempt to determine the influence molecular backscattering, we give in Figure 6 (left) the average range-normalized CPL relative backscattering, which is fitted with an appropriate pure molecular signal shown by the dashed curve [Sassen et al., 2001]. At right is shown the resulting δ profile when the assumed molecular signals are excluded in calculating laser aerosol depolarization. Although this is only an approximate solution affected by uncertainties in determining where the ‘pure molecular’ assumption is valid (assumed to be at 4.15 km MSL in this case), the δ values are not significantly elevated over the total ratios because of the relative aerosol backscattering strengths. The mean aerosol depolarization ratio is 0.18. Note that signals below about 1.0 km height are strongly influenced by the incomplete transmitter/receiver overlap function.
5. Discussion and Conclusions
 Previous lidar studies of volcanic aerosols are almost entirely limited to stratospheric measurements following major explosive eruptions in the 1980s and the 1991 Mt. Pinatubo event. The main finding has been that sulfuric acid droplets created during the stratospheric injection of sulfur dioxide, being spherical, do not depolarize laser light. One polarization lidar study implied the presence of lower stratospheric ash particles generating δ of ∼0.10–0.15 during the 2–5 month period following the Mexican El Chichón volcanic eruption in 1982 [Hayashida et al., 1984]. Similarly, airborne polarization lidar studies of the Pinatubo volcanic plume found several stratospheric scattering layers, one of which produced δ > 0.1 similar to those measured here in an ash plume [Winker and Osborn, 1992]. However, to our knowledge, only Raman lidar studies of a 2002 Mt. Etna, Italy, event yielded data from a fresh tropospheric eruptive plume, and no depolarization data were obtained in that case [Pappalardo et al., 2004].
 Thus, apparently for the first time, a tropospheric volcanic ash plume was sampled by a lidar with polarization diversity capable of assessing the degree of particle nonsphericity. In keeping with the nature of the ash particles sampled from the ground near the Augustine eruption site (Figure 2), the laser backscatter depolarization of 0.10–0.15 is significant for such aerosols. Only the aerosols from desert dust storms are capable of generating stronger aerosol depolarization with maximum δ values of 0.2–0.3 [Sassen, 2005]. Polarization lidars have the unique ability to detect this common type of volcanic aerosol after transport by virtue of their highly irregular shapes. This lidar technique also has the potential for identifying volcanic ash plumes from Earth orbit using the new generation of A-train satellites that includes a two-color polarization lidar [Winker et al., 2003].
 AFARS cloud and aerosol research is being supported by NASA grant NNG04GF35G. Puff dispersion modeling was supported by the U.S. Geological Survey as part of the Volcano Hazards Program, through the Alaska Volcano Observatory, a collaborative effort of the USGS, University of Alaska Fairbanks, and the Alaska Geological and Geophysical Surveys.