Figure 1: Comparison between measurements of SO2 column amounts measured by ASTER and DOAS in the plume of Etna on 3rd August 2006.
V24B-01
What more have we learned from thermal infrared remote sensing of active volcanoes other than they are hot? (Invited)
Thermal infrared (TIR) remote sensing has been used for decades to detect changes in the heat output of active and reawakening volcanoes. The data from these thermally anomalous pixels are commonly used either as a monitoring tool or to calculate parameters such as effusion rate and eruptive style. First and second generation TIR data have been limited in the number of spectral channels and/or the spatial resolution. Two spectral channels with only one km spatial resolution has been the norm and therefore the number of science applications is limited to very large or very hot events. The one TIR channel of the Landsat ETM+ instrument improved the spatial resolution to 60 m, but it was not until the launch of ASTER in late 1999 that orbital TIR spectral resolution increased to five channels at 90 m per pixel. For the first time, the ability existed to capture multispectral emitted radiance from volcanic surfaces, which has allowed the extraction of emissivity as well as temperature. Over the past decade ASTER TIR emissivity data have been examined for a variety of volcanic processes including lava flow emplacement at Kilauea and Kluichevskoi, silicic lava dome composition at Sheveluch, Bezymianny and Mt. St. Helens, low temperature fumaroles emissions at Cerro Negro, and textural changes on the pyroclastic flow deposits at Merapi, Sheveluch and Bezymianny. Thermal-temporal changes at the 90 m scale are still an important monitoring tool for active volcanoes using ASTER TIR data. However, the ability to extract physical parameters such as micron-scale roughness and bulk mineralogy has added tremendously to the science derived from the TIR region. This new information has also presented complications such as the effects of sub-pixel thermal heterogeneities and amorphous glass on the emissivity spectra. If better understood, these complications can provide new insights into the physical state of the volcanic surfaces. Therefore, new data processing algorithms, laboratory, and field-based TIR instrumentation have been developed to more accurately model and correct these data. This presentation will summarize the results from nearly a decade of ASTER TIR remote sensing of active volcanoes around the globe. It will also document the first results of a micro furnace designed to capture emission of molten surfaces in real time as well as a field TIR camera modified to extract emissivity of surfaces at the cm pixel scale. The integration of laboratory, field, and orbital TIR remote sensing of active volcanoes provide a more complete picture of processes operating a variety of spatial, temporal and physical scales.
V24B-02
Reconstructing recent basaltic fissure eruptions in the Afar Depression, Ethiopia, using satellite imagery
Since 2005, there have been numerous dike intrusions in the Manda-Hararo rift segment in central Afar, Ethiopia, as part of an on-going rifting event similar to that which occurred at Krafla, Iceland between 1975 and 1984. Two of the dikes have been associated with basaltic fissure eruptions, in August 2007 and in June 2009. A large fissure eruption also occurred in the Erta ‘Ale volcanic range in northern Afar in November 2008, which may be related to the activity further south. These eruptions took place in remote areas and lasted only a few days, making field observation almost impossible. We must therefore rely on the geomorphology and geochemistry of the erupted deposits, and syn-eruption satellite observations of thermally emitted radiance and gas emissions to reconstruct the evolution of these eruptions. Reconstruction of the eruptions is important for i) understanding the relationship between dike intrusion and fissure eruption dynamics ii) understanding the relationship between eruption dynamics and the resulting deposits, iii) interpreting the deposits of pre-historic eruptions from previous rifting events, and iv) estimating the hazard posed to the local population by further activity. In this study we use pre- and post-eruption high resolution ASTER and ALI images to map the morphology of the erupted material, and SEVIRI and MODIS time series of the eruptions to track evolution of the vent dynamics and lava flows. These datasets are augmented by oblique aerial and ground based photographic and FLIR surveys of the cooling lava flows and tephra. ASTER and ALI images are also used to map older erupted materials, which allow us to compare this rifting event with previous events to anticipate future activity.
V24B-03
Observations of Volcanic SO2 and HCl from Aura MLS
The Microwave Limb Sounder (MLS) on board the Aura satellite has been taking composition measurements of the Earth's upper troposphere, stratosphere and mesosphere for the past 5 years. During this time period, MLS has observed volcanic emissions from Manam, Anatahan, Soufriere Hills, Okmok, Kasatochi, Redoubt,and Sarychev eruptions. The eruptions from these volcanoes injected SO2 and HCl into the lower stratosphere. MLS makes vertically resolved measurements of these gases and therefore can determine the injection height of these volcanoes. We will provide a survey of the eruptions MLS has observed to date and compare results to SO2 columns seen by the Ozone Monitoring Instrument (OMI), also on the Aura satellite. Aura MLS however, can only make measurements along its orbit track twice daily which limits its usefulness for hazards detection or determining the amount of injected SO2. The utility of these measurements for hazard detection will be greatly enhanced in the next generation MLS instrument envisioned for the third tier decadal survey Global Atmospheric Composition Mission (GACM). The future mission will provide 50 km^2 near global coverage with 4--6 observations per day.
V24B-04
Simultaneous use of various sensors for thermal monitoring of active volcanoes - an application of Kalman filtering
Thermal remote sensing is a valuable tool for monitoring active volcanoes. One can detect thermal anomalies originating from a volcano by comparing signals in mid and thermal infrared spectra. Once a thermal anomaly is detected, it has to be characterized in order to evaluate the activity status of volcano. The usual procedure is to compute temperature and area of the thermal anomaly using the dual band method proposed by Dozier (1981). More recent work, however, has shown that radiant flux (can be derived from the temperature and area or directly from measured radiances) is a more robust parameter for volcano characterization. Atmosphere, satellite viewing angle and sensor characteristics have a significant influence on the thermal anomaly characterization. Some of the influences are easy to correct using standard remote sensing preprocessing techniques, however, some noise still remains in data. In addition, satellites in polar orbits have long revisit times and thus they might fail to detect short volcanic events. It would therefore desirable to use data from different satellites in order to a) reduce uncertainties and b) improve temporal resolution. If one tries to simultaneously use data from different instruments, the measurements are often not comparable. Fortunately, there are statistical techniques that allow the combination of measurements from different sources that have different levels of accuracy. A suitable method for such a data fusion is Kalman filter. It is a recursive operator that returns the optimal estimation of a system state in a chosen time. Here we applied the Kalman filter to reduce noise and increase the temporal resolution of volcano radiant flux measurements from simultaneous MODIS and AVHRR measurements. As the evolution of the volcanic activity is difficult to predict we did not apply a physical model for the system state transition; we used a stochastic based model. A main challenge was to define the process noise covariance matrix and the relation between process and noise measurement. We finally decided to weight the process noise with the pixel area and cloud coverage over the volcano. We applied this technique to an eruption of Etna in 2002 and found good agreement with data of better resolution (DLR micro satellite BIRD).
V24B-05
Measurements of volcanic SO2 with ASTER. Comparison with automated scanning DOAS measurements
Sulfur dioxide (SO2) emitted by volcanoes has an important impact on the environment and climate and is also a critical parameter for volcano monitoring. A number of satellites operating in the ultra violet and in the Thermal infrared can measure SO2. However a lot of work has still to be done towards a rigorous validation of SO2 measurements from space. ASTER (Advanced Spaceborne Thermal Emission Reflection radiometer) acquires images in the thermal infrared (TIR) with a resolution of 90m/pixel, which enables to quantify the SO2 fluxes emitted in small-scale tropospheric plumes. ASTER images are processed with radiative transfer simulations and a band ratio algorithm to produce maps of SO2 column amounts. The band ratios (B10+B12)/B11 and B14/B11 are used for their insensitivity to variations in ground altitude and atmospheric humidity, two variables that often complicate SO2 retrievals in the TIR. Their sensitivity to surface emissivity is also reduced. So far, the ground validation of satellite SO2 measurements has been complex due to logistics difficulties and the lack of strictly simultaneous measurements. Recently the development of permanent networks of scanning DOAS on several active volcanoes has provide a wealth of ground based SO2 measurements that can be exploited for validating satellite-based measurements. We will present the results of comparisons between SO2 Column Amount (CA) and fluxes measured by ASTER and by the FLAME network of Mt. Etna. The two independent measurements sets are in good agreement in magnitude. Fluxes range from 2000 to 5000 T/days and column amounts from 0 to 4 g/m2. CAs measured by ASTER present a 0.5g/m2 random dispersion and no systematic bias compared to DOAS measurements. However the CAs measured by DOAS are subject to increase at low-scanning angles. These results constitute a rigorous ground validation of ASTER SO2, and provides valuable insights into accuracy and precision on both methodologies.
Figure 1: Comparison between measurements of SO2 column amounts measured by ASTER and DOAS in the plume of Etna on 3rd August 2006.
V24B-06
Monitoring volcanic thermal activity by Robust Satellite Techniques: achievements and perspectives
Satellite data have been increasingly used in last decades to study active volcanoes and to monitor thermal activity variation in space-time domain. Several satellite techniques and original methods have been developed and tested, devoted to hotspot detection and thermal monitoring. Among them, a multi-temporal approach, named RST (Robust Satellite Techniques), has shown high performances in detecting hotspots, with a low false positive rate under different observational and atmospheric conditions, providing also a potential toward low-level thermal anomalies which may announce incoming eruptions. As the RST scheme is intrinsically exportable on different geographic areas and satellite sensors, it has been applied and tested on a number of volcanoes and in different environmental conditions. This work presents major results and outcomes of studies carried out on Etna and Stromboli (Italy), Merapi (Java Indonesia), Asamayama (Japan), Jebel Al Tair (Yemen) by using different satellite systems and sensors (e.g. NOAA-AVHRR, EOS-MODIS, MSG-SEVIRI). Performances on hotspot detection, early warning and real-time monitoring, together with capabilities in possible thermal precursor identification, will be presented and discussed.
V24B-07
MODVOLC2: A Hybrid Time Series Analysis for Detecting Thermal Anomalies Applied to Thermal Infrared Satellite Data
We developed and tested a new, automated algorithm, MODVOLC2, which analyzes thermal infrared satellite time series data to detect and quantify the excess energy radiated from thermal anomalies such as active volcanoes, fires, and gas flares. MODVOLC2 combines two previously developed algorithms, a simple point operation algorithm (MODVOLC) and a more complex time series analysis (Robust AVHRR Techniques, or RAT) to overcome the limitations of using each approach alone. MODVOLC2 has four main steps: (1) it uses the original MODVOLC algorithm to process the satellite data on a pixel-by-pixel basis and remove thermal outliers, (2) it uses the remaining data to calculate reference and variability images for each calendar month, (3) it compares the original satellite data and any newly acquired data to the reference images normalized by their variability, and it detects pixels that fall outside the envelope of normal thermal behavior, (4) it adds any pixels detected by MODVOLC to those detected in the time series analysis. Using test sites at Anatahan and Kilauea volcanoes, we show that MODVOLC2 was able to detect ~15% more thermal anomalies than using MODVOLC alone, with very few, if any, known false detections. Using gas flares from the Cantarell oil field in the Gulf of Mexico, we show that MODVOLC2 provided results that were unattainable using a time series-only approach. Some thermal anomalies (e.g., Cantarell oil field flares) are so persistent that an additional, semi-automated 12-µm correction must be applied in order to correctly estimate both the number of anomalies and the total excess radiance being emitted by them. Although all available data should be included to make the best possible reference and variability images necessary for the MODVOLC2, we estimate that at least 80 images per calendar month are required to generate relatively good statistics from which to run MODVOLC2, a condition now globally met by a decade of MODIS observations. We also found that MODVOLC2 achieved good results on multiple sensors (MODIS and GOES), which provides confidence that MODVOLC2 can be run on future instruments regardless of their spatial and temporal resolutions. The improved performance of MODVOLC2 over MODVOLC makes possible the detection of lower temperature thermal anomalies that will be useful in improving our ability to document Earth’s volcanic eruptions as well as detect possible low temperature thermal precursors to larger eruptions.
V24B-08
Radiative transfer retrievals for accurate UV-spectroscopic measurements of volcanic SO2 emissions
There is widespread use of passive remote sensing techniques to quantify trace gas column densities in volcanic plumes utilizing scattered sunlight as a light source. Examples include passive DOAS (ground-based or from satellite platforms), COSPEC, and the SO2 camera. In order to calculate trace gas concentrations or volcanic emission fluxes, knowledge about the optical path of the measured radiation through the plume is necessary. For ground-based measurements, a straight photon path through the plume has been assumed in the past although it was known that this is not always true. Satellite retrievals apply a variety of techniques to estimate the air mass factor (AMF) for a particular measurement. Recently, model studies were conducted to quantify the effects of realistic radiative transfer in and around volcanic plumes on ground-based remote sensing measurements of SO2. These have shown that possible errors span more than an order of magnitude, and that both, over and underestimation of the true column density can occur. Actual errors depend on parameters such as distance between instrument and plume, plume SO2 concentration and aerosol load, as well as aerosol conditions in the ambient atmosphere. Here, we focus on retrieving radiative transfer parameters directly from UV-spectroscopic measurements of SO2 between 300 and 325 nm. Data analysis techniques are discussed that allow the retrieval of SO2 vertical column density and aerosol extinction coefficient inside the volcanic plume. A three-dimensional backward Monte Carlo radiative transfer model is used to simulate the measurement results that would be obtained for a certain physical state of the atmosphere and volcanic plume. Thus, a solution to the forward problem is obtained. By varying the state vector until the simulations match the remote sensing observations, this problem can be inverted. Sample ground-based measurements and retrievals are presented to demonstrate the power of the new techniques for satellite validation (ground truth) as well as to show the necessity of applying such corrections, even under seemingly ‘ideal’ measurement conditions. The discussed retrievals could greatly improve the accuracy of UV-spectroscopic satellite measurements of SO2 in volcanic plumes.