1. FY10-11 GIMPAP Project Proposal Title Page • Title: Automated Volcanic Ash Detection and Volcanic Cloud Height and Mass Loading Retrievals from the GOES Imager • Project Type: Development • Status: New - continuing work from GIMPAP FY08-09 project • Duration: 1 year • Leads: • Michael Pavolonis (NOAA/NESDIS/STAR) • Other Participants: • Justin Sieglaff (CIMSS), Andrew Parker (CIMSS), and Greg Gallina (SAB/VAAC)
2. Project Summary • Under a previously funded GIMPAP project (FY08-FY09), an automated volcanic ash detection and retrieval algorithm was developed for GOES. • While these algorithms generally work well, there are a few important scientific issues that need to be addressed prior to an operational implementation. • The 13.3 m version of the algorithm must be improved. Difficulties with this channel include larger FOV (GOES-12 and GOES-13) and SRF uncertainties which adds significant uncertainty to the atmospheric correction procedure. • The current ash detection false alarm rate (0.01%) must be improved to allow for highly reliable automated warning to be sent to ash forecasters. • Expected Result: An operational implementation of these algorithms will be sought.
3. Motivation/Justification • Supports NOAA Mission Goal(s): 1). Commerce and Transportation 2). Weather and Water • Project Justification • Volcanic ash is a major aviation, health, and infrastructure hazard, and as such, must be monitored from satellite with the highest temporal resolution available in order to issue timely warnings. (Note: The aviation community has requested a 5-minute warning capability.) • Given their temporal resolution, spatial resolution, and spatial coverage, the GOES imagers are vital for monitoring volcanic ash. • Currently, all volcanic ash products in NOAA operations are based on image enhancements and hence require manual analysis. Also, specialized automated volcanic cloud height and ash loading products do not exist in operations. • Since it is not possible for an analyst to look everywhere all the time for volcanic clouds, some eruptions are not detected in a timely manner (e.g. Anatahan Volcano, 2003), and some eruptions are mistaken for meteorological cloud (e.g. Santa Ana Volcano, 2005). Also, manual height estimates are subject to large errors and mass loading information is simply not available. • Thus, it is vital that automated and quantitative volcanic ash products be developed for the GOES Imagers.
4. Methodology • The retrieval of ash height, effective radius, and mass loading is performed using a two channel (11 and 12/13.3 m) optimal estimation technique on pixels where ash was detected. • The volcanic ash detection algorithm utilizes the 0.65, 3.9, 11, and 12/13.3 m channels during the day and the 3.9, 11, and 12/13.3 m channels at night. • Even with background correction, the spectral information offered by GOES is not sufficient to perform automated ash detection with operational quality skill. Advanced usage of spatial information is needed. • We propose to utilize cloud objects to better detect volcanic ash and to improve the first guess used in the retrieval procedure. • A cloud object is defined as a group of spatially adjacent pixels that meet a certain criteria. • Cloud objects allow for ash/no ash decisions to be made by examining a distribution of spectral information, rather than the spectral information at a single pixel, which can be noisy or ambiguous. • This methodology, along with clear sky radiance bias correction, should help mitigate issues associated with using the 13.3 m channel in lieu of 12 m.
BTD(11 - 12 m) Ash Ash Meteo. Cloud Reflectance Ratio (3.9/0.65 m) Meteorological Cloud Ash Meteo. Cloud False Color Image (0.65, 3.9, 11 m) • The spectral signature of some pixels in the ash cloud overlaps with the spectral signature of pixels in the meteorological cloud. • However, the distribution of pixels that compose the ash cloud has a unique negative BTD tail and the reflectance ratio distribution peaks at very large values, which allows it to be uniquely identified.
Improved Ash Detection With Cloud Objects Without Objects With Objects
5. Expected Outcomes • The improved products will be generated in real-time at CIMSS, similar to our current real-time GOES processing (http://cimss.ssec.wisc.edu/geocat/). • Statistical estimates of ash detection skill (e.g. POD and POF) along with ash retrieval accuracy and precision will be determined. • If successful (given the validation statistics), a GOES PSDI proposal will be submitted to implement these new volcanic ash products into operations.
6. FY10-11 Milestones • FY10 • Implement robust cloud object capabilities into the Geostationary Cloud Algorithm Testbed (GEOCAT). • Implement dynamic clear sky atmospheric transmittance bias correction into processing. • Test ash algorithms in real-time at CIMSS. • Determine ash detection skill score. • Validate retrieval results using CALIOP.