Download presentation
Presentation is loading. Please wait.
Published byVirgil Ford Modified over 9 years ago
1
Diagnosing Aircraft Icing Conditions Using Geostationary Satellite Data P. Minnis, W. L. Smith, L. Nguyen LaRC, Hampton, VA R. Palikonda, D. Spangenberg, S. Houser, C. Yost, M. Khaiyer SSAI, Hampton, VA F.-L. Chang NIA, Hampton, VA P. W. Heck CIMSS, Madison, WI NASA Applied Sciences Weather Program Review 18-19 November 2008
2
Aircraft Icing Aircraft structures act as ice nuclei in supercooled clouds - ice collects, weight increases, plane falls Pilots need to know where and when icing can occur - PIREPS are first order - sparse, aircraft dependent, location uncertain - weather forecasts - freezing levels, cloud expectations - radar => precipitation All combined in NCAR/FAA/NOAA/NASA program to provide Current Icing Potential (CIP) & Future Icing Potential (FIP) products to pilots - some inadequacies remain - NWP uncertainties, intensity, altitude of icing, etc.
3
Objectives of LaRC Satellite-Based Cloud & Icing Products Generate & provide, in near-real time, cloud and radiation products from satellites for a variety of applications, especially aircraft icing diagnosis & prediction, using NASA Earth Science data &/or algorithms - apply cloud code developed for CERES-MODIS to GEO & LEO sats - improve & update code - develop I/O system to provide results to users in a timely manner - validate to provide error bars for users Contribute to aircraft icing program - nowcast: satellite only algorithm, cloud products => icing diagnosis - quick, simple, useful for non-model domains, provides insight into diagnosis, not as good at night - CIP: merge cloud products directly into CIP decision process - still fast, merges all info together, all times of day (Haggerty) - CIP/FIP: assimilate cloud products into RUC used by CIP ( next talk + 1) - provides forecast of weather, all times of day, icing diagnosed - High IWC hazard (future)
4
Satellite Remote Sensing of Icing Conditions ICING CONDITIONS ARE DETERMINED BY CLOUD liquid water content, LWC positive w/ intensity temperature, T(z)negative w/ intensity droplet size distribution, N(r)r positive w/ intensity SATELLITE REMOTE SENSING CAN DETERMINE CLOUD optical depth, effective droplet size, re liquid water path, LWP cloud top temperature, Tc Thickness, h IN CERTAIN CIRCUMSTANCES
5
Current Products 0.65 µm Reflectance3.7 µm Temperature6.7 µm Temperature 10.8 µm Temperature 12 or 13.3-µm Temp1.6 µm Reflectance Skin Temperature Optical Depth Eff Radius/Diameter Liq/Ice Water PathCloud Eff TempCloud Top Pressure Cloud Eff Pressure Cloud Top Height Cloud Eff Height Cloud PhaseCloud Bot HeightCloud Mask Cloud Bot Pressure Icing PotentialBroadband SW Albedo Broadband LW FluxInfrared Emittance New products Surface Flux (Gridded) Multi Layer Cloud Mask & Layer Retrievals LaRC Products from Geostationary & Polar-Orbiting Satellites http://www-angler.larc.nasa.gov/satimage/products.html
6
2-D Cloud Retrieval Maps From GOES Have a 3rd Dimension: Cloud Thickness
7
Comparison of cloud base heights from GOES retrievals & ASOS ceilometer data 1900 UTC, 18 March 2004 overlap regime ztop too low or cloud too thick
8
Primary Domain: CONUS Corresponds to RUC & CIP domains Apply CERES-MODIS analysis code to generate the cloud products
9
Analysis Applied to Two Satellites to Cover USA GOES-11 RGBGOES-12 RGB 1815 UTC, 8 Jan 2008 Each image is analyzed and the results are combined
10
Combined GOES-11/12 Retrievals, 1815 UTC, 8 Jan 2008 RGB Phase rere LWP Light Blue - Supercooled
11
LWP = LWC * h re = f[N(r)] Tc & h can yield depth of freezing layer z t is top of icing layer ceiling = z t - h IN MANY CASES, SATELLITE REMOTE SENSING SHOULD PROVIDE ICING INFORMATION CLOUD PRODUCTS VS. ICING PARAMETERS
12
Dependence of Icing on LWP and r e Major dependence on LWP, minor on r e Formulation developed for icing potential Limited Sampling
13
Icing Potential from GOES Data Alone 1815 UTC, 8 Jan 2008 Indeterminate areas (white) This method has been adopted for use in Europe with Meteosat (Francis, 2007)
14
Upgrading Icing Nowcast Method Repeat analysis using greater dataset (2 seasons) - only ~1 month of data used for initial parameterization PIREPS ICING INTENSITY Nov-Mar, 2006/2007 and 2007/2008 Frequency (%) 2855 812 12 10803 1689 4920 163 21 N=21,131 MOGlight Satellite Categories: Most reports light or mod Unfiltered dataset Negative icing reports undersampled Probability of icing always greater than 0.9
15
Filtered dataset but with negative icing reports tripled to account for sampling bias
16
488 g/m 2 LWP threshold chosen to discern light from MOG icing. Winter 2006/2007 and 2007/2008 dataset (c) Mean Cloud Parameters (Filtered data) (d) Standard Deviations (Filtered data) Filtered dataset This method is the starting operational algorithm for the NOAA GOES-R project.
17
GOES Example Jan 11, 2008 (1745 UTC) Terr a This method is the starting operational algorithm for the NOAA GOES-R project.
18
Main Problem Areas Overlapped clouds - false icing - indeterminate pixels Snow - overestimate of optical depth => LWP false icing, too severe Nighttime, terminator - discrimination of optical depth for thick clouds difficult - phase more uncertain
19
Finding More Icing in Indeterminate Areas Multilayer Cloud Detection & Retrieval Some indeterminate pixels are overlapped ice-over-water clouds - multilayered cloud detection used to find those areas where icing is a problem (Detect with CO 2 method, Chang & Li, 2005) Use a multilayered VISST to derive low cloud properties Minnis et al. JGR 2007 2 methods developed using Aqua & Terra data Upper cloud Lower cloud Single cloud SL VISSTML VISST
20
Multilayered Cloud Detection Using 11 & 13.3 µm Channels Use sounding to predict 11 & 13.3 BTs Use radiative transfer to obtain solution satisfying both channels by adjusting background temperature and upper layer cloud optical depth - OD(UL) - T(UL) - T(LL) Use UL cloud OD with 2-layer VISST to determine OD of LL If OD(LL) > 6 and T(LL) ML icing Only applies to GOES-12 over CONUS (Meteosat, MODIS)
21
Multi-layered Cloud Detection, 13.3/10.8 µm,1815 UTC, 8 Jan 2008 Magenta areas indicate multilayer ice-over-water
22
Icing Potential 1815 UTC, 7 Jan 2008 RGBStandard retrievals + ML results Multilayer retrievals pick up additional areas with icing that were formerly indeterminate … some areas remain undetected
23
Icing Potential 1915 UTC 10 Jan 2008 RGBStandard retrievals + ML results Multilayer retrievals pick up additional areas with icing that were formerly indeterminate … some areas remain undetected
24
ICING-Non ML ICING-ML enhanced PIREP PIREP ------------------------ --------------------------- S YES NO S YES NO A YES 933 66 A YES 1070 81 T NO 114 17 T NO 124 17 ---------------------- --------------------- ------ pody=yy/(yy+ny)=PODy=89.1% PODy = 89.6% podn=nn/(yn+nn)=PODn=20.5% PODn = 17.3% Ntot=1130 Ntot=1292 Indeterminate GOES icing excluded Summary of PIREPS Comparisons GOES-12, Jan, Oct, Nov 2008 Identifies icing ~ 90% of the time Multilayer method as accurate as single-layer technique No matching of altitudes performed
25
Single-Layer GOES (G) PIREPS (P) ----------------------------------------- Icing (G), Icing (P): 56.4% No Icing (G), No Icing: (P) 0.9% Indeterminate (G), No Icing (P): 1.8% Indeterminate (G), Icing (P): 30.9% No Icing (G), Icing (P): 7.3% Icing (G), No Icing (P): 2.6% ---------------------------------------- Multi-layer enhanced ---------------------------------------- Icing (G), Icing (P): 66.8% No Icing (G), No Icing: (P) 0.9% Indeterminate (G), No Icing (P): 1.2% Indeterminate (G), Icing (P): 19.6% No Icing (G), Icing (P): 8.1% Icing (G), No Icing (P): 3.4% Total number of compared events: 1614 ---------------------------------------- PIREPS vs GOES icing detection returns: frequency of occurrence Increases area of knowledge - 25% more cloudy area included in base Cuts indeterminate area by ~40%
26
Limitations on Multilayer Detection Multilayer retrieval depends on having a 13.3-µm channel - GOES-12 only (on Meteosat) - expand G-12 CONUS domain (high angles in west)? - use full disk imagery (limited temporal coverage)
27
NASA LaRC GOES Icing Product Nov 8, 2008 (1745 UTC) Satellite analysis indicates large area of icing conditions extend from the Dakotas eastward to PA, NY and WV Impact of Snow on Icing Diagnoses
28
NCAR Icing Analysis with PIREPS Overlay Nov 8, 2008 (1900 UTC)
29
HVY Note: HVY icing report in Ohio. Numerous MDT reports. Good correspondence with satellite analysis PIREPS confirm significant icing Aviation Weather Center PIREPS Display Tool Nov 8, 2008
30
NASA LaRC Cloud Analysis Input Nov 8, 2008 (1745 UTC) GOES RGB NOAA Daily Snow-Ice Map Magenta areas are snow covered Clouds identified properly Icing areas correctly identified Icing severity overestimated, LWP overestimated
31
Addressing Snow Problem More spectral channels - MODIS, future GOES (1.6, 2.2 µm) Improve retrieval with existing channels - use snow map to define snow areas - estimate surface albedo - retrieve using both CO 2 method and VISST
32
Cloud Phase, 1745 UTC, 9 November 2008
33
Retrievals over Snow Background, GOES-12, 1745 UTC, Nov 9, 2008 Predicted clr-sky reflectance, with model snowPredicted clr-sky reflectance snow correction - using snow-corrected background information reduces cloud optical depth & water path LWP using no-snow backgroundLWP difference (snow – no snow)
34
Snow Corrections Procedure in place for using snow map & surface-type snow albedos - regional albedos may differ => no retrieval - exception handling (IR for thin, adjust albedo for thick clouds) Test against MODIS retrievals using 1.6/3.7 µm method Snow map from yesterday - examine use of RUC snow amounts to guide snow cover & albedo Note: This problem primarily affects LWP and, hence only icing severity. Icing is still well-detected over snow!
35
Other Refinements Nighttime retrieval improvements - Detection of clouds in overlap conditions - More optical depth information (IR limitations) - affects severity - Refine phase algorithm (tune w/MODIS 8.5-µm channel) - Multilayer detection enhancement - Improved terminator cloud mask - use of hour-to-hour memory Cloud height improvements - low: new variable lapse rate approach - high: new rough ice crystal models (reduce OD) Daytime: better snow discrimination (use snow maps) higher resolution background (RUC scale, ~13 km)
36
Domain Expansion Algorithms operating on all 5 GEOsats, AVHRR, MODIS - leverage from ARM, CERES, & SMD R&A - in-house calibration program Icing in remote areas still difficult to predict - nowcast product would be valuable Cloud input to larger domain applications - CONUS results do not cover RR domain - expanded CIP needs more data - NASA GEOS-5 needs global data Full-disk imagery being processed at low resolution - 9 km, 3 hr, long lag - no refinement of input, spectral responses - shoestring effort
37
Expanded Domains Full-disk Products online Analyses of POES datasets also available
38
Full-Disk Icing Potential, GOES-12 14 Mar 2008, 1745 UTC Most icing occurs poleward of 30° latitude
39
Icing Potential, 1800 UTC, 8 Nov 2008 GOES-11GOES-12 Meteosat-9 MTSAT
40
Global GEOSat Composite, 15 Z 9-9-08 to 21Z, 11-9-08 RGBRGB Significant problems with FY2-C
41
Global GEOSat Composite, 15 Z 9-9-08 to 21Z, 11-9-08 Cloud Top Height
42
Progress Toward “Operational” Status Operational = running 24/7 support with short lag time - NASA Columbia supercomputer Porting had been completed at beginning of year - new compiler on Columbia would not run our software environment (McIDAS) New computer analyst to make McIDAS run with new compiler - 2-3 months required for security approval - began work in earnest 2 weeks ago - testing of McIDAS modules underway - developing backup system at LaRC to ensure 24/7 - Shuttle operations take over Columbia Expect CONUS retrievals “operational” in mid-to-late January 09
43
Summary & Future Research Multilayer method implemented for GOES-12, Meteosat - increases area of potential icing detection - maintains current accuracy of detection - will enhance CIP & RUC assimilation - will be applied to more of CONUS (full disk) Analysis problems identified & potential solutions in the works - snow, night, cloud heights New domains available “Operations” - CONUS retrievals should be 24/7 on Columbia at end of January - backup systems to be developed at LaRC - alter algorithm as improvements are completed - full disk dependent on other proposals (MAP 08)
44
Improvement of Operational Aircraft Icing Forecasts and Diagnoses by Assimilation of Satellite Cloud/Surface Properties in the RUC/WRF P. Minnis, W. L. Smith, L. Nguyen LaRC, Hampton, VA S. Benjamin, S. Weygandt NOAA ERSL GSD NASA Applied Sciences Weather Program Review 18-19 November 2008
45
Objectives Use NASA satellite products in a Decision Support System for Air Safety
46
Approach Evaluation - Compare RUC and GOES WP & other parameters - iterate with new code & input Verification - Establish LARC data accuracy criteria - Compare LaRC data w/other input & independent measurements - iterate w/retrieval improvements Validation - Develop/test LWP/IWP assimilation code - iterate with revised code Documentation - Present results at conferences - Summarize results in peer-reviewed papers - Document code as needed
47
Tasks Retrieval code improvements (Minnis) - discussed earlier in icing presentation Benchmarking, validation, evaluation - Bill Smith Assimilation tests - Smith & Benjamin Future - Benjamin
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.