400 likes | 594 Views
Satellite Cloud Data for Model Verification and Assimilation. P. Minnis 1 , W. L. Smith, Jr. 1 , L. Nguyen 1 , R. Palikonda 2 , D. A. Spangenberg 2 , T. L. Chee 2 , K. M. Bedka 2 , F.-L. Chang 2 , J. K. Ayers 2 1 NASA Langley Research Center, Hampton, VA 23681 USA 2 SSAI, Hampton, VA 23666 USA.
E N D
Satellite Cloud Data for Model Verification and Assimilation P. Minnis1, W. L. Smith, Jr.1, L. Nguyen1, R. Palikonda2, D. A. Spangenberg2, T. L. Chee2, K. M. Bedka2, F.-L. Chang2, J. K. Ayers2 1NASA Langley Research Center, Hampton, VA 23681 USA 2SSAI, Hampton, VA 23666 USA
Background • For many aspects of hazardous weather, clouds are involved - if a model or nowcasting system has an accurate depiction of the cloud field, vertically and horizontally, then prediction or diagnosis of a given hazard should improve • Cloud property retrieval systems have been developed for climate research - if they are used in real time, they can provide parameters that can be used to numerically characterize cloud fields they could help the accuracy of forecasting and nowcasting
Objectives • Provide accurate cloud and radiation properties from meteorological satellites for use by the meteorological, climate, energy, transportation, and other communities, e.g., - aircraft icing conditions (Bill Smith presentation) - surface radiation budget - NWP assimilation - fog, ceiling, etc. - Research: model validation • Make them timely and widespread for real-time applications - nowcasting & assimilation (Stan Benjamin presentation) - near-global at high temporal resolution - 3-D characterization of clouds
Data • Available geostationary satellites provide up to 1-hour monitoring between ~60°S and ~60°N (currently 3 hourly) • Sun-synchronous satellites (POES) provide coverage over polar regions and gaps in geostationary coverage • AVHRR on NOAA/Envisat can also be used for POES coverage
Approach • Normalize all imagers to well-calibrated" reference POES imager - Terra & Aqua MODIS • System to ingest, analyze, and store satellite & auxiliary data - U Wisc McIDAS, NASA, & NOAA NESDIS for data - apply 4/5-channel algorithms used for CERES • System to disseminate derived products - digital files available online - online gif imagery & loops - North American domain to NCEP - Global domain to NASA GSFC GMAO Minnis et al., JTech, 2002, 2008 Minnis et al., TGRS, 2008, 2011 http://www-angler.larc.nasa.gov
NASA Langley Cloud Products Standard, Single-Layer VISST/SIST 0.65, 1.6 µm Reflectances 3.7, 6.7, 10.8 µm Temp 12 or 13.3 µm Temp Broadband Albedo Broadband OLR Clear-sky Skin Temperature Icing Potential** Pixel Lat, Lon Pixel SZA, VZA, RAZ Cloud Mask,Phase Optical Depth, IR emissivity Droplet effective Radius or ice crystal Diameter Liquid/Ice Water Path Effective Temp, height, pressure Top/ Bottom Pressure Top/ Bottom Height Multi-Layer, CIRT, CO2 channel only (GOES-12 & later) Multilayer ID (single or 2-layer) effective temperature optical depth, thickness effective particle size ice or liquid water path height, top/base height pressure Upper & lower cloud
Small Domain Cloud Products ARM SGP Domain 1815 UTC, 30 Nov 2010 • Transfer of cloud code for CERES first sponsored by ARM program in late 1990’s (climate-to-weather transfer) • Longest record is for SGP - use both GOES-E&W • Excellent testbed area - ARM surface sites - NEXRAD
Field Experiment Satellite Support Products Example: ARM STORM-VEx Domain 1815 UTC, 30 Nov 2010
CONUS Domain Cloud Products:1815 UTC, 30 Nov 2010 Icing-related parameters Light Blue - Supercooled Phase RGB LWP gm-2 re (µm) 10
CONUS Domain Cloud Products:1815 UTC, 30 Nov 2010 Other aviation-related parameters: top & bottom heights, skin temperature Skin temperature RGB Ztop (km) Zbot (km) 11
North American Domain 1915 UTC, 1 December 2010 Pseudocolor RGB Scene ID • This combination of GOES-E&W data matches the Rapid Refresh domain • These data are ported to NCEP and available operationally (~ 30 min lag)
Global GEOSat Retrievals, 18 UTC, 30 Nov 2010 Cloud-top Height (km) • Results averaged on 0.25° x 0.3125° grid for GMAO • Nighttime information is limited because of lack of visible channel - e.g., optical depth • FY-2D lacks sufficient quality, requiring use of polar satellites • Results currently produced every 3 hours, 1 hour possible Clear Area Surface Skin Temperature (K)
Error Assessments of Cloud Products Cloud Top Height Single-layer cloud-top heights over ARM SGP, 2000-2004 Errors in GOES Single-layer Ztop, ARM SGP, 2000-04 All clouds • Low cloud-top height rms error only 0.8 km Smith et al., GRL, 2008
Cloud Thickness from CloudSat, CALIPSO & GOES CALIPSO/CloudSat ice cloud thickness (DZ_OBS) vs GOES thickness, April 2007 Cloud boundaries from GOES-12 over ARM SGP radar, 10 June 2009 • Triangle – physical top • diamond – effective top • square - base DZ_FIT based on empirical fit to CALIPSO/CloudSat data from different month
Cloud Base Heights Cloud thickness estimated as function of optical depth, temperature, water-path, Reff & phase: cloud base height = top - thickness Histogram, GOES-ASOS VISST Cloud Base vs. National Weather Service Automated Surface Observing System (ASOS) Central USA: June-July 2009 • Base height error only slightly greater than top error for water clouds • Base height error large for ice clouds (precip clouds included)
Comparison of ARM SGP radar-radiometer, in situ, & GOES Cloud Retrievals March 2000 Overcast stratus clouds only • GOES compares well with both in situ and surface-based MWR LWP • Effective radius, reff, too large, but expected because it is for the cloud top Dong et al., JAS, 2002
Comparison of ARM SGP radar-radiometer and GOES Cloud Retrievals March 2000 • GOES reff is overestimated (based on 3.9 µm) • GOES optical depth slightly low • GOES LWP a little high (4%) • SD(reff) = 38% • SD(OD) = 40% • SD(LWP) = 32% Dong et al., JAS, 2002
LWP (MODIS) Liquid Water Path Validation with Microwave Radiometer Data ARM SGP vs CERES MODIS, 2001 – 2004, Stratus • Unbiased over wide range of LWP (up to 500 g/m2) • Excellent correlation • Instantaneous Uncertainty ~30% • Note mean value ~150 g/m2 Dong et al., JGR, 2008
Ice Cloud Water Path (IWP) Validation with CloudSat GOES Cloud Water Path Comparisons with Cloudsat CWC-RO Dec 2006 – May 2007 • Excellent agreement between GOES & Cloudsat monthly mean total water path (TWP) for high thick ice clouds - GOES too high for snow scenes • Thin ice cloud IWP agrees well - large SD, matching? Summary for all months
Cloud Water Path Comparison July 11, 2010 (20 UTC) GOES RUC-OPS RUC-DEV RUC-OPS: • Operational RUC-13 at NCEP • Assimilates NESDIS CTP • ΔPcld = 40 mb in cloud building RUC-DEV: • Development RUC-13 at ESRL • Assimilates LaRC CTP, CWP • ΔPcld = adjusted in cloud building • based on LaRC CWP …not the only differences
Cloud Water Path Comparison April-June 2008, 2009 Domain Averages (gm-2) GOES Cloud Water Path over RUC Domain CloudSat Tracks Matched along CloudSat track (gm-2) • Model and observations in the same ballpark on average
North American Domain 1845 UTC, 21 March 2010 Cloud-top Heights (Kft, AGL) GOES Rapid Refresh Cloud data assimilated into NOAA Rapid Refresh NWP produce more realistic clouds
Rapid Refresh 1-hr Forecast vs. GOES Clouds 1915 UTC, 1 December 2010 Cloud-top Height (Kft, AGL) RR LaRC GOES • Heights generally compare well over continental North America • Some differences due to model, e.g., Pacific tropical convection • Others due to GOES analysis, e.g., NE Ontario twilight 24
Improving RUC Forecasts With LaRC Products IFR 1000’ ceiling - 1h forecasts - PODy - November 2008 Dev13 – RUC w/ NASA cloud RUC - NESDIS cloud % Positive Detection of 1 Kft Ceiling • Dev13 forecasts correctly predict ceilings below 1000’ ~9% more often than operational RUC 25
Assimilation – Aircraft Icing LaRC Satellite CONUS: FAA-NCAR Current Icing Potential (CIP) Product Radar RUC/RR Model Pilot Reports Surface Obs Lightning NCEP Match data to each 3-D model grid box ICING=0.0 SLD=0.0 Cloudy? NO YES Determine Vertical Cloud Structure and Weather Scenario Apply interest maps. Calculate icing probability and potential for supercooled large drop (SLD). ICING PROBABILITY and SLD POTENTIAL FIELDS 26
Developing 3-D Cloud Fields • Clouds typically interpreted as being a single uniform layer - vary vertically in LWC/IWC, T, re - finite thicknesses - often occur in separate layers • One challenge is how to distribute cloud water vertically - IWC/IWC affect precipitation - LWC, Temperature affect icing - need estimates of “safe” levels within clouds • Other is interpreting the clouds as being single or multilayered - need estimates of cloud properties of both layers - no hope for retrieving more than one layer
Method for Profiling Ice/Liq Water Contents • Develop avg normalized profiles of IWC/LWC from GOES-CloudSat data - Use variety of categories based on cloud water path CWP, Teff, DZ T3, DZ 2-4 Altitude Factor • Reverse process provides profile for a given pixel Mean Shape Factor RMS Error, %
Testing with CloudSat, 20 UTC, 7 Oct. 2009 GOES Cloud Water Path, CWP CloudSat Track
Following water content profiles for selected fixed transect 7 October, 2009 1845 – 2245 UTC
Testing with CloudSat, 19 UTC, 6 May 2008 GOES Cloud Water Path RUC Cloud Water Path CloudSat Overpasses • No LaRC cloud data assimilated in RUC
Multilayer Clouds Over Southern Great Plains GOES-12, 25 Nov 2010, 1845 UTC Multilayer Cloud Heights (km) over ARM SGP Domain Ztop (km) interpreted as SL cloud RGB Multilayer ID Lower-layer Ztop (km)
North American Domain 1515 UTC, 2 December 2010 Multilayer Cloud Detection Pseudocolor RGB Multi-layer Cloud ID Detecting & retrieving multilayered clouds improves characterization of 3-D fields 34
Covering the Poles Using New Spectra, 11 UTC, 6 July 2006 Terra MODIS Retrievals: Arctic Ocean, Svalbard RGB Phase opt depth droplet reff(µm)Zeff (km)
Summary and Conclusions • Cloud properties are valuable on their own for nowcasting - can be used to detect potential icing conditions - see W. Smith talk - provide approximate 3-D view of clouds - estimates of cloud base heights (ceiling) - most useful in remote areas with no other information - available for all GEOSats, soon in polar regions • Cloud properties more valuable for assimilation & verification - NOAA Rapid Update Cycle (RUC & Rapid Refresh) NWP model used for aviation - NCAR/FAA Current/Future Icing Potential (CIP/FIP) Products - NASA GMAO GEOS-5 • Research begun with CERES & ARM programs, developed under NASA ASAP & ROSES, future?
Future Work • Improve nighttime retrievals - test theoretical optical depth limits for various emission channels • Develop new techniques to deal with terminator & snow backgrounds - CO2 / IR / SIR combinations? regional memory? • Enhance thickness & WC profiles using different categories & other spectral channels (e.g., 1.6, 2.1 µm) on other satellites • Continue improvement of CO2 and other ML techniques • Work on improving retrievals over snow surfaces • Replace FY-2D for improved Asian coverage - MODIS & AVHRR, FY-2E, Korean COMS? • Continue working with users to meet their needs For references: -http://www-pm.larc.nasa.gov/ data: http://www-angler.larc.nasa.gov/
Data Product Formats • Pixel-level Product:Binary data retrieved at instrument nominal resolution. • Gridded Product: 0.5º or 1.0º average separated by cloud height / phase. - includes surface radiative fluxes & skin temperature • Surface Site, Aircraft, & CALIPSO/ICESat Matched Products Means within 10 or 20-km radius centered of selected (ARM & Euro) surface sites; single or weighted 4 pixels average along aircraft/satellite flight path. • Current lag times - 0.5 to 1.5 hrs depending on domain - Meteosat results under 24-h delay per ESA • Spatial/temporal resolution - GOES domains, 0.5 hr; others 1.0 hr, nominal 3-5 or sampled 6-8 km - Full disk, 3 hr, 9-10 km sampled