170 likes | 340 Views
Validation of the AMSU Snowfall Algorithm (over the US). Ralph Ferraro Matt Sapiano, Meredith Nichols, Huan Meng National Oceanic and Atmospheric Administration (NOAA) National Environmental Satellite, Data & Information Service (NESDIS) AND
E N D
Validation of the AMSU Snowfall Algorithm (over the US) Ralph Ferraro Matt Sapiano, Meredith Nichols, Huan Meng National Oceanic and Atmospheric Administration (NOAA) National Environmental Satellite, Data & Information Service (NESDIS) AND The Cooperative Institute for Climate and Satellites (CICS) College Park, Maryland
Objectives and Science Questions • NOAA has an operational precipitation product from AMSU • Includes a falling snow identification flag over land • Kongoli et al., 2003, GRL • Snowfall rates being developed by H. Meng • We know that it works in some cases and not in others • How do we quantify the accuracy of the detection algorithm? • Work was done in the algorithm development… • Under what meteorological conditions does it work and not work? • This is what we are really after! • Answer is crucial as we enter into the GPM era • Snowfall is important component of hydrological cycle • In some places, snowfall is the primary form of precipitation • What I plan on showing • Several attempts to validate • What we are hoping to accomplish (work in progress)
East Coast Snow/Ice Storm – 14 February 2007 NOAA-16 Precipitation Type/Rainfall Rate NEXRAD Reflectivity NOAA-16 Snowfall Rate • Corresponds with > 20 dBZ • Underestimates in heavy snow
February 5-6, 2010 Snow Event Courtesy H. Meng SFR (mm/hr)
February 5-6, 2010 Snow Event Courtesy H. Meng
Verification Issues • Hourly surface reports of snowfall are widely varying • “S-” can mean just about anything • Visibility, T-DP spread (RH) are better indicators of intensity • Hourly water equivalent are scarce and unreliable • ASOS, wind, etc. • Radar does not make rain/snow distinction w/o human interpretation • Virga, surface temperature are issues • Wide variety of conditions within MW satellite FOV • Previous work we have done show “lag” between surface and satellite signal • Snow fall slower than rain • Others
First Attempt – Climatology • We generated AMSU snowfall “climatology” • 7 years, 5 deg grids • NOAA-15 and -16 • Some assessments • Heaviest occurrences in “transition zones” • But values seem low • Large areas where retrievals don’t occur • Too cold and dry • Other features • E. Canada • Pacific NW/AK • Rocky Mountains • Himalayas SON DJF MAM
Comparison with Snow Cover Rutgers Snow Lab – Jan 2006 AMSU - Jan 2006
Verification • A. Dai (NCAR) • J. Climate, 2001; COADS climatology; 15,000 surface station reports • Can stratify by WMO weather codes • Grouping by all snow reports • Huge discrepancies. Why? • SW-, non accumulating snow • Filtered by S/SW, SW+/S+, temp/visiblity info from Rasmussen & Cole (2002) • Better qualitative agreement • Still not apples to apples comparison • AMSU 4 times/day; COADS, 24 times/day • Does imply that AMSU has skill in these type of events • Some recent work by Liu with CloudSat • Frequency of snow values comparable to these filtered data
AMSU (L) vs. COADS (R) SON DJF
Second Attempt – Daily Snow Reports • Are there denser surface networks that can be used? • Is there a better way to validate the ‘spatial’ patterns of the AMSU • Storm cases indicate ‘skill’; how best to quantify? • WMO/NCDC – “Global Surface Summary of the Day Data V7” • 9000 stations • Gives weather occurrences, max/min temp., precip. totals • Effort led by Matt Sapiano (now at CIRA) • 2000-08, N15, N16, N18 data
Nine Years of Comparisons • High resolution data still fairly sparse • 1 or 2.5 deg better for comparison • GSOD > AMSU just about everywhere • Probably due to very light snow • Let’s look closer…
Example – GSOD vs. AMSU • Extrapolate GOSD on 1 deg grid • Color coding • Green (Hit) • Blue (False on AMSU) • Red (Miss) • Could be rain • Gray (Hit – no snow) • Qualitative assessment • Overrunning snow – good • Upper low/backside snow • Missed • Attempted quantitative assessment at different locations • M. Nichols, HS student/intern • Different locations/regimes • Results inconclusive…time coincidence a limitation
Current Attempt – Hourly Snow Reports • Although we tried to avoid this, this is really the only way to go… • H. Mengs effort, HUGE data base of hourly synoptic reports colocated with AMSU for several years and all satellites. • Stratified data by location, weather conditions, precipitation totals, etc. • Still evaluating data….
Summary • Validation of AMSU snowfall algorithm is a difficult task • Algorithm has known limitations; we are trying to couple physical phenomenon to this • Temp./Moisture profiles, surface conditions, precip. Intensity, etc. • We know that the algorithm has ‘skill’, illustrating this has been a challenge. Why? • Incompatibility between satellite and ground data • More severe than in rainfall • Ground data fairly scarce, quality in question • Current method that should help answer is direct matches between satellite and surface reports • Emerging work with CloudSat (and GV) should also be pursued