Combining Satellite & Model Information for Snowfall Retrieval Yoo-Jeong Noh Cooperative Institute for Research in the Atmosphere Colorado State University CIRA/Colorado State University 5th Annual CoRP Science Symposium, Aug.12-13, 2008
Microwave radiation Rain vs. Snow; Emission vs. Scattering Low MW freq. (< 89 GHz) Emission High MW freq. (> 89 GHz) Scattering Rain Radiation * * * * * * * * * * * * * Radiation Snow Surface
Challenges The snowfall signals are not strong enough in lowfreq. microwave observations. High-freq. observations (>89 GHz) have not been available until recently and still quite limited (e.g., AMSU-B). The nonsphericity of the snow/ice particles, the scattering by snowflakes has not been fully understood - often sphere assumptions used. Complex land surface conditions
Objective Develop a Bayesian snowfall retrieval algorithm utilizing high-frequency satellite microwave measurements over the Great Lakes region
Data AMSU-A/B (Advance Microwave Sounding Unit-B) NEXRAD Level II data (from NCDC) AGRMET data Land surface model simulations at the Air Force Weather Agency Input for MEM (NOAA MW land Emissivity Model) run CloudSat products 2B-GEOPROF, 2B-CLDCLASS, 2B-CWC-RO Aircraft observations during the C3VP/CLEX10 Convair-580 measurements during the Canadian CloudSat/CALIPSO Validation Project and the 10 th Cloud Layer Experiment WRF model (ARW V2.2) Focused area The Great Lakes and surrounding areas Time period Oct. 2006-Feb. 2007
A Bayesian retrieval algorithm Bayesian theorem effectively combine atmospheric properties with radiative transfer simulations from a huge database The most important step is constructing a large database of T B s and snowfall profiles using radar observations and the radiative transfer model.
Algorithm Strategy Snowfall profiles CloudSat and NEXRAD radar Cloud Liquid water information CloudSat and C3VP/CLEX10 Sounding information C3VP/CLEX10 and WRF simulations Radiative Transfer Modeling Radiative properties of snowflakes Discrete Dipole Approximation Satellite Data (AMSU-B) a-priori database (Snowfall vs. T B ) Bayesian Retrieval Procedure Snowfall rate (profiles)
Radiative Transfer modeling Collect input data such as snowfall data from satellite, aircraft, and surface radars. Perform the DDA (Discrete-Dipole Approx.) simulations to accurately calculate the scattering parameters of snowflakes: Lookup table Radiative transfer modeling to generate T B s. Various data as input, which are from satellite/aircraft measurements and model results to diversify the database.
Two types of snowflakes in the DDA computation The DDA (Discrete-Dipole Approximation ) is a general method for computing the scattering and absorption of arbitrarily shaped particles. - divide a snowflake into a finite array of dipoles and calculate the combined scattering electric field Sector snowflake Dendrite Snowflake
Microwave Radiative Transfer Model MWRT: a radiative transfer model developed by Dr. Liu (Liu 1998 and 2004) Discrete Ordinate Method, 32 Streams Fresnel ocean surface boundary 5 types of hydrometeors: raindrop, cloud liquid drop, cloud ice crystal, snowflake, and graupel The single-scattering properties by assuming sector-like and dendrite-like particles with random orientations CIRA/Colorado State University 5th Annual CoRP Science Symposium, Aug.12-13, 2008
Ze-S relationships for snow Ze (mm 6 m -3 ) 10 4 10 3 10 2 10 1 10 0 10-1 10-2 0.1 0.2 0.4 1.0 2.0 4.0 Precipitation (mm/h) Sector snowflake-13.4ghz Sector snowflake-35.6ghz Sector snowflake-94.0ghz Dendrite-13.4GHz Dendrite-35.6GHz Dendrite-94.0GHz Aonashi et al. (2003) Puhakka (1975) Sekhon & Srivastaya (1970) Imai (1960) Ohtake & Henmi (1970) Carlson & Marshall (1972) Fujiyoshi et al. (1990) Boucher & Wieler (1985) Ze-S relationships for snow from the DDA calculation, compared with several previous studies 1.083 Z e = 250S Z e = 88.97S Z e = 38.06S 1.04 1.057 at 13.4 GHz at 35.6 GHz at 94.0 GHz
Algorithm Strategy Snowfall profiles CloudSat and NEXRAD radar Cloud Liquid water information CloudSat and C3VP/CLEX10 Sounding information C3VP/CLEX10 and WRF simulations Radiative Transfer Modeling Radiative properties of snowflakes Discrete Dipole Approximation Satellite Data (AMSU-B) a-priori database (Snowfall vs. T B ) Bayesian Retrieval Procedure Snowfall rate (profiles)
The a-priori database Constructed through radiative transfer modeling with combinations of snowfall profiles from the NEXRAD, CloudSat, airborne radar data analysis atmospheric sounding profiles from WRF simulations and C3VP/CLEX10 measurements Surface emissivity from MEM (NOAA Land emissivity model) using AGRMET and AMSU A/B CIRA/Colorado State University 5th Annual CoRP Science Symposium, Aug.12-13, 2008
CloudSat Data - To obtain snowfall profiles and liquid water information. - CloudSat products: 2B-CWC-RO, 2B-GEOPROF - Ze-S relationship (Ze: mm 6 m -3, S: mm/hr) for 94 GHz 1.04 CPR from DDA calculation, Z e = 88.97 S 22 Jan 2007 07:30 UTC Reflectivity LWC and IWC from CloudSat 2B- CWC-RO data (intervals: 0.1 g/m 3 )
Snowfall profiles from NEXRAD data Interpolated by using NCAR SPRINT & CEDRIC utilities Ze-S relationship by Super and Holroyd (1998) obtained from snowfall cases in KCLE, OH. Z e =180.0S 2.0
Aircraft measurements during C3VP/CLEX10 - CLEX10: the 10th Cloud Layer Experiments - C3VP: the Canadian CloudSat/CALIPSO Validation Project - 31 Oct 2006-1 Mar 2007, Southern Ontario, Canada - Obtain additional snowfall data from 35GHz airborne radar (ongoing work) ex) 31 Oct 2006
Aircraft measurements during C3VP/CLEX10 - Provide liquid water profiles and sounding information as input of the radiative transfer model 02 Dec 2006 1820-1844UTC 22 Jan 2007 0600-0630UTC Temp. LWC & IWC Temp. LWC & IWC
WRF simulations to get more sounding information for RTM dx = dy = 9 km 162 x 162 x 31 (one example of grid sizes) Time step (dt) = 60 seconds Thompson s cloud microphysics RRTM long-wave radiation scheme Dudhia short-wave scheme Noah land-surface model YSU PBL scheme Forecast frequency and length: 24~48 hr forecasts Initialization & Boundary conditions : GFS model analysis Ten snow cases simulations
WRF 1hr-accum precip & NEXRAD (dbz) 03 Nov 2006 22 Jan 2007
Microwave Surface Emissivity Land emissivity is highly variable and complex making it more problematic to use microwave data over land. Employ the NOAA Microwave Land Emissivity Model (MEM) using AMSU A/B and AGRMET (by US Air Force). (e.g.) 1100UTC 01/22/2007 CIRA/Colorado State University 5th Annual CoRP Science Symposium, Aug.12-13, 2008
Algorithm Strategy Snowfall profiles CloudSat and NEXRAD radar Cloud Liquid water information CloudSat and C3VP/CLEX10 Sounding information C3VP/CLEX10 and WRF simulations Radiative Transfer Modeling Radiative properties of snowflakes Discrete Dipole Approximation Satellite Data (AMSU-B) a-priori database (Snowfall vs. T B ) Bayesian Retrieval Procedure Snowfall rate (profiles)
The retrieval algorithm using Bayesian Theorem Input Satellite-observed T B s Search through snowfall profiles and simulated T B s from the a-priori database Compute probability that each profile fits the observations Output: retrieved snowfall profiles Weight snowfall profiles by computed probability and add to sum E( x) = T 1 exp{ 0.5[ y 0 y ( x )] ( O + S) [ y y ( x ˆ s j x 0 j j The normalization factor, Â s j )]} A ˆ T 1 = exp{ 0.5[ y y ( x )] ( O + S) [ y y ( )]} j 0 s j 0 s x j
AMSU-B retrieval The retrieval algorithm is applied to AMSU-B. Snow scattering signatures: T B depressions (ΔT B s= T B -T B_ ) are calculated from the bg background T B s obtained by analyzing the histograms of AMSU-B data from November 2006 to February 2007. Most frequently occurred T B at each frequency for every surface type and every local zenith angle range during this period is used as a standard background T B.
Histogram analysis to find background brightness temperatures at each frequency using four month AMSU-B data from Nov. 2006 to Feb. 2007 Brightness temperature ( o C) 280 260 240 220 89GHz Water Land Coast Scan angle (degree) Brightness temperature ( o C) 280 260 240 220 Latitude < 42 o N 150GHz Scan angle (degree) Brightness temperature ( o C) 280 260 240 220 183+1GHz 200 200 200 0 10 20 30 40 50 60 0 10 20 30 40 50 60 0 10 20 30 40 50 60 Scan angle (degree) Brightness temperature ( o C) 280 260 240 220 183+3GHz Scan angle (degree) Brightness temperature ( o C) 280 260 240 220 183+7GHz 200 200 0 10 20 30 40 50 60 0 10 20 30 40 50 60 Scan angle (degree) Brightness temperature ( o C) 280 260 240 220 89GHz Water Land Coast Scan angle (degree) Brightness temperature ( o C) 300 280 260 240 220 42 o N < Latitude < 45 o N 150GHz Scan angle (degree) Brightness temperature ( o C) 280 260 240 220 183+1GHz Scan angle (degree) Brightness temperature ( o C) 280 260 240 220 183+3GHz Scan angle (degree) Brightness temperature ( o C) 280 260 240 220 183+7GHz 200 200 200 200 200 0 10 20 30 40 50 60 0 10 20 30 40 50 60 0 10 20 30 40 50 60 0 10 20 30 40 50 60 0 10 20 30 40 50 60 Scan angle (degree) Water Land Coast Latitude > 45 o N Brightness temperature ( o C) 280 260 240 220 89GHz Brightness temperature ( o C) 280 260 240 220 150GHz Brightness temperature ( o C) 280 260 240 220 183+1GHz Brightness temperature ( o C) 280 260 240 220 183+3GHz Brightness temperature ( o C) 280 260 240 220 183+7GHz 200 200 200 200 200 0 10 20 30 40 50 60 0 10 20 30 40 50 60 0 10 20 30 40 50 60 0 10 20 30 40 50 60 0 10 20 30 40 50 60 Scan angle (degree) Scan angle (degree) Scan angle (degree) Scan angle (degree) Scan angle (degree) CIRA/Colorado State University 5th Annual CoRP Science Symposium, Aug.12-13, 2008
Preliminary retrieval results AMSU-B (ΔTBs) 1100UTC 12 Oct 2006 Retrieved snowfall (mm/hr) at 1.5km NEXRAD reflectivity
Summary A snowfall retrieval algorithm over land based on Bayes theorem is developed and applied to AMSU-B data for snowfall cases over the Great Lakes region. Ongoing work The a-priori database is constructed through radiative transfer modeling with combinations of Snowfall profiles from NEXRAD radar, CloudSat, and aircraft observations, atmospheric sounding profiles from aircraft observations and WRF simulations, Microwave land surface emissivity from MEM Single scattering properties for realistic nonspherical ice/snow particles calculated by the Discrete Dipole Approximation.
Future works More clearly detect scattering signatures of ice/snow over complex surfaces. Improve land surface emissivity calculations for snowfall cases Need deeper understanding of detailed structures of snow clouds in various regions and more validations to improve the retrieval algorithm. As more snowfall radar observations (e.g. CloudSat, EarthCARE, GPM) become available, a global database can be expanded in a similar fashion CIRA/Colorado State University 5th Annual CoRP Science Symposium, Aug.12-13, 2008
Thank you. CIRA/Colorado State University 5th Annual CoRP Science Symposium, Aug.12-13, 2008
Additional slides CIRA/Colorado State University 5th Annual CoRP Science Symposium, Aug.12-13, 2008
Discrete Dipole Approximation CIRA/Colorado State University 5th Annual CoRP Science Symposium, Aug.12-13, 2008
MW Radiometric properties of snowflakes and spheres (Liu, 2004)
Phase functions CIRA/Colorado State University 5th Annual CoRP Science Symposium, Aug.12-13, 2008
14 January 2001 snowfall case AMSU-B Retrieval (mm/hr) Radar (mm) CIRA/Colorado State University 5th Annual CoRP Science Symposium, Aug.12-13, 2008