Transcript Document 7864991
Global,
0000Z, 30Nov 2001
Microwave-based, Precipitation Analyses from Satellite at ½ Hr & 8 km Scales
0000Z, 30Nov 2001 AMSU
Bob Joyce
: RSIS, Inc.
John Janowiak
: Climate Prediction Center/NWS
Phil Arkin
: ESSIC/Univ. Maryland
Pingping Xie
: Climate Prediction Center/NWS 0030Z, 30Nov 2001 TRMM 0100Z, 30Nov 2001 SSM/I
Two primary types of precipitation algorithms: Infrared (GPI, convective-stratiform, OPI) (-) indirect - can only sense cloud-top temperature (+) very good sampling characteristics (time & space) Passive Microwave (SSM/I, MSU, AMSU): (+) considerably better estimate than IR – “sees” thru cloud and can directly sense information from hydrometeors (-) poor temporal sampling (polar orbit platforms)
Pessimist: “The food here is
terrible
!” Optimist: “Ah yes, but such
huge
portions!” Pragmatist: Meld together the IR & microwave data to take advantage of the strengths of each Vicente (U. Wisc.) Turk (NRL, Monterey) Adler and Huffman (NASA/GSFC) Kuligowski (NOAA/NESDIS) Microwave & IR data Combined statistically
Our Approach
Use the IR and microwave data but do NOT mix them Use the IR only as a transport and “morphing” mechanism Here we use precipitation algorithms developed by Ferraro (NESDIS: AMSU-B & SSM/I) and Kummerow (CSU: TRMM) but method is algorithm independent.
Enables the generation of spatially and temporally complete precipitation fields while maintaining a pure, albeit manipulated, microwave-based analysis
“Advection vectors” are computed from IR for each 2.5
o box gridbox and
all microwave pixels
contained in that grid
are propagated in the direction of that vector
2.5
o 2.5
o
2.5
o IR Spatial Correlation Domain for Computation of “Advection Vectors” 2.5
o IR (t+0) IR (t+1/2 hr) precip
Advection Rates for 00Z 30 Nov 2001 ZONAL MERIDIONAL |………………. EAST …………|………… WEST ………………| (pixels/hour) |………………. NORTH ……...|…………SOUTH ………………| (pixels/hour)
t+0 Actual Microwave Observations Interpolated “observations” t+2 hrs Time interpolation weights 0.75
0.25
IR 0.50
0.50
0.25
0.75
t+ 1/2 hr t+1 hr t+1.5 hr
“Validation”
Microwave estimates propagated byIR.
Microwave data from overpasses between the “initial” and “next” overpasses were withheld in this test to assess the performance of the technique. Validating analysis is in the lower-center frame.
Initial microwave pass Valid time is 5 hours after the “initial” pass and 2.5 hours before the “next” pass Next microwave pass Validating analysis, ie. the microwave pass between the “initial” and “next” microwave passes that was withheld.
Satellite Radar
Microwave-Advected GPCP “1DD” GPI (IR)
Potential Applications
Real-time
quantitative
Disaster mitigation global precipitation monitoring Provide timely updates for U.S. interests abroad Numerical model initialization & validation Improve diurnal cycle in the models Diagnostic studies: diurnal cycle in particular
Continuing Work
Account for precipitation that forms and dissipates between microwave overpasses Refine advection vector computation Continue validation effort Test/Include new precipitation products as they become available – method is not restricted to particular algorithms or sensors
Finis
mm/day Figure 5
Another test – this one is over The South Atlantic Convergence Zone (SACZ) Initial microwave pass Microwave propagated by IR Valid time is 1.5 hours after the “initial” pass and 6 hours before the “next” pass Next microwave pass Validating microwave data
Methodology
AMSU-B, SSM/I and, TMI derived rainfall is mapped to global, half hourly rectilinear grids, equivalent to 8-km at the equator. Between overpasses by a microwave sensor, features within each non-overlapping 2.5
o x 2.5
o lat/lon grid box are propagated by using IR data until the next microwave overpass occurs, as follows: 1. We compute the spatial lag correlations of 8-km pixel within 5 o x 5 directions from the “target” 5 2.5
o x 2.5
o .
o gridboxes that are displaced from1, 2, …, up to 12 pixels in the X and Y o x 5 o IR brightness temperatures gridbox. The X and Y vector fields are then interpolated to 2. “Advection vectors” are computed which are oriented in the direction of highest spatial lag correlation in the X-Y plane.
3. All microwave estimates in the “target” 2.5
direction of the advection vector.
o x 2.5
o gridbox are then propagated in the 4. Once a new microwave overpass occurs, the same process is repeated, but backward in time from the most current data. This process allows features to “morph” with time by linearly interpolating both propagations using each pixel’s temporal “distance” from analysis time as the interpolation weights.
Spatial Lag Correlation of IR nearby 5 o x 5 o pixel temperature among grid boxes to determine propagation direction (t + 1/2 hr) (t + 0 hr) 5 o + 12 pixels 5 o 5 o 5 o + 12 pixels