Mesoscale wind divergence profiles in convecting regions Brian Mapes, Jialin Lin, Paquita Zuidema data thanks to: CSU Radar met group, UW mesoscale group, TRMM ground validation office Outline 1. Motivations and methods 2. Results in some easy and hard cases 3. Statistical results 4. A well-measured oddity: JASMINE May 22 5. Summary and future plans Motivations Heating in tropical convective clouds drives larger-scale circulations (LSCs) of many scales Heating profiles are important to LSC structure, including feedbacks to convection Divergence profiles (inflow/outflow) are closely linked to heating; and affect layer cloudiness Clear-air (unheated) component of divergence is smaller, but especially important to feedbacks Divergent flow is hard to measure accurately, so better observations may lead to new discovery Continuing hope: better obs. glimpses of a unique aspect of deep moist atmospheric convection: it’s embedded in a stratified environment, where gravity waves disperse by vertical wavelength Wind divergence: interpretation are needed to see this picture. TIFF (LZW) decompressor QuickTime™ and a Rearranging, are needed to see this picture. TIFF (LZW) decompressor QuickTime™ and a are needed to see this picture. TIFF (LZW) decompressor QuickTime™ and a For Q >> dhT/dt (tropical convection scaling), w/the are needed to see this picture. TIFF (LZW) decompressor QuickTime™ and a = ∂/∂p( Q/s ) Mapes and Houze 1995 JAS Wind divergence : measurement The divergence theorem for an area A with perimeter P: Area averaged divergence = V dA = Vnormal dP perim A Defining the perimeter-mean velocity Vnormal, For a circular area, the overbar is an azimuth mean: A Vnormal x P/A = [Vradial] x 2pR/pR2 = [Vradial] x 2/R Velocity-Azimuth Display (VAD) At fixed horizontal range (radius) R and altitude, consider radial velocity Vr vs. azimuth angle Mean wind = sine wave Vr 0 [Vr] Area-avg divergence = [Vr] *2/R (NOTE: no uniformity wave-2 is deformation; real flows may have jets, etc. etc. ws assumption within circle) 0 azimuth (deg from N) wd 360 Background: Doppler radar Precipitation radar – Radio pulses bounce off hydrometeors – Z is 6th moment of the DSD, log scale dBZ – For applications, rainrate R ~some power law in Z Doppler (coherent) radar – Pulse pairs sent; Doppler spectrum P(Df) received – A mean Df gives along-beam velocity: Vr = Df/2p * l/Dt +/- nl/Dt =2nVnyq – For applications, remove fallspeed Vt(Z)cos(zenith) Cylindrical data binning for VAD Dr = 8 km (12 bins, 0-96 km) Daz = 15° (24 bins) Dz = 500 m (36 bins) -->Dp = 50 hPa Dt = 1 hour QuickTime™ and a TIFF (LZW) decompressor are needed to see this picture. 1 dBZ bins (for Z-R) 1 m/s bins 4 m/s bins (only need nadjust) Further pool data, e.g. height to pressure: simply sum the histogram arrays. (same for range or time pooling) VAD plot for each hour, layer, range (pool) unfolding guide for absolute Vr (after histogram compactification) From GOF, or ind. ws,wd guess sampling error bar [Vr] <0: convergence -2m/s *2/(40km) = -1e-4 /s wind: 7 m/s from 240 deg used for rel. weighting in least-square 3-param fit: [Vr], ws, wd Repeat, repeat, repeat… research radar deployments QuickTime™ and a TIFF (LZW) decompressor are needed to see this picture. TEPPS: July 1997 JASMINE: May 1999 EPIC: September 2001 several 10s of days = many 100s of hours each EPIC 2001: timepressure QuickTime™ and a TIFF (LZW) decompressor are needed to see this picture. Ivo Sparse 48h (Vr data everywhere, but sometimes just noise) An example hourly product convection developing near radar 48 glorious hours: Sept 23-24 RHB-C RHB-C QuickTime™ and a TIFF (LZW) decompressor are needed to see this picture. All points on all curves at all times are from independent data: Continuity in range, time, altitude encouraging 10x sonde div; decent mass balance ------> QuickTime™ and a TIFF (LZW) decompressor are needed to see this picture. RH profiles R=44km div mass flux (-omega) ice sat dQ/dp >0 ‘onion’ dQ/dp >0 RH soundings, sfc rain, mm cloud radar, VAD div: Paquita Zuidema, ETL Hard case: lots of Doppler folding Tropical Storm Ivo Sept. 10, 2001 QuickTime™ and a TIFF (LZW) decompressor are needed to see this picture. QuickTime™ and a TIFF (LZW) decompressor are needed to see this picture. QuickTime™ and a TIFF (LZW) decompressor are needed to see this picture. TS Ivo, Sept. 10, 2001 cyclone moves westward, N of ship wind 13Z 15Z QuickTime™ and a TIFF (LZW) decompressor are needed to see this picture. v<0 u>0 QuickTime™ and a TIFF (LZW ) decompressor are needed to see this picture. 17 QuickTime™ and a TIFF (LZW) decompressor are needed to see this picture. 18 19 a dna ™e mi Tk ciu Q QuickTime™ and a ro s se rp mo ced ) WZL ( F FI T TIFF (LZW ) decompressor .e rut cip siht ee s ot dedeen e rare a needed to see this picture. convergence 10-4s-1 x 7h = 2.5, ~10-folding of ~100km scale vorticity 21Z QuickTime™ and a TIFF (LZW) decompressor are needed to see this picture. v>0 u>0 Vr folding problem at 925mb looks OK QuickTime™ and a TIFF (LZW) decompressor are needed to see this picture. QuickTime™ and a TIFF (LZW) decompressor are needed to see this picture. clutter in S-W azimuths: Vr ~0. Spike hists dominate fit N E S W ~OK N Mesoscale wind divergence profiles in convecting regions Brian Mapes, Jialin Lin, Paquita Zuidema data thanks to: CSU Radar met group, UW mesoscale group, TRMM ground validation office Outline 1. Motivations and methods 2. Results in some easy and hard cases 3. Statistical results (EPIC, cloud model) 4. A well-measured outlier: JASMINE May 22 5. Summary and future plans EPIC simple-mean divergence Mapes and Houze 1995 COARE airborne (mid-troposphere) Doppler radar over W. Pac. smaller circles’ fluctuations have larger amplitude; convergent times are undersampled --> divergent bias (dQ/dp) mean s.d. QuickTime™ and a ice small circles liq have fallspeed overcorrection error (steep beam tilt) TIFF (LZW) decompressor are needed to see this picture. noisy - sea clutter QuickTime™ and a TIFF (LZW) decompressor are needed to see this picture. To isolate latent heat associated signal, regress div on R(Z) overshoot cooling? QuickTime™ and a TIFF (LZW) decompressor are needed to see this picture. QuickTime™ and a TIFF (LZW) decompressor are needed to see this picture. QuickTime™ and a standard TIFF (LZW) decompressor are needed error to see this picture. rain >0.5 all data QuickTime™ and a TIFF (LZW) decompressor are needed to see this picture. Regress div(p) to surface rainfall: reduces scale dependence, biases scale-dependent fluctuation intensities cancel out can pool or average ranges for more robustness Uses data when available. Undersampled times w/upper convergence & R~0 don’t affect slope Badness of crude ice/liq Vt(Z) assumption is ~uncorrelated w/ surface rainrate, since that mainly varies with coverage, not Z. div per rain (1e-6 s-1 per mm/h) recalibrated Moisture convergence per unit rainfall: R(Z) too great R(Z) too small re-calibrated QuickTime™ and a TIFF (LZW) decompressor are needed to see this picture. should be ~1 Gross variations of magnitude in rainconvergence regression slopes between experiments e.g. COARE-MIT Z-R rainrate is about half the linearly regressed moisture convergence, 5 dbZ too low? (Radar has always been low outlier of rain estimates Weller et al. 2004) Why is rain weakly associated w/ low-level convergence? (hourly, simultaneous, ~100km scale) Convective vs. stratiform by horiz. echo texture (thanks QuickTime™ S. Nesbitt, and a CSU) TIFF (LZW) decompressor are needed to see this picture. • QuickTime™ and a TIFF (LZW)sigma decompressor are needed to see this picture. all data rain >0.5 ° div>0 div<0 Multiple linear regression teases apart their pure signatures C,S rainrates are correlated but have independent variability too. QuickTime™ and a hours TIFF (LZW) decompressor are needed to see this picture. convergence ~525mb C autocorrelation C-S cross correlation Con-strat evolution in total precip lag regression Rapid rise of convergence to middle levels at time of precip max Trimodal cloud tops? (divergence) rising with time in advance of precipitation 1 mm per mm/h precipitable water 0 from microwave radiometer (ETL) 3D cloud model Marat Khairoutdinov KWAJEX case QuickTime™ and a TIFF (LZW) decompressor are needed to see this picture. Divergence of cloudy updraft mass flux Trimodal cloud tops, PW rises then falls 3D cloud model’s KWAJEX forcing QuickTime™ and a TIFF (LZW) decompressor are needed to see this picture. Sonde-array div vs. model rainrate Triple divergence levels, sounding PW rises then falls Note 55h time axis! internal variations in 3D cloud model divergence and rain in 64x64 km subset of domain (1/16 of area) (about VAD size) specific humidity in same area Tracing back to cases (fold) day 270 QuickTime™ and a TIFF (LZW) decompressor are needed to see this picture. QuickTime™ and a TIFF (LZW) decompressor are needed to see this picture. a dn a ™ emiTk ci uQ ro ss e rpm oc ed ) WZ L ( FFIT .e r ut ci p sih t e es ot d ed ee n e r a QuickTime™ and a TIFF (LZW) decompressor are needed to see this picture. note JASMINE oddity QuickTime™ and a TIFF (LZW) decompressor are needed to see this picture. Similar picture from other radar deployments (convergence rises up over several hours as convective clouds turn stratiform) Convective, stratiform,…? Are these archetypal components of convective rainfall rooted in physical processes…? detrainment ice physics QuickTime™ and a TIFF (LZW) decompressor are needed to see this picture. m=1/2 entrainment (…and is there a physical process behind it)? m=3/2 m=1 water physics …or in dynamical modes? (strictly, spectral bands) One clue: Does this exist? (A: Yes, and Maybe) A nice touch: interleaved tilts, for buttery-smooth vert. res. when pooled over an hour (2+ volumes) Mesoscale wind divergence profiles in convecting regions Brian Mapes, Jialin Lin, Paquita Zuidema data thanks to: CSU Radar met group, UW mesoscale group, TRMM ground validation office Outline 1. Motivations and methods 2. Results in some easy and hard cases 3. Statistical results 4. A well-measured outlier: JASMINE May 22 5. Summary and future plans 40N Eastern Ghats JASMINE Bay of Bengal May 1999 equator Dave Lawrence Mean diurnal cycle of 210K Meteosat-5 Infrared Imagery figs by Paquita Zuidema Mean diurnal cycle of 210K Meteosat-5 Infrared Imagery figs by Paquita Zuidema QuickTime™ and a TIFF (LZW) decompressor are needed to see this picture. JASMINE May 1999, spanning monsoon onset in the Bay of Bengal JASMINE squall (wave?) May 22, 1999 (figs from U. of Washington web pages on JASMINE) ~15 m/s Webster et al. 2003, Zuidema 2003 17-18Z JASMINE div(p) QuickTime™ and a TIFF (LZW) decompressor are needed to see this picture. 1604 Z mass flux QuickTime™ and a TIFF (LZW) decompressor are needed to see this picture. JASMINE wavy div(p): Double-decker convection? from UW JASMINE web pages 20-21Z JASMINE div(p) bad QuickTime™ and a TIFF (LZW) decompressor are needed to see this picture. QuickTime™ and a TIFF (LZW) decompressor are needed to see this picture. bad JASMINE May22 time-height section Or is it moist (dry air) processes? Is it dynamical wave# 3/2? x - old/ folded QuickTime™ and a TIFF (LZW) decompressor are needed to see this picture. note very low drying mm cloud radar data and overlay presentation by Paquita Zuidema, ETL Recalling the EPIC case… dry intrusion: a physical process, not simply related to a dynamical mode… but note deeper layer of descent drying in this case, onion RH min ~3km (725 mb) JASMINE case IS significantly unusual…. SUMMARY & CONCLUSIONS: Hourly VAD gives simple, automatic, good divergence measurements Sensibly consistent w/ surface rain & zenith cloud radar obs. (full EPIC and JASMINE overlay sections at http://www.etl.noaa.gov/~pzuidema) We see good ol’ convective & stratiform profiles (ho-hum), but also •Anvils snowing into dry layers, moistening and cooling their tops (cooling flattens the dry layer by diabatic divergence dQ/dp>0) •Steady convergence below 700mb in TS Ivo, to ~10-fold z in 7h •m=3/2 divergence profile in JASMINE propagating squall/wave: double-decker convection, and deviations from that pillar of tropical meso-meteorology, middle level convergence in stratiform rain •Wavy profiles of div regressed on shallow convective rainfall - are high and low tropospheric clouds/heating dynamically coupled on the mesoscale? If so, is random overlap a good assumption? Mapes and Houze 1992 airborne Doppler, EMEX QuickTime™ and a mean TIFF (LZW) decompressor are needed to see this picture. sd 1km res (subjectively categorized samples) •Better top! •Whole troposphere shorter in EPIC thanQuickTime™ in and a TIFF (LZW) decompressor Australian monsoon are needed to see this picture. •More complete life cycle: MH92 obs were in mature MCSs and rarely saw fresh development convective rain from 0-5km top cells >10km tops (like MH92 “intermediary”?) 5-10km tops stratiform EPIC MLR for 4 categories of precipitation: shallow, deep, very deep convection; stratiform ~~Vertical waviness?~~ ~~~~ convective rain from 0-5km top cells >10km tops 5-10km tops stratiform Corresponding omega EPIC MLR for 4 categories of precipitation: shallow, deep, very deep convection; stratiform Difficulty of measuring divergence: sounding arrays Erroneous barotropic part QuickTime™ and a of sonde-array TIFF (LZW) decompressor are needed to see this picture. divergence is comparable to signals sought in baroclinic component… low bar for our work! Hard case: a sparse time QuickTime™ and a TIFF (LZW) decompressor are needed to see this picture. N QuickTime™ and a TIFF (LZW) decompressor are needed to see this picture. QuickTime™ and a TIFF (LZW) decompressor are needed to see this picture. E (sea clutter?) S W N