Studies of columnar observations and model outputs Brian Mapes University of Miami with thanks to many data producers and sharers Looking under the hood of monthly mean data “points” • Time sections: f(submonthly time x pressure) – a vast, rich domain – complementary to typical climate model examination space (lat x lon x months) – Natural domain for observations – Sensible domain for physical processes Shortcomings • Free-running climate model time axis not directly comparable to observations – statistical comparisons of course • Models have climate-regime offsets in space, biasing comparisons at a fixed geo-location – non-mean statistics that is • Submonthly variations (dynamics-driven) a poor proxy for climate sensitivity (thermo-driven) – tough luck, we do what we can • Eulerian viewpoint a perverse view of weather – you’re free to leave inescapable programming programming tedium inescapable tedium COARE obs KWAJEX obs LBA obs EPIC, JASMINE KWAJEX CRM NCAR CAM GFDL AM2 NASA GMAO (super-CAM soon?) cross-forced SCMs •SCAM2 driven by CAM •SCAM driven by AM2 standard raw plots Common data format, variables, names, units, flux sign conventions, etc. standard stat plots Datasets All obs are from warm-wet tropics so far - what I have & know best COARE obs KWAJEX obs LBA obs EPIC, JASMINE (more: ARM? etc?) Cloud obs rare: satellite TOA rad in ~GCM-grid areas, but cloud profiles at only a few points (cloud radar) KWAJEX 3D CRM the most GCM-comparable cloud dataset NCAR CAM GFDL AM2 NASA GMAO (more: CAM-SP?) Driving the whole exercise -an AMIP, at the very least cross-forced SCMs weird experiments, not done real carefully so far. Future? •SCAM2 driven by CAM •SCAM driven by AM2 What variables? COARE obs KWAJEX obs LBA obs EPIC, JASMINE TOA radiation and cloud forcing cloud fraction and condensed water content (p) KWAJEX CRM humidity (p) NCAR CAM GFDL AM2 NASA GMAO (super-CAM soon?) vertical motion (p) structure within troposphere physical tendencies heating and drying (p) cross-forced SCMs •SCAM2 driven by CAM •SCAM driven by AM2 CLIMATE IMPACT SCALAR rain SCALAR LINK TO HYD CYC What plots? COARE IFA obs KWAJEX obs LBA obs EPIC, JASMINE KWAJEX CRM NCAR CAM GFDL AM2 NASA GMAO (super-CAM soon?) cross-forced SCMs •SCAM2 driven by CAM •SCAM driven by AM2 IFA OBS CRF: LW+, SW-, net ~0 [obs needclear-sky assumptions] note: net <0 at rainiest times 100 RH cld frac [ crude f(RH) here ] rain Q1, CWC (no obs) wind divergence and vertical grid % 0 What stats? COARE IFA obs KWAJEX obs LBA obs EPIC, JASMINE Lag regressions I: scalars vs. reference variable (here, ref = qbudget-derived IFA rain) TOA rad up: reflected SW net reduced OLR KWAJEX CRM NCAR CAM GFDL AM2 NASA GMAO (super-CAM soon?) cross-forced SCMs •SCAM2 driven by CAM •SCAM driven by AM2 Other rain vars: (here, other obs. estimates models, other rain types) What stats? COARE IFA obs KWAJEX obs LBA obs EPIC, JASMINE KWAJEX CRM NCAR CAM GFDL AM2 NASA GMAO (super-CAM soon?) cross-forced SCMs •SCAM2 driven by CAM •SCAM driven by AM2 regressions vs. IFA rainfall ocean shading* ~ -20% of latent heat in rain net longwave atm heating ~ 15% of latent heating ~6am diurnal rain peak * straight regression incl. nocturnal zeroes - not really right approach ~1 mm/h x 1d ~24 mm rain What stats? II. lag-height regressions of profile fields vs. surface rain COARE obs KWAJEX obs LBA obs EPIC, JASMINE KWAJEX CRM NCAR CAM GFDL AM2 NASA GMAO (super-CAM soon?) cross-forced SCMs •SCAM2 driven by CAM •SCAM driven by AM2 44 km radius “MCS-resolving” ~1/2 day timescale at a point What stats? regressions vs. surface rainfall SYNOPTIC SCALE (IFA) COARE IFA obs KWAJEX obs LBA obs EPIC, JASMINE KWAJEX CRM NCAR CAM GFDL AM2 NASA GMAO (super-CAM soon?) multiscale (incl. several-day timescales) cross-forced SCMs •SCAM2 driven by CAM •SCAM driven by AM2 but no cloud profile observations TOA: obs TOA: obs-forcedcondensate CRM can’t blow away What should areal cloud statistics look like? KWAJEX 3D CRM thanks Marat 4-8 units, vertical up to 16, vertical What stats? COARE IFA obs KWAJEX obs LBA obs EPIC, JASMINE KWAJEX CRM NCAR CAM - IFA GFDL AM2 - IFA NASA GMAO (super-CAM soon?) cross-forced SCMs •SCAM2 driven by CAM •SCAM driven by AM2 regressions vs. surface rainfall - AM2 july - 10 mm/d TOA about balanced, long time scales layers upward development time jitters sometimes zigzag layers AM2 - a different month What stats? COARE IFA obs KWAJEX obs LBA obs EPIC, JASMINE KWAJEX CRM nov - 11 mm/d TOA closely balanced, long time scale slow upward development NCAR CAM - IFA GFDL AM2 - IFA NASA GMAO (super-CAM soon?) cross-forced SCMs •SCAM2 driven by CAM •SCAM driven by AM2 zigzag layers What stats? COARE IFA obs KWAJEX obs LBA obs EPIC, JASMINE KWAJEX CRM NCAR CAM - IFA GFDL AM2 NASA GMAO (super-CAM soon?) cross-forced SCMs •SCAM2 driven by CAM •SCAM driven by AM2 regressions vs. surface rainfall - CAM jan - 7 mm/d CRF not well linked to rain events 3-day waves vertical thanks to ‘convective’ fraction but… …condensed water not coherent in vertical What stats? COARE IFA obs KWAJEX obs LBA obs EPIC, JASMINE KWAJEX CRM NCAR CAM - IFA GFDL AM2 - IFA GMAO 140E/8N July (super-CAM soon?) cross-forced SCMs •SCAM2 driven by CAM •SCAM driven by AM2 regressions vs. surface rainfall - GMAO a bit stark but about like IFA upward development, zigzag layers 2xKwajCRM ice content freezing level schism rain event met fields - obs <-KWX IFA-> rain event met fields - CAM-IFA <-Jan Dec-> rain event met fields - 3 models CAM AM2 GMAO But are these brief rain-event fluctuations a Climate Process? • hypothesis Divergence profile in tropical rain events linked to profile of divergent winds linked to profile of total winds? ( -> coupling issues…climate by any standard) ? Let’s look Top- stats of div regression sections (lag-height) Bottom- stats of total wind speed (time-height) each curve shows 1 month of data mean IFA AM2 stdev w/in 6-day lag section KWAJEX is 3rd line; LBA 4 CAM All 3 GCMs have unrealistic surface-intensified div fluctuations GMAOmidlevel swoop mean stdev GMAO 1000 vs 850 not right but no near- surface enhancement of wind fluctuations CAMjangly wind profiles wind profile results • hypothesis too simplistic • BUT – Systematic windspeed profile differences from observations are clear at low levels! Regressions wrt TOA CRF “events”? • Weird idea: submonthly TOA CRF has ~no feedback to the weather that causes it, so there’s less expectation of a repeatable “life cycle” • Proxy for climate sensitivities? – ? Maybe, if tau << {days, years, decades, etc.} ? • Anyway, try it (but I have no obs yet) – use 24h smoothing to kill diurnal cycle 24h-smoothed TOA -CRF as base time series -- CAM 85W 20S CRF<0, big slow changes SWup>0 Whiter w/clouds at 975 not 925 Another month at 85W 20S -- CAM faster changes, polymodal pdf? CRF<0 SWup>0 Complicated mixture of events AM2 - 85W 20S whiteness varies little; regression structure not stable steady -30ish CRF (Q1 data missing) PBL-based clouds rise to ~550mb here !? Minghua’s AM2 “midlevel” cloud (ISCCP) Cross-forced SCMs (CAM <-> AM2) • V. advection of SCM fields by prescribed vertical velocity from other GCM • H. advection prescribed (non-interactive) • Look at LBA (Amazon) just for fun LBA obs (via Minghua’s shop), & GCMs LBA obs CAM AM2 (warning: clear sky not careful) (sorry: Q1 missing) LBA obs CAM AM2 LBA models AM2 CAM SCAM driven by AM2 omega Cloud style CAMlike LBA models AM2 CAM SCAM2 driven by CAM omega Cloud style AM2 ish Future plans - team efforts? standard raw plots COARE obs KWAJEX obs LBA obs EPIC, JASMINE KWAJEX CRM NCAR CAM GFDL AM2 NASA GMAO (super-CAM soon?) Common data format, variables, names, units, sign conventions standard stat plots cross-forced SCMs clickable from Bony index plot •SCAM2 driven by CAM •SCAM driven by AM2 MORE DATA in (p,t) space SCMs! CRMs! EXPERIMENTS! OBS! co-ra.com/~bem more to come other people