Download presentation
Presentation is loading. Please wait.
1
How to Improve Mesoscale Atmospheric-Modeling Results Bob Bornstein San Jose State University San Jose, CA pblmodel@hotmail.com Presented at UNAM Mx City 9 March 2007
2
Acknowledgements Dr. H. Taha, Altostratus & SJSU Dr. H. Taha, Altostratus & SJSU Dr. F. Freedman Dr. F. Freedman Dr. Erez Weinroth, HUJI & SJSU Dr. Erez Weinroth, HUJI & SJSU my M.S. (ex) STUDENTS: Dr. C. Lozej- Archer, T. Ghidey, K. Craig, R. Balmori my M.S. (ex) STUDENTS: Dr. C. Lozej- Archer, T. Ghidey, K. Craig, R. Balmori Funded by: NSF, USAID, DHS, LBL, NSF Funded by: NSF, USAID, DHS, LBL, NSF
3
OUTLINE Introduction Introduction Getting Mesomet models to work better Getting Mesomet models to work better SYNOPTIC FORCING SYNOPTIC FORCING INPUT DATA INPUT DATA SFC/PBL FORCING SFC/PBL FORCING uMM5: Houston ozone uMM5: Houston ozone Conclusions Conclusions (Future) uWRF efforts (Future) uWRF efforts
4
Basic-Theme 1 e.g., O 3 -EPISODES OCCUR NOT FROM CHANGING TOPOGRAPHY OR EMISSIONS NOT FROM CHANGING TOPOGRAPHY OR EMISSIONS BUT DUE TO CHANGING GC/SYNOPTIC PATTERNS, WHICH BUT DUE TO CHANGING GC/SYNOPTIC PATTERNS, WHICH ENTER MESO-SOLUTION FROM CORRECT (we hope) LARGER-SCALE MODEL-FIELDS, & WHICH ENTER MESO-SOLUTION FROM CORRECT (we hope) LARGER-SCALE MODEL-FIELDS, & WHICH THUS ALLOW SCF MESO THERMAL-FORCINGS (i.e., UP/DOWN SLOPE, LAND/SEA, URBAN, CLOUDS/FOG) TO DEVELOP CORRECTLY (we hope) THUS ALLOW SCF MESO THERMAL-FORCINGS (i.e., UP/DOWN SLOPE, LAND/SEA, URBAN, CLOUDS/FOG) TO DEVELOP CORRECTLY (we hope)
5
CORRECT-ORDER OF FORCINGS IN A MESO-MET MODEL IS THUS FIRST: UPPER-LEVEL Syn/GC FORCING FIRST: UPPER-LEVEL Syn/GC FORCING pressure (the GC/Syn driver) Syn/GC winds Syn/GC winds NEXT: TOPOGRAPHY NEXT: TOPOGRAPHY grid spacing flow-channeling LAST: MESO SFC-CONDITIONS LAST: MESO SFC-CONDITIONS temp (the meso-driver) & sfc z temp (the meso-driver) & sfc z 0 mesoscale winds
6
Case Study 1: Atlanta Summer Thunderstorm (K. Craig) Obs: weak-cold front N of city Obs: weak-cold front N of city Large-scale IC/BC: front S of city Large-scale IC/BC: front S of city MM5 UHI-induced thunderstorm: MM5 UHI-induced thunderstorm: 5-km deep, w max 6-m/s, 8-cm precip 5-km deep, w max 6-m/s, 8-cm precip Should be: 9-km, 12-m/s, 14-cm Should be: 9-km, 12-m/s, 14-cm Source of problem: Source of problem: Wx model incorrectly put front S of City MM5-storm formed in stable-flow from N (& not in unstable-flow from S) MM5-storm formed in stable-flow from N (& not in unstable-flow from S)
7
ATLANTA UHI-INITIATED STORM: OBS SAT & PRECIP (UPPER) & MM5 W & PRECIP (LOWER)
8
Case study 2: SFBA Summer O 3 - episode (T. ) Case study 2: SFBA Summer O 3 - episode (T. Ghidey) Obs: daily max- O 3 sequentially moved from Livermore to Sacramento to SJV Obs: daily max- O 3 sequentially moved from Livermore to Sacramento to SJV Large scale IC/BC: Large scale IC/BC: shifting meos-700 hPa high shifting meos-sfc low changing sfc-flow max-O 3 changed location Results: good analysis-nudging in MM5 Results: good analysis-nudging in MM5 good mesoscale winds
9
H H L SAC episode day: D-1 700 hPa Syn H moved to Utah with coastal “bulge” & L in S-Cal correct SW flow from SFBA to Sac
10
L H SJV episode day: D-3 700 hPa Fresno eddy moved N & H moves inland flow around eddy blocks SFBA flow to SAC, but forces it S into SJV
11
Theme 2: MM5 Non-urban Sfc-IC/BC Issues Deep-soil temp: BC Deep-soil temp: BC Controls nighttime min-T Controls nighttime min-T Values unknown & MM5 estimation-method is flawed Values unknown & MM5 estimation-method is flawed Soil-moisture: IC Soil-moisture: IC Controls daytime max-T Controls daytime max-T Values unknown & MM5-table values are too specific Values unknown & MM5-table values are too specific SST: IC/BC SST: IC/BC Horiz coastal T-grad controls sea-breeze flow Horiz coastal T-grad controls sea-breeze flow But we usually focus only on land-sfc temps But we usually focus only on land-sfc temps IC/BC-SST values from large-scale model are IC/BC-SST values from large-scale model are too coarse & not f(t)
12
Theme 2A: MM5 deep-soil temp Calculated as average large-scale model-input surface-T during simulation-period Calculated as average large-scale model-input surface-T during simulation-period But this assumes zero time-lag b/t sfc & lower- level (about 1 m) soil-temps But this assumes zero time-lag b/t sfc & lower- level (about 1 m) soil-temps But obs show But obs show 2-3 month time-lag b/t these two temps 2-3 month time-lag b/t these two temps Larger-lag in low-conductivity dry-soils Larger-lag in low-conductivity dry-soils Thus MM5 min-temps always are too-high in summer & too-low in winter Thus MM5 min-temps always are too-high in summer & too-low in winter Need to develop tech (beyond current trial & error) to account for lag Need to develop tech (beyond current trial & error) to account for lag
13
Case Study 3: Mid-east 2-m air temp (S. Kasakech) July 29August 1August 2 July 31 Aug 1 Aug2 Standard-MM5 summer night-time min-T, But lower input deep-soil temp better 2-m T results better winds better O 3 obs Run 1 MM5:Run 4 Obs Run 4: Reduced Seep-soil T First 2 days show GC/Syn trend not in MM5, as MM5-runs had no analysis nudging
14
Case study 4: SCOS96 2-m Temps (D. Boucouvual) RUN 1: has No GC warming trend Wrong max and min T 3-Aug4-Aug 5-Aug6-Aug RUN 5: corrected, as it used > Analysis nudging > Reduced deep-soil T
15
Theme 2B: MM5 input-table zproblems Theme 2B: MM5 input-table z 0 problems Water z = 0.01 cm Water z 0 = 0.01 cm Only IC updated internally by Charnak eq = f(MM5 u) Only IC updated internally by Charnak eq = f(MM5 u * ) But Eq only valid for open-sea, smooth-swell, conditions But Eq only valid for open-sea, smooth-swell, conditions Obs for rough-sea coastal-areas ~ 1 cm Obs for rough-sea coastal-areas ~ 1 cm MM5 coastal-winds are over-estimated Urban z = 80 cm Urban z 0 = 80 cm too low for tall cities: obs up to 3-4 m too low for tall cities: obs up to 3-4 m Urban-winds: too fast Urban-winds: too fast Must adjust input-value or use GIS/RS input as f(x,y) Must adjust input-value or use GIS/RS input as f(x,y)
16
Case study 5: Houston GIS/RS z o (S. Stetson) Values up 3 m Values too large, as they were f(h) and not f(ơ h )
17
Theme 2C: MM5 SST-problems Current Wx Model SSTs MM5 Current Wx Model SSTs MM5 Only every 6 or 12 hours Only every 6 or 12 hours Lack small scale SST-variations Lack small scale SST-variations Thus produces poor land-sea T-gradients Thus produces poor land-sea T-gradients poor sea-breeze flows Solution: satellite-derived SSTs Solution: satellite-derived SSTs More frequent More frequent More detailed More detailed
18
Case Study 6: NYC SST+currents (J. Pullen) COAMPS input satellite: SST + sfc currents L
19
Theme 3: Model-Urbanization History Need to urbanize momentum, thermo, & TKE Eqs Need to urbanize momentum, thermo, & TKE Eqs At surface & in SBL: diagnostic Eqs At surface & in SBL: diagnostic Eqs In PBL: prognostic Eqs In PBL: prognostic Eqs Newest from veg-canopy model of Yamada (1982) Newest from veg-canopy model of Yamada (1982) But Veg-param’s replaced with GIS/RS urban param’s But Veg-param’s replaced with GIS/RS urban param’s Brown and Williams (1998) Brown and Williams (1998) Masson (2000) Masson (2000) Martilli et al. (2001) in TVM/URBMET Martilli et al. (2001) in TVM/URBMET Dupont, Ching, et al. (2003) in EPA/MM5 Dupont, Ching, et al. (2003) in EPA/MM5 Taha et al. (2005), Balmori et al. (2006b) in uMM5 Taha et al. (2005), Balmori et al. (2006b) in uMM5 Input: detailed urban-parameters as f(x,y)
20
From EPA uMM5: Mason + Martilli (by Dupont) Within Gayno-Seaman PBL/TKE scheme
21
Advanced urbanization scheme from Masson (2000) ______ _________ 3 new terms in each prog equation
22
New GIS/RS inputs for uMM5 as f (x, y, z) land use (38 categories) roughness elements anthropogenic heat as f (t) vegetation and building heights paved-surface fractions drag-force coefficients for buildings & vegetation building height-to-width, wall-plan, & impervious- area ratios building frontal, plan, & and rooftop area densities wall and roof: ε, cρ, α, etc. vegetation: canopies, root zones, stomatal resistances
23
Case study 7: SJSU uMM5 performance by CPU With 1 CPU: MM5 is 10x faster than uMM5 With 96 CPU: MM5 is only 3x faster than uMM5 (< 12 CPU not shown) With 96 CPU: MM5 is still gaining, but MM5 has ceased to gain at 48 CPU & then it starts to loose
24
Performance by physics sound waves & PBL schemes take most CPU in both urban/PBL scheme in uMM5 takes almost 50% of all time
25
Urbanization day& nite on same line stability effects not important mechanical effects are important Is it worth it?: Case study 7 (A. Martilli) MM5: uMM5
26
(Last) Case Study 8: uMM5 for Houston: Balmori (2006) Goal: Accurate uMM5 Houston urban/rural temps & winds for Aug 2000 O 3 -episode via Texas2000 field-study data Texas2000 field-study data Taha/SJSU modification of LU/LC & urban morphology parameters by Taha/SJSU modification of LU/LC & urban morphology parameters by processing Burian parameters processing Burian parameters modifying uMM5 to accept them modifying uMM5 to accept them USFS urban-reforestation scenarios USFS urban-reforestation scenarios lower daytime max-UHI-intensity & O lower daytime max-UHI-intensity & O 3 EPA emission-reduction credits $’s saved
27
GC influences are small Early-AM along-shore flow (from east) from N-edge of off-shore cold-core L Flow is then sequentially: from Ship Channel to Houston by Bay Breeze into Houston by UHI-convergence (time of O 3 -max) and finally beyond Houston to NW by Gulf Breeze Domain-5, episode-day, obs O 3 -transport: sea breeze + UHI-convergence influences
28
L H C Urban min + UHI Conv H Start of N-flow H L H L over-Houston: due to titration H Near-max O 3
29
uMM5 simulation for 22-26 August case Model configuration Model configuration 5 domains, with Δx = 108, 36, 12, 4, and 1 km 5 domains, with Δx = 108, 36, 12, 4, and 1 km (x, y) grid-pts: 43x53, 55x55, 100x100, 136x151, 133x141 (x, y) grid-pts: 43x53, 55x55, 100x100, 136x151, 133x141 full- levels: 29 (Domains 1-4) & 49 (Domain 5) full- levels: 29 (Domains 1-4) & 49 (Domain 5) lowest ½ level= 7 m lowest ½ level= 7 m 2-way feedback in Domians 1-4 2-way feedback in Domians 1-4 Parameterizations/physics options Parameterizations/physics options > Grell cumulus (D 1-2) > ETA or MRF PBL (D 1-4) > Grell cumulus (D 1-2) > ETA or MRF PBL (D 1-4) > Gayno-Seaman PBL (D 5) > Simple ice moisture > Gayno-Seaman PBL (D 5) > Simple ice moisture > Urbanization module > NOAH LSM > Urbanization module > NOAH LSM > RRTM radiative cooling Inputs Inputs > NNRP Reanalysis fields > ADP observational data > NNRP Reanalysis fields > ADP observational data > S. Burian urban-morphology LIDAR building-data (D-5) > S. Burian urban-morphology LIDAR building-data (D-5) > LU/LC modifications (from D. Byun )
30
Episode-day Synoptics: 8/25, 12 UTC (08 DST) H H 700 hPa Surface 700 hPa & sfc GC H’s: at their weakest (no gradient over Texas) meso-scale forcing (sea breeze & UHI convergence) will dominate
31
Concurrent NNRP fields at 700 hPa & sfc H H NNRP-input to MM5 (as IC/BC) captured GC/synoptic features, as location & strength of H were similar to NWS charts (previous slide) D p=2 hPa
32
MM5: episode day, 3 PM > D–1: reproduces weak GC p-grad & flow > D-2: weak coastal-L > D-3: well-formed L produces along-shore V L D-1 D-2 D-3
33
Domain 4 (3 PM) : L is off of Houston only on O 3 day (25 th ) L L Episode day day
34
Urbanized Domain 5: near-sfc 3-PM V, 4-days Episode day day Cold-L Hot Cool
35
Along-shore flow, 8/25 (episode day): obs at 15-UTC vs uMM5 (Domain-5) at 20-UTC Tx2000 HGA uMM5 (D-5, red box) cap-tured along-shore V HGA uMM5
36
1-km uMM5 Houston UHI: 8 PM, 21 Aug Upper, L: MM5 UHI (2.0 K) Upper, L: MM5 UHI (2.0 K) Upper,R: uMM5 UHI (3.5 K) Upper,R: uMM5 UHI (3.5 K) Lower L: (uMM5-MM5) UHI Lower L: (uMM5-MM5) UHI LU/LC error
37
8/23 Daytime 2-m UHI: obs vs uMM5 (D-5) H OBS: 1 PM uMM5: 3 PM Cold UHI
38
Along –shore V: due to Cold-Core L : D-3 MM5 vs. Obs-T MM5: produces coastal cold-core low Obs (18 UTC): > Cold-core L (only 1 ob) > Urban area (blue-dot clump) retards cold-air penetration C H H
39
UHI-Induced Convergence: obs vs. uMM5 OBSERVEDuMM5 C C
40
Obs speeds (D-5): large z speed-decrease over city Obs speeds (D-5): large z 0 speed-decrease over city - - - + + + + V
41
Current base-case Veg-cover (in 0.1’s), with an urban min of 0.2-0.3 Future case (from USFS) Increases in veg-cover (in 0.01’s), with max increases (in urban areas) of about 0.1 uMM5 urban reforestation & rural deforestation simulations
42
Soil moisture increase for: Run 12 (entire area, left) and Run 13 (urban area only, right)
43
Run 12 (urban-max reforestation) minus Run 10 (base case): near-sfc ∆T at 4 PM reforested central urban-area cools & surrounding deforested rural-areas warm
44
D UHI(t) for Base-case minus Runs 15-18 D UHI(t) for Base-case minus Runs 15-18 U1 sea Ru U2 UHI = Temp in Box-Urban minus Temp in Box-Rural UHI = Temp in Box-Urban minus Temp in Box-Rural Runs 15-18: different urban re-forestation scenarios Runs 15-18: different urban re-forestation scenarios D UHI=Run-17 UHI –Run-13 UHI (max effect, green line) D UHI=Run-17 UHI –Run-13 UHI (max effect, green line) Reduced UHI lower max-O 3 (not shown) Reduced UHI lower max-O 3 (not shown) EPA emission-reduction credits Max-impact of –0.9 K of a 3.5 K Noon-UHI, of which a 3.5 K Noon-UHI, of which 1.5 K was from uMM5
45
Overall Lessons Models can’t be assumed to be Models can’t be assumed to be perfect perfect black boxes black boxes If obs not available OK to make reasonable educated estimates, e.g., for If obs not available OK to make reasonable educated estimates, e.g., for Deep-soil temp Deep-soil temp Soil moisture Soil moisture Need data for comparisons with simulated fields Need data for comparisons with simulated fields Need good urbanization, e.g., uMM5 Need good urbanization, e.g., uMM5 Need to develop better PBL parameterizations Need to develop better PBL parameterizations
46
FUTURE WORK: uWRF uWRF with NCAR (F. Chen) for DTRA uWRF with NCAR (F. Chen) for DTRA Martilli-Dupont-Taha urbanization Martilli-Dupont-Taha urbanization Freedman turbulence Freedman turbulence Applications (current + will propose*) Applications (current + will propose*) Urban canyon dispersion for DTRA Urban canyon dispersion for DTRA Urban climate (with UNAM) Urban climate (with UNAM) *NYC ozone for EPA *NYC ozone for EPA *Calif ozone for CARB *Calif ozone for CARB *Urban thunderstorms for NSF *Urban thunderstorms for NSF *Urban wx forecasting for NWS *Urban wx forecasting for NWS
47
PROG* APPROACH FOR LENGTH SCALE Freedman & Jacobson (2002 & 2003, BLM) + Freedman at SJSU *2 prog Eqs.: TKE & DISSIPATION RATE ε Where ℓ = c ε E 3/2 /ε Values of σ ε & σ E are reversed in Mellor & Yamada reversed in all atm models K & TKE in upper PBL were wrong! __
48
CALIBRATION TO NEUTRAL ABL: ℓ vs. z Lines: various values of κ = c ε2 σ ε /σ E x = COLEMAN (‘99) DNS New (R-panel) best-fit κ = 1.3 (dashed line), w/ better results (ℓ↓) in upper PBL Standard approach (left panel) best fit with κ = 2.5, w/ poor results in upper PBL newold
49
Same, but for K (z) x = COLEMAN (‘99) DNS New (R-panel) best-fit κ = 1.3 (dashed line), w/ better results in lower PBL & K ↓ aloft Standard approach (left panel) best fit with κ = 2.5, w/ poor results in lower PBL new old
50
Thanks Any questions?
Similar presentations
© 2024 SlidePlayer.com. Inc.
All rights reserved.