Mesoscale Deterministic and Probabilistic Prediction over the

Northwest: An Overview

Cliff Mass

University of Washington

University of Washington Mesoscale Prediction Effort

• An attempt to create an end-to-end deterministic and probabilistic prediction system.

• On the deterministic side, examine the benefits of high resolution

• Identify major issues with physics parameterizations

Deterministic Prediction

• WRF ARW Core run at 36, 12, 4, and now 4/3 km grid spacing

• Extensive verification

• Variety of applications running off the deterministic forecasts.

Major Elements

• Two mesoscale ensembles systems UWME (15 members) and EnKF (60 members, 36 and 4 km grid spacing).

• Sophisticated post-processing to reduce model bias and enhance reliability and sharpness of resulting probability density functions (PDFs) for UWME.

• Stand-alone bias correction

• Bayesian Model Averaging (BMA)

• Ensemble MOS (EMOS)

Major Elements

• Psychological research to determine the best approaches for presenting uncertainty information.

• Creation of next-generation display products providing probabilistic information to a lay audience. Example: probcast.

Inexpensive Commodity Clusters

• This effort has demonstrated the viability of doing such work on inexpensive Linux clusters.

• Proven to be highly reliable

The Summary

Verification

Precip Verification

High Resolution

• Attempt to answer questions:– What is the payoff in getting the land-water

boundaries and smaller scale terrain much better

– Does ultra high resolution improve objective verification or subjective structures?

– Do physics problems get better or worse?

1.3

4 km

6-hr forecast, 10m wind speed and direction

1.3 km

4 km

Boundary Layer Physics: A Current Achilles Heel of

Mesoscale NWP

• Well known issues:– Winds too strong and geostrophic near

surface– Excessive low-level mixing– Inability to maintain shallow cold PBL

During the past few months we have continued our testing program of various PBL schemes, vertical

diffusion options, etc.

• A test case has been one in which the 4 and 1.3 km created unrealistic roll circulations.

http://www.atmos.washington.edu/~ovens/wrf_1.33km_striations/

1 km visible

Problem

• Instead of getting open cellular convection, there are these period cloud streets.

• Look like roll circulation, but of too large a scale (if you look at sat pics you can see hints of them).

• Sometimes apparent (but less so) in 4-km.

• Occurs only in unstable, post-frontal conditions.

Through the kitchen sink at it and consulted heavily with Dave

Stauffer at Penn State• Tried a range of PBL schemes (YSU, QNSE, ACM2,

MYNN, MYJ, MYJ with Stauffer mods)

• Added 6th order diffusion and played with diffusion coefficent.

• Fully, interactive nesting

• Upper level diffusion and gravity wave drag

• Monotonic advection

• Varying vertical diffusion, both more and less

Results

• ACM2 (Pleim PBL and LSM) was the only thing that helped reduce the rolls.

• It created this stratiform cloud mass that wasn’t very realistic.

Excessive Geostrophy and Mixing at Low Levels

• Tried every PBL option in ARW…not the solution!

• Trying other things: decreasing model diffusion and increasing surface drag by increasing ustar.

Example: Cut vertical diffusion 10 1/8th of normal value

Vertical Diffusion cut to 1/4

Standard Low Diffusion

Mesoscale Ensembles at the UW

UWME

Core Members• 8 members, 00 and 12Z• Each uses different synoptic scale initial and boundary

conditions from major international centers• All use same physics• MM5 model, will be switching to WRF.• 72-h forecasts

Resolution (~ @ 45 N ) ObjectiveAbbreviation/Model/Source Type Computational Distributed Analysis

avn, Global Forecast System (GFS), Spectral T254 / L64 1.0 / L14 SSINational Centers for Environmental Prediction ~55 km ~80 km 3D Var cmcg, Global Environmental Multi-scale (GEM), Finite 0.90.9/L28 1.25 / L11 3D VarCanadian Meteorological Centre Diff ~70 km ~100 km eta, limited-area mesoscale model, Finite 32 km / L45 90 km / L37 SSINational Centers for Environmental Prediction Diff. 3D Var gasp, Global AnalysiS and Prediction model, Spectral T239 / L29 1.0 / L11 3D VarAustralian Bureau of Meteorology ~60 km ~80 km

jma, Global Spectral Model (GSM), Spectral T106 / L21 1.25 / L13 OIJapan Meteorological Agency ~135 km ~100 km ngps, Navy Operational Global Atmos. Pred. System, Spectral T239 / L30 1.0 / L14 OIFleet Numerical Meteorological & Oceanographic Cntr. ~60 km ~80 km

tcwb, Global Forecast System, Spectral T79 / L18 1.0 / L11 OITaiwan Central Weather Bureau ~180 km ~80 km ukmo, Unified Model, Finite 5/65/9/L30 same / L12 3D VarUnited Kingdom Meteorological Office Diff. ~60 km

“Native” Models/Analyses Available

UWME– Physics Members

• 8 members, 00Z only• Each uses different synoptic scale initial and

boundary conditions• Each uses different physics• Each uses different SST perturbations• Each uses different land surface characteristic

perturbations

– Centroid, 00 and 12Z• Average of 8 core members used for initial

and boundary conditions

36 and 12-km domains

Post-Processing

• Post-processing is a critical and necessary step to get useful PDFs from ensemble systems.

• The UW has spent and is spending a great deal of effort to perfect various approaches that are applicable on the mesoscale.

Post-Processing• Major Efforts Include

– Development of grid-based bias correction– Successful development of Bayesian Model

Averaging (BMA) postprocessing for temperature, precipitation, and wind

– Development of both global and local BMA– Development of ensemble MOS (EMOS)

Grid-Based Bias Correction

• Use previous observations, land-use categories, elevation, and distance to determine and reduce bias in forecasts

-0.1

0.0

0.1

0.2

0.3

0.4

0.5

0.6

00 03 06 09 12 15 18 21 24 27 30 33 36 39 42 45 48

BSS

-0.1

0.0

0.1

0.2

0.3

0.4

0.5

0.6

00 03 06 09 12 15 18 21 24 27 30 33 36 39 42 45 48

*ACMEcoreACMEcore*ACMEcore+ACMEcore+Uncertainty

*UW Basic Ensemble with bias correction

UW Basic Ensemble, no bias correction

*UW Enhanced Ensemble with bias cor.

UW Enhanced Ensemble without bias cor

Skill forProbability of T2 < 0°C

BSS: Brier Skill Score

Bias Correction Substantially Improves Value of Ensemble Systems

BMA

BMA

• Testing both global BMA (same weights over entire domain) and local BMA (ensemble weights vary spatially).

EMOS

EMOS Test

EMOS Verification

Communication and Display

• Considerable work by Susan Joslyn and others in psychology and APL to examine how forecasters and others process forecast information and particularly probabilistic information.

• One example has been their study of the interpretation of weather forecast icons.

The Winner

PROBCAST

UW EnKF System

• Can we produce a superior 3D analysis?

• Can we use it to produce good short-term predictions…a major missing element in most systems?

• Can we use a significant proportion of the numerous observations that are now available?

UW EnKF Data Assimilation

• Now using a 36-4 km system with 3hr update• Previously, used 36-12 km and 6h update• Future: move to 1 hr update

EnKF 12km Surface Observations

EnKF 12-km vs. GFS, NAM, RUC

RMS analysis errors

GFS 2.38 m/s 2.28 KNAMRUCEnKF 12km

Wind Temperature

2.30 m/s2.13 m/s1.85 m/s

2.54 K2.35 K1.67 K

The END

Future Evaluation

• Improving PBL and surface drag may preferentially help 1.3 km (more later)

• Using 1.3 km as a testbed (some problems are more acute at higher resolution)

A Major Issue Has Been Excessive Wind Speeds Over Land and Excessive Geostrophy at the Surface –either

too much mixing in vertical or not enough drag. Winds over land and water too similar

• No magic bullet in PBL tests.• Recently, we tried something that really looks like

it has potential to help…increasing the friction velocity….ustar.

• Essentially adds drag, without messing other things up.

• Perhaps it is realistic, mimicking the effects of hills.