Forecast Techniques
ATMOS 5010: Weather Forecasting Forecasting Tools and Techniques
Jim Steenburgh Department of Atmospheric Sciences University of Utah
[email protected]
Successful Forecasting Requires n
– – – – – n
AKA: The Human-Machine Mix
Knowledgeable, well-trained, & engaged forecasters Meteorological knowledge and experience Local weather & climate knowledge User need recognition Model strength, weakness, and bias assessment Human cognition and interpretation
Skillful & reliable NWP guidance, forecast t ools, and other aids Image: whistlerdiaries.com
Human Cognition Automated Systems
The Forecast Process
Critical Forecast Questions n n
What has happened? Why has it happened?
n
What is happening? Why is it happening? What will happen?
n
Why will it happen?
n n
Easy to concentrate only on this
M orss an d Ralp h (2 0 07)
= NWP, tools, aids, etc. = Knowledgeable, well trained, & engaged forecasters
S o u rce : Bo sa rt (2 0 0 3 )
1
The Forecast Methodology
Critical Forecast Questions n
What has happened?
n
n
Why has it happened? What is happening? Why is it happening? What will happen?
n
Why will it happen?
n n
n
Important when NWP goes awry or cannot resolve local orographic effects
To answer these questions, use the forecast funnel – Begin at planetary scale – Focus attention on progressively s maller scales – In complex t errain, build in orographic effects
Planetary Scale Synoptic Scale Mesoscale Local Scale
S o u rce : Bo sa rt (2 0 0 3 )
The Forecast Methodology n
Answer the what and the why in the past, present, and future
n
Avoid “meso-myopia”
– Understand larger scales before progressing to smaller scales – When using high-resolution models, evaluate confidence in large-scale forecast before progressing to smaller scales – Expect limited local skill if large- scale is not well forecast
n
Forecast Funnel in Practice
Beware when the atmosphere is in outlier mode
Planetary Scale Synoptic Scale Mesoscale
– Generalizations break down
Local Scale
The Forecast Funnel in Practice
Evaluate past, current, and future planetary scale setting
The Forecast Funnel in Practice
Evaluate confidence in synoptic-scale forecast Funnel to synoptic scale
2
The Forecast Funnel in Practice
The Forecast Funnel in Practice
?
?
?
?
? Funnel to mesoscale Consider mesoscale, o rographic, and land-surface processes
Adjust for local e ffects
Real-Time Example Using IDV
Humans Make a Difference
S ou rce: NCEP/HPC
“This continuing skill advantage [indicates] that dedicated and trained forecasters can extract maximum advantage from improvements in operational weather prediction models” -Bosart (2003)
On the other Hand….
“Forecasters who grow accustomed to letting MOS and the models do their thinking…on a regular basis…are at high risk of “going down in flames” when the atmosphere is in an outlier mode” - Bosart (2003)
Don’t be on Autopilot
Although NWP is important, basic understanding, pattern recognition and climatology c ontinue t o play an essential role because of limitations in c urrent NWP s ystems, including inadequate t errain representation, initial condition uncertainty, and parameterization uncertainty
3
Forecast Tools
Bottom Line n
“Forecasters have a clear role in the forecast process, by contributing a wealth of knowledge, tools and techniques that cannot be duplicated by computers or NWP” – McCarthy et al. (2007)
n
But forecasters need to be engaged and increasingly need an advanced education to extract maximum benefit from today’s sophisticated forecast tools
n
This class begins that education
Forecast Tools
Forecast Tools Climatology Persistence n Observations n n
– – – – – –
A meteorologist knows their tools, including their strengths and weaknesses “All observations are bad, but some are useful”
n n
“All models a re wrong, b ut some are useful” n n
Your eyes In-situ surface and upper-air Wind profiler/RASS Satellite Radar Weather cameras
Manual analysis NWP Models
– Numerical analyses – Global and mesoscale models – Ensemble forecast systems
Model Output Statistics (MOS) Scientific analysis and visualization systems
S o u rce : Ga ry La rso n , Th e F a r S i d e
S o u rce : Ga ry La rso n , Th e F a r S i d e
Forecast Tools: Climatology
Forecast Tools: Climatology
n
The statistics of weather
n
More than just long-term mean – Mean, variance, extremes, probabilities – Impacts of ENSO and modes of climate variability § PDO, NAO, etc.
n
Local and mesoscale effects – Complex terrain results in large climatological gradients – Often poorly resolved by computer models – Climatology to used “downscale” or “bias correct” model forecasts for local effects – Can be overused § e.g., Not all storms have the climatological precipitation-altitu de relationship
Is this useful???? Means and Probabilities for Forecast Practicum Variables
4
Forecast Tools: Climatology
Forecast Tools: Persistence n
Persistence: What has happened recently – Including trends
n
Provides context for forecast
n
Relevance for forecast varies from high to low – High during slowly evolving patterns – Low during major pattern shifts
Think Beyond the Mean
Forecast Tools: Persistence
Forecast Tools: Your Eyes n
Context for forecast During this period is Different at LGU & WBB
Forecast T ools: Sfc/Upper-Air Data
Never underestimate the value of looking out the window or going outside to feel the weather
S o u rce : ca rto o n sti ck.co m , co l l a b o ra ti ve jo u rn e ys.co m
Forecast Tools: Satellite
ASOS, Springfield, IL ( NWS)
Weather Balloon ( NWS)
Wind Profiler ( Wikipedia)
Wind profilers provide more than wind!
Visible Imagery Visible radiation reflected back to space by clouds, aerosols, snow, land surface, etc.
5
Forecast Tools: Satellite
Forecast Tools: Satellite
“Window” IR Imagery Long-wave radiation emitted primarily by clouds, land-surface, etc. Cloud-top temperature and land-surface temperature
Water Vapor Channel (IR) Imagery Long-wave radiation emitted primarily by upper-tropospheric clouds and water vapor Upper-level flow, troughs, etc.
Forecast Tools: Satellite
Forecast Tools: Radar
Precipitable Water from Polar-orbiting microwave s ensors
NEXRAD Doppler Radar NEXRAD vs. TDWR
Now: Polarimetric NEXRAD GOES Fog D etection Longwave IR ( 10.7 micron) -Shortwave IR ( 3.9 micron)
MODIS
S o u rce s: S S E C, NE S DIS
S o u rce s: NOAA/S P C
Forecast T ools: Weather Cameras
Forecast T ools: Manual Analysis
Click for Animation
S o u rce s: Bo sa rt a n d S e i m o n (1 9 8 8 );; Ne i m an e t al . (1 9 8 8)
A manual surface analysis helps you “feel the weather in your veins”
6
Useful Sites for Observations n http://mesowest.ut ah.edu
n n
n n
– Surface observations & t ime s eries – Radar overlays http://weather.r ap.uc ar.edu – Satellite, radar, s urface m aps, upper-air m aps http://www.spc.noaa.gov/ ex per – Upper-air s oundings, upper-air m aps, s urface mesoanalysis http://www.wunder gr ound.com /w underm ap/ – You name it http://weather.c od.edu/satr ad/ – Satellite and radar
Useful IDV Bundles for Obs n
– KMTX-3DTopo – KMTX-2D-Obs+Anal n
Global Forecast System (GFS) – Medium range (out to 384 hours) global analyses a nd forecasts every 6 h – Effective grid spacing of ~13 km to 192 h (lower resolution thereafter) – Available on lower-resolution grids – Strengths relative to other NCEP models
Forecast T ools: NWP Models n North American Mesoscale Model (NAM) – Based on the “WRF-NMM” – Short-range (out to 84 hours) forecasts f or North America every 6 h – Grid spacing of ~12 km § Higher resolution 4-km CONUS nest available
– Available on lower-resolution grids – Strengths relative to other NCEP models
§ Accuracy of large-scale f orecast
§ Terrain representation, mesoscale detail
– Weaknesses
– Weaknesses
§ Terrain representation § Precip structure
§ Limited area, large-scale accuracy
Forecast T ools: NWP Models n Rapid Refresh (RAP) – Analyses for CONUS every hour – Very-Short-range (out to 18 hours) forecasts for CONUS every 3 h – Grid spacing of ~13 km – Available on lower-resolution grids – Strengths relative to other NCEP models § High frequency analyses and forecasts § Resolution, terrain representation, mesoscale details
– Weaknesses § Limited area, large-scale accuracy
Real-Time-WX>Analyses> – Global-10day – Global-2day – Supernational – Conus-East – Conus-West
Forecast T ools: NWP Models n
Real-Time-WX>Radar>
Forecast T ools: NWP Models n
High Resolution Rapid Refresh (HRRR) – Analyses for CONUS every hour – Very-Short-range (out to 12 hours) forecasts for CONUS every hour – Grid spacing of ~ 3 km – Strengths relative to other NCEP models § High frequency analyses and f orecasts § Resolution, t errain representation, m esoscale details
– Weaknesses § Limited area, large-scale accuracy
7
Forecast Tools: NWP Models n
Weather Research and Forecast Model (WRF)
Forecast T ools: NWP Models n Short Range Ensemble Forecast System
(SREF) – 21 members @ 16-km grid spacing based on differing models, model configurations, and initial conditions – Forecasts out to 87 h every 6-h (0300 UTC, etc.) – Strengths
– Run in various configurations at NCEP and other locations – Some configurations provide high resolution (