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   (