Transcript Document

Challenges and Directions for
NOAA’s Weather Modeling
for Renewable Energy
Stan Benjamin
Research Meteorologist
Chief, Assimilation and Modeling Branch http://ruc.noaa.gov/amb
NOAA Earth System Research Laboratory, Global Systems Division
Boulder, CO
NOAA/ESRL- John Brown, Jim Wilczak, Joe Olson,
Wayne Angevine, Melinda Marquis, Bob Banta
NOAA/NCEP - Geoff DiMego
AMS Summer Colloquium
12 Aug 2009, Norman, OK
NOAA plans for improving renewable energy forecasts
1. More accurate weather forecasts (esp. wind and solar) will
allow more accurate forecasts of transmission and storage
needs, carbon-based power, “load”, “ramps”, etc.
a. Costly inaccuracy exists currently in forecasts of all scales
(1h to 15 days) for energy.
2. What’s needed:
a. Better models, especially in treatment of 10-200m winds,
clouds, addition of chemistry/aerosols, including hourly
updating models (RUC, RR, HRRR)
b. Better global models for 1h-15d+ forecasts (and climate).
c. Better observations above surface to initialize models winds, moisture, temperature, clouds.
d. Higher-resolution models, ensemble forecasts/uncertainty
e. Much improved data assimilation - model initialization
f. Interaction between NOAA, DOE, NREL, other govt,
energy industry, other private sectors
Common users: Energy, aviation, severe weather, fire wx, AQ, hydrology
Hourly Updated
NOAA NWP Models
13km Rapid Refresh domain
RUC – current
oper model - 13km
Rapid Refresh
(RR) – replace RUC
Current RUC CONUS domain
at NCEP in 2010 - WRF,
GSI w/ RUC-based
enhancements
HRRR - Hi-Res
Rapid Refresh
-Experimental 3km
12-h fcst updated
every hour
2009 3km
HRRR domain
Hourly Updated
NOAA NWP Models
Experimental 13km Rapid Refresh
domain
• 13km RUC
• 13km Rapid
Refresh
• 3km Hi-Res Rapid
Refresh (HRRR)
Storm-resolving
(3-km) model init
from radarenhanced RUC or
RR
For aviation, severe
weather, energy,
fire weather,
emergency
situations, etc.
Current RUC-13 CONUS domain
Exp 3km HRRR
domain – Feb2010
National Airspace Planning in Presence of
Convective Weather
Playbook:
revised
at 1300
UTC
Initial Playbook
at 1100
UTC
13Z CCFP Forecast
Fcst valid : 15 – 21 UTC : 02 Aug 08
Air traffic management analogy to
regional/national power management
- both need accurate hourly updated
accurate, weather information
Rapid Refresh Hourly Assimilation Cycle
Cycle hydrometeor, soil temp/moisture/snow Hourly obs
plus atmosphere state variables
Data Type
~Number
Rawinsonde (12h)
150
NOAA profilers
35
1-hr
1-hr
1-hr
VAD winds
120-140
PBL – prof/RASS
~25
fcst
fcst
fcst
Aircraft (V,temp)
3500-10000
Background Analysis
TAMDAR (V,T,RH)
200-3000
Fields
Fields
Surface/METAR
2000-2500
Buoy/ship
200-400
GSI
GOES cloud winds
4000-8000
GSI
3dvar
GOES cloud-top pres 10 km res
3dvar
GPS precip water
~300
Obs
Mesonet (temp, dpt)
~8000
Obs
Mesonet (wind)
~4000
METAR-cloud-vis-wx
~1800
AMSU-A/B/GOES radiances
11
12
13 Time
Radar reflectivity/ lightning
(UTC)
1km
00z+6h
HRRR
No radar
Effective radar refl
assimilation
(via 13-km RUC)
critical
HRRR 3-km
forecast
00z+6h
HRRR
RUC
radar
NSSL verification
06z 16 Aug 2007
RUC Hourly Assimilation Cycle - fall 2009
18-h
fcst
Background
Fields
11
18-h
fcst
18-h
fcst
18-h
fcst
18-h
fcst
18-h
fcst
Analysis
Fields
3DVAR
3DVAR
3DVAR
Obs
Obs
Obs
12
13
14
3DVAR 3DVAR
Obs
15
Obs
16
Time
(UTC)
Obs impact for 312h forecasts for
near-sfc winds “Relative short-range
forecast impact from
aircraft, profiler,
radiosonde, VAD,
GPS-PW, METAR and
mesonet obs
within hourly
assimilation in the
RUC”
-Benj et al., 2010,
Mon. Wea. Rev., accepted
for publication
Sat,
, radar
Example:
11 Aug 09
http://Ruc.noaa.gov
RUC fcst –
80-120m wind power
potential
RUC
- cld
RUC fcst - cld
RUC fcst –
Solar rad
RUC 3h forecast of 1000’ ceiling (IFR flight
conditions)
-probability of detection – verification against
METAR observations over US every 3 h
2004
2005
2006
2007
2008
2009and METAR assimilation in Nov 2008
Improvements
in GOES-cloud
12
and March 2009 have led to major improvement in NCEP RUC cloud forecasts
Two main points on improved weather models for RE:
1. Atmospheric forecasts from weather models for both wind
energy and solar energy are outputs from the same model.
• Improved model and assimilation for better cloud/sun
forecasts also improve near-surface wind forecasts,
and conversely,
• Improve for better near-surface wind forecasts will
increase accuracy for forecasts of clouds, precipitation,
and even winds above the PBL.
Consistent forecasts for both wind energy and solar
energy generation potential are from the same physical
atmospheric system.
2. Equally complicated problems: Development of
a. Weather model and
b. Data assimilation (to initialize model)
Both are important for 1h to 15-day forecasts
Two main points on improved weather models for RE
1. Atmospheric forecasts from weather models for both wind
energy and solar energy are outputs from the same model.
• Improved model and assimilation for better cloud/sun
forecasts also improve near-surface wind forecasts,
and conversely,
• Improve for better near-surface wind forecasts will
increase accuracy for forecasts of clouds, precipitation,
and even winds above the PBL.
Consistent forecasts for both wind energy and solar
energy generation potential are from the same physical
atmospheric system.
2. Equally complicated problems: Development of
a. Weather model and
b. Data assimilation (to initialize model)
Both are important for 1h to 15-day forecasts
PBL research – NOAA
Winds and TKE - WRF
MYJ PBL
Aircraft Observations of TKE (m2 s-2)
MYNN PBL
Olson, Brown, Angevine – NOAA-ESRL
Upcoming PBL tests verifying w/
Industry tall tower 60m data
NOAA’s NWS Model Production Suite
Climate
CFS
Oceans
Hurricane
GFDL
HWRF
MOM3
HYCOM
WaveWatch III
~2B Obs/Day
Mostly
Satellite
+Radar
Global Data
Assimilation
Global
Forecast
System
North American
Mesoscale
WRF NMM
Dispersion
ARL’s HYSPLIT
Severe Weather
North American Ensemble
Forecast System
GFS, Canadian Global Model
Short-Range
Ensemble Forecast
WRF: NMM+ARW
ETA, RSM
NOAH Land Surface Model
WRF NMM
WRF ARW
Air Quality
NAM+CMAQ
Rapid Update
for Aviation, Severe
Weather, Energy
For
eca
st
Very Short-Range Ensemble Forecasts - VSREF
- Updated hourly w/ available members valid at same time
RR – hourly
time-lagged (TL) ensemble members
-2010 - 3km CONUS HRRR at ESRL
(incl. TL ensemble)
- 2012 - ensemble RR – 6 members
- 2012 - proposed HRRR at NCEP
future HRRR ensemble from RR-ens
NAM / NAM ensemble
VSREF –
Hourly
Updated
Probabilistic
Forecasts
= TL+
ensemble
GFS / GFS ensemble
SREF (updated every 6h)
VSREF to include subhourly
- merged RCPF/extrap
- similar for icing, turb, etc.
The vision for highfrequency weather
models in NextGen
Model
Ensemble
Members
- hourly (1h)
updated
HRRR
VSREF
members HRRR, RR,
NAM, SREF,
GFS, etc.
Explicit met variables
from each VSREF
member - V,T,qv,q*
(hydrometeors),p/z,
land-surface, chem,
etc.
Unified Post-processing
Algorithms (modularized!!) for
following: (multiple where
appropriate), built on current
WRFpost from NCEP
Turb (e.g., GTG)
Icing (e.g., FIP)
Ceiling
Visibility
Convection
ATM route options
Wake vortex
Terminal forecast
Object diagnosis
VSREF mems
output for
each AIV
variable
For turb
(line convection,
clusters, embedded)
4D-datacube
using
recent
VSREF mems
output - stat
obs
For turb
Optimal weighting
Others…
Potentially multiple variables
under each Avx-Impact-Var
(AIV) area
Stat correction
Stat
post- correction
processing
postusing recent obs
+
blending
processing
corrected
SAS
VSREF-
VISION: Toward estimating
and reducing
forecast uncertainty for
aviation applications
using high-frequency data
assimilation
Most-likely-estimate
single value
Probability/PDF output
Subset of full domain
NOAA/ESRL FIM global model
10km, 15km - horizontal resolution
(highest yet resolution real-time global
model runs in NOAA)
• New global model design
• icosahedral soccer-ball-like grid
• isentropic-sigma vertical
coordinate
• 3-5km global model now in design FIM extension - “NIM”
NOAA is increasing effort to improve
global model skill vs. ECMWF
10km FIM global model runs
out to 7 days daily
- NSF Texas supercomputer
since Aug 2008
-- NOAA experimental
Hurricane Forecast
Improvement Project (HFIP)
- http://fim.noaa.gov
Actual wind energy
potential - strong
regional and
temporal variations
Outstanding
RUC 1h fcst
80-120m wind energy
Valid 23z Wed 12 Nov 08
Wind energy - V3 - 80-120m wind / Solar energy (shaded)
6h RUC forecast - initialized 16z Wed 12 Aug 2009
http://ruc.noaa.gov/pig.cgi?13km_D2
13km development RUC
NOAA plans for improving renewable energy forecasts
1. More accurate weather forecasts (esp. wind and solar) will
allow more accurate forecasts of transmission and storage
needs, carbon-based power, “load”, “ramps”, etc.
a. Costly inaccuracy exists currently in forecasts of all scales
(1h to 15 days) for energy.
2. What’s needed:
a. Better models, especially in treatment of 10-200m winds,
clouds, addition of chemistry/aerosols, including hourly
updating models (RUC, RR, HRRR)
b. Better global models for 1h-15d+ forecasts (and climate).
c. Better observations above surface to initialize models winds, moisture, temperature, clouds.
d. Higher-resolution models, ensemble forecasts/uncertainty
e. Much improved data assimilation - model initialization
f. Interaction between NOAA, DOE, NREL, other govt,
energy industry, other private sectors
Common users: Energy, aviation, severe weather, fire wx, AQ, hydrology