Transcript Regional High Resolution Prediction

Determining the Local
Implications of Global Warming
Clifford Mass and Eric Salathe,
Patrick Zahn, Richard Steed
University of Washington
Project Support
•
•
•
•
King County
Seattle City Light
EPA STAR Program
Climate Impacts Group
Questions
What are the implications of global
warming for the Northwest?
How will our mountains and land-water
contrasts alter the story?
Do global models tell us the full story?
Regional Climate Prediction
• To understand the impact of global warming, one
starts with general circulation models (GCMs) that
provide a view of the global evolution of the
atmosphere.
• GCMs are essentially the same as global weather
prediction models but are run with much coarser
resolution and allow the composition of the
atmosphere to vary in time (e.g., more CO2)
• Even leading GCMs only describe features of
roughly 500 km or larger in scale.
•Northwest weather is
dominated by terrain
and land-water
contrasts of much
smaller scale.
•In order to understand
the implications of
global changes on our
weather, downscaling
of the GCM predictions
considering our local
terrain and land use is
required.
Model Topography and Resolution
MM5 Topo (15 km)
ECHAM5 Topo (150km)
Annual
Precipitation
Downscaling
Downscaling
• The traditional approach to use GCM
output is through statistical downscaling,
which finds the statistical relationship
between large-scale atmospheric structures
and local weather.
• Statistical downscaling either assumes
current relationships will hold or makes
simplifying assumptions on how local
weather works.
Downscaling
Such statistical approaches may be a
reasonable start, but may give deceptive or
wrong answers… since the relationships
between the large scale atmospheric flow
and local weather might change in the
future.
Downscaling
• There is only one way to do this right…
running full weather forecasting models at
high resolution over extended periods, with
the large scale conditions being provided by
the GCMs….this is called dynamical
downscaling.
• Such weather prediction models have very
complete physics and high resolution, so
they are capable of handling any “surprises”
Example of Potential Surprises
• Might western Washington be colder
during the summer under global
warming?
– Reason: interior heats up, pressure falls,
marine air pushes in from the ocean
• Might the summers be wetter?
– Why? More thunderstorms due to greater
surface heating.
Downscaling
• Computer power and modeling
approaches are now powerful enough
to make dynamical downscaling
realistic.
• Takes advantage of the decade-long
work at the UW to optimize weather
prediction for our region.
UW Regional Climate Simulations
• Makes use of the same weather prediction
model that we have optimized for local
weather prediction: the MM5.
• 10-year MM5 model runs nested in the
German GCM (ECHAM).
• MM5 nests at 135 km, 45 km, and 15 km
model grid spacing.
MM5 Model Nesting
• 135, 45, 15 km MM5 domains
• Need 15 km grid spacing to model local weather features.
Regional Modeling
• Ran this configuration over
several ten-year periods:
• 1990-2000-to see how well the
system is working
• 2020-2030, 2045-2055, 20902100
Details on Current Study: GCM
• European ECHAM model with resolution roughly
equivalent to having grid points spaced ~ 150 km apart.
Can resolve features of roughly 600 km size or more.
• IPCC climate change scenario A2 -- aggressive CO2
increase (doubling by 2050)
IPCC Report, 2001
IPCC Report, 2001
Global Forcing: Surface Temperature
First things first
• But to make this project a reality we needed to
conquer some significant technical hurtles.
• Example: diagnosing and predicting future deep
soil temperatures
• Example: requirements for acquiring GCM output
every 6 h and storing massive amounts of output.
• Evaluating the 1990-2000 simulations
Evaluating of Model Fidelity
• We have carefully evaluated how well the GCM
and the MM5 duplicated the 1990-2000 period.
• We previously had run the system using another
GCM…the Parallel Climate Model…with
unsatisfactory results….crazy cold waves during
the winter.
• ECHAM Model appears far better…but not
perfect.
Too Cold
• Cold episodes occurred 1-2 times per winter
with temperature getting unrealistically cold
(below 10F) in Puget Sound:
• Also a general cold bias to minima
• Better than previous attempts.
Why Cold Outbreaks?
• Widespread surges of arctic air originate in
ECHAM5, likely owing to poorly-resolved terrain
(Cascades and Rockies).
• Extreme cold air inherited by MM5.
• Results from previous experiments with lowerresolution (T42) GCM indicate that higher resolution
reduces frequency and severity of unrealistic cold
events.
• Also problem in model physics--probably more
important
The Fix
• Our research during the past few months
suggests the problem was a bug in the land
surface model.
• Fixed in the current version and will be
used in next production runs.
Evaluation of Future Runs
Because there are some biases in the GCM
runs, results for future decades (2020s,
2040s, and 2090s) will be evaluated against
the ECHAM5-MM5 1990-2000 baseline
Now, The Future
Why Such Strong Warming on
Mountain Slopes..Particularly in
Spring?
• Probable Answer: Snow melt
resulting in more solar heating.
Change in
Water
Of
Snowpack
(%)
Snow and Ice Reflect Much of
The Incoming Solar Radiation
Solar Radiation
Now
Global Warming Causes Snow level to Rise
Resulting In Absorption of Solar Energy on
Melted Slopes
Solar Radiation
Future
=WARMING
Why Cooling West of Cascades
in Spring?
• Low clouds due to more onshore flow from
off the cool, cloud Pacific.
• The Montereyization of the western
lowlands!
Precipitation
• Bottom Line: No Large Regional Trends
Summary
• The viability of the approach…using high
resolution numerical prediction models forced
by large-scale general circulation climate
models (GCMs)… has been demonstrated.
• Careful evaluation of the GCM output is
required…there are deficiencies.
• Although there is general warming over the
region for all seasons, the terrain and land
water contrasts of the region enhance or
weaken the warming in certain areas.
Summary
• Warming is enhanced on the upper windward slopes
due to snow melt.
• Springtime warming is lessened west of the Cascade
crest due to more low clouds.
• Many more hot days during the summer.
• Precipitation changes are more modest then
temperature changes.
• There will be a substantial loss of snowpack,
reaching catastrophic decreases by 2090.
Future Work
• We are just in the beginning of this work.
• Need to find and remove causes of biases and cold
outbreaks
• Need to test other global warming scenarios
• Will try to find higher resolution GCMs
• Try more sophisticated MM5 physics
• More analysis.
The END
• :
Climate Change in the Pacific
Northwest:
Do Global Models Tell the Whole Story?
Eric Salathé
CSES Climate Impacts Group
University of Washington
With: Cliff Mass, Rick Steed, Patrick Zahn
WSU, USDA Forest Service, NCAR
IPCC Scenarios for Pacific Northwest
Climate Change
IPCC Scenarios for Pacific Northwest
Climate Change
Downscaling Methods Used in CIG
Impacts studies
Empirical Downscaling
• Assumes climate model
captures temperature
and precipitation trends
Regional Climate Model
•Represents regional
weather processes
• May produce local
trends not depicted by
global models
Mesoscale Climate Model - ECHAM5 Climate Model used to force Mesoscale Simulation
MM5
 Based on MM5 Weather Model
 Nested grids 135-45-15 km
 Nudging on outermost grid by forcing global model
 Advanced land-surface model (NOAH) with interactive deep soil
temperature
More Rain
1950-2000 to 2050-2100 Nov-Dec-Jan
Shift
in Pacific
Storm
Composite
of 10 Global
ModelsTrack
“Observed” Climate
20th Century Model Composite
NCEP-NCAR Reanalysis
21st Century Model Composite
Salathé, Geophys Res Lett, 2006
MM5 Result for Sep-Oct-Nov
Increased
Westerly
Flow
Contours:
Change in
500-mb heights
Change in Sep-Oct-Nov Precip (mm/day)
1990s to 2050s
MM5 vs Statistical Downscaling
Statistical Downscaling
Precip only
Precip & Winds
MM5
Change in November Precip (mm/day)
1990s to 2050s
More Warming
Winter Warming in MM5
1990s to 2050s
Temperature Change
Difference between
MM5 and ECHAM5
Less
warming
In MM5
More
warming
In MM5
Change in Winter Temperature (degrees C)
Change in Winter Temperature (degrees C)
Loss of Snow cover and Warming
Temperature Change
Change in Winter Temperature (degrees C)
Snow Cover Change
Change in fraction of days with snow cover
January to April in MM5
-50
Comparison of MM5 and HCN
Observations
Consistent trend over 21st Century
2020s
2050s
Change in Winter Temperature (degrees C)
2090s
Winter Trends at Various Stations
Temperature Change (°C)
MM5 - ECHAM5
Winter Trends at Various Stations
10 IPCC Models
Temperature Change (°C)
MM5 - ECHAM5
1950
2000
2050
2100
Regional Model Compared to Global
Model
2020s
2050s
Change in Winter Temperature (degrees C)
2090s