Toward seamless prediction of weather and climate

The Impact of Land Surface on Sub-seasonal
Forecast Skill
Zhichang Guo and Paul Dirmeyer
Center for Ocean-Land-Atmosphere Studies
George Mason University, Fairfax, USA
World Weather Open Science – 17 August 2014
Background
•
Soil moisture is a potentially critical component of
subseasonal to seasonal prediction system, especially
over continental midlatitude areas where SST has limited
impacts on;
•
Previous studies of land-atmosphere interaction and land
surface impacts on atmospheric prediction have been
limited by a lack of observational data and by the model
dependence of the results;
World Weather Open Science – 17 August 2014
GLACE-1 and GLACE-2
•
Recently, two multi-institutional numerical modeling
experiments, GLACE-1 and GLACE-2 (Global LandAtmosphere Coupling Experiments) have been designed
and performed to study systematically impacts of land
surface on atmospheric variability and its prediction;
•
Advantages: it allows to synthesize commonalities among
various models which are less subject to problems in the
process parameterizations and allows the inter-model
comparison.
World Weather Open Science – 17 August 2014
GLACE-1 Models
GLACE-2 Models
1. GFS/OSU, NCEP, USA;
1. NCEP GCM, Princeton University, USA;
2. GEOS, NASA/GSFC, USA;
2. GEOS, NASA/GSFC, USA;
3. GFDL, USA;
3. GFDL, USA;
4. CAM/CLM, NCAR, USA;
4. CAM/CLM, NCAR, U. Gothenburg, Sweden;
5. CCCma, Canada;
5. ECMWF;
6. HadAm3, UK;
6. KNMI, Netherlands, ECMWF;
7. BMRC, Australia;
7. ECHAM GCM, IACS , Switzerland;
8. CSIRO-CC3, Australia;
8. CCCma, Canada;
9. CCSR, Japan;
9. NSIPP, NASA/GSFC, USA;
10. UCAL, USA
10. COLA AGCM, COLA/GMU, USA;
11. NSIPP, NASA/GSFC, USA;
11. FSU/COAPS. FSU, USA
12. COLA AGCM, COLA/GMU, USA;
World Weather Open Science – 17 August 2014
GLACE-1: Does atmosphere care
about anomalies in soil moisture?
• A question asked in GLACE-1: do changes in soil
moisture anomaly have impacts on atmospheric
variability? (commonalities among various
models and inter-model comparison).
• Obviously, if changes in soil moisture anomaly
are being ignored by the atmosphere, it won’t
contribute to the atmospheric prediction skills.
World Weather Open Science – 17 August 2014
Hotspots: regions with strong coupling strength
where 12 participating models are relatively consistent
This figure received
extensive attention due to
the following reasons:
1. It is found soil moisture
has strong impacts on
atmospheric variability
over some areas, namely
hotspots, ;
2. Hopefully, realistic soil
moisture initialization
could improve
significantly subseasonal
to seasonal prediction
over these areas.
World Weather Open Science – 17 August 2014
Topic of the GLACE-2
•
If soil moisture has impacts on atmospheric
variability, the realistic initialization of soil
moisture should contribute to the subseasonal to
seasonal forecast skill. Main topic of GLACE-2;
•
Will the stronger L-A coupling result in a larger
contribution to the subseasonal forecast skill?
•
whether soil moisture over hotspots contribute
significantly to the subseasonal to seasonal forecast?
World Weather Open Science – 17 August 2014
Series 1:
GLACE-2: Experiment Overview
realistic initial land
surface states
realistic initial atmospheric
states
Series 2:
realistic initial land
surface states
realistic initial
atmospheric states
World Weather Open Science – 17 August 2014
Perform
ensembles of
retrospective
seasonal forecasts
Evaluate
forecasts
against
observations
Prescribed, observed SSTs
Perform
ensembles of
retrospective
seasonal forecasts
Prescribed, observed SSTs
Evaluate
forecasts
against
observations
Forecast skill of air temperature(July, 1982-2006)
World Weather Open Science – 17 August 2014
Commonalities: realistic initialization
soil moisture contributes most over
Southwestern and northern USA; The
contribution is insignificant over the
Great Plains (one of the hotspots).
Why? stronger L-A coupling does not
guarantee a larger contribution to the
subseasonal forecast skill.
World Weather Open Science – 17 August 2014
Our explanation: stronger L-A coupling
can result in a larger contribution to the
subseasonal forecast skill only if the soil
moisture can be predicted properly. In fact,
forecast skill over hotspots are vulnerable
since the model biases in the land surface
there could be amplified into the atmosphere
through strong L-A coupling.
Soil moisture forecast skill and the role of soil moisture memory
Soil moisture forecast
skill heavily relies on
initial quality of soil
moisture. As the lead
time increases, it is
more dependent on
soil moisture memory
Soil moisture memory:
how long does the soil
moisture remember its
anomaly into the
future
World Weather Open Science – 17 August 2014
Forecast skill of temperature and soil moisture
Over hotspots, forecast
skill of air temperature is
low due to lack of forecast
skill in soil moisture,
though land-atmosphere
coupling is strong there
Forecast skill of air
temperature is high only
over the areas where L-A
coupling is strong and the
forecast skill of soil
moisture is large
World Weather Open Science – 17 August 2014
Forecast skill of
air temperature
Forecast skill of air
temperature is high
when the product
of forecast skill in
soil moisture and
land-atmosphere
coupling strength is
large
World Weather Open Science – 17 August 2014
Comparison between Region A and Region B
World Weather Open Science – 17 August 2014
Inter-model comparison: forecast skill of temperature
World Weather Open Science – 17 August 2014
Inter-model comparison: forecast skill of soil moisture
World Weather Open Science – 17 August 2014
Inter-model difference in SM forecast skill explains
inter-model difference in temperature forecast skill
World Weather Open Science – 17 August 2014
Conclusion
• Forecast skill of soil moisture relies on the quality of soil moisture
initialization and persistence of the soil moisture anomaly;
• Subseasonal forecast skill of temperature relies on the accurate
prediction of soil moisture and strong L-A coupling;
• Realistic initialization of soil moisture contributes most to the
subseasonal forecast over the Southwestern USA due to relatively
strong land-atmosphere coupling and the high prediction skill of soil
moisture there; its contribution to subseasonal predication of
temperature is insignificant over the Great Plains due to the relatively
low soil moisture prediction skill, even though the L-A coupling is
strong there;
• The inter-model differences in temperature forecast skills could be
explained by the inter-model difference in soil moisture forecast
Worldskills.
Weather Open Science – 17 August 2014
Thank You!
World Weather Open Science – 17 August 2014