EXPERIMENTAL CLIMATE PREDICTION CENTER
M. Kanamitsu, G. Auad, D. Boomer, L. DeHaan, H. Kawai, H. Li, J. Ritchie, K.
Yoshimura and E. Yulaeva
Scripps Institution of Oceanography
Climate, Atmospheric Science and Physical Oceanography
University of California San Diego, 0224
La Jolla, CA 92093-0224
phone:(858) 822-5176
fax: (858) 534-4163
email: mkanamitsu@ucsd.edu
http://ecpc.ucsd.edu/
ECPC progress report to NOAA CPO/CDEP: April 1, 2009-May, 2010
Reported by M. Kanamitsu (mkanamitsu@ucsd.edu)
0. Executive summary
1. Continued improvement of global to regional model system
1.1. Regional Ocean Model System (ROMS) coupling with G-RSM
1.2. Non-Hydrostatic Model
1.2.1. Case 1. Mountain Valley wind. Palm Springs area.
1.2.2. Case 2. Santa Ana event. Southern California region.
1.2.3. Case 3. Atmospheric river event. Northern California.
1.2.4. Case 4. Diurnal variation over coastal ocean. San Diego area.
1.3. Global ROMS
1.4. Marine stratocumulus parameterization
2. Application Studies
2.1 Intercomparison of global change downscaling over California
2.1.1. Implementation of the incremental interpolation into dynamical
downscaling of CMIP-type output
2.1.2. Climatic changes in surface diagnostic variables
2.1.3. Mechanism of future increase of evening precipitation in dry season
2.2. Multi-model Regional Ensemble Downscaling (MRED) project
2.3. Historical downscaling of reanalysis over the Far East
2.4. Experimental ensemble seasonal forecast for IRI and NCEP
2.5. Seasonal forecast of Southwest Monsoon
2.6. Regional ocean model applications
2.7. Application to inversion study
2.8. Application to wind power energy study
3. Basic research
3.1. Low frequency mode error in dynamical downscaling
3.2. Downscaling strategy for ensemble forecasts
3.3. Isotope reanalysis and research
3.4. Theoretical consideration of dynamical downscaling
3.5. Diurnal variation and monsoon
4. G-RSM outreach
5. References
5.1 References (External)
5.2 ECPC Refereed Publications 2008-2010
5.3 Other ECPC Publications 2008-2010
6. Figures (separate document)
0. Executive summary
The highlights of 2009-2010 are:
(1) Successful implementation of atmosphere-ocean coupled regional downscaling
system for the first time,
(2) continued improvement of the dynamical downscaling system,
(3) deeper understanding of the concept of downscaling and regional predictability,
(4) initiation of downscaling by a non-hydrostatic model,
(5) completion of the downscaling of several global change simulations and their
analyses,
(6) participation in several international and national intercomparison projects and,
(7) basic studies of water isotopes and the impact of diurnal variation on monsoons.
The ECPC Global to Regional Spectral Model (G-RSM) has been, and continues
to be a basis of the atmospheric and land surface component of the ECPC forecast system.
The Regional Ocean Model System (ROMS) was recently added to this system as the
official oceanic component. A parallel coupling of G-RSM and ROMS is now complete,
and extensive performance tests are ongoing. A non-hydrostatic cloud resolving model
operationally used at the Japan Meteorological Agency has been imported for the
downscaling of scales less than 10 km. The user interface to launch the models was
significantly improved. The system can launch the global model, regional model,
non-hydrostatic model, ocean model and atmospheric data assimilation with ease.
ECPC participated in a series of intercomparison experiments. Our downscaling
system continues to be one of the best performing and most reliable models among the
participants. In collaboration with California downscaling groups, we performed
downscalings of four global change simulations (CCSM, GFDL, ECHAM5 and MIROC);
10 years each for both present and future climate. Analysis of the simulations showed an
interesting increase in precipitation over southern desert areas during the summer monsoon
season in the global warming situations, which was common to all four global models.
A study of the low-frequency error of the downscaled analysis is now complete
and the publication is in the printing stage. The work clearly demonstrated that the
inter-annual variability is severely contaminated by regional model errors, but the errors
can be removed by applying spectral nudging. This finding is critical for climate research
when the downscaled analysis is used to study low frequency variability. As a by-product,
it was shown that the RSM does not suffer from reduction in high frequency kinetic energy,
which is common in other grid point based models, and the error in the low frequency
mode is not related to this particular error.
The work on parameterization, model numerics, seasonal time scale predictability
and forecast skill studies continues. A study of marine strato-cumulus was completed and
published. For regional predictability study, methods of downscaling ensemble mean
have been developed.
The stable water isotope study has been extended to the regional model and
applied to a case study of an atmospheric river event. Comparing with station
observations of water isotopes in precipitation, the causes of the isotope changes are
discussed. This work has been accepted for publication.
Work on the role of diurnal variation in large scale monsoon circulation is in
progress. Diurnal variation has a significant impact on daily mean and consequently on
seasonal mean via the nonlinear transport of heat and moisture by diurnal eddies. This
non-linear interaction between diurnal frequency and seasonal mean is an exciting topic for
further research.
1. Continued improvement of global to regional model system
The global and regional spectral models, originally developed at NOAA NCEP
and transferred to Scripps in the 1990s, have been constantly updated for research and
application during the past several years (e.g., Kanamitsu et al. 2005). The regional and
global models have now been merged and coupled, simplifying maintenance considerably
(see 2009 ECPC progress report available at http://ecpc.ucsd.edu/psreports/). Various
system tools have been developed to help students and new users utilize the model more
easily for their research. The models and their outputs have been used extensively by
Scripps graduate students. The system’s dynamical core, numerics, and physical
parameterizations are continuously being modified and improved (Hong et al., 1996;
DeHaan et al., 2007; Shimpo et al., 2008). This is a team effort involving national and
international collaborators. The model has its own wiki web page for communication and
consultation among collaborators (http://g-rsm.wikispaces.com).
A significant improvement to the user interface utility was made in 2009. This
upgrade makes it possible to launch atmospheric models, ocean models and coupled models
without knowing much about the details of the model. The model launcher system has
been redesigned to be as portable as possible, such that transfer of the system to a new
computer system can be easily incorporated with very few changes. The forecast system
can now launch the global model (GSM), regional model (RSM), reanalysis-2 data
assimilation (R-2, Kanamitsu et al., 2002), single column model (SCM, Seol et al., 2010),
regional ocean model (ROMS,
http://www.ocean-modeling.org/index.php?page=models&model=ROMS), JMA
non-hydrostatic model (NHM, Saito et al., 2006) and coupling of GSM and RSM as well as
RSM and ROMS. The system is ideal for use by beginning graduate students, and for
transfer to other institutions as well as to organizations in foreign nations.
1.1. Regional Ocean Model System (ROMS) coupling with G-RSM
The highlight of the model development this year is the coupling of RSM and
ROMS, attempted first by Haidvogel, et al. (2000). More work on the coupling was
performed by H. Seo for his PhD dissertation (SCOAR system, Seo et al., 2007a). His
continued effort resulted in several important papers on regional scale interaction between
ocean and atmosphere (Seo et al., 2007a, b). Similar work is now in progress by a new
PhD student, Dian Putrasahan. Seo and Putrasahan utilized climatological forcing rather
than the time varying observations, since these studies are strongly research oriented
without much consideration of practical application. Encouraged by the excellent results
of these studies, ECPC has started to couple ROMS with the G-RSM system incorporating
observed ocean and atmospheric initial and lateral boundary conditions. Our first objective
is to perform downscaling of historical ocean analysis and atmospheric analysis, both in
coupled and uncoupled mode, to study the impact of coupling on regional scale
atmospheric and oceanic analysis. This preliminary study has direct relevance to the study
of coupled downscaling of seasonal forecast and global warming simulations to be
performed in the future.
There are several sources for ocean initial and lateral boundary conditions. The
most popular method is to initialize the ocean model with climatology, and force the ocean
by observed atmospheric forcing for several years, until the ocean spins up. After many
trials, we decided to abandon this method since it is too costly and cannot incorporate ocean
observations. We then tried to use available ocean analysis. Simple Ocean Data
Assimilation (SODA, Carton et al., 2000) was selected as our first candidate. When
utilizing this analysis, the initial setting of the ocean model was found to be critical. The
placement of the model vertical levels, smoothing of the bathymetry and the coast lines and
elimination of lakes were all found to be essential for stable integration of the ROMS.
The original procedures which utilized MATLAB needed to be converted to Fortran
programs with objective smoothing programs to eliminate the human interaction process in
the initial setup. This made it possible to incorporate the ROMS into the G-RSM system
as well as to perform coupling within the G-RSM system. Using SODA, we are now able
to integrate the ROMS and coupled RSM-ROMS system very stably.
The regional ocean model also suffers from large scale error within the regional
domain. The error is mostly caused by the systematic error in the radiation fluxes
reaching the sea surface. ROMS has a built-in correction for radiation fluxes and fresh
water fluxes to avoid the development of large scale error. Our experiments with and
without correction in uncoupled mode indicated that large scale error in SST develops
within the domain if no corrections are applied. Figure 1 shows a comparison of the
ROMS integration with and without correction in uncoupled mode. For reference, the
uncorrected coupled integration result is also shown. The error in the RSM forcing was
found to be due to the shortcoming in the total heat flux, mainly the incoming long and
short wave radiation. In order to see this more clearly, we re-ran the RSM downscaling
with an improved cloud parameterization scheme and repeated the uncoupled experiment.
Figure 2 shows the comparison of the area averaged total heat flux from Reanalysis-2, an
older version of the California area downscaling of Reanalysis-2 and the latest version of
the downscaling. There are considerable differences among the models, but a decrease in
the heat flux of up to 25 watts/meter square during the summer months is apparent. The
radiation fluxes reaching the surface with the new cloud scheme compared better with the
Surface Radiation Budget (SRB, http://www.gewex.org/srb.html) analysis (not shown).
Figure 3 compares simulated and analysis SST averaged over the central domain. The
atmospheric forcing was computed from the old version of the RSM before 1994 and from
the new version thereafter. The sudden decrease in SST in 1994 indicates that the new
RSM forcing corrected the bias in SST but not entirely, indicating the need for further
improvement in the RSM total heat flux. Since significant work would be involved to
reduce the inconsistencies in the RSM total heat flux, we tentatively postponed the
improvement and instead utilized the ROMS built-in correction for coupled and uncoupled
integration experiments. As shown in Figure 4 the correction significantly reduced the
systematic error in the SST in the coupled model.
We have completed two 10 year uncoupled ROMS runs and one coupled run is in
progress. The two uncoupled runs consist of atmospheric forcing computed as monthly
mean or daily mean. Note that the coupling interval in the coupled model is one day.
We note that the spin-up time is of the order of two years, which is the same as that noted
by Marchesiello et al (2003) for their climatological forcing run. This fairly short spin-up
time makes it much easier to perform multiple experiments. More results are currently
coming out, and our progress and new findings will hopefully be documented in next year’s
report.
1.2. Non-Hydrostatic Model
As the downscaling product is becoming more useful and is being used in various
applications, it is clear that the 10-km resolution, which is the maximum resolution
theoretically permitted for hydrostatic RSM, is far from sufficient in some applications.
Recent PhD work by Mansbach (2010) showed that for the wind power application in
southern California, the resolution of 1-2 km is necessary. Although such a
high-resolution capable model is available in the U.S. through NCAR (WFR model),
considering the danger of depending on a single model for research, a decision was made to
import the operational non-hydrostatic model used at the Japan Meteorological Agency
(Saito et al., 2006). This decision is based on the fact that the performance of the model
has been checked thoroughly through use in operational forecast at JMA for the last several
years, and the model has been extensively used in the downscaling of reanalysis and global
warming simulations in Far East Asia with remarkable performance (e.g. Yanase et al.,
2004; Kanada et al., 2010).
A scientist (Mr. Sasaki) from the Meteorological Research Institute in Tsukuba
Japan (MRI) visited Scripps for three months and successfully implemented the NHM
model to our G-RSM system. In order to check the model performance, a downscaling of
four cases was performed for 48 to 96 hours and the results are summarized below.
1.2.1. Case 1. Mountain Valley wind. Palm Springs area.
A 2 km resolution covering 101x101 grid points, 50 layers in the vertical and the
model was integrated for 72 hours. This case is chosen to see the diurnal mountain-valley
circulation under weak large scale forcing, characterized by the near stationary high
pressure system over the area. Figure 5 shows near surface winds after 36 hours of
integration. Along the valley straddled by San Gorgonio Mountain and the San Jacinto
Peak, the westerly wind exceeding 10 m/sec is simulated. This wind turns southward and
extends to the Salton Sea. There are only two observations over this area, but the
comparison indicates that the turning of the westerly towards the northwest and colliding
with the southerly from mountain slopes near Palm Springs forms a stagnant area,
simulating the turning of wind to northwesterly at Palm Springs Airport well. Such
features are not well simulated by the 2.5 km RSM, but understanding the reason requires
further work. Figure 5 bottom panel shows the east-west cross section of temperature just
to the south of Palm Springs (33.8N). It is clearly seen that the desert area to the east of
San Jacinto Peak is characterized by a well mixed boundary layer. The temperature at the
same altitude is higher here than to the west. It is speculated that the strong westerly
winds between San Gorgonio Mountain and San Jacinto Peak is a result of a pressure
gradient created by the temperature difference, enhanced by the funnel-like geographical
features.
1.2.2. Case 2. Santa Ana event. Southern California region.
The area 700 km by 1000 km over Southern California is downscaled using 2km
resolution. The model was integrated for 96 hours. The large scale circulation is
characterized by a strong ridge over the Great Plains favorable for a Santa Ana with strong
northeasterly winds. The observation shows strong near-surface warming to the lee side
of the mountains along the coast during the day time. Figure 6 shows temperature and
winds after 48 hours of integration. The wind simulation shows a general northwesterly
over the Sierra Nevadas, colliding with onshore wind near the coast, agreeing well with
observations. The temperature is generally colder than observation, but the high
temperature area at the foothills and relatively cold area near the coast are well reproduced.
The average bias computed from about 50 observation stations available over the area at
this hour was found to be about minus 1.4 degrees Celsius. The main cause is due to the
model’s inability to reproduce very high near-surface temperatures higher than 36 degrees
Celsius in the desert area. Figure 7 is the cross-section of wind and potential temperature
approximately following the wind direction (north-north-east). This figure shows the
existence of a high potential temperature area to the west of the mountain range, below the
mid-point of the mountain slope. There is a downward motion to the east of this area,
indicative of high potential temperature air brought down from upper levels. To the west
of the high temperature zone, there is a sea-breeze front due to the cold maritime air
flowing onto land.
1.2.3. Case 3. Atmospheric river event. Northern California.
This is an atmospheric river event which produced heavy rain over Northern
California. This case is used by the Yoshimura et al (2010) water isotope study. At
Cazadero station (38.61N, 123.22W), precipitation started at about 22 UTC on 21March
2005, and continued until 10 UTC on 22 March 2005. RSM downscaling was started
from 00 UTC, on 15 March 2005, but NHM downscaling is started from 00Z 21 for 48
hours. Comparison is made between the 5km resolution with convective parameterization
and the 2km resolution without parameterization. Figure 8 is the time evolution of
precipitation at Cazadero station. The starting time of the precipitation is roughly the
same between the 5km and 2km resolution runs, but the ending time is improved in the
2km resolution model. The agreement of the model precipitation with observation is
reasonable.
1.2.4. Case 4. Diurnal variation over coastal ocean. San Diego coastal area.
This case is chosen to study the very small scale variability in the off-shore winds
for assisting the deployment of buoy observation. The model was run for 48 hours with 1
km resolution. There is a considerable geographical variability of wind even some
distance from shore (Figure 9). The wind is weaker and more chaotic during the offshore
wind hours. More study is in progress to validate the results from micro-scale
observations.
Although the number of cases tested is still not sufficient, the NHM was
successfully implemented into the ECPC G-RSM system, and is producing reasonable
results. Work will begin to apply this model to the downscaling of reanalysis, forecast and
global change simulations for a multitude of applications that require very high resolution.
1.3. Global ROMS
In anticipation of extending our RSM-ROMS regional coupled downscaling of
forecasts, we started to work on the global version of the ROMS, which allows the coupling
with GSM, making it possible to perform coupled global forecast and even a global change
simulation. The global version of ROMS has been tested by Dr. Auad with reasonable
success (See examples in Figure 10). The work is in progress to adopt the global version
of ROMS into the G-RSM system with parallel coupling. Since RSM is already coupled
with ROMS, and RSM and GSM have common I/O routines, it is not difficult to make this
new functionality.
1.4. Marine stratocumulus parameterization
Mr. H. Kawai from the Japan Meteorological Agency has completed the work on
climatological features of marine stratus using GOES satellite imageries. His main
purpose was to improve currently available marine stratus parameterizations. His work
mainly confirmed the works based on MODIS observation by Wood and Hartmann (2006).
The finding of his study is that there is a distinct deviation of liquid water in the clouds
from normal distribution, which places a moderate impact on the cloud-radiation feedback
mechanism. This deviation from normal distribution affects the ratio of cloud water
content to precipitation autoconversion rate and also reduces shortwave reflectance
(Figures 11), both computed as a function of cloud amount. These results can be
incorporated relatively easily into the marine stratus parameterization in the ECPC model
and a plan is in place to work on this subject in future. This work has been published in
Journal of Climate (Kawai and Teixeira, 2010).
2. Application Studies
2.1 Intercomparison of global warming simulation downscaling over California
The benchmark intercomparison of the dynamical downscaling of Reanalysis by
WRF-CLM, WRF-RUC at LLNL, RegCM at UCSC and a statistical method developed at
Scripps is now complete (Miller et al., 2009) and the project has advanced to the
intercomparison of global change simulations.
Our latest G-RSM is used for downscaling of global change simulations. The
main objective of this project is to determine the uncertainties of the downscaled global
change simulation. For this purpose, four different model simulations were downscaled.
The duration of the integration is set to 10 years. The 10 year period was determined by
comparing the temporal variability of a particular global simulation of near-surface
temperature with variability among the models.
2.1.1. Implementation of the incremental interpolation into dynamical downscaling of
CMIP-type output
Very few of IPCC/WCRP's CMIP3 simulation outputs can be directly used as lateral
boundary conditions for dynamical downscaling study because these data are archived with
coarse vertical and temporal resolution (at most 100 km and daily). We investigated the
impact of the resolution and use of incremental interpolation on downscaling. Japanese
T106 MIROC simulation results are used, from which all 23-level data are used for lateral
boundary for the control regional integration (CTL). We performed two experiments, one
applying a simple vertical interpolation from coarse vertical data (COA) and the other
utilizing incremental interpolation (INC; Yoshimura and Kanamitsu, 2009) from the lowest
9 levels (up to 200 hPa). The experiment specifications are summarized in Table 1. In
the experiments, we used the ECPC RSM over a domain covering the western U.S.,
Mexico, and the surrounding oceans in 10 km horizontal resolution. Integration was done
for a half year (January to July 2047 with A1B scenario) with identical initial conditions for
both experiments.
Table 1: Specification of regional experiments for incremental interpolation
Atmospheric Forcing Data MIROC-hi A1B in 2047
Forcing variables U, V, T, q, sfcP
CTL 23 vertical levels / 6 hourly
Forcing
COA 9 levels (1000~200hPa) / daily
resolution
INC COA + incremental interpolation
Period January to June, 2047
Figure 12 shows monthly precipitation distributions and their difference from CTL.
The coarse vertical information experiment (COA) resulted in excessive precipitation over
the Central Valley, whereas the INC experiment does not show this defect. Figure 13
shows the time evolution of domain-averaged RMSD (root mean square difference)
between COA or INC and CTL for monthly averaged surface air temperature, precipitation,
and surface wind. From this figure, it is clear that all three surface diagnostic variables were
improved by the incremental interpolation.
2.1.2. Climatic changes in surface diagnostic variables
In this section, we briefly show the climatological mean for the 20th and 21st
centuries with CCSM global forcings. First, surface air temperature is shown in Figures 14
and 15. Figure 15 illustrates that the large scale feature of the temperature is faithfully
reproduced in the downscaled simulations for both 20th and 21st centuries with small scale
features added. This is the unique characteristic of our downscaling by the use of the
Scale Selective Bias Correction. The averaged temperature increase due to increase in
CO2 over the domain is about 2 degrees. The downscaled simulation shows the increase
is larger in mountainous areas. The bottom panel of Figure 15 shows the difference
between the global background field and the dynamically downscaled field presenting
small scale details.
Figures 16 and 17 are for precipitation. The difference from the global model
simulations is more apparent, showing much greater spatial variability in the downscaling
results. Particularly, there is no big difference in the 21st century in the global model
simulation, but there is a significant increase of precipitation over narrow regions along the
Sierra Nevadas and the Southern California coastal region in the regional simulation.
2.1.3. Mechanism of future increase of evening precipitation in dry season at valleys
Because our output is in hourly time intervals, we can see how the diurnal pattern
changes due to global warming. By analyzing the hourly data, we found that evening
precipitation (17-20LT) increases greatly in future climate in dry area dry season (along the
Colorado River valleys) both in GFDL and CCSM runs (Figure 18). Since the average
precipitation amount is very small in this region and season (0.5~1.0 mm/day), this increase
may have a strong impact on ecological systems in the regions.
By further analyses, we found that this precipitation increase is driven by larger
moisture flux during evening. Figure 19 shows the change in the horizontal moisture flux.
The bottom panels show that the flux increase in the 21st century is more apparent during
evening in both simulations. This stronger moisture flux brings more precipitation to the
regions.
Now we divide the increase in moisture flux into three components, i.e., due to
humidity increase, wind increase, and residual, as expressed in the following equations.
Increase of vertically integrated moisture flux (zonal component) ΔQu is defined as
follows:
ΔQu = Qu _ 21c − Qu _ 20 c
= Δq ⋅ u 20c + Δu ⋅ q 20 c + Δq ⋅ Δu
ˆ ˆ ˆ ˆ ˆ ˆ
ˆ
where total column vapor q , and vertically averaged wind speed weighted by humidity
ˆ
u are defined as:
Qu = q ⋅ u
ˆ ˆ
q ≡ ∫ qdp , u ≡ ∫ qudp / q
ˆ ˆ ˆ
Figure 10 shows these three components (wind increase ( Δu ⋅ q20c ; top), humidity increase
ˆ ˆ
( Δq ⋅ u 20c ; middle), and residual ( Δq ⋅ Δu ; bottom)). This indicates that the humidity
ˆ ˆ ˆ ˆ
increase has the largest impact in increasing the moisture flux, which occurs both in the
morning and evening. However, because of stronger evening wind over the regions, the
impact of the humidity increase is enhanced, and accordingly, the moisture flux increase in
the evening.
2.2. Multi-RCM Regional Ensemble Downscaling (MRED) project
MRED is a project to downscale NCEP CFS ensemble seasonal forecast with
multiple models (http://rcmlab.agron.iastate.edu/mred/menu.html). The project is funded
by NOAA and managed by Dr. Ray Arritt of Ohio State University. ECPC is acting as a
focal point for data archiving and intercomparison.
We have completed 10 members of 27 years, 5 months each, of CFS downscaling.
Basic comparisons between the RSM and CFS have been made for 2m temperature,
precipitation, and 500mb height. Because of the use of our Scale Selective Bias Correction
Method, the 500mb height fields are quite similar. We have found a few specific areas
where the RSM improves on the skill of the CFS for 2mT or precip, but generally speaking
the results of the CFS and RSM are similar.
Several validation datasets have been compiled and made available to the MRED
group on: http://ecpc.ucsd.edu/http/MRED/VDATA/. All the data here are regridded to the
MRED grid and reformatted to NetCDF. The datasets are:
CFS01 forcing data
NARR reanalysis
ERA40 reanalysis
REA2 reanalysis
CMAP (monthly precip)
CRU (monthly precip)
GISS (monthly sfc temp)
GPCP (monthly precip)
GTS (daily precip)
HIGG (daily precip)
UDEL (monthly precip and sfc temp)
UNIF (monthly precip)
VASC (monthly precip and sfc temp)
CLDV (climate division monthly precip and sfc temp)
QLCD (station data: monthly pr,tas,uas,vas,ps,psl)
COOP (station data: monthly pr,tas,tasmax,tasmin)
SSMI (monthly snow depth)
SNOD (daily snow depth and swe)
To get an idea of the variation between verification datasets, we have made comparisons
between NARR and COOP, Higgins and GTS. Shown in Figure 20 are the temporal
correlations between the NARR and COOP (regridded station data) for low level temperature
and precipitation. It is interesting to note that the two analyses differ considerably over
complex topography areas for temperature, and over the eastern foothills of the Rocky
Mountains for precipitation. We need to take those uncertainties into account in the
observation analysis itself when validating the forecast.
Finally, we are in the beginning stages of model intercomparisons. We have data
from ISU, NCEP, and UCLA, as well as our own to begin the comparison. We have
attached a sample comparison for a single year in Figure 21. We intend to look at both
large-scale and fine-scale features, comparing biases and correlations with the regional
models as well as CFS for JFM.
2.3. Historical downscaling of reanalysis over the Far East
This project is nearing completion. A paper describing the validation was
prepared and is in the final stage for publication. We have participated in an additional
small intercomparison project among European and Japanese researchers. It was found
that our model result was comparable to other models over the Far East, except for a very
large overestimation of precipitation, the reason being known to be due to the use of
spectral nudging of humidity. The model was rerun with the latest configuration and
re-submitted for comparison. The model precipitation is now reasonable, but some cold
bias was noted (Figure 22). The intercomparison paper has been prepared for publication
by the scientists in Japan and Europe.
2.4. Experimental ensemble seasonal forecast for IRI and NCEP
This activity has been ongoing since 2002. A 22 ensemble member seasonal
forecast is performed every month to assist IRI and NCEP in their operational seasonal
forecast using the latest G-RSM applying an extension of AMIP type simulations as initial
conditions. The ensemble is made of a 12-member 7-month forecast using predicted SSTs
provided by IRI, as well as a 10-member 4-month forecast using persisted SSTs. The 22
member ensemble seasonal forecasts are also dynamically downscaled to finer 35 km
resolution over the continental U.S. to monitor the capability of the regional model and to
perform research on ensemble dynamical downscaling. All those forecast products are
available to the general public at http://ecpc.ucsd.edu/projects/GSM_seasons.html.
2.5. Seasonal forecast of Southwest Monsoon
We continue to participate in the real time seasonal forecast of Southwest
monsoon activity. Downscaling of our seasonal forecast to 10 km over the NAMAP
region is performed and the results are provided to NCEP. This is the third time in the last
3 years that we have collaborated with the Climate Prediction Center at NCEP.
2.6. Regional ocean model applications
RSM downscaling is conducted over Drake Passage. This project is in
collaboration with PhD student, Chuan-Li Jiang advised by Prof. Sarah Gille, to compare
observed fluxes over ocean with model-produced fluxes. The study showed that a high
resolution atmospheric model is essential in obtaining accurate fluxes due to their high
spatial variability. The paper summarizing the result is about to be submitted for
publication (Jiang et al., 2010).
2.7. Application to inversion study
The downscaling product, CaRD10 has been used for an inversion study
(Iacobellis et al., 2009). This study aims at predicting the increase/decrease of heavy
pollution events in future global change conditions. The California Reanalysis
Downscaling at 10 km (CaRD10) was extensively utilized to examine meso-scale
circulation features during typical heavy air pollution events. The near surface wind
circulation is found to be very different for strong and weak inversion cases (Figure 23),
for which offshore wind dominates for the former and onshore wind for the latter. The
CaRD10 was also instrumental in detecting the representativeness of near surface
observations in space and time, important for deriving conclusions from a very limited
number of surface observations.
2.8. Application to wind power energy study
Dr. David Mansbach extensively utilized the CaRD10 product as well as very high
resolution RSM simulations (up to 1 km resolution) to study the wind circulation in the Bay
Area and Southern California desert regions (Mansbach, 2010). He found from CaRD10
analysis that over the Bay Area, the land-sea breeze forms a front that travels inland and
squeezes through a narrow valley into the Central Valley, producing strong local wind
circulation (Figure 24). The formation and timing of strong wind can be predicted by
observing the passage of the front at the coastal area. For the southern desert region, he
found that the hydrostatic model is not capable of reproducing the strong wind circulation
through the narrow valley, due to the much smaller scale of the event. He successfully
reproduced the meso-scale wind circulation using the WRF non-hydrostatic model,
showing clearly the limitation of the RSM. The NHM simulation Case 1 presented in
Section 1.2.1 is an attempt to successfully reproduce the wind circulation using a
non-hydrostatic model, which was not possible with the hydrostatic RSM.
3. Basic research
3.1. Low frequency mode error in dynamical downscaling
The interannual variability of dynamically downscaled analysis and its error
relative to global coarse resolution analysis is examined (Kanamitsu et al., 2010). The
regional model error is shown to significantly contaminate the interannual variability of the
seasonal mean. The error occupies a significant part of the interannual variability,
particularly during the summer season. Accordingly, the leading modes of the Empirical
Orthogonal Functions (EOFs) of 500 hPa height in the region differ greatly from those of
the global analysis (Figure 25).
In this study, a variant of Spectral Nudging, the Scale Selective Bias Correction
(SSBC) method by Kanamaru and Kanamitsu (2006) is refined to further reduce the error
within the observational error. The application of this method in dynamical downscaling
reduced the error of the interannual variability of analysis fields (namely, height,
temperature and winds), and made the EOFs of seasonal mean 500 hPa height agree well
with those of the global analysis. The application of the SSBC had a modest impact on
model derived fields, such as precipitation and near surface air temperature. The
improvements in these fields are not as dramatic as those of the analysis fields, but the
increased simulation skill is evident.
A possible cause of the error in the interannual variability is discussed. No
apparent systematic reduction in the high-frequency variability is found (Figure 26), which
is common to other regional models (Rockel et al., 2008). The error in the interannual
variability is more likely due to the excitation of the stationary computational mode within
the domain forced by lateral boundary forcing and enhanced by the ill-posed lateral
boundary condition and its treatment.
3.2. Downscaling strategy for ensemble forecasts.
We also have a separate work on the downscaling of ensemble average. We have
analyzed two methods; one is to downscale one or two ensemble members of the CFS which
are closest to the ensemble mean, and the other is to downscale one member, but adding
constant forcing in such a way that the seasonal mean for the target season is the same as the
ensemble mean. The former method seems to work fairly well (Figure 27). The latter
method was found to reproduce ensemble mean large scale well, but the transient
disturbances are unaffected by the added constant forcing, resulting in very little change in
the precipitation skill (Figure 28, particularly the top panel). Further work is required to
improve this technique.
3.3. Isotope reanalysis and research
In this work, an isotope-incorporated regional model is developed and utilized for
simulations of an Atmospheric River event which occurred in March 2005 (Yoshimura et
al., 2010). A set of sensitivity experiments and comparisons with observations confirm that
the kinetic isotopic exchange between falling droplets and ambient water vapor below the
cloud base was mostly responsible for the initial enrichment and subsequent rapid drop of
the deuterium abundance in precipitation observed during the event. According to the
budget analysis, the increase in isotopic composition during the latter half of the event was
primarily due to horizontal advection (Figure 29). The contribution of condensation from
different heights in the air to the ground precipitation was not reflected in the precipitation
isotopes.
3.4. Theoretical consideration of dynamical downscaling
The use of spectral nudging for reanalysis and forecasts/simulations has very
different implications. For the latter cases, there is no “truth” toward which the nudging is
applied, leaving the possibility that the regional model can “improve” or modify the large
scale and hence produce “better” downscaling. Conceptually, the existence of the lateral
boundary implies that inconsistency will develop at the boundary when the large scale
within the domain is modified. This is apparently unfavorable for good simulation. In
addition, studies of the effect of resolution in a global model indicated no sound reasons for
the superiority of high resolution in forecasting/simulating the large scale. The regional
experiments produce mixed results; some do not show any improvements while others
show some improvements, but there is no clear demonstration that the large scale is
improved by the regional model. These discussions do not suggest any strong physical
basis for using or not using spectral nudging in the downscaling.
In addition to the lack of “truth,” simulations and forecasts produce multiple “truths,”
(often called internal variability, IV) due to the chaotic nature of the atmosphere (Alexandru
et al., 2007). This further complicates the downscaling, since the regional model must
also reproduce IV with reasonable accuracy. If the spectral nudging is applied, it
eliminates the IV of the nudged scale. This may not be desirable. However, it is possible
to impose the IV of the large scale forcing from ensemble global simulations. Since the
lateral boundary in a regional model also acts as a damping of IV, regional model IV is also
considered to be inaccurate. Thus, the relative merits or demerits of spectral nudging for
the forecasts and simulations are even more difficult to determine.
In order to resolve these difficulties at least conceptually, it is argued that these
problems are rooted in the definition of downscaling, which has not previously been given
in downscaling studies. We have attempted to establish an acceptable general definition:
“To produce the most likely estimate of small scale details of parameters and their
probability distributions given coarse resolution large scale circulation and its probability
distributions.” This definition includes uncertainties in the global forcing field, which is
the root of the problems mentioned above. However, if the global forcing uncertainties
are removed from the definition, only one large scale “truth” exists and many of the
problems disappear. Based on this argument, we introduced two types of dynamical
downscaling, prognostic dynamical downscaling (PDD) and diagnostic dynamical
downscaling (DDD). The PDD permits uncertainties in the global forcing, allowing the
downscaling to prognostically correct/modify the large scale. The DDD is considered as a
diagnostics to obtain small scale features when a specific large scale circulation and its time
evolution are given. Mathematically speaking, PDD and DDD are fundamentally
different. The PDD is an initial value problem and DDD is a boundary value problem
(including time evolution as a boundary condition). The PDD needs to avoid nudging for
allowing improvement in the large scale and not to damp large and small scale variability,
while DDD requires some form of spectral nudging to maintain the large scale. The DDD
suffers less from various problems associated with regional modeling, namely lateral
boundary condition, choice of area locations and domain size, if large scale is properly
maintained within the domain. DDD can still provide IV, although it is damped. This is
still consistent with the assumption that the large scale is given. DDD can also incorporate
the IV of global fields, by performing multiple downscaling of ensemble global simulations.
Downscaling of reanalysis is a DDD problem. PDD is much more difficult than DDD
due to the nature of the initial value problem, because it is sensitive to model configurations
and initial conditions, and is affected by predictability problems. However, it can
potentially improve the result of DDD. Many dynamical downscaling studies conducted
experiments without considering this difference in the basic assumption of downscaling.
A straightforward downscaling without spectral nudging (PDD) forced by Reanalysis
results in regional model error contaminating the large scale as shown in our work. The
same downscaling forced by a global change simulation will accordingly suffer from
similar regional model error. In this case, however, one may argue that the downscaling
may improve if the regional model "error" (or more properly "change") somehow acts to
correct the error of the global large scale simulation. This may occur when the regional
model physical processes are superior to those of the global model. Application of DDD
will make the regional simulation faithfully follow the large scale forcing. Therefore,
application of DDD to global change simulations or forecasts will suffer only from error
inherent in the global model, but not from regional model error. We feel that DDD is
much easier than PDD for downscaling study due to its simplicity and reduced sensitivity
to regional model configuration. The results are also easier to understand, although they
still present many difficulties. This view may be challenged by many regional modelers
using PDD.
3.5. Diurnal variation and monsoon
ECPC is continuing several projects with Prof. Song-You Hong at Yonsei
University in Korea. The most promising project is the study of the effect of diurnal
variation on the onset and evolution of the Asian monsoon. Yonsei University graduate
students perform most of the computations, and Prof. Song-You Hong’s frequent visits to
Scripps make it possible to rapidly advance the project. We have completed several RSM
integration experiments with and without diurnal variation of incoming short wave
radiation. The run without diurnal variation applied daily average short wave radiation
flux as constant. The main results are that the evaporation over ocean increases
significantly due to diurnal variation. Over land, sensible heat increases, particularly over
the Tibetan plateau. The total effect of diurnal variation is diagnosed from the non-linear
interaction among different frequencies, namely, diurnal and semi-diurnal with difference
in phase. We are finding that the difference in the phases of diurnal variation between
wind speed, temperature and moisture difference between atmosphere and land/ocean and
static stability (which affects exchange coefficient) resulted in the increase/decrease of
surface fluxes over land and ocean (Figure 30). These time integrated effects of diurnal
variation modulate the large scale monsoon circulations via thermally direct and
precipitation-induced indirect circulations.
4. G-RSM outreach
The ECPC has now sponsored 9 international RSM workshops including the one
this past year in Maui, Hawaii in collaboration with IRI and also in memory of the late Dr.
Roads. The central theme of the Maui workshop was an application of regional
downscaling to forest fire research. The ECPC also provides volunteer RSM model
masters, who respond to international queries about the model, provide training for the
model at IRI training workshops, and maintain the RSM home page as part of ECPC’s
extensive WWW outreach. The WWW outreach greatly expanded during the last couple
of years when wikispaces (http://g-rsm.wikispaces.com) were introduced. Now everyone
can be involved in writing and revising the documentation, and exchanging ideas and
comments. This community effort has been a great success. The number of subscribed
members has grown to 75 (7 more than last year), from a broad range of countries including
Australia, Bangladesh, Brazil, China, Hong Kong, Pakistan, India, Indonesia, Iran, Israel,
Italy, Japan, Korea, Peru, Senegal, Singapore, Spain, Sri Lanka, Sweden, Taiwan, Thailand,
Turkey, U.S.A., Venezuela, and Vietnam. ECPC is also providing basic datasets to run
hindcasts and forecasts for the users’ countries. Because of the expanded functionality of
the G-RSM system, model access requests are expected to increase further.
5. References
5.1 References (External)
Alexandru, A., R. de Elia, and R. Laprise, 2007: Internal Variability in Regional Climate
Downscaling at the Seasonal Scale. Mon. Wea. Rev., 135, 3221–3238. DOI:
10.1175/MWR3456.1
Carton, J.A., G. Chepurin, X. Cao, and B.S. Giese, 2000: A Simple Ocean Data
Assimilation analysis of the global upper ocean 1950-1995, Part 1: methodology, J.
Phys. Oceanogr., 30, 294-309.
DeHaan, L., M. Kanamitsu, C-H Lu, and J. Roads, 2007: A comparison of the Noah and
OSU Land Surface Models in the ECPC Seasonal Forecast model. J. Hydromet. 8,
1031-1048.
Haidvogel, D. B., H. G. Arango, K. Hedstrom, A. Beckmann, P. Malanotte-Rizzoli, and A. F.
Shchepetkin, 2000: Model evaluation experiments in the North Atlantic Basin:
Simulations in nonlinear terrain-following coordinates. Dyn. Atmos. Oceans, 32,
239-281.
Hong, S.Y., and H.L. Pan, 1996: Nonlocal Boundary Layer Vertical Diffusion in a
Medium-Range Forecast Model. Mon. Wea. Rev., 124, 2322–2339.
Iacobellis, S., J. Norris, M. Kanamitsu, M. Tyree and D. Cayan, 2009: Climate Variability
and California Low-Level Temperature Inversions - Final Report. PIER Project
Report. CEC-500-2008-017-F. Available from
http://www.energy.ca.gov/publications/displayOneReport.php?pubNum=CEC-500-2009-020-F
Jiang C., S. T. Gille, J. Sprintall, K. Yoshimura and M. Kanamitsu, 2010: Length scale of
the turbulent heat fluxes in the Southern Ocean. To be submitted for publication.
Kanada, S., M. Nakano and T. Kato, 2010: Changes in mean atmospheric structures
around Japan during July due to global warming in regional climate experiments
using a cloud-system resolving model. Hydrological Research Letters, 4, 11-14.
Kanamaru H. and M. Kanamitsu, 2006: Scale Selective Bias Correction in a Downscaling
of Global Analysis using a Regional Model. Mon. Wea. Rev. 135, 334-350. DOI:
10.1175/MWR3294.1
Kanamitsu, M., W. Ebisuzaki, J. Woolen, S. K. Yang, J. J. Hnilo, M. Fiorino and J. Potter,
2002: NCEP/DOE AMIP-II Reanalysis (R-2). Bull. Amer. Met. Soc. 83, 1631-1643.
DOI: 10.1175/BAMS-83-11-1631.
Kanamitsu, M., H. Kanamaru, Y. Cui and H. Juang. 2005: Parallel Implementation of the
Regional Spectral Atmospheric Model. CEC-500-2005-014. Available from
http://www.energy.ca.gov/publications/displayOneReport.php?pubNum=CEC-500-20
06-071
Kanamitsu, M., K. Yoshimura,. Y-B Yhang and S-Y Hong, 2010: Errors of Interannual
Variability and Multi-Decadal Trend in Dynamical Regional Climate Downscaling.
J. G. R. Atmosphere in print.
Mansbach, D. K., 2010: Synoptic and local influences on boundary layer processes with
an application to California wind power. PhD Dissertation. Scripps Institution of
Oceanography, University of California, San Diego.
Marchesiello, P., J. C. McWilliams, A. Shchepetkin, 2003: Equilibrium Structure and
Dynamics of the California Current System. J. Phys. Oceanography, 33, 753-783.
Miller, N. J., P. B. Duffy, D. R. Cayan, H. Hidalgo, J. Jin, H. Kanamaru, M. Kanamitsu, T.
O’Brien, N. J. Schlegel, L. C. Sloan, M. A. Snyder, and K. Yoshimura, 2008: An
analysis of simulated California climate using multiple dynamical and statistical
techniques. PIER Project Report. CEC-500-2008-017-F. Available from
http://www.energy.ca.gov/publications/displayOneReport.php?pubNum=CEC-500-2009-017-F.
Rockel, B., C. L. Castro, R. A. Pielke Sr., H. von Storch, and G. Leoncini, 2008: Dynamical
downscaling: Assessment of model system dependent retained and added variability
for two different regional climate models, J. Geophys. Res., 113, D21107,
doi:10.1029/2007JD009461.
Saito, K., T. Fujita, Y. Yamada, J. Ishida, Y. Kumagai, K. Aranami, S. Ohmori,
R. Nagasawa, S. Kumagai, C. Muroi, T. Kato, H. Eito and Y. Yamazaki, 2006: The
Operational JMA Nonhydrostatic Mesoscale Model. Mon. Wea. Rev., 134, 1266-1298.
Seo. H., A. J. Miller and J. O. Roads, 2007a: The Scripps coupled ocean–atmosphere
regional (SCOAR) model, with applications in the eastern Pacific sector. J. Climate, 20,
381-402.
Seo, H., M. Jochum, R. Murtugudde, A. J. Miller, and J. O. Roads, 2007b: Feedback of
Tropical Instability Wave - induced Atmospheric Variability onto the Ocean. J. Climate,
20(23), 5842-5855.
Seol, K.-H., S.-Y. Hong and M. Kanamitsu, 2010: Investigation of land surface process
over the ARM SGP in 1997 summer using a single-column model. Submitted to J.
Hydromet.
Shimpo, A., M. Kanamitsu, S.F. Iacobellis, and S.Y. Hong, 2008: Comparison of Four
Cloud Schemes in Simulating the Seasonal Mean Field Forced by the Observed Sea
Surface Temperature. Mon. Wea. Rev., 136, 2557–2575.
Wood, R., and D. L. Hartmann, 2006: Spatial variability of liquid water path in marine
low clouds: The importance of meoscale cellular convection. J. Climate, 19,
1748-1764.
Yanase, W/., G. Fu, H. Niino and T. Kato, 2004: A polar low over the Japan Sea on 21
January 1997. Part II: A Numerical Study. Mon. Wea. Rev., 132, 1552-1574.
Yoshimura, K., and M. Kanamitsu, 2009: Specification of External Forcing for Regional
Model Integrations. Mon. Wea. Rev. 137, 1409-1421.
Yoshimura, K., M. Kanamitsu and M. Dettinger, 2010: Regional Downscaling for Stable
Water Isotopes: A Case Study of an Atmospheric River Event. Accepted for
publication in J. G. R. Atmosphere.
5.2 ECPC Refereed Publications 2008-2010
2008
Anderson, B. T., G. Salvucci, A. C. Ruane, J. O. Roads, and M. Kanamitsu, 2008: A
New Metric for Estimating the Influence of Evaporation on Seasonal Precipitation
Rates. J. Hydromet. 9, 576-588.
Auad, G., 2008: Response of the Gulf of Alaska 3D winter circulation to oceanic climate
shifts: Ecosystem implications. Geophys. Res. Lett., 35, L02602,
doi:10.1029/2007GL031611.
Auad, G. and A. J. Miller, 2008: The role of tidal forcing in the Gulf of Alaska's circulation.
Geophys. Res. Lett., 35, L02602, doi:10.1029/2007GL031611.
De Haan, L. L., and M. Kanamitsu, 2008: Increase in Near-Surface Temperature
Simulation Skill due to Predictive Soil Moisture in a Numerical Seasonal Simulation
under Observed SST Forcing. J. Hydromet. 9, 48-60.
Haidvogel, D. B., H. Arango, W. P. Budgell, B. D. Cornuelle, E. Curchitser, E. Di Lorenzo,
K. Fennel, W. R. Geyer, A. J. Hermann, L. Lanerolle, J. Levin, J. C. McWilliams, A. J.
Miller, A. M. Moore, T. M. Powell, A. F. Shchepetkin, C. R. Sherwood, R. P. Signell,
John C. Warner, and J. Wilkin, 2008: Ocean forecasting in terrain-following
coordinates: Formulation and skill assessment of the Regional Ocean Modeling System.
J. Comput. Phys., 227, 3595-3624.
Hong, C.-C., M.-M. Lu and M. Kanamitsu, 2008: Temporal and Spatial Characteristics of
Positive and Negative Indian Ocean Dipole with and without ENSO. J. Geophys. Res.
113 D08107, doi:. 10.1029/2007JD009151.
Kanamaru, H. and M. Kanamitsu, 2008: Dynamical Downscaling of Global Analysis and
Simulation over the Northern Hemisphere. Mon. Wea. Rev. 136, 2796-2803.
Kanamaru, H. and M. Kanamitsu, 2008: Model Diagnosis of Nighttime Minimum
Temperature Warming during Summer due to Irrigation in the California Central Valley.
J. Hydromet., 9, 1061-1072.
Kim, H.-J., A. J. Miller, J. McGowan and M. Carter, 2008: Climate and coastal algal
blooms in the Southern California Bight. Progress in Oceanography, sub judice.
Kueppers, L. M., M. A. Snyder, L. C. Sloan, D. Cayan, J. Jin, H. Kanamaru, M.
Kanamitsu, N. L. Miller, M. Tyree, H. Du, and B. Weare, 2008: Seasonal temperature
responses to land-use change in the western United States. Global and Planetary
Change 60, 250–264.
Nunes, A. M. B., and J. O. Roads, 2008: Testing the impact of modeled surface water on
regional climate simulations. Submitted, J. Hydromet.
Ruane, A.C., and J.O. Roads, 2008a: Dominant Balances and Exchanges of the
Atmospheric Water Cycle in the Reanalysis 2 at Diurnal, Annual, and Intraseasonal
Time Scales. J. Climate, 21, 3951-3966.
Ruane, A. and J. Roads, 2008b: Diurnal to annual precipitation sensitivity to convective and
land surface schemes. Earth Interactions, 12, 1–13
Seo, H., R. Murtugudde, M. Jochum and A. J. Miller, 2008: Modeling of mesoscale
coupled ocean-atmosphere interaction and its feedback to ocean in the western Arabian
Sea. Ocean Modeling, accepted pending minor revisions.
Seo, H., M. Jochum, R. Murtugudde, A. J. Miller, and J. O. Roads, 2008: Precipitation from
African Easterly Waves in a Coupled Model of the Tropical Atlantic Ocean. J. Climate,
21, 1417-1431
Shimpo, A., M. Kanamitsu, S. F. Iacobellis and S.-Y. Hong, 2008: Comparison of Four
Cloud Schemes in Simulating the Seasonal Mean Field Forced by the Observed Sea
Surface Temperature. Mon. Wea. Rev. 136, 2557-2575.
Yoshimura, K., M. Kanamitsu, D. Noone and T. Oki, 2008: Historical Isotope Simulation
using Reanalysis Atmospheric Data. J. Geophys. Res. 113, D19108.
Yoshimura, K. and M. Kanamitsu, 2008: Dynamical Global Downscaling of Global
Reanalysis. Mon. Wea. Rev. 136, 2983-2998.
Yulaeva, E., M. Kanamitsu and J. O. Roads, 2008: The ECPC Coupled Prediction Model.
Mon. Wea. Rev. 136, 295-316.
2009
Anderson, B. T., A. C. Ruane, J. O. Roads, and M. Kanamitsu, 2009: Estimating the
Influence of Evaporation and Moisture-Flux Convergence upon Seasonal Precipitation
Rates. Part II: An Analysis for North America Based upon the NCEP–DOE Reanalysis
II Model. J. Hydromet., 10, 893–911. DOI:10.1175/2009JHM1063.1
Gutzler, D., L. N. Long, J. Schemm, M. Bosilovich, J. Chern, J. C. Collier, M. Kanamitsu, P.
Kelly, D. Lawrence, M.-I. Lee, R. Lobato, B. Mapes, K. Mo, A. Nunes, E. Ritchie, J.
Roads, S. B. Roy, S. Schubert, H. Wei and G. Zhang, 2009: Atmospheric Simulations of
the 2004 North American Monsoon: NAMAP2. Submitted to J. Climate.
Kanamitsu, M., K. Yoshimura, Y-B Yhang and S-Y Hong, 2009: Errors of Interannual
Variability and Multi-Decadal Trend in Dynamical Regional Climate Downscaling. J.
Climate, in review.
Kim, H. –J., A. J. Miller, J. McGowan and M. Carter, 2009: Coastal phytoplankton blooms
in the Southern California Bight. Progress in Oceanography, in press.
Nunes, A. M. B., and J. O. Roads, 2009: Dynamical downscaling of short-term climate
fluctuations: On the benefit of precipitation assimilation. J. Geophys. Res., 114, D12108,
doi:10.1029/2008JD010933.
Pryor, S. C., R. J. Barthelmie, D. T. Young, E. S. Takle, R. W. Arritt, D. Flory, W. J.
Gutowski, A. Nunes, and J. Roads, 2009: Wind speed trends over the contiguous USA.
J. Geophys. Res., in press.
Shimpo, A., and M. Kanamitsu, 2009: Planetary Scale Land-Ocean Contrast of
Near-Surface Air Temperature and Precipitation Forced by Present and Future SSTs.
Accepted in J. Met. Soc. Japan.
Yoshimura, K., and M. Kanamitsu, 2009: Specification of External Forcing for Regional
Model Integrations. Mon. Wea. Rev. 137, 1409-1421.
2010
Jimenez, C., C. Prigent, B. Mueller, S. I. Seneviratne, M. F. McCabe, E. F. Wood, W. B.
Rossow, G. Balsamo, A. K. Betts, P. A. Dirmeyer, J. B. Fisher,M. Jung, M. Kanamitsu,
R. H. Reichle, M. Reichstein, M. Rodell, J. Sheffield, K. Tu, K. Wang, 2010: Global
inter-comparions of 12 land surface heat flux estimates. Submitted to J.G.R.
Atmosphere.
Kanamitsu, M., K. Yoshimura, Y-B Yhang and S-Y Hong, 2010: Errors of Interannual
Variability and Multi-Decadal Trend in Dynamical Regional Climate Downscaling. In
print, J. G. R. Atmosphere.
Kawai, H., and J. Teixeira, 2010: Probability Density Functions of Liquid Water Path and
Cloud Amount of Marine Boundary Layer Clouds: Geographical and Seasonal Variations
and Controlling Meteorological Factors. J. Climate, 23, 2079-2092.
Yoshimura, K., M. Kanamitsu and M. Dettinger, 2010: Regional Downscaling for Stable
Water Isotopes: A Case Study of an Atmospheric River Event. In print, J. G. R.
Atmosphere.
8.3 Other ECPC Publications 2008-2010
2008
Vila, D., and A. Nunes, 2008: Impact of the Satellite-Gauge Based Precipitation
Assimilation on Regional Downscaling. Predicting Precipitation II Posters
(Presiding Session), American Geophysical Union 2008 Fall Meeting, San
Francisco, CA.
Nunes, A., 2008: A North American Climate Simulation: The ECPC-RSM Contribution to
NARCCAP. Regional-Scale Climate IV Posters, American Geophysical Union
2008 Fall Meeting, San Francisco, CA.
2009
Iacobellis, S., J. Norris, M. Kanamitsu, M. Tyree and D. Cayan, 2009: Climate Variability
and California Low-Level Temperature Inversions - Final Report. PIER Project
Report. CEC-500-2008-017-F. Available from
http://www.energy.ca.gov/publications/displayOneReport.php?pubNum=CEC-500-200
9-020-F
Jimenez, C., C. Prigent, S. Seneviratne, B. Muller, M. McCabe, W. Rossow, J. Fisher, K.
Wang, J. Sheffield, E. Wood, M. Jung, M. Reichstein, M. Bosilovich, M. Kanamitsu, A.
Beljaars, P. Dirmeyer and M. Rodell, 2009: Inter-comparsion of 1993-95 global land
surface monthly averaged heat fluxes. GEWEX Radiation Panel Report.
Nunes, A., 2009: Dynamical Downscaling of Short-Term Climate Fluctuations. 21st
Conference on Climate Variability and Change. 89th American Meteor. Soc. Annual
Meeting, Phoenix, AZ.
Nunes, A., and D. Vila, 2009: Dynamical Downscaling Using Satellite-Gauge Precipitation
Estimates. Precipitation: Measurement, Climatology, Remote Sensing, and Modeling.
European Geosciences Union General Assembly 2009, Vienna, Austria.
2010
None
total
Figure 1. SST of uncoupled and coupled simulation after 2 year integraion. QSC run is the
uncoupled run with flux correction, NON run is the uncoupled run without flux correction and
COUPLED run is the fully coupled run. Note small formation of small scale eddies, not resolved
by large scale SST analysis. Also see the narrow band of cold SST along the coast caused by the
strong upwelling, only resolved by high resolution ocean model.
Figure 2. Comparison of total heat flux at the ocean surface for three experiments. Black line:
R-2, Yellow: CaRD10V2, Green: CaRD10V3. Note reduced total heat flux for CaRD10V3 during
the summer months.
Figure 3. Time variation of monthly averaged area mean Sea Surface Temperature over the
central domain (33N-44N, 132W~-126W). Note the abrupt decrease in the deviation of SST from
CaRD10, SODA and QSC (uncoupled simulation with radiation correction) in uncoupled without
radiation correction (yellow) after 1994, when the atmospheric forcing is obtained from the latest
cloud scheme.
Figure 4. Comparison of SST simulated by RSM-ROMS coupled model with radiation correction
(top left), uncoupled ROMS forced by daily RSM forcing (top right), uncoupled with monthly
average RSM forcing without radiation correction (bottom left) and large scale SST analysis
(bottom right).
Figure 5. 36 hour simulated near surface wind (in vector) and surface elevation (in colored
contour) by NHM, valid at 12 UTC, 20th July 1999 (top) and east-west cross section at a location
slightly south of the Palm Springs at the same time (bottom). The highest surface altitude peak is
3200 meters.
Figure 6. Wind vector and temperature after 48 hour simulation (left) and corresponding station
observation. Valid date and time is 00UTC, October 26, 2003.
Figure 7. Cross section of wind and potential temperature along the wind direction
(north-north-east) centering on the area of warmest temperature (31.3N, 116.1W).
7
6
5
4
2km
3 5km
2 obs
1
0
19 21 23 1 3 5 7 9 11
18 20 22 0 2 4 6 8 10 12
Figure 8. Time evolution of precipitation at Cazadero for 5km and 2km resolution model.
24 hour forecast 36 hour forecast
Figure 9. Near surface (10 meter above ground) wind at 24 and 36 hour simulation by NHM at
1km resolution.
Figure 10. An example of Global ROMS simulation.
Figure 11: Left panel: The correction ratio of autoconversion rate calculated by assuming
Gaussian distribution of total water content. Lines show cases in which different formula for
autoconversion rate on l are assumed (red line: autoconversion rate proportional to l 2 , green
7/3 2.47 3
line: l , pink line: l , blue line: l ). Horizontal axis shows cloud amount [%].
Right panel: Reduction factor of effective shortwave reflectance obtained from LWP PDF converted
from total water content Gaussian PDF using conceptual model. Horizontal axis is cloud amount
[%]. Three lines correspond to the cases in which the average value of LWP PDFs is 40[g/m2]
(green line), 60[g/m2] (red line), and 80[g/m2] (pink line).
Total Precipitation
*Precipitation intensified
over central valley.
improved
Figure 12: Monthly precipitation from 10 km regional dynamical downscaling of MIROC's global
future projection results. Upper panels show monthly distribution and lower panels show difference
from the control (CTL).
Error decreased!
Figure 13: Time evolutions of root mean square differences (RMSD) of COA and INC
experiments relative to CTL.
CCSM 2mTmp
20C
Background Downscaled
data data
A2
Figure 14: Climatological annual mean of surface air temperature from global models (CCSM; left
column) and dynamically downscaled runs (right column). Top and bottom rows show the results
for 20th century and 21st century (A2 scenario), respectively.
CCSM 2mTmpDif
A2–20C in A2–20C in
Background downscaled
data data
Added value by DS
Figure 15: Difference of surface air temperature between 21st and 20th century runs (top two
panels) and the difference between the two top panels (bottom panel).
CCSM Total Prcp
Figure 16: Similar to Figure 14, but for precipitation.
CCSM Total PrcpDif
Figure 17: Similar to Figure 15, but for precipitation.
17-20LT Total Prcp (JJA) 6-8LT
GFDL CCSM GFDL CCSM
20C
21C
The Increase in JJA
is mainly derived from
afternoon precipitation.
DIF
Figure 18: Precipitation in 20th and 21st centuries for JJA (top: 20th . middle: 21st, and
bottom: difference). The left 6 panels show evening period and the right 6 panels show
morning period. Results with the GFDL (CCSM) forcing are placed on the left (right) side of
the 6 panels.
17-20LT Moisture Flux (JJA) 6-8LT
GFDL CCSM GFDL CCSM
20C
21C
Stronger m.flux in the evening than in the morning
DIF
compare
Figure 19: Similar to Figure 18, but for horizontal moisture flux.
Figure 20. the temporal correlations between the NARR and COOP (regridded station data) for
low level temperature and precipitation.
Figure 21. Difference in seasonal average (JFM) near surface temperature between NCEP RSM
and UCLA ETA model, valid in 1983. Considerable difference, up to more than 8 degrees C is
noted.
Figure 22. Comparison of the RSM downscaled temperature (left) and precipitation (right)
between original downscaling with (blue) and without moisture nudging (red) over Far East.
Y-axis for temperature (precipitation) is the difference between downscaled temperature
(precipitation) and observation. X-axis is the observed temperature and precipitation. It is clear
that the positive bias in precipitation is corrected, but the cold temperature bias increased slightly.
DJF JJA
Figure 23. Average values of CaRD10 Dθ850 and 10-meter wind anomalies during the 30 days with
the largest (top row) and smallest (bottom row) observed Dθ850,SJV during time period 1979-2005 for
DJF (left column) and JJA (right column). The white line denotes the approximate boundary of
the San Joaquin Air Basin. The color filled contours indicate the CaRD10 Dθ850 anomalies while
the vector arrows show the 10-meter wind anomaly. (From Iacobellis et al, 2009).
Figure 24. Climatological near surface wind
corresponding to July at 00UTC (4 pm PST).
Note the front like features protruding into the
Central valley through the Bay area.
R2 NOSSBC SSBC NEWSSBC
Mode 1
32% 38% 32% 32%
Mode 2
19% 20% 15% 18%
Mode 3
16% 11% 19% 16%
Figure 25. Leading three modes of summer time 500 hPa geopotential height EOF for analysis
(R2) and experiments during summer. The percent variance is indicated by percent in each panel.
NOSSBC: without correction, SSBC: with original correction, NEWSSBC: new version of
correction.
day-to-day variance of 500-height
R2 NOSSBC SSBC NEWSSBC
Summer
(1984)
Winter
(2001)
Figure 26. Daily variance of 500 hPa height for R2 and experiments. Top row is for summer
1984 and bottom row is for winter 2001. These two years are selected as the amplitudes of the
low-frequency errors are large. Note the increase in daily variance for summer and winter for
NOSSBC experiment as compared to R2.
2mT Spring, Summer, & Fall Forecast Skill 2mT Winter+ Forecast Skill
0.8 0.5
0.7 0.4
0.6 0.3
0.2
C o rre l a tio n w / N A R R
0.5
0.4 RSM Ens Mean 0.1 RSM Ens Mean
RSM hi cor mem RSM hi cor mem
0.3 Global 0 Global
OND-200609 JFM-200610 JFM-200611 JFM-200612 FMA-200701
0.2 -0.1
0.1 -0.2
0 -0.3
ASO-200607 SON-200608 AMJ-200703 SON-200708
-0.1 -0.4
Figure 27. T2m Anomaly Correlations for Single Member and Ensemble Mean of downscaled
forecast. A comparison of the ensemble mean of downscaling (blue) and single member
downscaling of the global simulation closest to the ensemble mean (purple). The skill of the global
ensemble mean is also shown (yellow). For these limited cases, the high correlation member has
the best skill 3 out of 4 times in the low variability seasons and the 12-member RSM has the best
skill 4 out of 5 times in high variability seasons.
CTL-DS Run.Mean of Sgl
Run.Mean of Ens.
COR-DS
Figure 28. Time variation of various parameters in CTL and COR at Bay Area, for the second
month of the simulations. From the top, precipitation, 10m meridional wind, 10m zonal wind, 2m
air temperature, and 500mb height are shown. In 500mb height, 1-month running means of
ensemble mean and the single member of global forcing fields are also shown.
(a) (b)
(c)
Figure 29: Accumulated increments in isotopic composition (δD) of vapor by different processes in
the model at 940 hPa (a), 890 hPa (b), and 820 hPa (c) levels during the five isotopically distinct
periods. The processes are divided in three parts, namely physics (closed diamonds), dynamics
(closed squares), and diffusion (closed triangles) processes. Total increments are also presented by
lines with crosses. The impact from the physics process is further divided into planetary boundary
layer (dashed line), deep convection (dot-dashed line), and large scale condensation (dotted line),
and they are all denoted by open diamonds. Note the dominant contribution of large scale
dynamics to the total change in isotopic composition.
Diurnal variation
SH = ρC p Ch V Δθ Ch is calculated by SH, Cp, |V|,
Δθ
LH = ρC L Ch V Δq (ρ is ignored….).
CTL
NDI
SH 10m wind Ch △θ
LAND
LH 10m wind Ch △q
OCEAN
Figure 30. Diurnal variation of each component of the sensible and latent heat fluxes over land
and ocean. Note that the in-phase relationship between the parameters seen in above figures
implies that fluxes are larger than non-diurnal fluxes if integrated over a day.