Modelers’ Workshop
Danish Climate Centre
Danish Meteorological Institute
Lyngbyvej 100, DK-2100 Copenhagen Ø
DENMARK
22-23 May 2000
Monday 22 May 2000
The workshop opened with a welcome by Jens Christensen, Senior Advisor of the Danish Climate Centre of the Danish Meteorological Institute. Self introductions were given by participants. Bill Gutowski of PIRCS gave a brief overview of the International Institute of Theoretical and Applied Physics (IITAP), partial sponsor of the workshop and home institution of PIRCS. Acknowledgements were given to IITAP and the US Electric Power Research Institute for supplying funding for the workshop and the project.
DISCUSSION OF PIRCS EXPERIMENT 1b
Ray Arritt gave a brief overview of results to date for PIRCS Experiment 1b, a 30-day simulation of the continental US for June and July of 1993. John Roads points out that the reanalysis shows a daytime max which, compared the regional models, shows that the regional models are superior.
Richard Jones raised the question about variability in output that might occur if the initial conditions were perturbed at the bit level. [RJ was thinking of some work that Giorgi has done perturbing initial or lateral boundary conditions, which showed surprisingly large, ongoing differences between runs, even if only IC perturbed.] How much should we expect weather systems to evolve as observed? We are not doing weather prediction, but climate simulation. Thus simulating statistics of climate is important, but simulating precise sequence of weather events is not. Ole Christensen commented on repeated 1-yr runs that showed variability from year to year due to evolving and variability of soil moisture.
Comments:
S-Y Hong: Plot large-scale and convective precipitation separately.
Publication plans were discussed
R. Arritt: Need documentation of all the models!
J. Roads: suggested using coarser resolution (100 km) and running longer to examine land surface iterations.
R Arritt: Consistent patterns emerged among models that are similar (e.g., common convection mechanisms).
J. Roads: spectrum of energy - is that a reasonable diagnostic? What should it be compared against? [Spectrum should transition smoothly from the reanalysis spectrum to the RCM spectrum.] Eta model analysis (EDAS) would be interesting to look at — even if time period is not the same, could get a climatology.
Need for additional analyses was discussed.
J. Roads: suggested a difference between 93 and 88 would be informative.
Possible use of an ensemble of models was discussed.
Jens Christensen suggested we focus on areas where regional models add value to global models, and that we need to determine when and where the added value is realized.
NCEP reanal II should be coming out now. Soil moisture is improved because actual observed precip (rather than model) is used for soil moisture.
DISCUSSION OF A PIRCS EXPERIMENT 1c.
This would build on PIRCS 1a and 1b, but simulate several years.
G. Takle: Pan at ISU re-ran a short simulation with grid extended in north, south, east or west directions. Precipitation changed, even in central US.
M. Fox Rabinowitz: Boundaries should not be in a data-poor area for the reanalysis.
J. Christensen: Western boundary seems ok but lack of skill is most important in our target region.
C. Jones: For 1993, extending to south is of interest because of low-level jet
J. Roads: Practical consideration is that groups already are set up from having run 1a,b. He has run a 10-year simulation for PIRCS domain.
R. Jones: Need to let science issues drive choice of domain
M. Fox-Rabinowitz: Include North American Monsoon. Extend the domain south and west
B. Gutowski: GCIP extension (GAPP) includes emphasis on SW U.S. as well as Mississippi R. basin.
J. Christensen: Enlarging domain increases freedom within RCM domain.
J. Roads: ENSO cycle may be reasonably represented already as it is a winter phenomenon with signal in large-scale flow.
R. Jones: Build on what we have done. Two extremes have been done; simulate a longer period.
R. Arritt: Need to explore deficiencies in predicting cool season precipitation.
J. Roads: Need to test for sub-period. Different models may require different domains.
M. Fox-Rabinowitz: Specify minimal domain.
J. Roads: PIRCS domain probably satisfies minimum
C. Jones: Is the goal to improve results or understand what’s happening?
C. Jones: Double nested domain. Internal mesoscale convective complexes require higher resolution (~ 20 km)
R. Laprise: Tradeoff between time and space. Support climate change experiments. Few have run climate change experiments at < 50 km. Extremes were a prototype for interannual variability. 10 years is minimal for statistics. Look at interannual variability of model vs .observations
C. Jones: People use RCMs for seasonal prediction, not just climate change.
R. Jones: Again, scientific question will dictate emphasis.
C. Jones: Basic question is whether RCMs provide added value
J. Christensen: GCMs can’t produce extreme events.
B. Gutowski: Models aren’t capturing amplitudes of extremes.
J. Roads: Applications community wants longer time scales. Three basic points:
U. Busch: Effect of vertical resolutions
M. Fox-Rabinowitz: 20 layers may be too low. 70 layers may not even be enough based on some of his work.
J. Christensen: In PIRCS 1, vertical resolution was left to modelers, e.g., to match ongoing applications.
B. Gutowski: Specify minimal period and encourage modelers to do more if resources allow.
R. Jones: For long simulations, even a small set of models would be useful. Getting people involved is more important than strict requirements.
S. Y. Hong: Each participant needs to optimize domain. Climate forecast is an anomaly forecast; ability to capture mean doesn’t imply ability to capture anomalies. Comparison w/statistical downscaling and GCMs in ability to get extremes.
J. Roads: How many will extend or do higher res.?
B. Gutowski: Be more careful about what we mean by "go beyond". - higher res. or larger domain (to south).
D. Caya: Keep resolution coarse for long simulation, then higher res. for a shorter period if there is something interesting.
R. Arritt: Need to coordinate for providing data sets.
J. Roads: Do 88, 93 at higher resolution as a subset of the longer run by implementing a one-way nest during those years.
J. Christensen: Using entire ERA 1979-93 for verification in MERCURE.
M. Fox-Rabinowitz: Community interest in 1997-99 ENSO.
Conclusion: Minimal domain will be the same as for PIRCS 1a,b with encouragement to go beyond and experiment with (a) higher resolution, (b) expanded boundaries, and (c) longer time series.
FINAL PROPOSAL FOR PIRCS EXPERIMENT 1C
Minimal simulation:
minimal period: 1987-1993 (with appropriate spin-up as determined individually by modelers)
minimal domain: Same as PIRCS 1a,b
maximum grid spacing: ~ 50 km
Recommended extensions
Extend the period to 1978-1999
Start in summer of 1986 or do multiple-year spin-up for soil moisture
Nest a hi-res domain of 20-25 km for ’88 and ’93 60-day periods previously studied or more
Move southern boundary south by 5 deg (~ 10 grid points)
R. Laprise: Start before 87 for land surface spinup
J. Roads: Started in summer 86
Decide spinup "as needed"?
Sensible, latent heat flux, sfc temp.
C. Jones: Look at AMIP
DISCUSSION OF A PROPOSED PIRCS EXPERIMENT 2
This would focus on simulating South America and be timed to take advantage of the special observations and other data sets being created for the Large-Scale Biosphere-Atmosphere (LBA) Experiment in the Amazon
Vicky Ballester (University of Sao Paolo) presented an overview of the LBA experiments and data
Introduction
Data Information system
Data use
LBA started in early nineties and has attracted global interest. Issues addressed include exchange of energy water, carbon, trace gases and nutrients. An underlying question is how conversions of tropical forest will alter these exchanges. It is of considerable interest to predict the impact of deforestation.
Questions
How does Amazonia function as regional entity?
How do land use changes and climate variability affect this regional functioning?
Field campaigns started in 1998 and 1999. LBA integrates data and processes across many disciplines and includes:
Ecological studies
Physical climate and hydrology
Atmospheric chemistry
Database compilation
Diagnosis
Integrated data system is used for data archive. Data generated by LBA will become public after 2 years
Web based search system for data includes a metadata search system
Land cover/land use (soil) database is available
Half hourly values are available of all variables measured
3 major sites: Rondonia, Santarem, Manaus
The LBA web site is http://www3.cptec.inpe.br/lba/index.html and can be linked to from the PIRCS web site (under Copenhagen meeting)
Campaigns
WETAMC: Wet Season Atmospheric Mesoscale Campaign
Jan — Feb 1999 in Rondonia
LBA and TRMM (also Jan-Feb 1999). Also involved tower measurements, Doppler radar, etc.
Reanalysis using the RAMS model is planned
- 80 km resolution domain with 20 km inner nest
- target period: Jan-Feb 1999
Future plans
Dry AMC Sept-Nov 2002 availability of data will be accelerated
A low-level jet campaign planned for Dec. 2002 — Jan/Feb. 2003
Chemistry campaign — trace gases measurements
Time windows
Sept 2000
July/ Aug 2003 data from all campaigns will be available.
Main web site: www.cptec.inpe.BR/LBA
DISCUSSSION
B. Gutowski: What about S. American monsoon
C. Jones: implications of ENSO for domain choice
LBA data sets 2001-2002. C. Nobre says we may be able to get some data earlier. Pre-LBA campaigns may be held.
V. Ballester: Can work out data sharing through LBA PIs.
Launch tracers from source regions?
J. Christensen: How to treat surface parameters will be a considerable challenge; role of land surface process schemes in RCMs will be tested.
How large are deforested areas?
M. Fox-Rabinowitz: quality of reanalyses lower over S. America due to lack of observations; this will affect the accuracy of verification data (could we run w/different reanalyses?)
J. Roads: IRI program has also done South American simulation. PIRCS can build on that.
L. Sun: Showed transparencies with IRI domain size. About 6 models are part of the IRI program.
J. Roads: Regional model precipitation is similar to reanalysis but with more detail. Results show a drying out of soil compared with reanalysis
Data archive has monthly means of various fields.
Provided a subset of reanalysis data.
Driving conditions available through ecpc.ucsd.edu/projects/brazilcomp.html
4/1/97 — 3/31/99, Soil drying is a problem
A very deep root zone needs to be considered for the soil submodel
Bill Gutowski presented the proposal for PIRCS 2
A domain of 150x150 gridpoints at 50 km is proposed. Land use type is available at this resolution.
Some key considerations:
ENSO events
South American Monsoon
Availability of LBA datasets:
observing periods 1999 and forward
some in 1996 and 1997.
data collected in 1999 can be available from specific investigators
Chemistry and soil-atmosphere interaction are key concerns. Question was raised about how much independent verification data we have?
R. Jones: (1) Why S. America? (2) Are we involving any S. American modeling groups?
B. Gutowski: Wanted to do a different (tropical) region; newer opportunity; IITAP connection promotes interaction with the developing world.
Also:
It is a tropical region
Amazon atmospheric heating is important for global climate
Comparison of low level jet, monsoon regions between North and South America
Sharp terrain is important to the flow and offers challenges to models
PIRCS has corresponded w/modeling groups at CPTEC, others (Tanajura, Argentina).
M. Fox-Rabinowitz: Others in U.S. have interest, e.g. Rutgers, Utah, U. MD.
J. Roads: An opportunity to test soil moisture; all models dry out during El Nino.
Trees may root to 20 m from which roots will draw moisture during dry periods. This will require a very long spin-up period.
Width of the Andes will be described by about 4 grid points at the proposed resolution.
R. Jones: Difficult to say where the boundary should be. One group should run a model to experiment with different boundaries to test sensitivity.
David Hassell — RCM SIMULATIONS OF THE SOUTH ASIAN MONSOON: Issues important to a tropical simulation
Domain size experiments: 3 different domains with experimentation with southern and western boundary. Domain size did not relate to anomaly correlation, as was the case for Europe. There seems to be an insensitivity to domain size. Results seem to suggest that use of a small domain is possible. In the tropics physical incompatibility between global and regional model is not such a problem. No evidence that a small domain restricted the ability of a mesoscale model.
RCM mean precipitation is consistently ~ 20% higher than GCM (all 3 domains). Four individual summers in 1980s were simulated. There are break and active regimes of the Monsoon. The regional model did better than global model . Large portion of the heavy precipitation events in the regional model come in the break periods. During active periods cyclones track more northerly than during break periods. The 850 mb flow anomaly shows strengthening of westerlies in break periods and deeper monsoon trough in the active period. Cyclones initiate in Bay of Bengal, but not produced in GCM. The RCM develops cyclones but they are too large (resolution dependent?). Important factors include locally important regions of cyclogenesis.
Summary
A 20-year integration of the smallest regional model gave large improvement over the driving GCM in local mean precipitation features of break and active region
This is a result of the RCM’s ability to generate cyclones passing over the SE with realistic frequency and intensity.
DISCUSSION
Boehm: REMO simulation showed the importance of capturing ITCZ migration effect on dry/wet year.
B. Gutowski: Define analysis area as shown, to overlap w/IRI. Then vary placement of forcing. Get experience of S. American modelers as input. On resolution, P. de Silva Dias says it is necessary to resolve squall lines. Initial proposal will be for 50 km, with possible inner nest. Possible choice of domain: keep the region of analysis as the "white region" in the IRI map shown by J. Roads. We should get input from SA modelers on this. F. Meisinger and Mike Fenissey on Cola are running experiments with the ETA model which will be useful to evaluate domain size.
Tentative experiment:
Resolution: 50 km (option — 20 km inner nest is a possibility)
Period of simulation: Not the same as IRI. (which is Apr 97-Mar 99).
Period: Jan 98 - December 99 (Max El Nino -> current La Nina)
Coincides with LBA data sets.
V. Ballester: there are some max and min temps but there are large gaps in the datasets.
How many would do a PIRCS Expt 2: Christensen (HIRHAM), M. Fox-Rabinowitz (stretched grid), S.-Y. Hong (RSM), J. Larson (Argonne + ISU, MM5) indicated interest.
Modelers need justification. Up to 7 groups are willing.
C. Jones has tried using both the ECMWF and NCEP reanalysis for PIRCS 1and finds ECMWF better.
M. Fox-Rabinowitz: New DAO reanalysis will be much better for ENSO cycle than GEOS-I. It will provide data at 1ox 1o, 48 levels for 1997-99. It has not been started yet.
GEOS I reanalysis stopped in 1996 (1979-96). ERA 40 is being planned.
J. Roads: NCEP-II reanalysis? Is it at Lawrence Livermore National Lab? How can we get the data?
S.-Y. Hong: Format is the same as NCEP-I
Question was raised about the vertical interpolation
C. Jones: It is better to use raw data than interpolated values.
Surface boundary conditions were discussed. Vegetation and topography data are available. The 2o degree sea-surface temperature (SST) data are probably consistent with the reanalysis (1o SSTs may not be consistent). In the tropics we will want to use the best SSTs available because of their importance for convection and precipitation.
M. Fox-Rabinowitz showed results of a 40-km simulation that captured the precipitation well in the SE US. The University of Delaware (Willmot) has assembled a dataset with much better precipitation. Michael encourages use of 40 km resolution in simulations over the US.
POSTER DESCRIPTIONS
(Some titles are missing or have been approximated; please correct)
Analysis of the drying out problem over south-eastern Europe using the limited area model HIRHAM4
Stefan Hagemann, M. Botzet, and B. Machenauer
A spectral nudging technique for dynamical downscaling purposes
Frauke Feser
Variability of Regional Climate Models due to Soil Moisture.(?)
Ole B. Christensen
Time and space dependent biophysical land surface parameters from AVHRR NDVI data
R. Stockli and Pier Vidale
Investigations of influence of domain size and soil moisture (?)
Song-You Hong
Title
Daniel Caya
An Analysis of Precipitation variability in the Rossby Centre RCM contribution to PIRCS experiment 1B.
Colin Jones
RCM on various locations around the globe
Liquian Sun
Influence of two different land surface coupling techniques
Jan Pert Schultz
Distinct circulation states of the Arctic by natural variability.
Klaus Dethloff
Combining satellite data and a high resolutions meteorological model for estimation of
surface fluxes and resistances
H.Soegaards, E.Boegh, M.S. Rasmussen, N.Woetmann, J.H.Christensen, C.B.Hasager and N.O.Jensen,
Ensemble simulations with perturbed physics in a nested regional climate
Model
Z. Yang and R.W. Arritt
Tuesday 23 May 2000
FURTHER DISCUSSION OF PIRCS 2
It has been decided to use NCEP 1 reanalysis.
Target observations/reference fields
~ 500 precipitation stations in Brazil, far fewer outside
Goddard 1o daily precipitation (J. Roads)
GPCP monthly means; at higher resolution
Quick SCAT (M. Fox-Rabinowitz) over oceans. Winds and precipitation. The previous such dataset was called NSCATT.
LBA tower obs.
Topographic dataset: 30 sec resolution
SATOBS and TRMM
Gridded half degree data set of temperature and precipitation (Vorosmarty and Wilmot)
Satellite data of water vapor (NVAP)
OLR
Convective cloud cover (highly reflective)
PROPOSAL FOR STANDARD DIAGNOSTICS AND DATA INTERCHANGE FOR PIRCS
Jay Larson
Mathematics and Computer Science (MCS) Division
Argonne National Laboratory
A utilities wish list for an intercomparison project would include the following:
Common utilities for
Reading and interpolating model input data format
Computation of diagnostic fields
Quality control
Writing and re-reading the agreed-upon output format
Visualization and analysis tools that work with the output format
MCS has software design as one of its core competencies, so they approach this PIRCS need as a software project.
The proposed schematic for PIRCS diagnostic output and applications program:
Further development will include:
An example can be seen at www.mcs.anl.gov/~larson
A demo of ANL’s MM5 data viewer can be seen at www-climate.mcs.anl.gov/
Discussion
Ballester: data are half hourly or hourly frequency and are transmitted daily.
More output are available on-line.
C. Jones: ISCCP H/M/L cloud corresponding to above 440 mb/ 440-680 / 680 mb and below.
Use maximum random overlap.
Storage: 120x120x20x4 = 1.15 MB
Allows study of vertical distribution of diabatic heating
Tucker: vertical interpolation (via M. Fox-Rabinowitz)
M. Fox-Rabinowitz will send a list of diagnostics (email to remind)
Jens: Suggest groups do a 1-month simulation as a test
M. Fox-Rabinowitz: Will the diagnostics software handle staggered fields?
Meta-data base describing more extensive data fields should be available at modeling institutions.
Project to Intercompare Regional Climate Simulations
Iowa State University
OUTPUT For Use in PIRCS 1C AND PIRCS 2 Diagnostic Package
(this is what the modeling groups will run into Jay Larson’s black box)
Three-hourly cumulative
Input Output
Convective precip Convective precip
Stratiform precip Stratiform precip
Surface energy fluxes Surface energy fluxes
Incident solar Incoming solar
Absorbed solar Reflected solar
Up and down longwave Up and down longwave
Downward flux into the soil Downward flux into the soil
Sensible latent Sensible latent
Latent heat Latent heat
Runoff Runoff
OLR (top of the atmosphere), have modelers specify the bands of longwave they assume.
Reflected (outgoing) shortwave at top of atmosphere (cut this out unless compelling reason to keep)
Three-hour average
Precipitable water
Clouds (ISCCP) ISCCP cloud layers
CAPE (2-D) ((need to supply algorithm to individual models))
Three-hourly instantaneous
2-m air temp
Vertically integrated horizontal water transport precip water
Surface press (from which we will calculate sea-level pressure)
Omega at 500 hPa
LLJ statistics
Daily
Min/max temperature
Soil moisture
Monthly averages at 850, 700, 500, 300, and 200 hPa (of 3-hourly samples)
East-west wind speed
North-south wind speed
Vertical wind speed/omega.
Temperature
Diabatic heating on the mandatory levels
Moisture — specific humidity
Relative humidity
Be sure that the interpolation is done before the monthly averaging is done.
Include all mandatory levels.
J. Roads: GCIP MOLTS approach (profiles of heating at radiosonde sites)
Calendar items
October — climate diagnostics meeting at Lamont
Meet at AGU in SF in Dec
Starting date for Experiment 1c April 2001
Meeting ~ 1 year from now. At CPTEC?
Jay will get diagnostic codes available as early as Sept.
Colin Jones (Rossby Center) will start PIRCS 2 this summer.
Start running PIRCS 2 at start of 2001
Perhaps there should be more model summary information available to modelers online.
Contact information:
Webpage: http://www.pircs.iastate.edu
Ray Arritt: rwarritt@iastate.edu
Bill Gutowski: gutowski@iastate.edu
Gene Takle: gstakle@iastate.edu
| Copyright/Trademark Legal Notice |