PIRCS Dallas Meeting

January 12, 1999
Dallas, Texas

Attending: Michael MacCracken (National Assessment Coordination Office, US Global Change Research Program), Daniel Caya (Université du Québec à Montréal), Sebastien Biner (Université du Québec à Montréal), Pete Robertson (NASA Marshall), Jan Paegle (Univ. of Utah), John Roads (Scripps Institute of Oceanography), Uri Stein (Univ. of Maryland), Michael Fox-Rabinovitz (Univ. of Maryland), Jack Katzfey (CSIRO), Chuck Hakkarinen (EPRI), Roger Pielke Sr. (Colorado State Univ.), Tom Wigley (NCAR), Rick Lawford (NOAA), Zaitao Pan (ISU), Chris Anderson (ISU), Song-You Hong (NCEP), Jinwon Kim (Lawrence Berkeley National Laboratory), Bill Gutowski (ISU), Gene Takle (ISU), and Ray Arritt (ISU)

Introduction

The meeting was preceded by a buffet dinner courtesy of EPRI. Gutowski thanked attendees for their participation and Hakkarinen for making arrangements. Attendees introduced themselves and the agenda was briefly viewed.

Review

The group reviewed various past and upcoming meetings relevant to PIRCS:

Gutowski first visited the WGNE in 1995 and discussed the planning for PIRCS. Andrew Staniforth then served as WGNE liaison to PIRCS until his move to the UK Met Office in 1998. Gutowski and Arritt met with WGNE in November 1998. While some on the WGNE expressed skepticism regarding the general validity of nested regional climate models (RCMs), the meeting provided valuable feedback to PIRCS and prompted some on the WGNE to educate themselves about regional modeling. The WGNE is interested in testing basic aspects such as whether an external signal can be propagated through the RCM domain, and whether the RCMs recover large scale features of the driving GCM or analysis (as advocated by R Jones). Other concerns are the appropriate domain size and location for RCM integration, as well as implications of the lack of conservation at the RCM boundary for energy and water balances. The WGNE members were very interested in the Dickinson simulations which compare a RCM to a set of CCM3 simulations differing in their horizontal resolution (including uniform high resolution). Some at the WGNE meeting (such as Phil Merilees) pointed out importance of specific atmospheric phenomena. Fox-Rabinovitz noted that the WGNE’s concerns are legitimate and must be addressed. Kim noted that depending on domain size and forcing, the regional model can drift into its own climatology and can to some extent correct large scale features if dependent on terrain, etc. Paegle mentioned that this depends on the specific field and that analyses differ for divergent part of the field. Gutowski pointed out that the large scale behavior to be reproduced has to be behavior that the analysis captures well. Pielke raised concerns about the mismatch of scales at the RCM boundaries.

Gutowski attended the CLIVAR in Paris during early December 1998. The CLIVAR Science Plan promotes evaluation of downscaling to provide useful information for ultimate users and projections of extreme events. They have expressed a desire to promote evaluation of regional models driven by reanalyses along with comparison to comparable simulations driven by GCMs and comparison of regional climate models with statistical downscaling.

The PIRCS team continues to maintain contact with the European Union MERCURE program. Jens Hesselbjerg Christensen is a participant in both MERCURE and PIRCS.

Arritt is co-convenor (with Filippo Giorgi) of a session on regional modeling at the IUGG meeting to be held July 1999 in Birmingham, England. He encouraged the group to attend and present their results.

Filippo Giorgi is conducting a colloquium on "The Physics of Climate Change: A Regional Perspective", to be held in Trieste, Italy, 7-25 June1999. The colloquium is divided into 3 parts: a conference on Global Climate Change (7-8 June); a conference on "Detection and Modeling of Regional Climate Change" (9-12 June); and a training workshop on "Theory and Modeling of Regional Climate Change" (14-25 June). The second of these is of most direct relevance for PIRCS.

NCAR CSM Runs

Hakkarinen provided background on recently completed runs of the NCAR CSM on a NEC SX-4 in Japan. Wigley then gave an extended description of these runs, which were completed in December 1998. The CSM was run at horizontal resolution T42. The period 1870 to about 2000 was run at NCAR and then the period 1980-2100 was run in Japan for two scenarios: business-as-usual, and CO₂ stabilization around 550 ppm. The runs included direct sulfate forcing only, which corresponds to radiative forcing of about –0.6 W m^-2. It is planned that a future run will include solar variability. The ocean was initialized to representative SSTs for the late 19th century. The SO₂ emissions are very different from IS92a and most previous scenarios in that emissions reduce by about 60 percent by 2075.

Two 20-year data records were saved at 6-hourly intervals for driving regional models. It will be at least another month before the runs have been sufficiently evaluated to release. Pielke noted possible consequences of the fact that the CSM runs don’t include land use changes.

Overview of PIRCS 1b (1993 flood)

The group discussed diagnostics for PIRCS 1b. The ISU group has constructed a 0.5 degree precipitation analysis from NCDC data and compared this to predicted precipitation for 1993 for several models. All models reproduced the precipitation maximum over the north-central US and did not have the spurious precipitation maximum in the southeastern US that appears in the NCEP/NCAR reanalysis. (Hong mentioned that this was corrected in Reanalysis-2 with the new convection and PBL schemes adopted in 1995.) Magnitudes of predicted precipitation maxima corresponded reasonably well to observations. A presentation in the GCIP meeting earlier that evening noted that 12-24 hour precipitation forecasted in the ECMWF reanalysis gives error of 10-20%, to which absolute errors of PIRCS models compare favorably.

The question of "value added" by mesoscale simulations was raised. It was suggested that we look at the intense rainfall events in comparison with time history of rainfall over the Upper Mississippi River area to see if models again are able to capture individual events. The stationary front over the central US was important for the flood and may represent value added for mesoscale models. We also should check if the nocturnal precipitation max shows up in 1993. Roads suggested that since the LLJ was stronger in 1993 the nocturnal maximum may be better captured. Lawford asked what is the added value for the surface data and can we understand the process. Hong suggested we look at individual heavy precipitation events, e.g. 7 July 1993. Roads recommended comparison to daily gridded precipitation.

Publications

Reviewers of the JGR paper on the Experiment 1a results say there needs to be more analysis. Analysis of the 1988 case is hindered somewhat by the lack of profiler observations. We need to explain common problems and collective biases among the models. There is a European model comparison report that provides an example of ways to do analysis. A specific addition to the analysis is cumulative precipitation maps compared to the 0.5 degree precipitation analysis produced by the ISU group. Presently the comparison focuses too much on the large scale flow. Gutowski suggested diagnosis of mesoscale circulation for the precipitation episode on Day 160 in 1988, which apparently was an organized mesoscale system. Another diagnostic is predicted precipitation in the southwest US monsoon for 1988. While all models show an indication of monsoon onset, some give unrealistically large precipitation.

There was discussion regarding a follow-up paper on Experiment 1b. It was suggested that we look at the intense rainfall events in comparison with time history of rainfall over the Upper Mississippi River area to see if models again are able to capture individual events. Kim mentioned that the models can produce most events in western U.S. where precipitation is anchored to the terrain. There was discussion as to whether models can be expected to reproduce specific events. Katzfey suggested we look at whether mesoscale processes generally can be produced, not necessarily specific events. We have profiler data available for 1993 for evaluating the predicted wind field, especially the low-level jet.

Results from a stretched grid model

Michael Fox-Rabinovitz showed results for several simulations of the 1988 case. These included both a uniform resolution GCM and a stretched grid model driven with observed SSTs. Output was also shown for data assimilation with data withdrawn over target area as an "imitation" of the nested grid approach. This method shows very good results. There was general agreement that PIRCS should include variable-resolution models.

Future activities

Roads mentioned that Rick Lawford is asking for observations for a climatology of the Mississippi River basin. PIRCS may be able to contribute evaporation simulations or energy budget information.

The group briefly discussed some sensitivity experiments. Results were shown from grid extension experiments for the 1993 flood, done at ISU by Zaitao Pan. Extending the domain to the south or east produced a similar or slightly improved precipitation field compared with the standard PIRCS domain while extending the domain to the west produced worse results. Arritt mentioned that extending the domain westward pushes the inflow boundary into a data-poor region (the Pacific Ocean). Hong showed simulations comparable to Seth and Giorgi (1988, J. Climate) which gave different results for soil moisture sensitivity. Notably, Hong found there always was positive feedback from soil moisture to precipitation amount.

Possible follow-up experiments were briefly discussed. Among the ideas presented were a multi-year simulation using the present domain; repeating the runs for 1988 and 1993 at higher resolution; and intercomparison for the Amazon region for comparison with forthcoming LBA results. It was agreed that PIRCS should develop closer affiliation with AMIP.

The next meeting of PIRCS modelers is not likely before Fall AGU. Hakkarinen encouraged planning for a next meeting. A paper describing 1993 results should be prepared by Fall AGU, preferably sooner. Continuing to publish results is important for modelers to justify participation in PIRCS.