Close Window

Univ. of Quebec (Montreal): "Operational and training uses of numerical model diagnostics"

Final Report


1.1 Introduction

With the continuing improvement and proliferation of numerical models, as well as communications and visualization software, operational meteorologists will increasingly be making use of detailed output from a number of numerical models such as NGM, ETA, MRF, RUC, RAMS etc. In order to choose the appropriate model for a particular situation as well as use the model output intelligently, it will be important to provide the operational meteorologists with a better understanding of the models available as well as tools for understanding complex model processes.

The goal of this project was to implement and validate the operational usefulness of a numerical model diagnostic package, DIONYSOS, developed at the Université du Québec à Montréal. This involved 1) modifying DIONYSOS so that it would run on NCEP's meso-ETA model, 2) installing the DIONYSOS along with a visualization software on the Denver office workstation (HP), 3) programming the workstation to automatically access and run the diagnostics, 4) train Eric Thaler, the Denver SOO, in the use of the visualization software, 5) train all the Denver meteorologists in the interpretation of the output of the diagnostics, 6) experiment with the use of the diagnostics operationally, and 7) perform case studies.

The university partner was responsible for 1-4 and the Denver office was responsible for 6. Items 5 and 7 were the joint responsibilities of both the university and Denver office.

The version of DIONYSOS that was implemented in the Denver office diagnoses output from NCEP's meso-ETA model. It calculates the instantaneous contributions of standard individual forcings (sensible heating, latent heat release, orography, friction, temperature advection, and vorticity advection) to the surface and upper air geopotential tendencies, temperature changes, vorticity tendencies, and vertical motion. These diagnostic results can be compared to the model values, which are either directly available or calculated from the standard model output.

A significant part of this project involved training operational meteorologists on 1) how they can better understand and visualize modern synoptic-dynamic theory using DIONYSOS, 2) how DIONYSOS can be used to better understand model processes, and 3) how such understanding can help them in the making better use of model output.

1.2 Description of Research/Development Accomplishments

1.2.1 Software and Hardware Issues

The first objective of the project was to adapt DIONYSOS to diagnose output from the Meso-ETA model on the Denver Weather Office Science Applications Computer and to make it available operationally. This was a major effort and included installing DIONYSOS and the visualization software on the Denver Weather Office workstation, automatically retrieving the meso-eta output from an ftp site, converting the grib format to the DIONYSOS format, and interpolating the fields to a polar stereographic projection. Finally, the DIONYSOS was programmed to run automatically every night.

For the first year, the interpolations were performed on UQAM's computer because the interpolator required a more recent version of the HP operating system. The interpolated data were then sent by ftp to the Denver Weather Office Workstation on which the diagnostics were run and visualized. With the upgrade of the HP operating system in Denver, the entire numerical processing was run on the Denver computer.

Much of this effort could have been avoided if the model output were available on a variety of projections. (NCEP has software to convert model output to a number of projections, which could be made available on ftp sites.) The second lesson learned was that the visualization software (WFO advanced / AWIPS) at the Denver office is not very flexible in ingesting new fields. By appropriately designing the visualization software, the operational testing of new techniques such as DIONYSOS could be facilitated in the future,

1.2.2 Diagnostics studies

As is indicated below, there were unfortunately very few cases of interest to the Denver Weather Office. As a result, the UQAM group adapted the DIONYSOS software to diagnose tropical weather systems. Because of lack of tropical cyclones in the Meso-ETA grid in 1997 and the unavailability of appropriate AVN or MRF output, this diagnostic software was run on output from the Canadian Global Model Forecast of two hurricanes (Erica 1997 and Edouard 1996). The results of the diagnostics indicate that the hurricane simulations represented balanced systems and much of hurricane behavior can be understood in that framework. These results (Zwack et al, 1998) point to a potentially different approach to the understanding of hurricane motion and intensification.

1.2.3 Diagnostics for training purposes on the WEB

To make DIONYSOS available to the educational and operational communities, an interactive web page is now available at UQAM at: The user name is sca and the password is hiv98. This page makes available daily diagnostics from ETA-32 00z model. This page originally used output from the Meso-ETA. However, case studies showed that this model had spurious gravity waves, which resulted in spurious values of vertical motion and pressure tendencies, which was confirmed through correspondence with Eric Rogers at NCEP. Apparently, the existence of spurious behavior was already known and described in NCEP documentation. The NCEP team is planning to correct the numerical methods, which are responsible for the behavior.

NCEP has just made available (March 1998) at an ftp site the necessary output from the ETA-32 model in which the numerical methods have been corrected. However, recent tests show that there are still spurious vertical motions in the lower half of the troposphere in the output of this model. UQAM scientists are continuing correspondence over this matter with NCEP scientists.


The first order of business was familiarizing the Denver SOO on the use of the visualization software that UQAM installed on the Denver HP workstation. This involved a conference call of 4 hours while all participants were running visualization software on their local computer. During this conference call, the UQAM group went through a case study of model diagnostics of a system affecting the Denver area. In addition, the Denver SOO collaborated in the diagnostics and presentation of results for two non-classical weather systems (Zwack et al, 1996c).

The UQAM group developed and tested a one-day training workshop on DIONYSOS and diagnostics of weather systems at the Canadian Meteorological Center. After fine tuning the workshop, the UQAM group visited the Denver Weather Office in the fall of 1996 and gave the workshop to all of the forecasters over a two-day period. In addition, extensive documentation on DIONYSOS was developed and provided to the Denver SOO (Zwack et al, 1996a).

This workshop was followed by numerous telephone calls and e-mail exchanges between the Denver SOO and UQAM to clarify the diagnostics. During the second year, the Denver SOO and UQAM scientists discussed the implications of modifying the DIONYSOS software to be able to diagnose successfully hurricanes. In September 1997 the Denver SOO visited UQAM at which the hurricane cases were discussed as well as the future use of the diagnostic software in the Denver SOO`s Ph.D. dissertation.


Zwack, P., C. Pagé, Tardif, R., N. Ben Hadj, 1996a: DIONYSOS: A Diagnostic Package for the Interpretation of Numerical Weather Prediction Model Output. On Line Scientific and Technical Documentation. Department of Earth Sciences, University of Quebec at Montreal,, 57 pp.

Zwack, P., E. Thaler, and C. Pagé, 1996b: DIONYSOS: Diagnostics of Numerical Model Forecast Output, Proceedings of the 11th Conference on Numerical Weather Prediction, Amer. Meteor. Soc., Norfolk, Virginia, 3 pp.

Zwack, P., O. Hamelin, P. Santurette, E. Thaler, and C. Pagé, 1996c: Diagnosing Weird (Non-Textbook) Weather Systems in Numerical Model Output, Proceedings of the 11th Conference on Numerical Weather Prediction, Amer. Meteor. Soc., Norfolk, Virginia, 3 pp.

Zwack, P., P. Roucheray, C. Pagé, 1997: Diagnosing the motion of hurricanes Edouard (1996) and Erika (1997) as Simulated by the Canadian Spectral Forecast Model, 10th Extratropical Cyclones Workshop, Val Morin, Québec.

Zwack, P., 1997: Diagnosing the motion of a hurricane in the Canadian Global Forecast Model, 31st Annual Congress, Canadian Meteorological and Oceanographic Society, June, Saskatoon.

Zwack, P., M. Desgagné, P. Roucheray, C. Pagé, 1998: Diagnosing tropical cyclone motion and intensity change in numerical model simulations, Proceedings of the Symposium on Tropical Cyclone Intensity Change, Amer. Meteor. Soc., Phoenix, Arizona, 6 pp.

Additional reference

Caya, A., R. Laprise, P. Zwack, 1998: On the consequences of using process splitting for implementing physical forecings in a semi-implicit semi-Lagrangian model, Mon. Wea. Rev., in printing.


4.1 University's Perspective

Students have been exposed to examples of non-textbook type cases that have been suggested by the Denver Forecasters. The university group has modified the diagnostic package DIONYSOS in order to perform a diagnostic of the movement and intensity change in a Canadian numerical model simulation of hurricanes Eduoard (1996) and Erika (1997). This modification of the software makes DIONYSOS much more general so that it can be used for diagnosis of both tropical and extratropical weather systems.

The major problems encountered have been numerical problems in the output of NCEP`s models. Such numerical problems have also been identified through our diagnostics and have been corrected in the Canadian non-hydrostatic community research model MC2 (Caya et al, 1998). These numerical problems, not found in the French operational model, ARPEGE, or the Canadian operational regional and global models, can have significant negative impact on the operational use of the Eta model output.

4.2 NWS Perspective

Unfortunately, due to major operational changes in the forecast office during the time of the research, full implementation of the software into forecast operations was not possible. Additionally, the weather did not cooperate well, with a rather benign pattern over the Denver area of interest during most of the time that the software has been available. Nonetheless, there were times that several of the forecasters had the opportunity to look at a few cases and were quite enlightened by what they were able to see. Namely, they were able to better visualize what physical processes were taking place in the model, which in turn helped lead to a better physical understanding of what was going on in the atmosphere. This in turn has helped lead to better use of the numerical model data in the forecast process. Again, due to the lack of time in the forecast office due to the AWIPS installations, taking complete advantage of the diagnostics was not completed.

However, not all has been lost. The software continues to be available on the SAC computer and there have been some recent cases that will be used as training cases in the future. Fuller implementation into the operations is also planned as the AWIPS implementation issues slow down. In addition, the SOO at Denver also plans on using the diagnostic software as part of his Ph.D. research and he will be making continuing efforts to migrate the software onto the AWIPS platform, perhaps also as part of the research. Implementation into AWIPS will make operational use of the system more amenable to the forecasters as well.

In all, this has been an excellent project. A tool now exists that our forecasters can use to help understand model output data and the real atmosphere. This in turn aids in more intelligent use of model output and hence improved forecasts to the user.