For those of you familiar with the original white paper, here are the significant changes to the original (29 September 2003) white paper
(1) We have added spectra from a simulation over the ocean, found at the end of section 2.1. This addition is in response to concerns in the NCEP response that the WRF-mass core forecasts and spectra might be contaminated by truncation errors associated with discretization of the terrain transformation. The spectra for forecasts over the ocean look very similar to that over the US, hence contamination associated with terrain appears not to be problem with the WRF-mass core.
(2) We have added a brief discussion of scale-selectivity in spatial filters, found near the end of section 3. This addition is also in response to statements in the NCEP response arguing that the WRF-mass core was as dissipative as the NMM and Eta models based on evaluation of their filtering at a single wavelength. The additional discussion in the revised white paper points out the much more scale selective nature of the WRF-mass high order filters as compared to the lower order filters used in the NMM and Eta models, hence the much greater damping in the NMM and Eta compared to the WRF-mass model.
(3) We have reworded the statement concerning model efficiency in section 4.1 to more clearly pose the central question raised by the white paper, which, in reference to the operationally configured Eta and NMM models, is "What is the scientific justification for running high resolution forecasts using numerical filters that systematically remove the resolution gained by the refined grid? ".
(4) Our investigation into filtering has led us to modify the filtering we use in WRF (specifically the 10 km simulations). We have turned off the deformation-based 2nd order spatial filter. We now use the spectra from this formulation in all the plots. This new configuration is what we are running in all our experimental forecasts.
Comments concerning the revised white paper are welcome. We would also be happy to post to this webpage any comments from individuals or groups, and an NCEP response if and when it becomes available.
Bill Skamarock (NCAR/MMM, 303-497-8161, skamaroc@ucar.edu)
Mike Baldwin (NOAA/NSSL, 405-366-0504, Mike.Baldwin@noaa.gov)
20 November 2003
--------------------------------------------------------
Comments on the spectra white paper:
black_janjic_spectra_response.pdf
(18 December 2003)
--------------------------------------------------------
In the fall of 2004, I published a paper on the spectra produced by
the ARW (WRF-Mass) model -
---------------------------------------------------------
Since the appearance of the white paper in September of 2003,
significant modifications have occurred to the filtering in the
NMM. Specifically:
(1) The lower and upper bounds on the deformation-based second order
horizontal diffusion have been removed.
(2) The coefficient multiplying the deformation term has been reduced
by factor of 10
(this is equivalent to reducing the Smagorinsky constant in the lateral
diffusion by 70%, as noted in the NCEP change logs).
(3) The horizontal diffusion of scalar variables is turned off if the
terrain slope is greater than 0.001.
There have also been a number of other smaller changes that may
affect the filtering characteristics of the NMM, but these are the
major ones.
These changes were introduced before the WRF Developmental Test
Center's 4-month (one month each season) retrospective forecasts
tests. The raw model output from these tests were difficult to
acquire, but
the few forecasts that were available suggested that the filtering
changes in the NMM have had a substantial impact on the
spectra.
More recently, using the results from the WRF Developmental Test
Center's (DTC) Winter Forecast Experiment (DWFE), Dave Dempsey and I
have examined the small-scale structure and spectra produced by the
WRF-ARW (WRF-Mass) model and the newly configured WRF-NMM models.
The results are described in a 2005 WAF/NWP conference preprint titled
To summerize, we have found that the filtering in the NMM still
needs to evolve. In the 2003 white paper we (Mike Baldwin and I)
argued that the filtering needed to be reduced and the horizontal
divergence damping should be removed from the NMM. The newly
configured NMM has significantly reduced horizontal filtering. In
fact, it appears to have been reduced too much based on results from
the DTC tests showing too much unresolved structure in the NMM
solution and too much energy in the NMM low-level spectra at the
gridscale. Conversely, the fields are too damped at upper levels
because of the continued use of horizontal divergence damping. We
have tested an external mode filter in the NMM and have found that it
controls noise that was previously controlled by the horizontal
divergence damping. The external mode filter should be used in
place of horizontal divergence damping. We are testing
higher-order (more-scale selective) horizontal diffusion
formulations. See the preprint for further details.