For the evaluation phase in December and January, the SPC concentrated on the winter weather evaluation of the RUC II output. Our goal was not to compare the old RUC with RUC II, but rather to see how it could aid the winter weather forecaster and to see how it verified against actual observations. A formal winter evaluation (complete with evaluation forms) is still planned for the SPC during the remainder of the winter. Also, a convective weather evaluation is also planned.
I remain extremely interested in the RUC and most certainly will continue looking at it, but I am shifting gears to fire weather forecasting and looking in particular at lightning and so-called "dry thunderstorms". Please contact Russ Schneider and he can tell you who is doing the on-going work.
Personally, I do feel the RUC will be a very big weapon in our SPC arsenal for storm prediction. You have made some very important advances with RUC II. With all the off-time observations incorporated into one data set (ACARS, satellite data, VADs, Profiler, etc.) I also think that the analysis (F00) every hour may be one of the "best things to come down the pike" in quite some time. This valuable analyzed data set may even help my fire weather work by providing a valuable aid for understanding today's weather better (our Day 1 Thunder forecast).
We appreciate all the work that has gone in to providing the RUC II forecast out to 12 hours (currently on the Cray 5). Many changes/updates and even a few corrections to errors that were found were quickly implemented in response to our input during the past two months.
We eagerly await the RUC II going to the Cray 3. We realize that for the evaluation, the porting to the Cray 5 was necessary, but the timely production and delivery of the model grids are crucial to the SPC. The 12 and 00 UTC runs were often arriving about the same time as the early eta, rather than earlier than the eta.
Unfortunately, we were not able to get the GEMPAK RUC II data before the testing/evaluation period began. We were several weeks into our testing and then we discovered what I'll term the error in the navigation of the gridded output (which was quickly corrected). Throughout the test, because of disk space, we were only able to ingest the 3 hourly runs (00, 03, 06 UTC, etc.) for testing. Our routine generation of some RUC II graphics for the meteorologists has begun this week and the rest will hopefully be on line very soon now that we have the extra disk space on a new machine. With the switch over to a machine with more disk space, we also hope to being ingesting and using the "off" hours (01, 02, 04, 05 UTC, etc.). With the isobaric grids, we did use GEMPAK to build selected isentropic surfaces for analysis.
I have archived some selected cases for more detailed analysis as time permits.
When I worked the mesoscale assistant desk (winter weather forecast), I did use the RUC II in my mesoscale discussions and found it very valuable. The problem was I could manually generate some graphics, but they were not "operational" for the other forecasters (and we had to watch the disk space over the holidays). I have since shared some of my results with other mesoscale forecasters.
Greg Carbin is a SPC winter weather forecaster (mesoscale assistant) who provided the following input. He has also been working on generating graphics using the RUC II data. Thanks to Greg for his input.
From Greg Carbin, SPC "Although the events are very limited, I did see a few cases where the RUC2 appeared to have some difficulty in estimating the strength of the lee-side troughing that occurs east of the Rockies during episodes of strong zonal flow aloft. On a number of occasions the RUC2 MMSL forecasts were off (high) by several mb from east-central MT southward into CO. Although the error appeared to diminish as the F00 hour approached the model did not completely catch up to reality.
[Stan: Could this be from the fact that a different SLP reduction is used in RUC-2 (the MAPS SLP reduction - see http://maps.fsl.noaa.gov/vartxt.cgi)? The obs use the NWS reduction.]
The RUC2 also seems to have trouble catching rapid radiational cooling in the boundary layer during the evening. I noticed a few events in the Plains when the forecast surface temperatures were several to tens of degrees too warm when compared with the verifying analysis. This usually coincided with mostly clear sky conditions during the evening when surface temperatures dropped rapidly, and the forecasts just did not indicate such rapid cooling."
[Stan: We've noticed this also for some time. Dan Petersen at MLB also pointed this out. Our prime suspects are radiation and night-time mixing near the surface. Unfortunately, there is no quick fix that we've been able to figure out. We think the performance of the soil temperature is very good, so that is not the culprit. Expect a fix here within a few months.]
GENERAL COMMENTS
Our evaluation took place using the National Center AWIPS (NAWIPS) software which uses GEMPAK software to display model output. During the evaluation , cross-sections of selected relative humidity, isentropic surfaces, temperature and circulation vector (composed of wind parallel to the cross section and the scaled vertical velocity) were produced. The cross-sections were all from 25 degrees to 50 degrees north latitude and were produced along every 2 degrees of longitude. They ran from 1025 mb to 400 mb. The documentation packet that was mailed (to Stan) shows a US radar mosaic with an example of how the cross section lines appear every 2 degrees (starting at latitude 25 and ending at latitude 50). Two sample cross-sections (along 90 and 86 degrees West longitude) illustrate what appears on the cross- sections). As a side note, it is obvious that with the more dense data, the circulation vectors make an already "busy" looking plot even busier. This was compounded on the cross-sections produced because the grid points were closer together. Also the scaled vertical motion which is often quite large in the RUC II can often produced very large circulation vectors. Temperature is contoured in degrees C at +4, +2,(solid red) 0, -2, and -4 (solid cyan) and again from -12 to -18 (solid cyan ) (to highlight the area of large crystal growth). Humidity is contoured at 10, 20, 30 (yellow dashed) and 70, 80, 90 (green dashed). Isentropic surfaces are contoured every 3 degrees (dashed magenta).
1. Vertical velocity
The addition of vertical velocity to the output fields was the single biggest enhancement.
Vertical velocities (both positive and negative) produced are larger than those produced by other models. Spatially smaller (yet larger absolute value) centers of positive and negative vertical velocities are visible in the output whether viewed as a cross-section or plan view when compared to other models.
There is no "spin-up" problem with this RUC (in viewing the 00 through 12 hour forecast data). Areas of large upward vertical motion are well correlated with nearly saturated air starting at the analysis (or F00) time and develop and/or decay in a consistent manner through the 12 hour forecast.
Within precipitating systems, these large (upward) vertical velocities correlate well with the precipitation. The large downward vertical velocities are quite often noticed in the lee of the mountains. At times, the downward velocities in the lee of the mountains can be quite extreme and can cause the forecaster concern. John Brown and Stan have addressed some of what is going on within the model. Forecasters will likely express concern about the large negative vertical velocities and or the general flow in the lee of the mountains so this should be well addressed in any additional future RUC II documentation.
The patterns of upward motion will often show multiple areas of upward motion rather than one large area. The very first set of graphics with a cover sheet labeled RUC II Cross-sections shows two distinct areas of upward motion from the 3 hour forecast valid at 18 UTC on 1/22/98 (90 degrees west). This correlates well with the national radar display at 1759 UTC.
2. Terrain
A single page labeled 12/2/97...Single time comparison of RUC II and eta illustrates the better terrain features (which the meteorologists find a
very attractive feature), the higher vertical velocities.
The meteorologists that have used the cross-sections have commented that the terrain field displayed on the cross-sections is very good. While simply displaying the terrain field on a cross-section may not seem that significant, the winter weather forecasters that have used the data have benefited from the increased features that are displayed.
3. Great Lakes and Gulf of California ...resolution by the model.
In displaying fields of near freezing surface temperatures and dew points of 50, 60 and 70 degrees F, the RUC II does an extremely good job of detailing the warmer waters of the Great Lakes and the much higher moisture over the Gulf of California. It was primarily these fields that led us to discover that the model GRIB output had an error which resulted in the data being plotted incorrectly because of a navigation error. During the coming year, it will be interesting to see how the model performs with moisture return from the Gulf of California. Hopefully you have received some input from other offices about the RUC II and the Great Lakes/Lake effect snows.
[Stan: Got some good comments from United Airlines and WFO BTV on lake effect snows.]
4. Sounding analysis via UNIX SHARP program
The SPC was not ingesting the point forecast soundings. However, we were able to use our interactive sounding analysis program "SHARP" to display soundings from any grid point. The three shown are from San Francisco, Tulsa and Midland. The model vertical velocities are shown on the right side of the figure and the meteorologist has the ability to modify the sounding temperature and dew point. We can also use this sounding analysis program to calculate other fields not provided by the model. Any site with a valid three letter ID can be specified and RUC II data retrieved and displayed for that site.
5. Precipitation fields
Our work in generating all the required graphics is only now nearing completion, but I have been impressed when I look at the precipitation fields and types of precipitation on a 4 panel display from 4 models all valid at the same time (i.e., and of the 00, 03, 06, 09 or 12 UTC runs). There is great consistency from run to run. This is certainly preliminary and subjective, but for example, when you compare a 12, 9, 6 and 3 hour forecast all valid at the same (the 3 hour total precip), there is usually no big changes or swings from run to run. Three hour forecast amounts generally remain consistent and in light precip areas of around 0.01 inch, if the 3 previous model runs tend to reduce the spatial extend from run to run, I've noticed that the 4th and most current run generally continues that trend. Bottom line. Consistency with no wild swings.
6. SNOW (binary 0,1), snowfall amount and precipitation type
I've only recently starting using the snowfall from SWEM03 (contouring it as inches per three hours. It looks very promising. Also the general rain/snow line is generally good although that is very preliminary. I have one very good case of freezing rain over Kansas, but have not had the time to look at the model performance for that, except to note that the model did not develop the very deep layer of warm air over the cold that was actually observed (see 1/4/98 case). I've not been able to evaluate the ice pellet cases, plus the routine has been changed within the model. More work needs to be done on that evaluation.
7. Surface temperature fields
In the case of the near freezing surface temperatures (28, 32, 36 degrees F), I have noticed that the forecast near freezing surface temperatures lag the "observed" objectively analyzed data by 1 to 2 degrees of latitude at the 12 hour forecast, gradually catching up to the actual data by the f00.
CASES
As a National Center, we have had the opportunity to look at several cases from across the US.
12/4/97 IOWA SNOW EVENT
Included in the figures are three radar images from DSM on the morning of 12/4/97 as snow developed over Iowa. The cross sections show on the 9 hour forecast that by 12 UTC, the model was predicting greater than 70 percent relative humidity with a well developed area of vertical motion over central Iowa. (12 hour was not yet available at that time). They received 3 to 5 inches of snow.
12/8/97 MIXED PRECIPITATION EVENT IN KANSAS (EVENING OF 12/7)
Included in the packet are a base reflectivity subset of the national mosaic and several cross-sections comparing the eta and RUC II. Several more distinct centers of vertical velocities can be noted in addition to the stronger vertical velocities in the RUC II. (Please note that it was not the intent to perform an eta and RUC II comparison, these figures are to highlight the differences in vertical velocities). The surface plots from 02 and 03 UTC show snow across northeastern Kansas and freezing rain in south-central Kansas and rain in southeast Kansas. The remainder of the cross-sections shows the well developed circulation and the temperature stratification forecast over the area of mixed precipitation. We were not able to display fields of RUC II precipitation type during this event, however since the precipitation type algorithms were changed during the course of the evaluation (snow/ice pellets), no further work has been done on this case.
12/9/97 NORTHERN KANSAS SNOW EVENT
A small scale band of precipitation oriented from west to east across north-central and northeast Kansas set up on the morning of the 9th as shown in the radar images. The cross-sections show a well developed circulation over the area with good vertical velocities through the region of good crystal growth from -12 to -18 degrees C.
12/14/97 8 INCH SNOWFALL AT JACKSON MS
A visible satellite photo shows the heavy snow band across central Mississippi. Surface maps from 12 and 16 UTC show the temperatures just at freezing with snow across the area. Vertical velocities contours from the 6 and 9 hour forecast along with the cross-sections illustrate the ability of the model to correctly forecast the small scale nature of the event. The temperature structure shown in the cross section also shows that only over the central part of the state was the entire column forecast to be below freezing. Subsets of the national radar mosaic also show the snow band in the developing stages from 10 and 13 UTC. The RUC II did predict 3 hourly amounts of .1 inch or more (liquid) for 2 consecutive three hour periods over central Mississippi. The model correctly predicted snow. As a side note, no heavy snow statements were issued prior to the event.
12/23/97 HEAVY SNOW (24 TO 30 INCHES MAXIMUM) OVER MASSACHUSETTS.
The first figure shows the location of heavy snow reports from 12 to 20 UTC on the 23rd. One report from the snow event reported 8 inches of snow in just one hour!!! As mentioned earlier in this report, the model "started out" with very strong vertical velocities over the area and they continued through the 6 hour forecast valid at 18 UTC over the area as shown in the cross-sections.
See comments from BTV on this event.
1/4/98 FREEZING RAIN OVER KANSAS
This case is presented to illustrate a significant freezing rain event where the 00 UTC 1/4/98 RUC II model did not correctly forecast the temperature stratification valid at 12 hours. The 12 UTC upper-air data was objectively analyzed and a cross-section was produced from the objectively analyzed data. As shown in the observed gridded data cross-section labeled only 980104/1200 (no Vxxx for valid time), the plus 4 degrees C at 825 and 800 mb extended much farther to the north than predicted. The surface plot valid at 10 UTC shows freezing rain and "4-dot freezing rain" over Kansas. This "4-dot" rain came from automated stations with temperatures in the 20s and the decoding of the weather type as rain. In an area of northwestern Oklahoma, the model correctly had convective precipitation forecast and there was thunder observed with much of the freezing rain.
Thanks again for the opportunity to participate in this evaluation. Feel free to contact me regarding any questions/omissions, etc..