New SREF Snow Products and Diagrams

The National Centers for Environmental Prediction Short-Range Ensemble Forecast System (SREF) provides forecasts for the contiguous U.S. at 16-km grid spacing four times a day out to 87 hours.  The ensemble is comprised of 26 forecasts, with 13 from the Advanced Research version of the WRF model (ARW) and 13 from the Nonhydrostatic Mesoscale model on a B grid (NMB).  NOAA does not stand for the National Organization for the Advancement of Acronyms for nothing.

We have been downscaling the SREF forecasts to higher resolution for some time, but Mike Wessler, a student in my group, has recently upgraded our processing significantly.  Specifically, he's decreased the grid spacing of the downscaled forecasts to 800 meters, added estimates of snow-to-liquid ratio, and added the capability of converting the downscale precipitation forecasts into downscaled snowfall forecasts.  He has also improved the graphics considerably. 

We went "live" with these improvements yesterday.  Below is an example of the downscaled precipitation (i.e., water equivalent) forecast from the 0300 UTC 19 February initialized SREF.  Included is the ensemble mean, maximum (i.e., highest value of all 26 members at each location), minimum (i.e., lowest value of all 26 members at each location), and the percentage of members at each location producing 0.01, 1, and 2 inches of precipitation during the 87-hour forecast period.  I've chosen Colorado as it shows a great deal of structure in this forecast period. 

With snow-to-liquid ratio estimates, we can convert that water into snow.  Below is an example of the downscaled snowfall product that includes the ensemble mean precipitation (liquid equivalent) and snow, followed by the percentage of ensemble members producing 1, 6, 12, and 24 inches of snow during the forecast period.  Mammoth Mountain users can perhaps laugh at the 24" top category, at least this year. 

Finally, there are upgraded diagrams for specific locations.  Below is an example for Wolf Creek Pass.  Readers of this blog will be familiar with the plume diagrams at left, with water equivalent on top and snowfall on the bottom.  In addition to a line for all 26 members, colored by core, means for the ARW, NMB, and SREF as a whole are provided. 

On the right are "violin plots" of 3-hourly precipitation and snowfall.  These violin plots provide information about how many members lie within each 3-h precipitation or snowfall amount range.  Their width is proportional to the number of members.  Where they are fat, there are more members, and where they are skinny there are fewer.  This is sometimes referred to as a probability density.  Black bars denote the middle 50% of the forecasts and red lines the middle 90%. 

We are also plotting information on the snow-to-liquid ratio used.  We are currently using a very simple snow-to-liquid ratio algorithm for a variety of reasons.  I think by plotting the snow-to-liquid ratio we will see we need to improve it.  In this instance, all the members are using a snow-to-liquid ratio of 15:1 due to the cold temperatures that prevail.  However, a grey shade region will appear above and below the grey line during periods when we are estimating different snow-to-liquid ratios from the ensemble members.  That shade region will encompass about 68% of the ensemble members. 

This is an experimental product, available on weather.utah.edu.  Look for "SREF-Downscaled" in the left hand navigation bar.  For each region, you will find a PQPF and PQSF product, the former for liquid equivalent and the latter for snow.  Here, PQPF stands for probabilistic quantitative precipitation forecast and PQSF for probabilistic quantitative snowfall forecast. 

We have a minor bug in the automated processing that we should be able to fix shortly, but there might be some delays in product production for another day or two.  Beyond that, the snow-to-liquid estimate is something that we will need to work on improving.  A challenge with using the ensemble data is that that it takes a great deal of time to download and process three-dimensional data, so we need to take shortcuts compared to the algorithms we use for the Little Cottonwood Guidance at http://weather.utah.edu/text/COTTONWOODS.txt.

If you use this product and find it helpful, buy Mike a beer or send him a free lift ticket.