Friday, October 5, 2018

Forecast Tools on

With winter approaching (conditions currently and over the next few days are really just a teaser), it seems a good line to discuss some of the forecast products that we have available on

All products available on are based on computer models run by the National Centers for Environmental Prediction (NCEP), a part of the National Weather Service.  We download and process this data, producing a variety of plots, graphs, and tables that can assist in the preparation of a weather forecast.  Really, what we provide on are not forecasts but guidance, which is defined by the National Weather Service Glossary as computer generated materials used to assist the preparation of a forecast. 

Output is available from the Global Forecast System (GFS), North American Mesoscale Forecast System (NAM), High Resolution Rapid Refresh (HRRR), North American Ensemble Forecast System (NAEFS), and the Short Range Ensemble Forecast System (SREF).

Data from these models is downloaded anywhere from 2 to 24 times per day depending on run frequency.  We also download the data at different grid spacings (a.k.a., resolution) depending on the application.  Here is a breakdown:

GFS-0.25 degree: Global Forecast System data provided on a 0.25º by 0.25º latitude–longitude grid.  Used to produce all plan view (horizontal) plots under the GFS-0.25deg tab in the left-hand nav bar.

GFS-13 km:  Global Forecast System data provided on a grid with cells approximately 13 km on a side.  This is essentially a native resolution product.  I like this to examine the precipitation and wind forecast at the highest resolution possible. 

FV3-13km: These are experimental forecasts produced by the next-generation version of the Global Forecast System, based on what is known as a Finite-Volume Cubed-Sphere Dynamical Core.  The FV3 acronym comes from Finite-Volume (FV) and Cubed (3).  Essentially, this is the next generation GFS.  This is essentially a native resolution product with grid cells approximately 13 km on a side.  The FV3 will become the operational GFS probably in early 2019 and this product will go away. 

NAM-12km: Essentially a native resolution product from the operational NAM with grid cells approximately 12 km on a side. 

NAM-3km: Essentially a native resolution product from the high-resolution NAM conus nest, a high resolution grid covering the continental United States with grid cells approximately 3 km on a side. 

HRRR: High Resolution Rapid Refresh data provided on a native-resolution grid with cells 3 km on a side.  Forecasts are produced to 18 hours every hour.  Although operational HRRR forecasts are extended to 36 hours at 0000, 0600, 1200, and 1800 UTC, we're not accessing data past 18 hours yet. 

NAEFS-Downscaled: This is a unique product based on the North American Ensemble Forecast System (NAEFS).  The NAEFS is based on the GFS-based Global Ensemble Forecast System (GEFS), which produces 21 forecasts with an effective horizontal grid spacing of about 33 km, and Canadian Ensemble Forecasts produced by their GEM model (also 21 forecasts).  Precipitation forecasts produced by the NAEFS are downloaded on a 0.5ºx0.5º latitude–longitude grid, but downscaled to much higher resolution based on climatological precipitation analyses.  We also estimate snow density from additional NAEFS forecast fields to produce a snowfall forecast. 

SREF-Downscaled:  Based on the Short Range Ensemble Forecast System with 26 members, with precipitation downscaled similar to the NAEFS.  We haven't incorporated a snow density estimate yet to produce a snowfall forecast, but hope to do that in the future. 

Time Heights and Soundings: Time height diagrams and soundings are available for some locations under the GFS-0.25deg and NAM-12 tabs.  These are produced using special vertical profile data (known as BUFR) from the nearest model grid point at the highest vertical and temporal resolution possible.  The time-height sections are fairly unique.  I don't know of any other sites that provide them, but I stand to be corrected.  They are quite useful for precipitation forecasting in complex terrain.

Little Cottonwood Guidance: A tab at the top of provides access to tabular guidance for upper Little Cottonwood Canyon based on the nearest model grid point from the NAM-12km, GFS, and NAM-3km.  This includes the wet-bulb zero level (height where the wet-bulb temperature drops below 0ºC and useful for estimating snow level); snow-to-liquid ratio; snow water content; temperature, relative humidity, and wind at the elevation of Mt. Baldy (11,000 feet); precipitation amount (1-hour and total); and snowfall amount (1-hour and total).  The wet-bulb zero, Mt. Baldy temperature, and Mt. Baldy relative humidity are based on the model forecast temperature and moisture profile.  The Mt. Baldy wind is derived from model winds based on a simple algorithm trained with past data.  The 1-hour and total precipitation amounts are taken directly from the model, but snow density, snow water contents, and snowfall amounts use the snow-density algorithm described by Alcott and Steenburgh (2010) and developed using past observations at Alta. 

Note that meteograms for Alta available under the GFS-0.25deg and NAM-12 tabs are based on this data. 

Lake-Effect Guidance:  A tab at the top of provides access to tabular guidance for lake-effect precipitation that is based on model data, recent Great Salt Lake temperature observations, and techniques described by Alcott et al. (2012).  Emphasis is placed on the likelihood of lake-effect precipitation and its location.  Amounts are not predicted.

Caveats: With a few exceptions noted above, provides access to model guidance without bias correction.  This is one reason why these products are called guidance and not forecasts.  Even the downscaled products, while accounting to some degree for terrain effects, exhibit biases, especially at individual locations.  Model bias is also highly dependent on location and model grid spacing, so the 3-km NAM has different biases than the 12-km NAM and those biases vary depending on location.  Gowan et al. (2018) examines the performance of many of these models at high altitude SNOTEL sites, but does not provide statistics for individual stations.  Caveat emptor (buyer beware).  A good carpenter knows their tools.  For official forecasts, use the National Weather Service, and for snowpack information, the Utah Avalanche Center. 


  1. I love the forecasting tools (as well as your insights about weather and climate), but I have one suggestion for improvement. Most computer screens are wider than tall, so when the visual is taller than wide I need to shrink it down to see, and as a result don't see the smaller scale picture as clearly. Would it be possible to move the speed controls from the top to the side? I'm thinking particularly of the northern hemisphere tropopause pressure that you show in your post. Thanks!

  2. Hi Jim,

    Thanks for sharing this - I had often wondered what the different forecast (guidance) tools were on the website. It would make it a lot more accessible to me (and presumably others) if the time the forecast or retrocast was for was displayed readily at the top of the maps. Trying to parse the time information out of the yellow long form at the bottom is challenging for me and makes me just want to give up and go to instead. This might be a great challenge for the students in the CS department that are looking for a capstone software project!


    1. Any good computer scientist will thumb their noses at the software we use to generate those plots!

      Really, static plots of this type is pretty old school, but I don't have the stomach, time, or money right now for an end to end rebuild. Maybe down the road, but "skibatical" is approaching...