21 February 2015

Understanding How VERY Difficult Forecasting Snowfall Is

Posted by Dan Satterfield

This forecasting snow Info graphic made for me by Ilissa Ocko.

This forecasting snow Info graphic made for me by Dr. Ilissa Ocko a Post-Doc at EDF.

Dan's pic.

Dan’s pic.

In the warm season, if we forecasters are off by two degrees, and get the rainfall off by a tenth of an inch, not one person will notice. In a snow event, this error is often the difference between nothing, and an icy mess on area roads. This happened today in Northern Alabama, where the models missed a very light amount of precip. but that one tenth of an inch caused serious road problems. The temp. forecast was off by only three degrees as well, so that added to the problems.

Snowfall is so extremely dependant on so many factors, many of which are very difficult to precisely forecast, that I’m actually surprised we do as well as we do! It’s especially difficult in areas where snow is rare, because most events there are right on the line between rain vs snow. In these cases, a very slight change in the variables can make a major difference.

Another factor is the snow to liquid ratio. You may have heard that one inch of rain gives ten inches of snow, but this is not really correct. It actually varies from as much as 20:1, to as little as 6:1. A recent study has shown that a good way to predict this ratio (which is vital to make a decent snow prediction) is to use what is called the 850 to 700 millibar thickness.

Let me explain what this is, and you need to think in three dimensions to really understand it. The pressure drops as you go above the surface, and at around 1500 meters the pressure is around 850 millibars. Climb to 3.5 kilometers above the ground, and your atmospheric pressure is around 700 millibars. I say around, because the height above the ground depends on two things, the pressure and the temperature. The vertical distance between the 850 millibar level and the 700 millibar level is what meteorologists call the 850/700 mb thickness. This distance can be shown with thermodynamic equations to be dependent on the mean temperature of that layer of air.

It turns out that this thickness can be used to estimate the snow to liquid ratio, and I received some info on this sent to forecasters from Jeff Orrock the Meteorologist in Charge at the NWS office in Wakefield, Virginia. Now that you understand thickness, it should make sense! A word of caution: This tool is valid for the Mid-Atlantic and it will likely not work as well in other regions.

AKQ_SnowAmt_from_QPF_and_Thickness Tool

This smart tool produces a SnowAmt grid based from the QPF grid and derived snow-to-water ratios (product of both). A linear relationship between snow-to-water ratios and 850-700 mb thicknesses used in this smart tool was derived from a local study at WFO Wakefield. The project included a total of 53 snow events, from January 1999 through March 2005, at three co- located climate and RAOB stations in the mid-Atlantic region (Dulles/Sterling Virginia, Blacksburg Virginia, and Greensboro North Carolina). The goal was to ensure consistency between observed snow-water ratios and upper air analysis (from which thickness was calculated).

Observed snowfall and liquid equivalent data were gathered from the local climatological data (LCD) summaries for each site, while thicknesses were derived from the RAOB data. Observed liquid precipitation was retrieved through the hourly data, not 24 hour or daily summary, during the period of accumulating snow only. This was done to obtain the most representative snow-to- water ratio, as in many events the snow had either mixed with or changed over to another precipitation type. Thicknesses were calculated using the RAOB(s) coinciding to the period of accumulating snow. Since many of events in this study lasted 12 hours or longer, multiple soundings were required (averaged) in order to obtain a representative thickness value, including special 06Z and 18Z soundings when available.

Comparisons were then made between observed snow-to-water ratios and low-level thicknesses (1000-850 mb and 850-700 mb). The results indicated poor correlation of snow-to-water ratios to the 1000-850 mb thickness; however, correlation to the 850-700 mb thickness was by far more favorable (linear), with much less variance noted. It is hypothesized that the best correlation was found from the 850-700 mb thickness, since cloud microphysical processes for snowflake growth typically occur in this layer. A “best fit” linear regression equation between snow-to-water ratios and 850-700 mb thicknesses was derived, and then subsequently incorporated into this smart tool.

AKQ_SnowAmt_from_QPF_and_Thickness also incorporates an optional reduction adjustment factor, selectable at the user’s option, designed to lower snow accumulations when near surface and/or ground temperatures are above freezing. This adjustment factor is independent of the snow-water-ratio calculated in this smart tool, as it takes into account surface temperatures and not the thermal regime above the surface (i.e. 850-700 mb layer). The reduction factor is a fixed value, and is multiplied along with both the QPF and derived snow-to-water ratio to yield a SnowAmt grid.

The SnowAmt grids are returned to GFE only where snow or a mixture of snow and sleet is present in the weather (“Wx”) grids. In other words, this smart tool will not assign snow amounts to grids where just sleet or the following mixtures are occurring: sleet and freezing rain, snow and rain, or snow/sleet/freezing rain.
Summary of Snow-Water Ratio to 850-700 mb Thickness Relationship, based from the best-fit linear regression:

SW_Ratio = (-0.1621 * ThickH85toH70) + 257.98 

 

Ctsy NWS Wakefield,Va.

Ctsy. Larry Brown, Senior Forecaster NWS Wakefield,Va.