Unifying and Simplifying Winter Precipitation Accumulations in NCEP Models

The implementations of the Rapid Refresh Forecast System (RRFSv1) and GFSv17/GEFSv13 in the upcoming years present a much-needed opportunity for winter precipitation accumulation parameters in the NCEP Production Suite to be revisited, simplified, and unified. The current products are inconsistent across the modeling suite, leave much interpretation to the users, and create much confusion. The goal of this presentation is to describe the various scientific issues associated with winter precipitation accumulation products and discuss planned paths towards a more streamlined and consistent set of model output for winter precipitation parameters.

Other than the RAP and HRRR, no other operational NCEP model predicts an explicit snowfall accumulation in meters/inches. The primary parameter available for all other NCEP modeling systems is the water equivalent of snow depth (WEASD), in which snow and sleet are tallied together. The snow depth parameter does allow for an explicit accumulation but also has limitations, particularly in conditions with marginal temperatures and warm soil. In addition, other than in the RAP and HRRR, there is no attempt to predict accumulations of sleet or freezing rain, leading to many users to generate their own accumulations based on invalid assumptions.

The major problem with the accumulated WEASD field is that it is a liquid equivalent and not an amount of inches, so to display snowfall, the user must choose a snow-to-liquid ratio (SLR). No SLR is generated by the models, so the common user approach is to apply a very generic 10:1 ratio. This works fine in “cold” storms, which typically have SLRs of at least 10:1 (and sometimes higher, which can even lead to an underprediction of snow totals), but it performs terribly in storms with marginal temperatures (and accompanying low SLRs) and in storms with predominantly sleet or other mixed precipitation.

To increase confusion, there can be a disconnect between snow amounts and precipitation type. The WEASD parameter comes from the model microphysics and is based on the composition of the falling hydrometeors. The precipitation type field is computed entirely in the post processor using the wet-bulb temperature profile (note that the precipitation type is computed by a “dominant” approach, as four algorithms based on the wet-bulb temperature profile are used; the NAM has 1 extra member (out of 5) that is based on the microphysics), so it is possible for the WEASD field with a 10:1 ratio to indicate several inches of snowfall while the precipitation type shows rain or freezing rain. The RAP and HRRR do not use this method and compute precipitation type in the post processor using output from the microphysical scheme and temperature data at the lowest model level.

The upcoming RRFSv1 and GFSv17/GEFSv13 implementations provide an opportunity for unification of winter precipitation products. The RRFS will subsume the NAM, Hi-Res Windows, RAP/HRRR, and HREF. As noted, the RAP/HRRR system has a more sophisticated set of winter products, and matching those products in the RRFS and GFS/GEFS is a desirable goal. Both the RRFS and GFS/GEFS should provide snowfall totals (in actual depth of snow) and accumulations of freezing rain and sleet. The dominant precipitation type code can be replaced with code based on the output from the microphysical scheme. Finally, snow-to-liquid ratio (SLR) will be computed directly and become a direct model output. A plan forward for this project and a proposed timeline will be discussed.