Last updated on January 25, 2020
Snow Maps - love them or hate them, they're everywhere in winter! Confusion reigns when it comes to different snow map algorithms. While this is a messy topic with few easy answers, our goal is to clear up some of the confusion in one central place.
What is snowfall?
In Pivotal Weather, "snowfall" refers to snow that is like thisreachedEarth's surface over the specified period. If a given ground surface is warm enough to melt, the accumulated snow pile you see on that surface at the end of a storm can be noticeably less than what we call snowfall. Suppose you had a snowboard that you kept well below freezing and you diligently went outside every hour to measure and clear new snow. Not much melting, sublimation, or densification would take place during these hourly intervals, regardless of weather conditions. The sum of all snow you cleared from the chilled snowboard during the storm would represent the observed "snowfall" that our 10:1 and Kuchera* snowfall maps attempt to forecast.
Predicting the ultimate accumulation on a ground surface at the end of a prolonged snowstorm is more complex. It depends on the surface type, in addition to the weather conditions on the ground and their development during the storm. Even solar radiation passing through clouds and thus the time of day can have an impact on melting. We don't attempt to forecast this explicitly, but our snowfall products are still often a useful proxy for ultimate accumulation on pristine natural surfaces. However, this may not be true when ground temperatures are warm, air temperatures are above freezing, or when a storm is particularly long-lasting and compaction is a major factor.
*The Kuchera method was originally formulated to fit a sample of observed snow depth measurements (e.g., a ruler reading after a storm), so even our attempt to define "snowfall" has caveats - more below.
Evaluation of the model precipitation type
Our primary snowfall product types, 10:1 and Kuchera, apply specific snow-to-liquid ratios (SLRs) to the precipitation in the model that we believe fell as snow between data output times. But wait a minute... do we really know how much snow has fallen?
The ECMWF, UKMET and Environment Canada models track precipitation type precisely when the model is integrated, so we know how accurately precipitation falls in the form of snow (at least based on the model's internal diagnosis). This removes any concerns about including sleet, sleet, or rain when we calculate snowfall for these models.
With NCEP models, the accounting for precipitation types is less accurate, so miscategorizing some precipitation that fell between data output times as mixed precipitation or precipitation that changes rapidly is always a risk.We've taken an approach that typically avoids treating sleet incorrectly as snow for NCEP models, so you shouldn't see a "false snow" shield extending well beyond the actual sleet line in a large mid-latitude cyclone, for example extends. Still, it is inevitable that we sometimes overestimate the fraction of mixed precipitation that falls as snow in borderline and transitional (usually small-scale) environments.
Snow to Liquid Ratio (SLR)
Now on to the topic of SLR (often colloquially called "snow ratio" or just "ratio"). After snow falls, we can melt it and measure the liquid precipitation it contains. Dividing the snowfall by this liquid equivalent amount gives the SLR. Since current numerical weather forecast (NWP) models directly predict liquid-equivalent precipitation, some SLR must be applied to the predicted amount of liquid in order to obtain a meaningful snow forecast.
For over a century, weather forecasting has commonly used a 10:1 SLR (1 inch liquid = 10 inches snow) as the standard. A large climatology of SLR cameras in the USA byBaxteret al. 2005found a distribution of scores centered around 12:1, with scores between 10:1 and 12:1 occurring more frequently than any other bin (see fig. 9). Values in the range of 6:1 to 18:1 are fairly common in the US and can occasionally reach 2:1 on the low end and 44:1 on the high end. It's obvious that while 10:1 is a reasonable 'default' if you had to pick one, this approach regularly results in SLR errors of >50%!
From a physical point of view, SLR boils down to the structure and density of the snow crystals, whose formation mechanisms are quite complex (cfTakahashiet al. 1991). However, like any such pinpoint detail, current NWP models can only parameterize (estimate) this based on larger variables such as predicted air temperature, humidity and wind. Within the model, these variables could theoretically be used in a nuanced way to estimate SLR with considerable accuracy, but this is rarely done in the current operational NWP. Instead, outside users like Pivotal Weather must estimate the SLR themselves based on the more limited publicly provided data.
In late 2004, then Air Force Weather Agency graduate student Evan Kuchera developed what has since become widely known as the Kuchera SLR method. It is an attempt to relate model-predicted variables to SLR and is a linear function of only one value at each horizontal grid point: the warmest temperature in the air column from the surface to 500 mb. Describing the origins of his approach, Evan said:
"Basically, I am manually correcting data from various snow events that I was aware of at the time [2004] until I was satisfied. Notably, the bifurcation at 271.16 K should attempt to account for melting effects after the snow lay flat for warmer events. So I really tried to target all of the storm's snowfall that a COOP observer or member of the public would measure, rather than pure, literally correctly measured and snowboard-cleared snowfall.”– Evan Graben
Shortly after developing the method, Evans' colleague Earl Barker (www.wxcaster.com) implemented it for its online NWP graphs, and the rest is history – “Kuchera snowfall” is now part of almost every winter weather enthusiast's vocabulary and is computed by numerous NWP graph providers! Although this method has not been published in a peer-reviewed scientific journal, its simplicity of formulation and subjective usefulness have made it increasingly popular, and verification papers presented at academic conferences have also confirmed its usefulness. Air temperature does not exclusively determine SLR in the real world, but several published studies have shown a fairly strong relationship between low to medium temperatures and observed SLRs (e.g.Röbber et al. 2003;Alcott and Steenburgh 2010). If NWP users want to implement a simple approach to SLR that doesn't break their data processing or requires obscure model diagnostics that they don't have access to, they probably won't do much better than Kuchera.
Although Kuchera can deviate significantly from the observed SLR in some cases, it should still offer a first-order improvement over assuming a blanket 10:1 ratio. In fact, its advantage can be most apparent when temperatures are borderline, a situation where it correctly reduces snowfall below a 10:1 estimate, as Evan intended. Nevertheless, we emphasize that Kuchera is highly imperfect, as true SLRs depend on cloud and precipitation physics, which is far more complex than a single column temperature distribution statistic. In the future, we hope that NWP models can start tracking snowfall internally, using more physically-based diagnostics to estimate SLR at sub-hourly intervals, increasing the snowfall forecasts versus those derived from 10:1 and Kuchera-derived SLRs could improve significantly.
Model snow height
Finally, we would like to discuss the use of model snow depths. We record snow depth for many models; In addition, for some, we record an estimate of snowfall using "cumulative positive snow depth change".The name "Snow Depth" seems to imply that it is a highly accurate representation of the snow that accumulated on the ground during the model run, removing non-snow precipitation and applying highly accurate SLRs. Unfortunately, this is not usually the case, leading to misunderstandings in the weather community.The details of how snow depth is calculated vary from model to model, even within the NCEP suite. Below are links to technical descriptions of snow cover, density and depth for three different classes of NCEP models:
NCEP RAP/HRRR:https://sites.google.com/a/ucar.edu/model-encyclo-determ/deterministic/rap/snow-initial-condition-and-forecast
Environment Canada GDPS/RDPS: see Eq. 24 fromBelaire al. 2003
Note: Modeled precipitation is typically counted as snow if more than half of the precipitation is frozen, even if some or all is actually sleet. Furthermore, the SLR applied to this frozen precipitation is based solely on surface air temperature for NAM and RAP/HRRR; It is similar with the GDPS/RDPS, except that the surface wind speed is also taken into account. Essentially, these SLRs are just simpler and less accurate versions of what Kuchera does, only considering one temperature level instead of many. Therefore, we cannot assume that the snow depth in the model is accumulated with an accurate SLR, nor that it only contains snow. [Note that a particularly common error in using surface temperature to derive SLR is having a "warm nose" over very cold air at the surface, in which case it may be SLRveryoverestimated; Kuchera approaches this scenario more realistically].
On the other hand,the snow depth variable tries to account for melting, compaction and sublimation on a representative ground surface,and is even able to take advantage of up-to-the-minute changes in the soil model state. So in that regard, it may be more useful for estimating soil accumulation at the end of a snowstorm than our 10:1 and Kuchera snowfall products. However, this advantage is offset by the significant pitfalls of using very inaccurate SLR cameras and typically treating sleet as snow.
Conceptually, users should recognize that the snow depth variable is just a by-product of internal model considerations about surface flows; This is an area of physics where the exact snow depth may not be quite as crucial as the total mass of frozen precipitation covering the ground. Therefore, using model snow depths to predict snowfall is subject to limitations and errors that are similar in size to 10:1 or Kuchera, and it can perform even worse in some situations!
Explicit model snowfall
HRRRv4 and RAPv5 (implemented at NCEP in December 2020) started to provide explicit snowfall forecasts; To our knowledge, this is a first in the established operational NWP. These forecasts can be found on Pivotal Weather as the Total Accumulated Snowfall parameter. Although the HRRR/RAP use a very simplified SLR, which is a rigorous function of the 2m AGL temperature, they are able to assess SLR and melting at each model time step, providing far greater temporal precision than our post-processed ones Products (e.g. , Kuchera).
In addition, the NWS National Blend of Models applies relatively sophisticated SLRs to the QPF of each input model. Although NWS NBM precipitation forecasts are essentially post-processed ensemble means and therefore can tend to smooth peaks in forecasts beyond the first 12 to 24 hours, the NBM's SLR approach is more advanced than any individual NWP model for Pivotal Weather.
Summary and practical recommendations
Our snowfall products generally try to predict the snow that will fall on the surface. not necessarily the pile of snow you see on your lawn, highway, roof, or anywhere else after a long storm. There are some caveats with Kuchera (partially penalizes warm temperatures to explain bottom melting) and accumulated positive depth changes (explicitly explains melting, albeit with the frequency of model data files as a confounding factor) - but neither of these products will do so consistently provide an accurate prediction of the final ruler-measured snow depth, even when the model's QPF and vertical profile are spot on!
We recommend using Kuchera snowfall products in most situations. Kuchera Snowfall is certainly imperfect, but in our view it's the least flawed practical option on the table for most models right now.
Model snow depth can also be very useful when your main prediction problem is the final ruler measurement on a natural surface, especially in events where melting is a major concern. However, this product sometimes contains sleet, and the implied SLR cameras used are probably less accurate than Kuchera most of the time.
The 10:1 ratio snowfall products exist primarily as a very conservative estimate for legacy purposes and because they are easy to compute. For certain datasets, we may not be able to calculate Kuchera, leaving 10:1 as the only practical option. In the future we may consider removing some or all 10:1 products, but they may still have value for some users as a baseline approach that can be easily compared across all datasets.
In our view, the best route to more accurate and less confusing NWP snow forecasts is for modeling centers to internally track snowfall during integration, rather than just liquid-equivalent frozen hydrometeors. The current situation leaves it up to end-users like us to apply SLRs (and in some cases infer precipitation types) based on limited and time-sparse data. Even a simplified in-model SLR algorithm that estimates the expected crystal type and edge (with the benefit of full-grid data and microphysical parameters), applied during integration much more frequently than publicly available data files, can lead to drastically better snow forecasts than today's.UPDATE: RAP and HRRR have started providing explicit snowfall forecasts as of December 2020, so hopefully more models will follow in the near future!
(Video) A Look at When Winter Will Unleash Arctic Blasts and Major Snowstorms - Direct Weather Channel
Freezing rain
The freezing rain situation reflects snow in many ways: we know that a certain amount of liquid precipitation will fall as rain into below-freezing ground-level air, but that doesn't mean that the accretion on a given surface is the liquid equivalent. In fact, during significant ice storms, ice accumulation is almost always less than Freezing Rain QPF (QPF)—sometimes less than half when precipitation falls in violent bursts or when temperatures are near freezing.Our Freezing Rain QPF charts should never be used literally as predictions of accretion; they are merely a starting point for such a prognosis.
For Plus subscribers and some models, we recently added the Freezing Rain Accumulation Model (FRAM;Sanders and Barjenbruch 2016) as a rough estimate of accretion thickness on elevated horizontal surfaces. This product predicts the ice-to-liquid ratio (ILR) empirically from a large climatology and is quite analogous to Kuchera for snowfall, except that wind speed and precipitation rate are taken into account in addition to (wet-bulb) temperature. Keep in mind that ice accumulation is even more difficult than snowfall accumulation and varies widely depending on surface type, shape, exposure and orientation. As such, we urge accountability and context when using even FRAM for publicly available predictions.