Numerical Methods For Determining Precipitation Type

This section describes numerical methods models, algorithms, and humans can use to determine precipitation type. These use a variety of approximations and physical assumptions which vary based on the thermodynamic setup. Given the various sensitivities to precipitation type -- e.g., precipitation rate, convection, hydrometeor size, presence of a warm nose and/or cold dome, etc. -- ideally a multi-method ensemble approach would be used in a probabilistic framework.

Rudimentary Thermodynamic Profile Techniques

These techniques are generally not used any more in numerical models and post-processing algorithms, but are still sometimes used by human forecasters for a quick judgment on possible precipitation type.

Pressure Layer "Thickness" Techniques:

• These simple techniques based on temperature profiles were created from rules-of-thumb in times of limited computing capabilities to quickly delineate regions of snow, so the user can then go on to estimate snow amounts. These can be generally categorized into two types:
  • Deep Layer (1000-500 hPa): Areas with deep layer thickness less than a certain depth are often approximated to be completely below freezing, so snow is assumed to be the primary precipitation type (usually, 5400 m is used near sea level)
  • Partial Layer (usually 1000-850 hPa or 850-700 hPa): Different thickness values for each layer can indicate what type of precipitation (snow, sleet, rain, freezing rain, or a mix of types) may occur since thickness relates to temperatures in each layer
    • Was developed for southeastern Canada but is sometimes used in the northeastern US (See Cantin and Bachand 1993)
• Limitations:
  • Uses a single thickness value to approximate the layer's maximum temperature, not accounting for the actual temperature profile - so it can fail to account for shallow atmospheric layers above freezing, overestimating areal snow coverage
  • Does not account for the moisture profile; for example it would miss evaporative cooling potential in a dry air mass
  • Critical thickness values can vary by elevation

"Top-Down" Methods:

• Various techniques using the thermodynamic profile (dry-bulb temperature and dewpoint, wet-bulb temperature, or relative humidity) above a given point, starting from the level of precipitation particle development and going down, to determine the surface precipitation type at that point.
• Limitations:
  • Doesn't account for vertical motion which can change the local thermodynamic profile
  • If a dry-bulb temperature profile is used, would not account for evaporative cooling in a dry air mass
  • Does not account for melting and refreezing from smaller cold layers, it is solely based on maximum temperature of a warm nose and minimum temperature of the cold dome below.

Physical Methods

Dominant Precipitation Type Based on Wet Bulb Temperature Profiles (Manikin 2005 and Model Evaluation Group 2021):

• This technique uses the wet bulb temperature profile in model post-processing to create a mini-ensemble of precipitation type outcomes at the surface based on different methods.
• The majority of techniques in the mini-ensemble determines the surface precipitation type
  • If there is a tie, the precipitation type is determined in order from most to least "hazardous" as follows: Freezing Rain > Snow > Sleet > Rain
• Limitation: Although this uses a mini-ensemble approach in post-processing, this method does not account in any way for synoptic/mesoscale uncertainty. This approach only attempts to capture the uncertainty associated with how different schemes handle wet bulb temperature profiles to determine surface precipitation type.

Thompson Bulk Microphysics (Thompson et al. 2008):

• Advanced technique that predicts mixing ratios of five liquid/ice species: cloud water, rain, cloud ice, snow, and graupel AND the number concentration of cloud ice
• Provides much more accurate precipitation type and QPF amounts by breaking out different types of precipitation within the model output, yielding a better snow forecast
• Limitation: it is still a parameterization, therefore it is not fully representing actual physical processes

Global Systems Division (GSD) Thompson scheme (for HRRR/RAP) (DTC/CCPP Reference)

• More advanced than the original Thompson 2008 scheme by including aggregation of liquid and freezing precipitation in cloud
• Includes/allows for supercooled water droplets in-cloud improving growth/aggregation rates and types
• Limitation: Though slightly improved, it is still a parameterization, therefore it is not fully representing actual physical processes

Bourgouin Method (Bourgouin 2000)/Revised Bourgouin Method (Birk et al. 2021):

• Based off of Derouin (1973), Cantin and Bachand (1993), Ramer (1993), and Baldwin and Contorno (1993) methods
• The original method looks at the vertical temperature profile to determine areas above and below 0°C which, along with mean thickness temperatures, are used to determine magnitude of melting and refreezing, and then uses surface temperature and dewpoint to determine a final categorical precipitation type/mixture among these choices: Snow, Rain, Freezing rain, Sleet, Mixed rain/snow, Mixed freezing rain/sleet
• The Revised Bourgouin Method implements several improvements:
  • A larger developmental and independent dataset
  • Use of wet-bulb temperature profiles
  • Ability to diagnose freezing precipitation in situations without ice nucleation, improving overall prediction of freezing rain events
  • Probabilistic output that accounts for uncertainties in the spectrum of hydrometeor sizes, forecast or observed soundings, and space and time
  • Allowance for all combinations of wintry mixes
• Limitation: Both methods will struggle with rapid temporal and/or spatial variations in the thermodynamic profile (often the case around convective precipitation)