For years WeatherTAP has offered a set of "winter" mosaic radar images that show not only the location of precipitation, but also the form, whether it be rain, snow, or mixed precipitation. Radar is definitely useful for determining the location and intensity of precipitation, but it is not as useful for determining its form. If radar data alone isn't sufficient for determining whether precip is rain or snow, how does WeatherTAP produce the winter mosaic? Good question....
To produce the winter mosaic products, we first build a "mask" to accompany the radar imagery that contains the most likely form of precipitation in all of the same areas covered by the radar.
To build this mask, the mosaic area is subdivided into a grid consisting of thousands cells. For each gridpoint, an analysis is done on a fictional mass of water dropped from high altitude. Using vertical temperature profiles derived from model data, the amount of energy entering and exiting the droplet is quantified.
As energy is transferred to or from the ambient environment, the temperature of the droplet will change. If the temperature passes through 0 degrees Celsius, and there is enough additional energy to overcome the latent heat of fusion, the droplet will experience a change in state. Ice or snow will melt into rain, or rain will freeze into ice or snow. As the droplet falls through the atmosphere on its way to the surface, it may pass through many layers of air at different temperatures and indeed may change state multiple times.
In theory, we might be able to do a complete energy balance and apply a microphysics model to the droplet, however, this approach is not currently practical as the computational demands would be excessive. In addition, it would likely require knowledge of parameters that are not normally observed or computed. Instead, we adopt a simplified approach that still quantifies the energy flow but does so more indirectly than a true energy balance.
The exact method involves computing a quantity that is proportional to the temperature and layer thickness for a number of pressure levels corresponding to the fall to the surface. The mean temperature of each level relative to 0 degrees Celsius is used to assign a positive or negative sense to the computed quantity.
The higher the temp above 0 degrees, the more energy moving into the droplet (tending to transform ice into rain). The lower the temp below 0 degrees, the more energy moving out of the droplet into the environment (tending to transform rain into ice). Thicker layers usually result in more energy exchange than thinner layers for the same mean temperature.
Once the areas and their signs have been computed for each vertical layer within the cell, they are summed and compared to known statistical data to determine the most likely state that the droplet will be in when it reaches the surface. This procedure is repeated for each of the thousands of cells that comprise the precip type mask.
Once the mask is built, it is paired with the radar data during the rendering process to produce the images that we call the "winter" mosaics. The precipitation form is indicated on the radar by shifting between three distinct color scales. This technique easily and conveniently depicts the location and intensity of rain, snow, or mixed form precipitation.
For those who want to know more about the science behind WeatherTAP's winter mosaics, we suggest referring to "A Method to Determine Precipitation Types" by Pierre Bourgouin in the October 2000 issue of the American Meteorological Society's Journal Weather and Forecasting. Incidentally the technique itself is referred to as the "Bourgouin Method" or "area method" and is also used officially by the Canadian Meteorological Center. A similar technique (but with different empirical thresholds) is employed by the NWS here in the States.
- Rob P.
Comments (2)
I think this is a very decent solution, though maybe unnecessarily complex. One problem is that the models are not always right--there seems to be a trend in meteorology today to use model initializations as an analysis, which can be incorrect.
As an alternative, why not simply start with METAR data and work off that first, then supplement with the Bourgouin method where there are insufficient data? At the very least, your method might include METAR data as a check against the precipitation type from your model.
Posted by Robert | January 6, 2007 6:14 AM
Posted on January 6, 2007 06:14
So far it seems reasonably accurate. In a sense, metar data is incorporated into the system because it is used as part of the model initialization. Of course that's somewhat indirect.
Thanks for the suggestions!
Posted by Rob P | January 8, 2007 3:39 PM
Posted on January 8, 2007 15:39