# Interpreting adjectives in weather forecasting

1. Apr 16, 2012

### BarryS

I have seen the Probability of Precipitation defined as

$\mathrm{PoP} = C \cdot A$

where $C$ is the confidence that measurable precipitation will occur somewhere within the forecast area during the forecast period, and $A$ is the percentage of the forecast area that will experience measurable precipitation assuming that any occurs at all (i.e. the conditional probability of a randomly chosen point in the forecast area experiencing precipitation given knowledge that precipitation occurred somewhere within the area).

I have also read the following descriptions for adjectives attached to nouns like "showers" or "thunderstorms"

"isolated" means less than 15% areal coverage
"widely scattered" means 15-25% areal coverage
"scattered" means 25%-55% areal coverage

I just want to get confirmation from an expert that "areal coverage" this means this terminology is describing precisely a range of values for $A$ in the above formula (and not PoP directly), and that furthermore it indicates that $A$ is in the respective ranges.

2. Apr 17, 2012

### wsnell

I work with the NWS in Puerto Rico. We are one of the few offices that uses "areal coverage" exclusively in the United States. When we refer to POP, or the "Probability of Precipitation" we refer to the amount of area within a specific predefined zone that we expect to be covered by 0.01" or more of rain, usually over 12 hours from 0600L - 1800L (for daytime) or if in an update for the remainder of the period. Currently adjectives allowed are:
Isolated 5 to < 25% coverage (Slight chance when referring to probability)
Scattered 25 to < 55% coverage (Chance)
Numerous 55 to < 75% coverage. (Likely)
The next higher category is considered "categorical" and various adjectives are used, including, Brief, Frequent, Occasional, Periods of, Intermittent, and Widespread. Omission of the adjective implies categorical eg. "Showers" with no adjective.
Point forecasts are for specific points and are different from area or zone forecasts.
Terms like "Widely scattered, few scattered" are not used because they tend to confuse rather than illuminate. Hope this helps.

3. Apr 18, 2012

### BarryS

Thank you. My I presume you use areal coverage exclusively because you are in a tropical zone, and so your confidence of rain somewhere in any 12 hour period is usually near 1?

It sounds like you are saying that "areal coverage" means precisely the value of A in my post (that was part of my uncertainty), just not as a conditional probability, and that the words you use are chosen based on the value of A. Is that right?

I didn't realize that there was a separate term "point forecast". Is this different than a zone forecast where the specific predefined zone is just very small?

4. Apr 18, 2012

### wsnell

"Thank you. May I presume you use areal coverage exclusively because you are in a tropical zone, and so your confidence of rain somewhere in any 12 hour period is usually near 1?"

Yes, but there are times when we are uncertain whether our zone will receive rainfall at all. (Timing of a front for example: if it arrives the whole zone gets wet. If it doesn't none of it gets wet.) If the front arrival is 50% certain our areal forecast morphs into a probability forecast (C dot A). In the long run, areal forecasts translate into a forecast of probability. That is, for a particular point in the zone, rain falls on average 20% of the time when all the 20% area forecasts' results are added together. This is assuming that one chooses a point in the zone where rainfall is average and the forecasters have no bias. (Additionally, no zones are completely homogenous, and, San Juan, for example, is the driest area of our Zone 1.)

Our zones are small enough so that certain areas (depending on wind flow) are likely to not receive any rain except in the wettest periods (confidence <<1 even for 20% coverage). This changes daily. Climatologically, over the course of several years the chance of rain at San Juan is about 42% in a 12 hour period. In our wettest months I have seen this increase to 70%. In some areas of Puerto Rico I have seen average zone coverage drop to 20% during the driest months (as assessed by radar). In March 2005 no measurable rain fell at the airport in San Juan. Our zones encompass areas that are rather large and, due to widely varied topography, somewhat diverse. This is part of the reason that the NWS migrated to gridded forecasts. The POP given inside our 1.25 km square grid is the chance that 0.01" of rain will fall at any point within that grid and not an aerial forecast. Most other offices have 2.5 km square grids.

For an example of a point forecast in text form please see:
http://www.srh.noaa.gov/productview.php?pil=SJUPFMSJU&version=0 [Broken]
For an example of an area forecast in text form please see:
http://www.srh.noaa.gov/productview.php?pil=SJUAFMSJU&version=0 [Broken]
For an example of a gridded forecast please see:
http://graphical.weather.gov/sectors/puertorico.php
For an example of our zone configurations please see:
http://www.nws.noaa.gov/mirs/public/prods/maps/map_images/state-maps/zone/pr_zone.pdf

"It sounds like you are saying that "areal coverage" means precisely the value of A in my post (that was part of my uncertainty), just not as a conditional probability, and that the words you use are chosen based on the value of A. Is that right?"

Yes, but often the conditional probability influences the POP. For example widespread rain is expected but we're only 60% sure. I am sure you understand that we are dealing out some inconsistencies here. We could use the probability terms I mentioned above, rather than the aerial terms, but we have to follow specific rules. Terms can't be mixed like "showers likely, with isolated thunderstorms". Our formatting tools favor using the aerial terms even when they aren't particularly applicable. Using, showers likely with a slight chance of thunderstorms, works much better, but change is slow and difficult and it is weak in conveying the whole picture (drive across the whole zone and showers may be a given). Forecasters are also reluctant to forecast 70% chance of rain and use 'numerous showers' when it will only rain for perhaps 10 min and sometimes hedge with 20% when real chances are less. We're working on that, but no obvious solution has been forthcoming.

"I didn't realize that there was a separate term "point forecast". Is this different than a zone forecast where the specific predefined zone is just very small?" Not really. Point forecasts are always verified at a particular rain gauge. It doesn't make much sense to consider areas smaller than the mouth of a rain gauge. Zone forecasts have to be verified by multiple gauges (in a dense network) or radar (which also averages to very small areas, but it is still an average.)

Stated more clearly: POP is C dot A but whether the forecast is 30% = 1 dot 0.30 or 30% = 0.5 dot 0.60 is simply not specified and usually not known to the current state of the art. And, to tell the truth, even if it were known and stated, it would create a lot of confusion.

Last edited by a moderator: May 5, 2017