Friday March 8, 2013:

The following is an evaluation of Groundhog's day forecast.

First, I define the normal dates of an "early spring," and "six more weeks of winter."

Then, I determine what is needed for it to be spring, or winter.

To determine this consider two categories: the date of the last snowfall, and the average temperature during the first half of March.

# When is Spring and when is Winter

First, since the terms are somewhat subjective, we have to set the parameters for "spring" and "winter".

As far as timing, the saying goes "six more weeks of winter". So, a non-early spring is February 2ed (Groundhog's Day) + 6 weeks = March 16th. Coincidentally, since 2005, this has also been the average date of the last snowfall here in Kingston. Since an early spring is "just around the corner," we can safely assume that this indicates about a month after Groundhog's Day, or two-thirds of the "six more weeks" of winter. * So, timing-wise, we'll say that March 16th is the normal end of winter, and that an early spring is roughly 2 weeks prior… March 2ed.* So we have our timing parameter.

# Defining Spring and Winter

Now what about determining what conditions qualify as winter and spring? There are two ways to do this: using snowfall, and using temperature.

Let's use both.

**Snowfall:** Let's assume that during the winter we regular receive series of snowfall events, and that as we move into spring, they become further and further apart. Looking at the actual pattern over the years for Kingston reveals that, during the later portion of the season, any snowfall occurring more than two weeks later than the previous snowfall can be considered an outlier. **So, once we get to the end of winter, and it stops snowing for more than two weeks, we can consider the last snow fal lto be the end of winter.**

**Temperature:** The average winter (Dec-Feb) temperature is 38 degrees. The average spring (Mar-May) temperature is 60. If we average the two, we can assume a winter/spring threshold temperature of about 49. So, once the temperature begins to get above 49 degrees, we can assume it is becoming spring. But first we have to determine the length of time that we have to be at 49 degrees. To rule out random "spike" days, we can consider the average March 1st - 16th temperature, and assume that **years in which the average temperature exceeds 49 degrees in the first half of March are characteristic of spring.**

Yay! We can now assign spring or winter designations to both the snow, and temperature categories. So obviously, if we scored a spring in the snow category, and a spring in the temperature category, then it was spring. If we got winter in both categories, then it was winter.

But what if we get spring in one category and winter in the other?

Let's say it's a dry, but cold winter, and it doesn't snow, so our snow category is indicative of spring via the parameters we established. But the temperatures fall short of the 49 degree parameter we set, indicating winter. In this case, we have to evaluate our temperatures more closely to determine how indicative of winter they are.

The average February temperature is 39 degrees. The average March temperature is 48 degrees. If we blend these two, giving more weight (75%) to the February side, since the category had a winter bias to begin with, we can get a reasonable spring/winter criteria to apply to this situation. That is, (39*.25)+(48*.75) = 42 degrees. **So, when snowfall indicates spring, but temperatures indicate winter, we can assume that if the March 1 to 16th average temp was above 42 degrees, both categories fell into the spring category.**

In the opposite case where temperatures rise to warm levels during the day, but at night we get a lot of light nuisance snowfalls that just stick to the grass, the categories would reflect snowfall= winter, temperatures= spring. Like before, we look at the magnitude of the temperatures. This time, because the temperatures had a spring bias, we use 25% of the April average of 60, and 75% of the March average of 48 to get a spring/winter threshold of 51 degrees. (We use an April:March of 1:3 because we are focusing on the beginning of March.) **So, when snowfall indicates winter, but temperatures indicate spring, we can assume that if the March 1 to 16th average temp was above 51 degrees, the spring temperature category supersedes the winter category, and the year was an early spring.**

# Summary:

Temp. (T) ……. Snow (Sn) ……… Outcome

> 49 …………… end by ~3/2 …… Early Spring

< 49 …………… end after ~3/2 … Normal Winter

> 49 ………….. end after ~3/2 … T>51, then Spring

< 49 ………….. end by ~3/2 …… T>42, then Spring

Question for self: Does the average temperature refer to just highs or overall temperature??

Would it be more accurate to include overall temps?

And if so, does the climatological data include overall temps or do you need to go back and modify it??

**This only incorporates HIGH temperatures— more accurate assessment would be of overall average

Average temperature^{(March 1 - 16)} |
Date of last Snow^{(Omitting additional snows more 2+ weeks later)} |
Outcome |
---|---|---|

> 49 | end by ~March 2 | Early Spring |

< 49 | end after ~March 2 | Normal Winter |

> 49 | end after ~March 2 | If temp. above 51, then Early Spring If temp. below 51, then Normal Winter |

< 49 | end by ~March 2 | If temp. above 42, then Early Spring If temp. below 42, then Normal Winter |

^{*Note: ~March 2 indicates "around March 2," or generally within about two days.}

Or

Temp. Below 49 | Temp Above 49 | |
---|---|---|

Last snowfall By ~3/2 | If temp. above 42, then Early Spring If temp. below 42, then Normal Winter |
Early Spring |

Last snowfall After ~3/2 | Normal Winter | If temp. above 51, then Early Spring If temp. below 51, then Normal Winter |

^{*Note: ~3/2 indicates "around March 2," or generally within about two days.}

# Evaluating

Applying this to the last eight years, we get:

2005… T= 39.5, Sn= March 23 … Winter

2006… T= 47.1, Sn= March 02 … Spring

2007… T= 42.5, Sn= March 16 … Winter

2008… T= 48.3, Sn= March 01 … Spring

2009… T= 45.1, Sn= March 02 … Spring

2010… T= 49.7, Sn= March 03 … Spring

2011… T= 46.8, Sn= March 23 …. Winter

2012… T= 55.5, Sn= March 02 …. Spring

2013… T= 44.9, Sn= March 19 …. Winter

2014… T= 38.8, Sn= February 19 .. Winter

2015… T= 41.9, Sn= March 03 …. Winter

Thus, in evaluating some of the Nation's groundhogs, we get the following accuracy ratings for the 2005-2013 time period:

—Punxsutawney Phil…

Correct prediction(s) made in: 2005, 2014. 20% accuracy.

—Buckeye Chuck…

Correct prediction(s) made in: 2008, 2010, 2012. 33% accuracy.

(Chuck started in 2006, there was no 2005 forecast)

—Staten Island Chuck…

Correct prediction(s) made in: 2008, 2009, 2010, 2012, 2014. 71% accuracy.

(Chuck started in 2008, there were no 2005-2007 forecasts)