Over the past several months, I have been working to interpret snow data from the past several winters.

The purpose of this research has been to identify coherent correlations relating to winter snow data.

Some of the correlations that I have looked at have included the yearly trends in season snow totals, winter rainfall vs winter snowfall, and seasonal temperatures vs season snowfall. All of these comparisons have shown interesting results, but none of them really helped explain any trends in snowfall data over the years.

This week I've been working with a new correlation: Total Season Snowfall vs. Season Start date.

The assessment was done under the general premise that earlier starting seasons are more likely to have higher snow totals than later starting seasons.

Kingston snowfall data for the past nine seasons (going back to 2004-2005) was used in evaluating this correlation. Incorporating only nine seasons, the conclusions reached are far from absolutes, but they represent the strongest correlations that I have been able to identify relating to winter trends thus far.

## Here's what I've found:

# I.

When total season snowfall is plotted against the date of the first sticking snowfall, we get the following chart:

Now, there are a couple of different ways that we can interpret this image, so we'll do them one at a time.

First, we can draw a "Best fit line" to give us an idea of the overall trend.

That looks somewhat like this:

Simply using a best fit line, it would appear that seasons that begin later, tend to have higher snowfall totals.

However, there are other more prominent trends in the chart that we should observe as well.

For starters, there is the main band of dots.

These are the five dots right around the 35 inch mark.

Together these dots form a pretty notable chart axis.

The second notable feature is the cluster of three dots higher up on the snow-scale in the month of December.

Both of these features are represented in the following chart:

The red oval encompasses the main axis dots, and the blue triangle encompasses the higher total cluster.

The black dashed line on this chart represents roughly the beginning of the cluster of higher season totals.

Interestingly, the beginning of this cluster correlates roughly with the beginning of December.

Conversely, the only season in the nine year time frame that was significant below the main axis (below average) occurred before the start of November. Unfortunately, having only one pre-November season be below average isn't enough to draw any conclusions. However, with all of the above average seasons occurring after November, and three of the five December-starting seasons being above average, we can start to make * general* conclusions about later starting seasons.

From these assessments combined, we can say that:

- The average seasonal snowfall is about 38 inches.
- Most significant (above average) winters start after the month of November.
- Of winters that start after December 1st, about 60% (3 out of 5) are significant.

# II.

I was also able to incorporated interpretations of season duration into this evaluation.

The following chart is the same as the very first one, except it shows new data.

The first new numbers (before the slash) are the number of days that the given winter season lasted, from the first measurable snowfall to the last. No restriction was placed on significantly deviant first or last snowfalls.

The second new numbers (after the slash) are a division of the seasonal total by the number of days in the season, giving an average daily snowfall during the season.

The numbers are presented in a xxx / # format with the first number being the length in days, and second number being the average daily snowfall in tenths of an inch.

What we can interpret from this chart is that the start date strongly affects the season length.

Almost without fail, winters in Kingston during the past nine years have ended within the month of March.

The average end date has been March 16th.

This means that the main influence on the length of a season tends to be the start date.

The chart reflects this with the longer seasons starting in October/early November, and the shorter seasons beginning in December.

Specific examples:

- The 2011-2012 season (earliest starting) was 154 days long

- The 2004-2005 season (latest starting) was 100 days long

Focusing on the second new number, the average daily snowfall total, the earlier starting seasons consistently have lesser values.

Specific examples:

- The 2011-2012 season has an average daily snowfall value of 0.1 inches

- The 2004-2005 season has an average daily snowfall value of 0.7 inches

Even when we look at seasons in that main average axis (the red oval), there is a gradient with later starting seasons having higher daily averages.

What this implies is that even if two seasons record an average amount of snowfall, the season with the earlier start date is more likely to have a lesser average daily snowfall, and *appear* to be less significant than a later starting season that has a higher average daily snowfall.

This can be observed in comparing the average daily snowfall value for Kingston to the number of impacts on the Kingston City School District. Impacts are quantified by the total number of school delays + cancellations.

Year | Avg. daily snowfall ^{(In.)} |
Number of delays+cancellations |
---|---|---|

10-11 | 0.49 | 15 |

07-08 | 0.72 | 13 |

09-10 | 0.43 | 11 |

08-09 | 0.42 | 11 |

06-07 | 0.26 | 8 |

05-06 | 0.39 | 6 |

12-13 | 0.29 | 4 |

11-12 | 0.14 | 4 |

*_{Accurate cancellation data does not exist for the 2004-2005 winter season.}

The comparison between a season's total snow amount vs. its actual impact can be related to the difference between actual temperature and windchill temperature. Just because the temperature actually is one measurement, doesn't mean that it has to *feel* that way. Similarly, just because a certain amount of snow is recorded, doesn't necessarily mean that the season's significance will correlate.

This interpretation of the data can also be used to explain why the recent 2012-2013 winter season subjectively felt less significant to many people, when compared to previous seasons such as 2009-2010, although the two seasons had virtually the same amount of snow overall.

From this section, the conclusion is that:

- Earlier starting seasons do not necessarily end earlier
- Earlier starting seasons are more likely to be greater in length
- Earlier starting seasons are more likely to feel less significant, due to lower average daily snowfall values
- Average Daily Snowfall is more reflective of a season's significance than total snowfall

_{Note: Additional investigation needs to be done to evaluate at what point the enhanced affects of a short season are over shadowed by the short length of the season. Example: at 89 days, the 2009-2010 season was long enough for the 0.43 inch average daily snow value to make the season relatively significant. If, however, a hypothetical season only recorded 5 inches of snow in a 5 day time frame during the whole winter, then the short duration of the season (5 days) would make the average daily snowfall of 1.0 inches largely irrelevant. There must be some form of cut off at some point.}

# III.

This final section offers another look at the relationship between start-date and impacts.

Again, we have the plot of seasonal totals plotted against the season start date.

On this chart, however, the red numbers indicate the number of snowfalls that were at least one-inch during the given winter season.

Essentially, what this chart strongly indicates is that the later a season starts, the more one-inch snowfalls there are likely to be.

**Why is the one-inch snowfall threshold important?**

First, most of a season's total snowfall is due to 1+" snowfalls.

This is evident in that seasons with higher totals, * generally* tend to have higher incidences of one-inch-plus snow events.

More importantly, however, the 1+" snow event incidence is important, because this is the threshold when we start to see storm impacts on travel. Specific examples of this can be found in the very strong correlation between 1+" snowfall frequency and the combined number of Kingston City School District delays+cancellations:

Year | Number of 1+" snowfalls | Number of delays+cancellations |
---|---|---|

10-11 | 15 | 15 |

09-10 | 14 | 11 |

07-08 | 12 | 13 |

08-09 | 10 | 11 |

06-07 | 8 | 8 |

12-13 | 8 | 4 |

05-06 | 6 | 6 |

11-12 | 6 | 4 |

*_{Accurate cancellation data does not exist for the 2004-2005 winter season.}

Having already established the strong correlation between later starting seasons and increased 1+" snowfalls, we can now extend that correlation to say that later starting seasons have more travel impacts in the form of school cancellations.

Again, this is only nine years of data, and there seems to be one or two seasons that are exceptions to each these general trends (giving about a 77% to 88% correlation rate), BUT

# Overall,

the following conclusions can be drawn from this entire analysis:

- The average total per season is about 38 inches.
- Most significant (above average) winters start after November.
- Of winters that start after December 1st, about 60% (3 out of 5) are significant.
- Earlier starting seasons do not necessarily end earlier
- Earlier starting seasons are more likely to be longer
- Earlier starting seasons are more likely to feel less significant, due to lower average daily snowfall values
- Average Daily Snowfall is more reflective of a season's significance than total snowfall
- Later starting seasons have higher incidences of one-inch-plus snow events, which provide more opportunities for travel impacts

The conclusions can be used to explain the following:

- The lack of snow in 2011-2012 in context of the season starting notably early
- The apparent insignificant of the 2012-2013 season, despite average an average snow total
- The record number of snow cancellations (9) in 2010-2011
- The record number of two-hour delays (7) in 2007-2008
- The lack of snow cancellations in 2012-2013

So, to answer our heading question, "Are earlier-starting winters more significant?", if there is any conclusion to be drawn from this analysis, the answer must be "no". If anything, later starting seasons, seasons starting after the end of November, are more likely to experience higher snow totals, and to appear **more** significant.

## Practical Application

The current weather forecast (11/23) calls for a 30% chance of snow showers tonight.

Some very light accumulations are possible.

If there are no accumulations tonight, then it looks like we should make it to at least Friday (11/30) with no snow.

*Strictly applying the findings* to the 2013-2014 winter season, we get the following:

If any snow sticks tonight, we might expect a winter along the lines of 2006-2007 and 2008-2009:

- Roughly 40 inches of snow (about average to slightly above)
- Roughly a 60% chance of an average winter, 20% of an above average winter, and 20% of a below average winter
- Roughly 113 days of winter
- A daily average snowfall of about 0.4, which is characteristic of a standard winter and more significant than the last two winters
- About 8 to 10 1+" snowfall events

If it snow does not stick until the first week of December, we might expect a winter closer to a blend of 2007-2008, 2008-2009, 2009-2010:

- Roughly 48 inches of snow (above average)
- Roughly a 50% chance of an above average winter, 40% of an average winter, and 10% of a below average winter
- Roughly 100 days of winter
- A daily average snowfall of about 0.5, which is above the average
- About 10 to 14 1+" snowfall events

**Edit: November 25th**

Kingston picked up about a tenth of an inch of snow Saturday night, which was enough to coat the ground and cars.

That combined with the frigid temperatures that have moved in, and the fact that some patches of snow are still lingering around, will qualify November 23rd as the start date of the 2013-2014 winter season.

Below is what I am calling my official season forecast, based off of the study:

- 38 to 45 inches of snow (about average to slightly above)
- 60% chance of an average winter (33 to 43 inches),
- 30% of an above average winter (more than 43 inches),
- 10% of a below average winter (less than 33 inches)
- Roughly 110 days of winter
- A daily average snowfall of around 0.4, which is characteristic of a standard winter

<0.4 correlates with insignificant winters, >0.4 correlates with significant winters.

- About 8 to 10 snowfall events of at least one inch

This is my first time attempting a seasonal forecast, and obviously the first time using this start-date method.

We'll see how it pans out in March!!