This website describes the new version 3.0 of the Spatial Synoptic Classification, which was developed in late 2019. The concept of the SSC is the same as previous versions, but with a few different methods that allow a much more flexible and automated classification process. (It should be noted that the SSC that has appeared in publications over the past 20 years is actually the SSC2, although it often gets referred to just as the SSC since the original version’s data was never widely available. The first projects with the SSC v3.0 are underway, but there are no published manuscripts using it at this point. I am eager to explore uses of the data so please be in touch if you have any ideas.
First, if you're unfamiliar with the SSC, please check out here first, as this contains a full summary of what the SSC is and its uses.
The SSC has been popular with a number of applications, but there have been some limitations to it, which were largely a function of the fact that the previous version was developed as part of my dissertation and never originally conceptualized to be expanded, updated, and used as widely as it has been. Namely:
-Seed-day selection is a time-consuming process, as seed days were identified at each station based on the nearest station with a similar climate. This process is particularly difficult when trying to deal with topography or microclimates. As the seed-day selection process required a long time series (to make sure there were enough seed days for each SSC type for each year), stations with shorter data sets, and stations with climates dissimilar to locations around them, were particularly difficult to work with.
-Because of the above, occasionally stations identify a seed day that was out of typical character, or have multiple seed days just along one side of the seed-day selection criteria (e.g., within a 5-degree range, it would identify only days in the coldest 2 degrees), making it inappropriately different from its neighbors. This is particularly true for rarer types.
-While the SSC has been reliable at the station level, interregional comparisons could be difficult. Namely, the original SSC was developed just for the US and Canada. When it was developed for Europe, the co-developer wanted to have a greater representation of all SSC types, and thus polar and tropical conditions were more moderate in character there (holding all other things consistent). This did not negatively impact any of the studies that took place, but it made broader comparisons across regions more difficult.
-The program to identify the seed days was never automated, and thus could not be made public. It therefore could not guarantee reproducibility and tended to make the SSC more of a black box than would be ideal.
-There was no way to take advantage of modern gridded data sets effectively.
-Given climate change, there was no way to allow SSC types' character to change over time.
SSC v3.0 stations
While the specific seed-day criteria changes a lot from station to station, the difference between the seed day criteria and the station climatology is relatively homogeneous. That is, Dry Polar typically has a mean temperature around 1 standard deviation below the mean for much of the year. Regressions were run on all of the original stations categorized, and the seed day criteria for each station could be well predicted by just five variables: a station's latitude, month of the year, mean and standard deviation of the variable for the month, and annual temperature range. R-squared values with the regression approached 0.9 for the temperature variables for most SSC types.
The development of these regression models effectively means the SSC can then be automated. All that is needed is the station's latitude and weather data, and then the SSC self-calculates based on the guiding regression equations.
The changes do not have substantial changes on the classification across North America, where the regression was based. There are somewhat greater differences between the SSC2 and SSC3.0 across Europe, where the calendars are now more in line with the original concept in North America.
Given the increasing focus on climate change studies, there are now more statistics showing the trends in SSC weather types over time already calculated. You can access these on the left.
All versions of the SSC have always categorized daily weather conditions into one category. The SSC2, and SSC3.0 both also produce a percentage likelihood that any given day falls into each of the types. To explore the idea of weather ‘regimes’, these percentages are aggregated to the weekly level, centered on the day shown. Thus it represents the broader conditions over a period of time and may be of use to certain applications. The weather types all represent the same conditions as with the daily classifications, although a regime must last for more than one day. One day regimes identified are thus reclassified as transitions, unless the days before and after are identical, in which case it is reclassified to match. Only the six weather types and transitions are used, no subsets are shown.
Both the SSC2 and this new version will continue to be updated in parallel for the foreseeable future, and there is nothing 'wrong' with the old calendars, and so either could be used.
All versions of the SSC have always categorized daily weather conditions into one category. The SSC2, and SSC3.0 both also produce a percentage likelihood that any given day falls into each of the types. To explore the idea of weather ‘regimes’, these percentages are aggregated to the weekly level, centered on the day shown. Thus it represents the broader conditions over a period of time and may be of use to certain applications. The weather types all represent the same conditions as with the daily classifications, although a regime must last for more than one day. One day regimes identified are thus reclassified as transitions, unless the days before and after are identical, in which case it is reclassified to match. Only the six weather types and transitions are used, no subsets are shown.
The SSC's seed days in this version are based on climate normals from 1981-2010, but this can change. Thus, future projects can examine changing character over time.
This automated process lends itself nicely to being used with gridded data - either reanalysis or GCM - which is one of the ultimate goals.
If you have any questions, comments, or notice any errors, please e-mail me. Thanks!