UFO Reports over Time
A time series examination of UFO reports.
One thing I have been neglecting to do in my posts on geographical UFO reports is to look at reports over time. This is called a time series, and there’s a whole statistical and machine learning subfield called time series analysis which covers this. I’m not that experienced in this subfield, however, I have a basic understanding of it. So let’s look at the UFO reports from NUFORC sorted by date. A seasonal decomposition using GPT and statsmodel can find the various trends and seasonal aspects of the time series.
The trend component notices the general trend the data is taking. The seasonal component is seasonal effects like weather, climate, holidays, and other attributes that might affect the sightings. For example, the 4th of July might increase sightings due to fireworks and cloudy weather in August might obscure sightings. The peaks in the trend can indicate when certain important events took place that caused a higher number of UFO reports.
- April 2014: 775.63
- March 2014: 773.58
- February 2014: 771.17
- January 2014: 764.92
- May 2014: 764.79
- December 2013: 753.58
- June 2014: 743.71
- November 2013: 743.08
- October 2013: 728.08
- July 2014: 726.50
The high point really is in 2014, especially spring of 2014. After that, there was a decline until around 2020 when another peak of reports occurred, coinciding with the start of COVID. What’s interesting is that 2014–2015 is the same timeframe multiple pilots reported large numbers of UFO sightings off the Atlantic coast.
Another interesting peak is 2004 when the Nimitz sighting occurred. It seems the peaks of the trend correlate with the public sightings by naval aviators. The seasonal component has a periodicity of around one year indicating that UFOs are seasonal within a year. This may be due to holidays as well since July has a higher number of sighting reports on average. However, the trend component is not affected by seasonal factors and so should not have any seasonal attributes like the effects of holidays. One way we can deseasonalize reports is by subtracting the mean for that month which removes any seasonal effects like holidays and climate patterns.
The anomalies are the deviation from the mean for that month and so indicate periods when there were more reports than usual for certain months and years. Certain time periods stick out, for example, April of 2020 and July of 2014. July 2014 is very close to when a pilot observed the “sphere encasing a cube” that zipped between two jets. This is probably not coincidental and something was causing more sightings which some pilots observed. These are the highest peaks for the anomalies.
- April 2020: 772.74
- July 2014: 669.47
- November 2015: 569.79
- July 2013: 531.47
- July 2012: 501.47
Notice they are all nearly around the timeframe of 2013 to 2015, except the peak in 2020. There was something very unusual about that timeframe. One way of determining how reports reoccur is to look at the periodicity. This means we find where there are trends that reoccur time after time. We can do this by decomposing the time series into a power spectra with a Fourier transform and then finding peaks less than 10 years.
- 3.02 years: Power = 1.16×10⁷
- 5.83 years: Power = 8.51×10⁶
- 5.91 years: Power = 6.21×10⁶
- 8.12 years: Power = 5.97×10⁶
- 2.11 years: Power = 5.78×10⁶
The most important period is 3 years followed by around 6 and then 8. The 8-year cycle can be observed from 2020 to 2012 and then to 2004. The 6-year cycle from 2008 to 2014 and then 2020. The 3 year cycle is more obscure but from 2014 to 2017 and then to 2020. If we predict into the future we can surmise that the 6 and 8-year cycles will get close around 2027.
6-year cycle: 2008–2014–2020–2026
8-year cycle: 2004–2012–2020–2028
The months can also give an indication of some trends. Even after deseasonalizing, certain months stand out as trend points in UFO reports.
May in particular stands out as does July and January. I wondered still whether the effects of holidays could be at play even after deseasonalizing and it’s possible. However, May has no significant holidays with fireworks and yet has a higher number of report anomalies. This can’t be due to the Chinese New Year, or 4th of July, or any other holiday. May for some reason has more sighting anomalies than the average would suggest.
One thing that is learned from the time series is that supposedly isolated events like Nimitz and sightings in 2014 are not isolated at all and are part of a larger trend in sightings at large. Reports are not isolated but part of a larger trend in sightings that occurs at certain intervals. Another thing that we learn is that reports occur in cycles of around 3, 6, and 8 years. This can be used to make predictions. For example, with the 8-year cycle, we can predict that there may be more UFO reports in 2028 which is around the same time as the supposed “event” will happen according to John Ramirez and it’s also when SETI “magically” says they will find aliens 😆. In the future using a model like Prophet to predict anomalies would be useful but for now, we know there are definitely patterns to UFO reports.
