U-Sphere

Spatial Point Pattern Analysis, Identity-Based Voting and UAPs: The Unknown Connection

Fig. 1 : Identity Vote intensity (2012), Fig. 2 : UAP residual intensity (2015)

Can spatial statistics reveal a structure behind UAPs?

In 2012, while analyzing GEIPAN data (The GEIPAN is an Unit of the French National Space Center in charge of investigating UAP for the governement), I found a connection so sensitive I waited 10 years to publish it.

If you’ve followed my work—from my earliest publications to my most recent—you know that a recurring theme in my research is the relationship between UAPs (Unidentified Anomalous Phenomena) and Systemic Environmental Risks through pattern analysis.

This is why, as far back as 2005, I began building the U-Sphere Project, a website designed to aggregate data and support cross-analysis across environmental layers. I developed my own custom GIS and started testing hypotheses.

From a computer scientist’s point of view, this approach is one of the key entry point to understanding the UAPs. The reason is very simple: from a purely logical standpoint, unexplained phenomena can always be grouped into one of the three categories:

In two out of three cases, the phenomenon—whatever his origin is—has something to do with the environment.

So yes : spatial data.

It’s the reason why in 2009, as a consultant and after 3 years working on the GEIPAN database, I proposed carrying out a study using this unique dataset—whose data were homogeneously distributed across space and collected since 1977 (the "space" is French territory in this case)—and testing whether environmental correlations existed for what we call "PAN D" in French. "PAN D" are UAPs which are categorized as "unexplained" after an investigation.

Once the project was accepted by the Space Center, I reached out to researchers at GREMAQ (a research group in mathematical and quantitative economics) in 2010. My goal was to leverage state-of-the-art mathematical tools for our study, titled “Spatial Point Pattern Analysis of UAP Phenomena in France”. Amidst administrative delays and university constraints, the study was largely completed by 2012, though it was only shared online in 2015.

During this time, we tested many different covariates to check if the UAPs in the dabatase may have a connection (direct or indirect) with one of the following categories: polluted sites, nuclear sites, wetlands, sunlight, airports, forests, and all explained phenomena (UAPs A) from the database itself.

The results we obtained were highly encouraging and we were struck by the significance of these new findings:

Results of the modeling with all covariates (2015)

To my knowledge, this was the first spatial p-value evidence establishing a statistical link between UAPs D (aka "Unexplained cases after investigation") and nuclear-related activity or polluted sites. It prefigures, by nearly a decade, the 2023 study by R. M. Medina et al. published in Scientific Reports (Nature Portfolio), though with fundamental differences: Medina focuses on the witness's 'opportunity to see' (sky visibility, light pollution), my approach uses identified phenomena (UAPs A) as a control group to isolate a signal that is independent of reporting bias and directly correlated with strategic environmental attractors.

Using our model, we managed to reconstruct the UAPs distribution over the French Territory. But there was an intriguing glitch deeply hidden behind the veil of the data : the model left unexplained clusters—called residuals—which were still unexplained by the three variables above (Fig. 2). These areas were not scattered randomly. They were mostly distributed along the Mediterranean coast and in the north of France.

The curious and concentrated shape of these residuals left me with one question: If there’s still a structure visible in the residuals, what variable did I miss?

And I started thinking about that question during several weeks...

Then one day, while driving home from my daughter's school, the answer struck me: what if the same logic underlying these wave patterns was that of a supervised 'spaced retrieval' learning process? A process I studied previously where the phenomenon appears to conserve energy to maintain a low signal-to-noise ratio.

If there were a similar optimizing effort over time—trying to inject information with minimal energy—how would I organize targets in space?

To help, you can formalize this question differently:

If I wanted to raise the average level of a population on a topic (say: UAPs), who should I prioritize? The most informed—or the least exposed?

The answer, from learning theory, is counterintuitive but simple: the least exposed. People with low prior knowledge often show the fastest early learning gains; later gains get harder and more “expensive.”

So the next question naturally followed, mechanically, in terms of how knowledge about UAPs might spread: 

Geographically, where are the populations that are least exposed to UAP ideas—or most culturally resistant to them? 
Is there any territorial variable that—however imperfect—could serve as a proxy for these dynamics 
and match the residual map?

I ended up thinking in two broad sociological directions (always speaking in aggregates, not individuals):

With this strong hypothesis in mind, I proposed to my team to include data from the Front National (an identity-based, far-right movement) and the vote shares for Marine Le Pen in the second round of the 2012 election, to test whether this could be a good proxy. This was a robust way to align our findings with the sociological dimensions found in electoral geography literature. Let me be very explicit: this is not a moral judgment, nor should it be interpreted at an individual level (the ecological fallacy is real). It’s a statistical proxy at the territory level, nothing more.

We had many discussions, and finally we chose not… to include the data. The UAP topic was already struggling for scientific legitimacy; adding a political variable would likely have killed the paper on arrival—and possibly even caused issues for the GEIPAN itself.

Of course, one could argue that it might be a psychosocial effect. For that precise reason, one of our key methodological steps was to include explained cases (UAPs A) as a covariate in the model: if a region simply produces more reports overall, competent investigations should yield more A and more D together. And that’s not what we saw: the spatial distribution of A cases was a poor predictor of D cases.

That implies something quite important as well: D observations aren’t "decided" by witnesses. Whatever produces D cases seems to have a structure that isn’t reducible to reporting behavior alone.

Since, as the global context has steadily shifted toward institutional recognition—marked by the pioneering release of the GEIPAN archives in France (2007) and followed by significant transparency efforts in the UK, Chile, Brazil, and Uruguay—the environment is now far more open. And more recently, with the momentum in the US, starting with the UAPTF in 2020 and leading to the permanent AARO office, the focus has shifted from "Whether UAPs are worth the investment" to "How to gain insight into the UAP Phenomenon?". I think it is finally time to contribute my own provocative findings to this global discussion.

So what happens when you add the covariate?

Well the data shown exactly what I expected. If we consider only the significant covariates, the identitary vote absorb part of the pollution variable and surpass the nuclear significance level:

Updated modeling including Identitary Vote (Results of Second Turn of French Presidential Election, See Fig 1 above)

Through these new results, two things stood out to me:

In other words, the findings are consistent with the idea that both ecological and societal dimensions can coexist within the spatial structure of UAP sightings—without necessarily interfering with one another.

The evidence is clear: a non-random, organized structure exists within the unexplained. Now, it is up to us to decide how we interpret this signal. To support transparency and further research, I am sharing the software on GitHub, inviting anyone to replicate these tests or apply the methodology to other datasets.

For those who wish to explore the broader context of my work—including the temporal analysis and the 'spaced retrieval' learning process—you can find a full article which will complete this one, exploring the Where as much as the Where and When on u-sphere.

If you’d like to leave a comment, please join the discussion on LinkedIn