They remove things not germane to the purpose of the data they publish. For example, USGS seismic data is noticeably bereft of most seismic events that are not geological in nature or sometimes related to mining (though in some sources they often scrub the mining ones too). Events of military interest like weapons tests, some target getting blown up in the middle of nowhere that may never make the news, etc is removed.

There are a ton of artifacts that show up in other sensing systems that are indicative of interesting or sensitive things that are outside the scope of their purpose, and these too may be edited from the data.

The people deciding what constitutes an event that should be scrubbed is pretty opaque AFAIK. It is official policy and sensing companies that do a lot of work with the government seem to follow similar guidelines.

Due to the proliferation of crowd sourced and alternative sensing platforms, I would argue that this is increasingly an exercise in futility. Nonetheless people still view many of the US sensing data sources as authoritative for all practical purposes. There are countries with laws dictating that some alternative data they control must be treated as authoritative for all purposes for their country, but that US data is sitting out there.

That's a lot of words but still no source.

Well, you could go to the website where they clearly state that some data is reviewed before publication and may be removed or modified. It is a frequently asked question. Or you could find an obvious counter-examples in the data, since it is public. The detection and flagging of anomalous events for review has been automated for decades, also publicly mentioned. I don’t assume everyone knows, I’ve been working with government sensing data for 20 years, but they are quite explicit if you look.

What they remove is a secret AFAICT but if you are an expert in the sensing modality it becomes obvious what should be in the data but isn’t. There are now businesses that specialize in differentially finding or reconstructing things that have been removed or modified in sensing feeds, so the effectiveness has diminished greatly.

> That's a lot of words but still no source.

Less flippantly, I literally went to the USGS FAQ to find this question https://www.usgs.gov/faqs/does-latest-earthquakes-map-show-n... where it says mining events are not reviewed, but if they are identified, they are still included! So ... I'll ask more explicitly:

Can you provide an explicit LINK to anything that supports your claim? Thanks.

Like others nearby, I’d like to see something more specific.

I’m willing to believe that in some silos (like relatively high-cadence seismographs) there might be some censoring. For example, it’s believable that siting of permanent stations is nudged away from some sensitive areas. Also, more believable in the past (say, 1980s) than the present.

Related, I’m sure that some sensors aren’t allowed to be flown over some areas (e.g., certain military bases) in the US.

However, you are claiming a broad based program of censoring US scientific data - gathered by the government or by government contractors. Like you, I’ve worked in this space for a long time. But I have not seen what you describe.

I wonder if we are working under different definitions of “censor” (see military base remark above)?

For people’s reference, the US-sponsored seismograph network is under EarthScope (https://www.earthscope.org/gsn/).

Your remarks caught my notice because I have personally worked with GNSS (lower cadence than seismograph) data, and personally know people who placed the sensors, wrote the data assimilation algorithms it uses, and set up the data pipeline. This data is not censored. (Although, famously, it was, before GPS was opened up.) I’m trying to find a way to rectify these two viewpoints.

That raised my eyebrow also.

In the past I've heard similar statements and had people point at 'cooked' raw data as evidence of editing.

'Cooked" generally means raw data with warts removed, the raw data is still available, the cooked data is what's on offer as the primary data of record - typically it may have had sensor errors and saturated bursts removed, undergone light smoothing filtering, and perhaps been geolocated to earth coords rather than retaining raw instrument attitudes, etc.

'Censoring' can mean 'no longer linked on public webpages for easy downloading' - generally the raw and cooked data is still on servers and accessable by direct FTP.

I'd be interested to know what specifically the GP actually meant by that throwaway assertion.