Every week the Resources Regulator emails out a bulletin. Its own description of it is flat: The Resources Regulator publishes a reportable incidents summary each week that provides information on reportable incidents and safety advice for the NSW mining industry. Each edition opens with a line of the form "40 reportable incidents, 3 summarised below", then describes those three, with a short set of recommendations to industry for each.

We have covered three of those weeks one at a time. This is the first time anyone, including us, has put the year together, and doing it changes what the weekly number is worth.

The year, week by week

Reportable incidents in NSW mining, as counted by the Resources Regulator, 2026 to 7 August
Week endingReportable incidents
16 January 202633
23 January 202628
30 January 202626
6 February 202644
13 February 202639
20 February 202634
27 February 202658
6 March 202635
13 March 202637
20 March 202643
27 March 202654
3 April 202642
10 April 202629
17 April 202654
24 April 202643
1 May 202651
8 May 202649
15 May 202635
22 May 202648
29 May 202640
5 June 202646
12 June 202635
19 June 202620
26 June 202642
3 July 202631
10 July 202637
17 July 202638
31 July 202639
7 August 202640

Every figure is the count stated in that week’s own edition. 29 editions, 1,150 incidents. There is no edition for the week ending 24 July on the regulator’s index page, which is the only 14-day gap in the year.

What the series says

The mean is 39.7 incidents a week and the median is 39, so the average week is genuinely average. The range is the interesting part: from a low of 20 in the week ending 19 June to a high of 58 in the week ending 27 February. That is nearly a threefold spread between the quietest and busiest weeks of the same year, in the same industry, under the same rules.

Which means the weekly number, on its own, tells you almost nothing. A week of 31 and a week of 48 are both ordinary. We should say plainly that this is a correction to our own coverage. We published three single-week pieces in July, each leading with that week’s count. The counts were right and the incidents were worth describing, but presenting a single week’s figure as news implies a signal that 29 weeks of data does not support. The right unit for the number is the year; the right unit for the incidents is the incident.

And the year is flat. Split it in half and the first fifteen weeks average 39.9 against 39.4 for the last fourteen. Nothing is getting better and nothing is getting worse, at least not at the resolution a count of notifications can show.

The 86

Twenty-eight of the 29 editions describe exactly three incidents; one describes two. So the number of NSW mining incidents that get a public account is fixed by the format at around three a week, whether the week reported 20 or 58. Across the year that is 86 of 1,150, or one in thirteen.

What does get described skews hard. Of the 86, our reading counts 79 classified as a dangerous incident and six as a high potential incident. By site type the most common settings are underground metals mines (22), open cut coal (21) and underground coal (13), with the remainder spread across construction materials, surface coal, quarries and extractives.

None of that is concealment and we are not suggesting it is. The bulletin is explicitly a safety-advice product: it picks incidents that carry a transferable lesson and publishes recommendations to industry with each one. A regulator is not obliged to narrate every notification, and 40 short accounts a week would be read by nobody. The point is narrower and worth holding: the weekly summary is a teaching tool that happens to carry a count, and the count is the only public trace most incidents leave.

We went looking for a fuller public source and did not find one on the regulator’s site: the safety section links to incident reporting obligations rather than to published incident data, and two plausible data paths returned 404. We are not asserting that no fuller dataset exists anywhere, only that we could not find one from the regulator’s own navigation, which is where a member of the public would look.

What we would do with this next

Three things follow, and we are recording them so they can be checked against later. Whether the 2027 series stays near 39.7, which is the only way to know if anything actually changes. Whether the missing week ending 24 July was published elsewhere or skipped. And whether the site-type mix in the described incidents matches the mix in the count, which is not answerable from the bulletin and would need the regulator to say.

Sources

  1. NSW Resources Regulator, Weekly incident summary (index page, read 17 August 2026): the regulator’s own description of the bulletin quoted above, the list of every edition, and the absence of an edition for the week ending 24 July 2026.
  2. NSW Resources Regulator, the 29 individual weekly incident summaries published in 2026 to the week ending 7 August, each fetched from the link given on that index page and read 17 August 2026: the stated count of reportable incidents and the stated number summarised in each edition, and the incident classification and site type of each summarised incident. The most recent, week ending 7 August 2026, opens “40 reportable incidents, 3 summarised below”.
  3. NSW Resources Regulator, Incident reporting (read 17 August 2026): the obligations that generate a notification, checked while looking for a fuller published dataset.
  4. The Duty Holder, our three single-week pieces on this bulletin: week ending 3 July, 10 July and 17 July, whose published counts of 31, 37 and 38 match this dataset exactly and were used as an independent check on the extraction.

Methodology

The regulator’s index page lists every edition of the weekly incident summary. We took the 29 links dated in 2026 up to the week ending 7 August, fetched each one, and read the count stated in its own opening line. Nothing here is modelled or estimated: each weekly figure is the regulator’s, and the total is their sum. The extraction was checked against our own three published weekly stories, whose counts of 31, 37 and 38 for the weeks ending 3, 10 and 17 July match the parsed values, which is an independent test of the parser rather than of the source. The classification and site-type tallies for the 86 described incidents were parsed from the same pages and are the weaker numbers here, because they depend on the bulletin’s labelling being consistent across the year; treat them as indicative. The year is incomplete by design: it runs to the most recent edition published at the time of writing, so it is 29 weeks and not 52, and the totals should not be read as an annual figure. We have not contacted the Resources Regulator.

See an error? Request a correction and we will check it against the sources above and log the outcome.