Corrections

Twelve named California incidents carried the wrong date

Rows affected. the twelve rows whose evidence is 'sha256 59be0de2 comments' (incidents named in CARB's comment prose, carrying no quantity of their own)

What was published

became_public for each of the twelve incidents, in the file's own month/day/year form, equal to the first date in the first comment cell naming the incident.

Amended: the remedy below replaces one that was withdrawn

Amended 9 September 2026. The remedy this entry published on 5 September was wrong in kind and is withdrawn. It prescribed the date of a CARB action, the approval of a verification or the retirement of buffer credits, as the date the incident became public. Those are different dates. A regulator's approval is an action, and the date of an action is not the date it became publicly readable. Three parts of the entry survive unchanged: that the twelve published dates are wrong, the twelve dates CARB states for its own action, and the three class labels corrected from CARB's own word, which do not depend on the date question. What was never established, and is not established now, is when any of the twelve became public. A correction to this entry's own first amendment, made the same day: it said the workbook does not state when it was published. It does, on its Introduction sheet. What it does not date is the sentence inside it. The register's specification now separates four date concepts, and the twelve values are carried under it as the date of the CARB action.

What is right

For each of the twelve incidents, the date CARB states for its own action on that incident: the date CARB approved the verification of the reversal where the project's own row states it, otherwise the date CARB retired buffer credits or retired surrendered instruments for it. Where both are stated they fall on the same day in nine of ten cases. That is an action date and the register carries it as one. The date each incident became public is not established, and the reason is precise. The workbook carries an update label: its Introduction sheet reads "Updated August 26, 2026", and it states a cadence, "typically updated on the second and fourth business Wednesdays each month". An update label is what the publisher says about the file. It is not a release timestamp, it does not date any sentence inside the file, and it establishes nothing about when a particular incident first became readable. What we can state is an observation of our own: this exact byte content, sha256 59be0de2 and 560,860 bytes, was retrieved from CARB's public address https://ww2.arb.ca.gov/sites/default/files/2022-07/nc-arboc_issuance.xlsx at 2026-08-31T17:12:39Z, recorded in snapshot/evidence/2026-08-31/data/manifest.psv against that same hash. So each of the twelve carries one bound and no publication date: the statement was publicly readable at that address by 31 August 2026, because we read it. Whether it was readable earlier, and when each incident first became public, is unknown. Only one version of this workbook has been retrieved, so nothing here establishes what any earlier version contained, and reading the 2026 cadence statement back onto a 2017 incident would be inference rather than evidence. Three class labels are corrected in the same rows from the word CARB uses in the approval comment.

How it was found

The layer-file audit of 5 September 2026 (kyroq-engine scripts/resolve/audit_layer_files.py) flagged the twelve dates for being in month/day/year form. Checking the values against the sentence-level incident series the same day showed that every one of the twelve was also the wrong date: the builder had taken the first date in the first comment cell that names the incident, and a CARB comment cell often narrates several incidents in sequence.

Basis

CARB workbook, column AI, cells named per row; dates read as month/day/year. Each row's basis is the cell's own words. CARB writes the typographic apostrophe in "project's"; it is transcribed here as a plain apostrophe and nothing else is altered. Where a row's basis spans more than one cell, each cell is named and the elision is marked.

The rows

Two California reversals are stated in CARB's comments and not in its columns

Rows affected. CAFR5204 (one row, reversal_intentional 6,471) and CAFR5170 (two rows, reversal_intentional 681,970 and reversal_unintentional 119,018)

What was published

Quantities read from CARB's columns V and W. For CAFR5204 the columns carry only the 6,471 tCO2e intentional portion; for CAFR5170 they carry 681,970 (V) and 119,018 (W), 800,988 in all.

What is right

CAFR5204: CARB's comment on the project's own row (AI782) states a reversal of 82,728 tCO2e in the seventh reporting period, 76,257 unintentional and 6,471 intentional, and states that CARB retired 76,257 credits from the buffer pool; cell AI753 records that retirement on 6/11/2024. Column W on row 782 reads n/a. The register's quantity rows therefore omit 76,257 credits of unintentional reversal that CARB states twice. CAFR5170: CARB's comment (AI650) states a reversal of 820,725 tCO2e (698,765 intentional, 121,960 unintentional) approved 8/13/2024, of which 681,970 instruments were to be surrendered under section 95983(c)(3), 119,018 credits were retired from the buffer pool, and 16,795 intentional plus 2,942 unintentional (19,737) are to be compensated through the Early Action Offset Program. The columns record the two compliance-programme portions and not the 19,737 remainder; 681,970 plus 119,018 plus 19,737 equals 820,725. The register's rows are right on their stated basis (the columns) and understate the tonnes reversed by 19,737.

How it was found

Reading the comment cells behind the audit's second warning (retirements in the incident series exceeding the stated magnitude) on 5 September 2026.

Corrected rows

• project CAFR5204
• add class reversal_unintentional; quantity 76,257; became_public 2024-06-11; basis AI782 and AI753; column W n/a

• project CAFR5170
• restate reversal_total_tco2e 820,725; intentional 698,765; unintentional 121,960; compensated_by_surrender 681,970; compensated_from_buffer 119,018; compensated_through_early_action_programme 19,737; became_public 2024-08-13; basis AI650; columns V and W of row 650

What changes in the totals

Edition zero's California quantity rows undercount reversed tonnes by 76,257 (CAFR5204) on a column basis and by a further 19,737 (CAFR5170) against CARB's own statement of the reversal: 95,994 credits in all against the 12,174,135 credits of reversal the 29 quantity rows carry (20 intentional, 3,634,831; 9 unintentional, 8,539,304), 0.79 per cent.

Four terminations: the instruments surrendered are 1.20 times the tonnes reversed, and one joint obligation was attributed twice

Rows affected. the surrender_operator rows attributed to CAFR5038 (12 cells), CAFR5058 (the same 12 cells), CAFR5423 (1 cell) and CAFR6313 (2 cells)

What was published

In edition-0-events.csv, the tonnes reversed on termination (column V): CAFR5038 662,284; CAFR5058 302,354; CAFR5423 28,048; CAFR6313 440,956. In incident-series.csv, the compliance instruments surrendered by the operators and retired: 1,157,566 attributed to CAFR5038 and the same 1,157,566 attributed again to CAFR5058; 33,658 to CAFR5423; 528,967 to CAFR6313. Both files are right about what they measure and the series counted one obligation twice.

What is right

On termination of a compliance offset project, section 95983(c)(4) requires the forest owner to submit compliance instruments equal to the ARB offset credits issued multiplied by the compensation rate the Forest Protocol sets by years since the project's start. CARB states the obligation in the termination row of each project: CAFR5038 794,741 (row 178), CAFR5058 362,825 (row 264), CAFR5423 33,658 (row 1310), CAFR6313 529,147 (row 1636). Each is 1.2000 times the tonnes reversed (794,741 / 662,284; 362,825 / 302,354; 33,658 / 28,048; 529,147 / 440,956), which the Forest Protocol's compensation rate table sets for a termination in the sixth to tenth year after commencement; all four projects were first credited for vintages 2015 or 2018 and terminated in 2024 or 2025. The register relies on CARB's stated obligations, not on its own reading of the table. The instruments surrendered are therefore not the loss; they are the loss plus a 20 per cent compensation charge, and the register carries them under their own name. CAFR5038 and CAFR5058 have one joint obligation: the twelve cells (AI604, AI605, AI814, AI860, AI861, AI871, AI883, AI1282, AI1484, AI1548, AI1549, AI1619) each say the credits were surrendered "to meet the obligation for the CAFR5038 and CAFR5058 reversal/terminations", and their sum, 1,157,566, equals 794,741 plus 362,825 to the credit. The series attributed all twelve to each incident; summed by incident it doubles the obligation.

How it was found

The audit's second warning: for these incidents the incident series' retirements exceed the reversal magnitude in edition-0-events.csv.

Corrected rows

• incident CAFR5038 and CAFR5058 (joint)
• instruments_surrendered_and_retired 1,157,566
• obligation_stated CAFR5038 794,741; CAFR5058 362,825
• tonnes_reversed CAFR5038 662,284; CAFR5058 302,354
• compensation_rate 1.2
• rule one obligation, twelve cells, attributed jointly; never summed per incident

• incident CAFR5423
• instruments_surrendered_and_retired 33,658
• obligation_stated 33,658
• tonnes_reversed 28,048
• compensation_rate 1.2

• incident CAFR6313
• instruments_surrendered_and_retired 528,967 by the comments (370,796 in AI1635 plus 158,171 in AI1482); 529,147 by column AA of row 1635 (370,976) plus AI1482; see C4
• obligation_stated 529,147
• tonnes_reversed 440,956
• compensation_rate 1.2

What changes in the totals

Nothing in the reversal totals. The incident series' surrender rows are relabelled as compensation obligations and the joint obligation is counted once; the audit rule becomes retirements at most the magnitude times the compensation rate, joint cells counted once.

CAFR6313: CARB's comment and CARB's column disagree by 180 credits

Rows affected. the CAFR6313 surrender_operator row from cell AI1635 (370,796)

What was published

370,796 credits, as the comment in AI1635 states: In June 2024, 370,796 credits surrendered by the offset project operator to meet the obligation for the CAFR6313 reversal/termination were retired.

What is right

Column AA (Retired for Intentional Reversal) on the same row states 370,976. The obligation CARB states on row 1636 is 529,147 instruments; 370,976 plus the 158,171 retired per AI1482 equals 529,147 exactly, and 370,796 plus 158,171 falls 180 short. The column reconciles and the comment does not, which reads as a transposition of two digits in the comment. Under the register's rule both readings are published; the series' row keeps the comment's figure with the column's beside it and the reconciliation stated.

How it was found

Reading the CAFR6313 rows behind the audit's second warning on 5 September 2026.

Both readings

• comment_AI1635 370,796
• column_AA_row_1635 370,976
• difference 180
• reconciles_with_stated_obligation column

The buffer pool's remaining balance was published as 27.97 Mt

Rows affected. the buffer pool bar's right-hand label

What was published

27.97 Mt remaining, beside 36.23 Mt contributed and 8.97 Mt consumed at 24.8 per cent, as at the regulator's file of 9 December 2025.

What is right

Contributions sum to 36,228,767 and retirements to 8,967,106, which implies a closing balance of 27,261,661, or 27.26 Mt. The file itself states 27,254,352, or 27.25 Mt. The difference of 7,309 units is the break already published, from three arithmetic breaks in the regulator's running balance column on 23 December 2024, 8 April 2025 and 22 April 2025. Both readings stand; neither is preferred silently.

How it was found

An outside review of the homepage on 8 September 2026 noted that the three figures in the buffer bar do not subtract. Checking them against the mirrored file the same day showed the cause was not a presentational one: the remaining balance was the arithmetic of the superseded pair. 36.91 minus 8.94 equals 27.97. The correction that replaced 36.91 and 8.94 and 24.2 per cent with 36.23 and 8.97 and 24.8 per cent fixed three figures and left the fourth, and because 27.97 remained consistent with the numbers it came from it survived every review since. The three buffer charts, which are generated from the file, carried the right figure throughout; only the hand-written label was wrong.

What changes in the totals

Nothing in any loss total, any rate and any denominator. The remaining balance is a presented figure and is not an input to the loss table, the incident series or any published rate. What changes is the figure itself, by roughly 710,000 credits, and the standing of the entry in the fact register, whose verdict read VERIFIED while carrying it.