Global Ocean Accounts Partnership Secretariat · GOAP Technical Guidance on Ocean Accounting · circular
After working through this Circular, you will be able to show that an ocean asset account is complete, balanced and fit to publish, and to hand a reviewer the evidence that confirms it. You will be able to:
The Circular sets no accounting rule of its own. Every rule it tests is owned by the circular named beside the check, most often TG-3.1 Asset Accounts; what this Circular adds is the procedure, the tolerance and the record.
Essential prerequisites:
Helpful background:
This Circular addresses the stock side of the Ocean Accounts Framework, at the same edges as TG-3.1: the asset accounts that record SNA produced assets (SG1) and environmental assets (SG3). In the Ocean Accounts Framework (TG-0.1 Figure 0.1.2):
| Edge | Direction | Description |
|---|---|---|
| E3 | FG1↔SG1 | Economic flows interacting with SNA assets (Applied) |
| E9 | SG3→FG1 | Environmental assets supplying ecosystem services to economy (Applied) |
| E11 | FG3↔SG3 | Environmental assets supporting intermediate ecosystem services (Applied) |
Scope. TG-3.16 owns the QC procedure, the tolerances and the evidence for the asset account tables of TG-3.1, and the reconciliations between those tables and the accounts beside them. It covers asset accounts only. The QC of the other Section 3 accounts — flows (TG-3.2, TG-3.4), economic activity (TG-3.3), social (TG-3.5) and governance (TG-3.7) accounts, and combined presentations (TG-3.8) — sits with TG-0.7 and the circular that owns each account; TG-3.16 meets them only where they reconcile with an asset cell (Section 3.6). It sets no accounting rule of its own and records no account entry, so it shares none.
An asset account can balance and still be wrong. The identity opening + additions — reductions = closing holds in any account whose compiler has closed it on a residual, and a residual can absorb a dropped conversion, a double-counted loss or a unit slip without trace. The worked example in Section 3.8 carries nine seeded errors, and every column in it balances. The identity is necessary, and it is the first check; it is not sufficient, and the rest of this Circular is about what else has to be tested.
Four conventions this Circular depends on are not yet settled. Until they are, it uses these defaults, and comments on them are invited: a two-field flag of grade and status (C2, Section 3.7); the uncertainty ladder of Section 3.9 at 95% (C3); exchange values only, one carbon price per account (C4, AQ-VA6), which departs from TG-6.2 Section 3.6, where the social cost of carbon is accepted for carbon retention; and the indicative tolerances of Tables 3.16.4 and 3.16.5 (C7).
The three terms below are used in the IPCC sense throughout (Table 3.16.1). They separate who performs a check from what it is compared against, and the separation is what makes a review independent.
Table 3.16.1: Terms, after IPCC (2006) Volume 1 Chapter 6 Box 6.1.
Scope. The checks apply to the asset accounts of TG-3.1, physical and monetary, as the Framework Position sets out: SEEA CF individual environmental assets (coastal aquifers only for water), SEEA EA extent, condition and monetary ecosystem assets, and produced assets only where they touch environmental assets.
Basis. No SEEA publication gives a quality assessment framework of its own. The SEEA EA sets out the quality dimensions and the uncertainty categories (paras 2.86—2.95), points compilers to the IMF Data Quality Assessment Framework (para 2.89), and lists the articulation of data quality assessment frameworks in its research agenda as work still to do. The SEEA Central Framework treats quality as data confrontation: putting estimates into an accounting framework exposes their inconsistencies (paras 1.54, 6.7, 6.86), and revisions are “standard practice” (para 6.68). This Circular therefore draws the procedure from general frameworks the SEEA points to: UN-NQAF requirement 15.1, that source data, intermediate results and outputs are regularly assessed and validated; IMF DQAF element 3.4, validation of intermediate data and outputs; and GSBPM v5.2 sub-processes 5.3 (review and validate) and 6.2 (validate outputs).
Layers. The Circular follows the pattern the IPCC inventory guidelines use for the same problem: general procedures are written once, and category procedures sit with the category. TG-0.7 holds the principles; for this consultation TG-3.16 also carries the generic QC core (Section 3.2). TG-3.16 holds QC for the asset-account category, shared by TG-3.1 and the Section 6 circulars that populate asset accounts. TG-3.1 and each Section 6 circular keep their rules, and list in their own quality subsection only the checks distinctive to them. Figure 3.16.1 places each kind of check in the statistical production process.
· Figure 3.16.1
Source IPCC (2006) 2006 IPCC Guidelines, Vol. 1, Ch. 6, Box 6.1 and Sections 6.5–6.10; UNECE GSBPM v5.2, sub-processes 5.3 and 6.2; TG-3.16.
Check identifiers carry the prefix AQ- (asset QC), so a QC log can cite AQ-ID3 without ambiguity against another circular’s numbering.
The plan, the roles, the records, QC intensity (this section), independent review, the archive, the release gate and the generic evidence pack (Section 3.10), and the uncertainty ladder (Section 3.9) apply to every ocean account, not only to asset accounts. They are set out here so that this Circular can be read and used on its own during the consultation. After the consultation closes they move to TG-0.7, and this Circular will keep only the asset-account application. Comments on where this core should sit are welcome.
A QA/QC plan is written before compilation starts and names the checks, who runs them and when, the independent reviewer, the release gate, the tolerance register and the data-quality objectives. It is the document a reviewer reads first, and it is what TG-0.8’s checklist item 8 asks an institution to have adopted. For an asset account the plan also says what each data-quality objective means for assets, which checks each tier runs, and which parameters the parameter register holds (Section 3.3).
Roles. Six roles carry the plan. One person may hold several in a small office, with one exception: the independent reviewer may not also be a compiler.
Records. Three records are kept from the first day: the QC log (check, cell, result, action, initials, date); the issue register (open or closed, and whether each issue blocks release); and the parameter register, the single place every parameter is set (Section 3.3).
QC intensity. QC effort scales with tier and with materiality. A representative sample of derived cells may be checked in any one cycle, provided every part of the account is covered over successive cycles. Headline figures and every figure quoted in narrative text are recomputed in every cycle, whatever the tier (AQ-UN4).
Data-quality objectives. The plan sets the seven data-quality objectives of the IPCC guidelines. For an asset account each means the following (Table 3.16.2), and each needs a national value in the plan.
Table 3.16.2: Data-quality objectives for an asset account, after IPCC (2006) Volume 1 Section 6.5.
Tiers. Tier 1 is the minimum viable approach and higher tiers add rigour, as in TG-0.7 Section 3.8. The tiers here are proposed for consultation:
The worked example measures what the minimum misses: Tier 1 catches six of the nine seeded errors, and the three it does not catch (net recorded as gross, a condition index applied twice, and a loss valued twice) are the case for Tier 2.
QC log template. Tables 3.16.4 to 3.16.6, with columns added for the result, the action, initials and date, are the template for the QC log. For condition cells the log also records the expected range, its source in the reference-value register, and the range verdict under AQ-CO7 (inside the usual range; unusual but within the plausibility range; or outside it and investigated).
Parameter register. Every parameter that scales a result is set once, in one register, with its source, vintage and owner. For an asset account it holds: the discount rate and horizon, the carbon price and its basis, unit prices and rents, live-weight conversion factors, the carbon-to-CO₂ factor of 44/12, reference levels, the minimum mapping unit and the CRS pair. A value typed into a formula is a finding in itself, because it cannot be changed consistently.
Source-data acceptance. Data from monitoring, research and administrative sources often concentrate on edge or problem cases, and need review against the measurement concepts and scope of the account before use (SEEA EA paras 2.87—2.88). That review belongs to the Section 4 circular that delivers the variable, with its definition, source, unit, coverage, reference value, uncertainty and quality tier (TG-0.4 Section 4.5). The account’s test is only that each value arrives with those attributes, a flag and its location in the primary source (a table cell or a quotation); a value without them goes back to its Section 4 owner. The location matters because values copied through secondary syntheses drift: an index is re-labelled as a stock, a biomass figure as carbon, or a value measured in another country is attached to a local place name. A value that cannot be traced to its primary source is not entered.
Double counting. Four pairs cause most of it, and each has a check: fish-stock value and provisioning flow (AQ-ID6; TG-6.7 Section 3.5), produced and ecosystem protection (AQ-PR1), rent and government receipts (AQ-RC4), and nursery services (AQ-VA7).
Time series and revisions. Every revision is documented (SEEA EA para 10.40), breaks are back-cast where feasible (TG-0.7 Section 3.2.4), and the asset account revises when and as the national accounts do, so that the two revisions policies stay aligned (SEEA CF para 6.70). Table 3.16.3 sets the two cases. A back-revision that moves a published indicator by more than 5% is re-published, the threshold TG-4.5 Section 3.5.4 sets.
Table 3.16.3: Treatment of revisions, aligned with national accounts benchmark revision and rebasing.
Several checks below name TG-3.1 as the owner of their rule. The TG-3.1 worked example does not yet pass all of them, and the corrections are scheduled for the revision after this consultation: a reclassification is not shown in Table 5 (AQ-ID4); storm damage is booked as catastrophic loss within the regrowth horizon (AQ-RT1); restoration is valued as enhancement before the survival check (AQ-RT4); the seagrass linear-shortcut test is applied below its 70% condition (AQ-VA3); and depletion is equated with extraction (AQ-PM3). Read the TG-3.1 rules, not its example, until then.
The general checks apply to every asset account, whichever asset type it holds (Table 3.16.4). They are written in the style of the IPCC general inventory QC table: each is a test with a pass condition, and the rule it tests belongs to the circular in the right-hand column. Tolerances belong in the national QA/QC plan, and this Circular strongly recommends that they be set there, taking account of regional conditions: the regional products and assessments a country verifies against, the regional fisheries body’s assessment practice, and regional differences in ecosystem variability and data. The tolerance column gives indicative values only, for use until the plan sets its own; plan marks a check with no indicative value. The evidence each check leaves is listed in the evidence pack (Table 3.16.8).
Table 3.16.4: General QC checks for asset accounts. Indicative tolerances are for national plans to replace; those not taken from a rule owner are GOAP proposals.
Two of the rules these checks test are easier to follow as diagrams. Figure 3.16.2 is the routing rule behind AQ-RT1: which row a change enters, depending on what the event did. Figure 3.16.3 is the trace behind AQ-PM3, AQ-PM4 and AQ-VA2: how each monetary depletion and degradation entry is recomputed from a physical one, and the single point at which the condition index enters.
· Figure 3.16.2
Source TG-3.1 Sections 3.5.1 and 3.5.3; SEEA EA paras 4.15, 10.22–10.23, 10.24, 10.37.
· Figure 3.16.3
Source TG-3.1 Sections 3.3 and 3.5.3; TG-2.1 Section 3.4.1.
An asset-type check is an acceptance test at the interface between the account and the method that feeds it, as TG-3.1 frames the requirements a stock assessment must meet. The account does not re-run the method. It tests whether what arrives is usable, and sends it back to the owning circular if not (Table 3.16.5).
Table 3.16.5: Asset-type acceptance checks. Each tests a rule set by the circular in the right-hand column; the extent thresholds are TG-4.1’s.
Area estimates. Where the area-adjusted estimate of AQ-EX2 (Olofsson et al. 2014) differs from the mapped area by more than the plan’s tolerance, the account records the estimate and says so. This rule is proposed, for the owners of TG-4.1 and TG-2.12 to confirm.
Reconciliation tests the asset account against the accounts compiled beside it. Each pair in Table 3.16.6 must agree within the plan’s tolerance, or the difference is explained in the log. None of these pairs is new: each is required by the circular that owns the neighbouring account. What this Circular adds is the requirement to run all of them and record the result (Figure 3.16.4).
· Figure 3.16.4
Source TG-3.1 Step 6; TG-3.2; TG-3.7; TG-3.11; TG-6.2; IPCC (2006) Vol. 1, Ch. 6, Section 6.10; IMF DQAF element 3.4.
Table 3.16.6: Reconciliation matrix.
Verification compares the account with data from outside it, . It cannot confirm a figure, because the comparator has its own error, but a large gap shows where to look. The checks are:
Trend checks. Each cycle is compared with the previous one, following IPCC practice. Implied unit prices and unit rents are compared across years, and any extent change beyond the national threshold in the plan is investigated.
TG-0.7 Section 3.6 defines one flag scheme. This Circular applies it to asset cells in two fields, so that a cell can say both how good it is and how it was obtained (Figure 3.16.5):
· Figure 3.16.5
Source TG-0.7 Section 3.6 (flag scheme and lowest-flag rule); TG-3.16 Section 3.7 (AQ-FL1 to AQ-FL3).
[A], or [C·E] for a grade-C estimated value.Where no standard error can be computed, which is common where condition variables come from published studies and non-probability surveys, the condition grade rests on the evidence behind each input instead: A for a full-text primary source that reports n and a labelled dispersion type; B for a full-text source without them, or a statistic the compiler computed from non-probability survey records; C for a value read from an abstract, from grey literature or digitised from a figure. A regional or global proxy takes status E. The index takes the lowest grade of its inputs under AQ-FL3. A monetary cell takes the lower of its physical grade and its valuation grade, and never a higher one.
The interim thresholds, adopted as a proposal only pending decision on the flag scheme, are a relative 95% half-width of at most 10% for A and at most 30% for B, with C above that.
Local scales map onto the scheme. High, medium and low confidence become A, B and C. TG-4.1’s “conditionally accepted” becomes C. Provisional, indicative and fallback values take status P, and modelled, benefit-transfer and TK-derived values take status E. “Data held but publication restricted” becomes X.
This example is synthetic, and every number in it comes from one script run on one set of inputs. It is independent of the TG-3.1 worked example. An accounting area of 1,200 km² holds five ecosystem types, one demersal fish stock and a mangrove condition account; the period is one calendar year, and money is in USD thousand at current prices. The submitted accounts contain nine deliberate errors, E1 to E9, of kinds common in compiled accounts (Table 3.16.7).
Table 3.16.7: Seeded errors and the checks that catch them.
Two things in Table 3.16.7 are worth noticing. First, AQ-ID1 catches none of the errors: every submitted column balances. Second, several errors trip more than one check. E4 is the clearest case: a welfare price for carbon breaks the exchange-value rule, and by inflating carbon’s share of the mangrove’s annual flow value it also takes the asset outside the conditions under which the linear degradation shortcut may be used.
Three tables are printed here: the submitted monetary ecosystem account, which holds three of the errors; the QC log; and the same account corrected and flagged. All 22 tables of the example — the parameter register, the extent, change-matrix, condition and fish-stock accounts, every submitted and corrected table, the reconciliation and the uncertainty results — are in a companion page, with the QC log also as CSV.
Table WE-6a is shown as a compiler submitted it. Do not copy it as a model.
Table WE-6a: Monetary ecosystem asset account, as submitted (USD thousand)
Subtidal sediment and Shelf waters: not valued (”..”), so not shown. Aquaculture ponds: a confirmed zero, not a gap.
Table WE-6a: Monetary ecosystem asset account as submitted, containing errors E3, E4 and E5. Structure after TG-3.1 Table 4.
The QC log is the record the reviewer reads (Table WE-7). One line is written per finding, naming the check, the cell and what was found; the fix is recorded against the same line.
Table WE-7: QC log for the submitted set
Table WE-7: QC log. Thirteen checks fail; AQ-ID1 is not among them.
Each correction follows the rule its check names; here the condition index enters once, carbon takes the exchange-value price, and the seagrass entry is re-derived from its unit value.
Table WE-8c: Monetary ecosystem asset account, corrected and flagged (USD thousand)
Subtidal sediment and Shelf waters: not valued (”..”), so not shown. Aquaculture ponds: a confirmed zero, not a gap.
Table WE-8c: Monetary ecosystem asset account, corrected. Every valued cell except the confirmed zero for ponds is C·E, because the valuation ranges are wide (companion Table WE-10b).
What the example shows. An account can be internally consistent and still wrong in six ways. It can be incomplete (E1); recorded net where gross is required (E2); in breach of a valuation rule (E3, E4); carrying an arithmetic slip (E5, E9); letting a total or a row take in an item that belongs elsewhere (E6, E7); or computing a derived figure on the wrong concept (E8). Each is caught by a check that recomputes from evidence outside the row being checked: the change matrix, the parameter register, the unit value, the physical account, or the stock assessment.
TG-0.7 Section 3.4 names the four categories of uncertainty from SEEA EA paras 2.90—2.94, with their usual sources in ocean accounts, and this Circular does not repeat them. For asset accounts three physical sources dominate, and each has its method elsewhere: the area confidence interval from the accuracy assessment (AQ-EX2), the biomass and sustainable-yield intervals from the stock assessment (TG-6.7 Section 3.6), and the standard error of the condition index (AQ-CO5). Valuation and future-price uncertainty are carried by the discount-rate grid (AQ-VA5) and by the range reported with every value.
Method. Uncertainty is quantified up a ladder, going as far as the data allow: a qualitative statement; an analytic, first-order combination in the manner of the GUM; Monte Carlo propagation with at least 1,000 draws; and a Bayesian treatment where priors are defensible. Intervals are reported at 95%. For an asset valued as the present value of its flows, the analytic step is
V=t=1∑T(1+r)tQtpt,(VuV)2≈(QuQ)2+(pup)2
where Qt is the quantity, pt the unit value, and u a standard uncertainty. The quadrature form assumes the quantity and the price are independent. Where they are not, as when a sustainable yield and a unit rent both depend on the same biomass, Monte Carlo over rent, sustainable yield, discount rate and horizon replaces it.
Reporting. Every monetary asset value is published with its range. The discount-rate grid (AQ-VA5) and the aggregation-rule sensitivity (AQ-CO4) go beside the account, not in an annex a user may not open.
Modelled flows. Where a service flow or a stock comes from a model, it is validated as TG-4.11 Section 3.3 sets out. Until it is, it is flagged E (TG-4.5 Section 3.3.3). A biophysical model also conserves mass and area, the check the IPCC 2019 Refinement adds in Section 6.12.
AQ-RV1: review. A reviewer not involved in compilation reviews the evidence pack, recomputes a sample of cells and every headline figure, and logs each finding as closed or accepted. Independent review is recommended, not a condition of release. For an asset account the reviewer should include a subject specialist for each part compiled: a stock-assessment scientist for the stock, a remote-sensing accuracy specialist for extent, and an environmental economist for valuation. Tier 3 makes this the standard (Section 3.2).
AQ-RV2: documentation. The four documentation categories of TG-0.7 Section 3.5 are completed for the account. Beyond these, an asset account records the biomass concept, the assessment vintage and the valuation method; every other setting is in the parameter register (Section 3.3). Where an SDG indicator is derived from the account, the validation and quality fields of its metadata (4.d and 4.i to 4.k) are completed from the same record.
AQ-RV3: archive. Inputs, version-controlled code, the parameter register, the QC log and the outputs are archived together, sufficient for anyone to reproduce every published cell, .
AQ-RV4: release gate. The account is released when four conditions hold:
Initial monetary releases are labelled experimental (AQ-EV4).
The evidence pack. Every account keeps a generic evidence pack: the scope and boundary statement (accounting area, period, boundaries, residence or territory, BSU level); the QA/QC plan with its tolerance register; the source register (every source, its owner, licence, date, method and coverage); and the archive manifest. Table 3.16.8 lists what an asset account adds; each item is produced during compilation, and none is written for the review.
Table 3.16.8: What an asset account adds to the generic evidence pack.
Four limits apply to everything above. First, a future SEEA quality framework may supersede parts of this Circular (Section 3.1). Second, the tolerances here are indicative. Several are GOAP proposals, not SEEA rules: the 0.20 decline and 70% share of the linear shortcut, the 80% and 75% accuracy thresholds, and the interim grade thresholds. Third, the monetary ecosystem checks inherit the status the UN Statistical Commission gave SEEA EA chapters 8 to 11: internationally recognised statistical principles and recommendations, and not a full statistical standard. Fourth, coherence is not validity. An account can pass every check here and still rest on an ecological model that is wrong, and only verification against independent observation, repeated over cycles, tests that.
Authors: [To be confirmed]
Reviewers: [To be confirmed]
The structure of this Circular draws on the IPCC inventory guidelines’ treatment of quality assurance, quality control and verification, on the UN National Quality Assurance Frameworks Manual, and on the IMF Data Quality Assessment Framework.
Cite this report
Quality Assurance and Quality Control of Ocean Asset Accounts (2026) (GOAP Technical Guidance on Ocean Accounting, TG-3.16, v0.3.0). GOAP Secretariat. https://guidance.oceanaccounts.org/circulars/section-3/3-16-asset-accounts-qaqc/
guidance.oceanaccounts.org/circulars/section-3/3-16-asset-accounts-qaqc
© 2026 Global Ocean Accounts Partnership Secretariat · CC BY-NC 4.0 · The report, its PDF and this deck are built from the same source.
Swipe on touch screens. Print gives one landscape page per slide. The address keeps the slide number, so a link opens on the right slide. On a phone in portrait the deck becomes a scrolling document.