Taxonomic Documentation Under Uncertainty: Data Quality and Provenance in a Pirabas Formation Fossil Collection from the Brazilian Amazon

1. Introduction

The evolution of the Equatorial Atlantic Ocean led to the development of numerous offshore sedimentary basins, such as the Foz do Amazonas, Pará-Maranhão, and Barreirinhas basins, as well as onshore sedimentary sequences, including those of the Bragantina Platform [29,30]. The Neogene of the Bragantina Platform is mainly represented by the marine carbonate-siliciclastic deposits of the Pirabas Formation [31], a lithostratigraphic unit that occurs discontinuously across the states of Pará, Maranhão, and Piauí in northern and northeastern Brazil [1,29,32]. The abundant and diverse fossil assemblage of the Pirabas Formation comprises predominantly paleoinvertebrates, along with paleovertebrates, microfossils, and plant fossils [1,2,32,33]. These fossil assemblages document the remarkable paleobiodiversity of northern Amazonia during the early to late Miocene, representing one of the best-preserved fossil records of the Brazilian marine Cenozoic [24,32,34].

Despite its recognized significance, the Pirabas Formation remains comparatively under-documented compared with better-studied Cenozoic assemblages from southeastern and southern Brazil, a disparity consistent with the concentration of paleontological research infrastructure in the country's principal research centers [3]. Many fossiliferous outcrops occur in active limestone-mining areas in the municipalities of Primavera, Capanema, and Salinópolis, northeastern Pará, where industrial extraction exposes and destroys fossil-bearing layers. Brazilian law treats fossil deposits as property of the Union and requires prior authorization for their collection [4]. In practice, however, the survival of specimens exposed during quarrying depends on the rescue and curatorial capacity of nearby research institutions working together.

1.1. Geological and Paleontological Context of the Pirabas Formation

The sedimentary rocks of the Pirabas Formation overlie Precambrian basement rocks and are overlain by the continental siliciclastic deposits of the Barreiras Formation. The genesis of the Pirabas Formation deposits is associated with the development of marine carbonate platforms in what is now the Amazonian coastal region during the geological evolution of the Equatorial Atlantic Ocean in the Oligocene – Miocene [17,30]. The carbonate-siliciclastic sedimentation of the unit varied in response to fluctuations in relative sea level and tectonic subsidence [35]. During the Miocene, the progradation of deltaic-fan systems led to the cessation of carbonate sedimentation of the Pirabas Formation, driven by a substantial increase in siliciclastic input resulting from the combined effects of sea-level fall and tectonic reactivation [35].

The age of the Pirabas Formation was initially regarded as Oligocene – Miocene based on the occurrence of the gastropod Orthaulax pugnax [36]. Subsequently, different ages were proposed by other authors, including the early Miocene based on palynological studies [34] and the late Oligocene – early Miocene based on ostracod biostratigraphy [37]. More recent biostratigraphic studies based on palynomorphs, foraminiferans, and other marine microfossils, such as dinoflagellate cysts and acritarchs, from five fossiliferous localities of the Pirabas Formation (Praia do Atalaia, Pedreira de Capanema, Baunilha Grande, Ponta do Castelo, and Grupo Fazenda) indicate that the sedimentary successions of the unit were deposited from the early to the late Miocene [24].

Within the Pará state, the best-known fossiliferous exposures are concentrated in the northeastern region, particularly around Salinópolis, São João de Pirabas, Capanema, and adjacent localities (Figure 1). The discontinuous surface exposure of fossiliferous rocks of the Pirabas Formation occurs along coastal cliffs, road cuts, beaches, and open-pit mines at 27 known localities [38]. The Natural Heritage Collection (Coleção de Patrimônio Natural, CPAN) of the Federal University of Pará (UFPA), where the specimens discussed herein are curated, is in the city of Belém (Figure 1), southwest of the main fossiliferous localities of the Pirabas Formation. This spatial arrangement underscores the documentary role of locality data in reconnecting material housed in the capital with its geological source [25]. The formation should therefore not be treated as a single homogeneous fossiliferous horizon. Locality and stratigraphic information are essential whenever specimens are compared across their geographic distribution.

Lithologically, the Pirabas Formation comprises a diverse succession of carbonate rocks interbedded with sandstones, mudstones, and shales. The carbonate rocks include massive and stratified biocalcarenites, biocalcirrudites, marls, calcilutites, and bioherms [17]. The depositional model initially proposed for the Pirabas Formation combined a shallow carbonate platform with lagoonal, tidal, and mangrove-influenced environments [17]. Facies and paleontological analyses of the Pirabas Formation indicate substantial paleoenvironmental heterogeneity, including shallow platforms, reefs, beaches, lagoons, mangrove environments, tidal channels, and tidal deltas within a marine-coastal system that was periodically influenced by storms [23,24,32,39,40].

The paleontological record of the Pirabas Formation is exceptionally diverse, comprising both microfossils and macrofossils [21,33]. Paleoinvertebrate macrofossils are the most abundant groups and include corals, bryozoans, abundant mollusks, crustaceans, and echinoderms [25,32,41,42,43,44,45,46]. Paleovertebrate macrofossils are less frequent, comprising chondrichthyans, bony fishes, chelonians, crocodilians, and sirenian mammals [20,32,47,48,49]. Microfossils are represented by foraminiferans, ostracods, palynomorphs, poriferans, and calcareous nannofossils [18,24,30,32,37,50]. Other fossil groups include trace fossils, plant remains documenting terrestrial influence, and marine coralline algae [19,21,22,39].

The lateral and vertical alternation of sedimentary facies within the Pirabas Formation explains why fossils collected from different fossiliferous localities may represent distinct energy regimes, substrates, and spatial positions within the broader marine-coastal system [17,23]. Repeated transgressive–regressive cycles controlled the distribution of fossils throughout the succession and demonstrate that taxonomic occurrence cannot be interpreted independently of facies [24].

This geological heterogeneity directly affects the scientific value of specimens housed in the CPAN. Taxonomic classification alone does not establish whether two specimens originated from the same depositional environment or stratigraphic interval. Whenever possible, collection records should preserve information on the municipality, precise locality, geographic coordinates, lithology, collector, and field association. When such information is unavailable, the catalogue should explicitly indicate the limitation rather than reconstructing a depositional context based solely on regional knowledge. The Pirabas Formation thus provides a clear example of why paleontological identification and provenance control must be integrated into museological documentation.

1.2. Institutional History of the CPAN

The Natural Heritage Collection (Coleção de Patrimônio Natural, CPAN), also referred to institutionally, before 2022, as the Natural History Collection, was established between 2017 and 2018. It is housed in the Laboratory of Preventive Conservation of Movable Heritage (Laboratório de Conservação Preventiva de Patrimônio Móvel, LCPPM), a teaching, research, and extension^[1] facility linked to the Museology Program of the Federal University of Pará (UFPA). The initiative was led by Dr. Sue Anne Regina Ferreira da Costa in collaboration with Museology and Biology students to safeguard fossils recovered through authorized fieldwork at a limestone quarry in Primavera and at other localities in Pará, including Capanema and Salinópolis [5,6,7,8].

The collection subsequently expanded through donations and associated research and extension projects. Its scientific and educational mission is to create an institutional resource that supports plural discussions about heritage and nature in the Amazon. Paleontological specimens predominate and consist mainly of fossils from the Pirabas Formation. Management is undertaken by an interdisciplinary team of faculty members, students, volunteers, and graduates, particularly from Museology, Conservation-Restauration and Biology. Through planned activities in the collection storage, the team links conservation, documentation, research, teaching and public communication [5,6,7,8].

This paper examines taxonomic documentation as an essential but technically demanding component of paleontological collection management in the CPAN. The specimens are maintained in the collection storage facility of a Museology Program rather than in a Biology, Geology or Paleontology department, which makes the distinction between curatorial documentation and specialist taxonomic authority especially visible. The study does not undertake a systematic revision, estimate taxonomic richness, or infer paleoecological abundance. Instead, it asks five questions: (i) how complete are key taxonomic, provenance, location, and conservation fields in the available record-level inventory snapshot, (ii) at what documentary granularity are taxonomic entries recorded, (iii) where do direct cross-field contradictions and identifier problems occur, (iv) how can a curatorial protocol preserve uncertainty and provenance while facilitating later specialist validation, and (v) can that protocol be operationally applied to inventory records and related to an established biodiversity-data standard without implying taxonomic verification. The specific objectives are to describe the collection's institutional development, quantify record-level data quality, examine the relationship between taxonomic precision and documentary reliability, operationally apply the proposed workflow to representative records, and identify potential correspondences with Darwin Core terminology.

2. Materials and Methods

2.1. Fossil Provenance and Study Collection

The study concerns the paleontological specimens of the CPAN, currently housed at the LCPPM in Belém, Pará. Most specimens were obtained through authorized field collection at a limestone-mining operation in Primavera, with additional material recovered from outcrops in Capanema and Salinópolis. The collection also grew through donations and associated research and extension projects [5,6,7,8]. Field rescue at the Primavera quarry was organized through collaboration between UFPA researchers and the mining operator, allowing exposed fossils to be recovered before destruction [6]. The record-level data-quality analysis reported below is restricted to the 106 paleontological entries present in the inventory snapshot dated 2 September 2026 and should not be equated with the total institutional specimens.

The three cited municipalities lie within the principal exposure area of the Miocene Pirabas Formation in northeastern Pará. Its carbonate deposits record shallow-marine to estuarine environments and contain a diverse assemblage of marine invertebrates and vertebrates, together with terrestrial plant remains [1,2,32,33,39]. The locality recorded for each specimen is therefore essential for connecting an object in storage to its geological source. However, the degree of contextual resolution is uneven. Material collected during organized field rescue may retain the municipality, collecting locality and geological formation, whereas donations or older entries may lack stratigraphic position, geographic coordinates or a detailed description of the collecting horizon. In this study, provenance was treated as documentary evidence and was not reconstructed when the institutional record did not support a more precise assignment.

2.2. Specimen Preparation, Measurement and Identification

Fossil material is transported from the field in plastic bags placed inside cardboard boxes, a procedure that can retain moisture and compromise specimen integrity. Wet samples are therefore left without handling for approximately one week in a dedicated preparation room. This room forms part of the air-conditioned collection storage environment and is maintained at approximately 24 °C with continuous climate control^[2]. The stated temperature describes the controlled collection space and not the ambient tropical climate of Belém. After drying, excess sediment is removed mechanically with small spatulas, dental instruments, brushes, pins, and needles. Staff take care to preserve natural morphology, particularly that of mollusk shells [7].

Each prepared specimen is measured in three dimensions with a caliper and then submitted for taxonomic identification using specialized literature on the Pirabas Formation. For mollusks fossils, identification seeks to reach genus level, although this is not always possible. Diagnostic soft-tissue and shell-surface structures may be lost during fossilization, and comparative references for the formation remain limited. The identification recorded in the catalog is therefore treated as provisional when diagnostic evidence is insufficient [7].

2.3. Digital Documentation and Data Compilation

Before 2022, CPAN documentation combined a physical registration ledger with individual catalog sheets prepared in word-processing software. The ledger recorded biological identification, provenance, geological formation, geological age, acquisition mode, and specimen dimensions. Catalog sheets were organized into identification, conservation, and historical-analysis groups, but inconsistencies prevented automatic migration to a unified system [5]. In 2022, in partnership with the UFPA Museums and Collections Network, the CPAN adopted Tainacan (https://acervosonline.ufpa.br/index.php/historia-natural/), a free and open-source repository implemented as a WordPress plugin. The current repository schema contains 55 metadata fields, of which 14 are mandatory, arranged into six groups: property characterization, preservation, communication, acquisition data, life and research, and registration data. It also uses an alphanumeric registration code and conservation categories for good, regular, or poor condition, with an additional quarantine designation for biological infestation [5]. Further details on the complete documentation workflow are provided by Lima, Sepúlvreda, and Costa (2023) [5].

The cataloguing manual consulted for this study requires the physical registration ledger to be completed before the virtual record. It defines a unique identifier in the format RTM.II.1.0000 for paleontological material and prescribes record-level fields for previous numbers, segments, location status, biological classification, museological classification, title, descriptive summary, measurements, provenance, acquisition and preservation. Photographs should show more than one view and include a numerical scale, a color scale, and the registration number. These requirements were examined alongside institutional records to describe the documentary workflow.

2.4. Inventory Audit and Data-Quality Assessment

Data were compiled from CPAN institutional documentation, published accounts of the collection's history and repository [58], and the institutional inventory workbook dated 2 September 2026. This workbook was provided by the current curatorial team as the most up-to-date inventory available, already incorporating the ichthyology collection. The worksheet “Paleonto” contained 106 records. Because a separate institutional diagnostic reported 630 catalogued fossil items [10], these 106 entries were treated as the record-level subset currently represented in the most recent inventory supplied by the collection's curators, rather than as a complete census of CPAN's paleontological collection.

Fifteen fields were assessed: current registration identifier, scientific-name field, associated term, previous number, taxonomic classification, phylum, shelf, location status, collector, collection date, acquisition date, origin, acquisition mode, conservation status, and observation. Completeness was calculated as the number of non-empty cells divided by 106, with blank and whitespace-only cells treated as missing. Missing values were not reconstructed from other institutional sources for the purpose of this calculation.

The scientific-name entries were normalized only for surrounding whitespace and assigned to four documentary granularity classes based on the form of the entry, without validating taxonomic correctness: genus-level or genus-style nomenclature, formal suprageneric taxon, descriptive or non-standard label, and blank. The third category included vernacular names, anatomical or fossil-material terms and lithological descriptions entered in the scientific-name field. A direct cross-field contradiction was counted only when populated broad categories were mutually incompatible under a conservative rule, for example “Echinodermata” in the scientific-name field with “Mollusca” in the phylum field, or a record marked “Vertebrado” while the phylum field contained “Arthropoda” or “Echinodermata”. Current and previous identifiers were also screened for duplication and obvious pattern deviations. The audit evaluates documentary structure and internal consistency, not the biological validity of individual determinations. The original 15 inventory fields and the derived audit categories are provided in SupplementaryTable S1. Records meeting the conservative contradiction rule are listed in Supplementary Table S2, and identifier anomalies are listed in Supplementary Table S3.

2.5. Operational Protocol Application and Darwin Core Crosswalk

To assess practical applicability without implying taxonomic validation, the seven-stage protocol was operationally applied to seven records purposively selected from the audit to represent different documentary conditions: a genus-style entry without a detected cross-field contradiction, two types of direct taxonomic-field conflict, a material-based label with incomplete classification, a duplicated current identifier, a blank scientific-name field with a descriptive associated term, and an ichnofossil label with unresolved classification. The exercise used only the inventory metadata. No specimen was re-examined, no taxon was corrected, and no specialist determination was inferred. Each case was assigned a documentary response focused on preservation of original values, explicit review status, provenance visibility and identifier control.

To examine interoperability, the information required by each protocol stage was compared with current Darwin Core terminology [26,27]. Potential correspondences include scientificName, taxonRank, identificationQualifier, verbatimIdentification, identificationVerificationStatus, identifiedBy, dateIdentified, identificationReferences, locality, recordedBy, eventDate and lithostratigraphic terms such as formation, member and bed. The mapping is functional rather than a claim of one-to-one semantic equivalence, and it does not imply that the present CPAN repository already implements these fields. The complete crosswalk is provided in Supplementary Table S4.

3. Results

3.1. Collection Development and Scope of the Audited Inventory

In August 2019, the CPAN paleontological specimens comprised 467 fossils entered in the physical registration ledger. Of these, 123 were mollusks rescued from the Primavera mining operation [6]. A 2023 institutional account estimated more than 700 fossil specimens, predominantly from the Pirabas Formation [5]. Between that report and the subsequent UFPA-wide diagnostic, an inventory-review procedure reassessed the specimens and established 630 catalogued fossil items [10]. The earlier value was an extrapolated approximation rather than a consolidated count. The record-level workbook available for the present audit contained 106 paleontological entries dated 2 September 2026. Because the 630-item institutional count and the 106-row spreadsheet derive from different documentary sources and scopes, the latter was analyzed as an auditable subset and was not used to infer the total size or taxonomic composition of the collection.

3.2. Taxonomic Structure, Granularity and Internal Consistency

The phylum field was populated in 104 of the 106 audited records (98.1%). As recorded in the spreadsheet, 44 entries were assigned to Mollusca, 31 to Echinodermata, 23 to Chordata, five to Arthropoda, one to “Pteridofta”, and two were blank (Table 1). These values describe the inventory fields exactly as entered and do not constitute a validated taxonomic census of the collection. The scientific-name field was populated in 105 records (99.1%), but its documentary depth was uneven. Twenty-eight entries (26.4%) used genus-level or genus-style nomenclature, 45 (42.5%) contained formal suprageneric taxa, 32 (30.2%) used descriptive, vernacular, anatomical, lithological, or fossil-type labels, and one entry was blank.

The audit also identified direct contradictions between broad taxonomic fields. Eight records entered “Echinodermata” in the scientific-name field while assigning “Mollusca” in the phylum field. Four additional records combined the classification “Animalia, Vertebrado” with either “Arthropoda” or “Echinodermata”. At least 12 of 106 records (11.3%) therefore contained an explicit broad-level contradiction under the conservative audit rule. Other entries also require review because the scientific-name, associated-term and classification fields do not always describe the same object consistently. These cases were retained as documentary evidence rather than corrected in the dataset.

The distinction between field completion and taxonomic reliability is central to the collection's research usability. A populated scientific-name field may contain a phylum, an order-level term, a vernacular label, or a genus-style entry, and these forms carry different evidential weight. When diagnostic structures or specialist validation are unavailable, the record should retain the broadest defensible rank, state the determination status, and preserve the history of later revision. The inventory audit therefore supports treating uncertainty as structured information rather than allowing a provisional or inconsistent entry to acquire authority through repetition.

3.3. Metadata Completeness, Provenance Gaps and Identifier Control

The audited spreadsheet shows a marked asymmetry between taxonomic descriptors and provenance or acquisition information. Current registration identifiers and previous-number fields were populated in all 106 records. The scientific-name field was complete in 105 records (99.1%), the associated-term field in 103 (97.2%), taxonomic classification in 105 (99.1%) and phylum in 104 (98.1%). By contrast, collector, collection date, acquisition date, origin and acquisition mode were empty in every audited row. The dedicated shelf field was populated in 13 records (12.3%), conservation status in 10 (9.4%) and observation in 11 (10.4%) (Figure 2).

Location status was populated in all records, but its level of precision varied. Fifty-one entries (48.1%) contained a shelf-level location such as “Estante 5, Prateleira E” or “Estante 5, Prateleira D”, 54 (50.9%) contained only the generic value “Localizado”, and one record (0.9%) was marked “Não localizado”. Identifier screening detected one duplicated current registration code (RTM.II.1.0068-2), one duplicated previous number (RTM.II.1.07) and one current-code pattern deviation (RTM.II..1.0021). The conservation fields also illustrate the need for cross-field reading: all ten records with a conservation-status value were marked “Bom” (Good), yet six of those records contained observations documenting quarantine or preventive treatment related to biological contamination. These findings show that presence of a value in one field does not by itself guarantee completeness or consistency of the curatorial record.

3.4. Digital Repository Outcomes

The transition to Tainacan produced a structured, searchable repository schema with 55 fields per catalog record and replaced the former combination of paper and word-processing files. The system distinguishes mandatory and optional metadata, links registration numbers to specimen images and integrates identification, acquisition, preservation, research and communication information. Public access makes catalog descriptions and images available for remote consultation, while private administrative fields remain restricted. The 2026 audit analyzed a 15-field inventory worksheet rather than a full repository export, so its completeness rates cannot be generalized to every Tainacan metadata element. The contrast nevertheless demonstrates that defining a field in a schema and consistently populating that field are separate curatorial tasks. This implementation forms part of the wider UFPA digitization effort that culminated in the establishment of the Museums and Collections Network and online repositories for university assets, still ongoing [5,10].

3.5. Proposed Protocol for Taxonomic and Provenance Documentation

The proposed protocol organizes taxonomic and provenance information as a revisable documentary record (Table 2). First, the curator assesses whether diagnostic structures are visible after preparation and imaging. Second, the specimen is entered at the broadest rank supported by the preserved evidence and consulted source. Third, the determination is marked as provisional, specialist-validated or requiring review. Fourth, the record identifies the determining person, date and bibliographic or comparative basis. Fifth, provenance is classified as documented, partial or unavailable, with locality, coordinates, formation, horizon and source recorded only when supported by evidence. Sixth, cases that exceed the expertise available to the curatorial team are referred to a paleontologist or another relevant specialist. Finally, a later determination is added without deleting the preceding entry, preserving authorship, date and reason for revision. This sequence also requires explicit retention of missing data rather than reconstruction from general regional knowledge.

3.6. Operational Application of the Documentation Protocol

Applying the protocol to representative inventory cases produced differentiated documentary actions without requiring a taxonomic correction (Table 3). Records with internally compatible genus-style nomenclature could be retained as documented while remaining explicitly unverified until determiner, date and reference information are recovered. Direct conflicts between populated taxonomic fields were instead assigned review-required status, with the original values preserved and no corrected taxon inferred. Material-based or fossil-type labels were treated as descriptive information rather than as equivalent to a biological determination. Identifier anomalies triggered reconciliation of the digital record with the registration ledger and physical specimen rather than silent renumbering. Blank taxonomic fields remained unresolved even when an associated descriptive term was available.

The exercise therefore demonstrates operational use of the protocol at the metadata level. It does not test whether any recorded biological determination is correct. Across the seven cases, the protocol consistently separates three decisions that are often conflated in collection databases: what the inventory currently says, whether the record is internally coherent, and whether a biological determination has been verified. The Darwin Core crosswalk in Supplementary Table S4 further shows that most of the proposed documentary elements can be represented through established biodiversity-data terms, although revision history may require more than a flattened single-row export.

4. Discussion

The central result is not a new taxonomic composition for the Pirabas Formation, but a record-level demonstration that nominal field completion and scientific reliability are not equivalent. In the audited subset, 99.1% of records contained a value in the scientific-name field, yet only 26.4% used genus-level or genus-style nomenclature and at least 11.3% contained direct contradictions between populated broad taxonomic fields. The collection therefore illustrates a problem relevant to paleontological research beyond the institutional case: a searchable taxonomic field can still transmit uncertain, heterogeneous or internally inconsistent information. Lieberman and Kaesler argued that the scientific value of natural history collections cannot be reduced to taxonomic completeness [11]. Here, that value also depends on making the evidential status of each determination visible and revisable.

The documented workflow aligns with the principle that digitization should be planned as part of curation rather than treated as the final transcription of isolated labels. Karim et al. emphasized the need for documented workflows in paleontological digitization [12], while Contreras connected field recovery, preparation, curation and public databasing in fossil collections [13]. CPAN follows a comparable sequence from rescue and controlled drying to cleaning, measurement, identification, registration, imaging and online publication. The present audit adds a further requirement: metadata design must be accompanied by record-level quality control. A repository may provide fields for provenance, acquisition or preservation, but those elements contribute to scientific reuse only when they are consistently populated and their evidential basis is clear.

The operational application also clarifies what can be achieved before specialist taxonomic support is available. Curatorial quality control can identify conflicting fields, ambiguous documentary categories, duplicate identifiers and absent provenance without deciding which biological name is correct. This distinction is important because a database-cleaning exercise should not silently become a taxonomic revision. The selected cases show that the protocol can stabilize the evidential status of records by preserving original wording, assigning review status and separating descriptive labels from determinations, while explicitly reserving biological correction for later specimen-based assessment.

Interoperability provides a second route by which the protocol can extend beyond the institutional case. Darwin Core was developed to standardize the sharing of biodiversity occurrence and collection data [26], and its current terminology includes fields for original and structured identifications, uncertainty, verification status, determiners, dates, references, collecting events, locality and geological context [27]. These terms correspond closely to the information gaps revealed by the CPAN audit. In particular, verbatimIdentification can preserve the original wording, identificationVerificationStatus can distinguish unverified from reviewed determinations, identifiedBy and identificationReferences can expose attribution, and formation, member and bed can prevent broad geological context from being conflated with documented stratigraphic provenance. The proposed crosswalk is therefore intended as a pathway toward interoperable revision, not as evidence that CPAN records are already Darwin Core-compliant.

The uneven identification depth is scientifically and museologically significant. Loss of diagnostic structures during fossilization and limited comparative resources can restrict lower-rank assignments [7]. The audited spreadsheet captures this heterogeneity directly. Formal suprageneric taxa accounted for 42.5% of scientific-name entries, while 30.2% used descriptive, vernacular or material-based labels. Such entries may still be useful for object retrieval and preliminary management, but they should not be treated as equivalent to specialist determinations. The repository should preserve the rank actually supported by evidence, identify the person or source responsible for a determination, mark provisional assignments and retain revision history. This approach separates documentary reliability from taxonomic precision and reduces the risk that an uncertain label becomes authoritative through repeated reuse.

This distinction also clarifies professional roles. Museologists are responsible for documentation standards, terminology control, record integrity, preservation history and access, while paleontologists and other biological specialists validate determinations that require disciplinary expertise. The inventory results show why these responsibilities must remain connected. Direct contradictions between the scientific-name, classification and phylum fields can be detected through curatorial quality control even before a specialist undertakes a systematic revision. Recording determination status, author, date, source and referral outcome makes the cooperation auditable and allows later correction without erasing the documentary history of the object.

The scientific utility of a fossil specimen depends not only on its taxonomic identification but also on the quality and traceability of its geographic and stratigraphic provenance [28]. A fossil name without a traceable collecting context has more limited value for stratigraphic, paleoenvironmental and biogeographic research. In the audited spreadsheet, collector, collection date, acquisition date, origin and acquisition mode were absent from all 106 rows. This result must be interpreted narrowly: it demonstrates absence from the analyzed inventory fields, not necessarily absence from field notebooks, earlier catalog sheets or other institutional documentation. Even with that limitation, the finding identifies a concrete data-integration problem. Provenance information should be transferred with its documentary source and degree of certainty, while unavailable elements should remain explicitly marked as missing rather than reconstructed from general knowledge of the Pirabas Formation.

Digital access expands the potential audience and reduces the need for preliminary handling of fragile material. Hedrick et al. described digitized natural history collections as infrastructure that enables remote and cross-disciplinary use [14], and Popov et al. documented research, educational and broader public value arising from accessible collection data [15]. These benefits depend on persistent identifiers, images, provenance and preservation records. The audit detected one duplicated current registration code, one duplicated previous number and a formatting deviation in the identifier pattern. Although limited in number, such anomalies are important because identifiers connect physical specimens, legacy documentation and digital records. Routine validation rules and periodic duplicate checks should therefore form part of the repository workflow.

The institutional context remains both a strength and a vulnerability. A national assessment of Brazilian university geological and paleontological collections found uneven curatorial practices and associated better performance with dedicated heritage-conservation expertise [16]. The UFPA diagnostic likewise identified dependence on students and the need for repeated training as recurring conditions in university collections [10]. At CPAN, interdisciplinary work by Museology and Biology students connects curation to education, but turnover can affect continuity. The seven-stage protocol proposed here converts tacit decisions into a repeatable sequence and adds provenance status to the earlier taxonomic workflow. Written procedures, stable supervision, specialist partnerships and periodic record-level audits are necessary if teaching participation is to reinforce rather than fragment scientific documentation.

This study has five principal limitations. First, the 106-row inventory snapshot is an auditable subset and should not be treated as a complete census of the 630 catalogued fossil items reported by the institutional diagnostic [10]. Second, the audit evaluates documentary content and internal consistency without re-examining specimens or validating individual taxonomic determinations. Third, the operational protocol application is a metadata-level test. It demonstrates how records can be triaged, preserved and referred, but it does not demonstrate that any biological identification has been corrected or verified. Fourth, missing provenance or acquisition values in the spreadsheet may exist in other institutional sources, so the reported percentages describe this inventory snapshot rather than the total documentary archive. Fifth, the analyzed records use different levels of taxonomic and locational granularity, which prevents interpretation of the dataset as a measure of species richness, relative abundance or paleoecological structure. Future work should reconcile the inventory with the public repository and legacy documentation, recover provenance wherever documentary evidence exists, and submit selected specimens for specialist review when such collaboration becomes available. The present workflow is designed so that those later determinations can be added without erasing the existing documentary history.

5. Conclusions

This study evaluated taxonomic documentation and data quality in an auditable subset of the CPAN paleontological collection. Of 106 inventory records dated 2 September 2026, 105 contained a scientific-name entry and 104 contained a phylum value, but only 28 used genus-level or genus-style nomenclature. Forty-five records contained formal suprageneric taxa, 32 used descriptive or non-standard labels and one scientific-name field was blank. At least 12 records showed direct contradictions between populated broad taxonomic fields. At the same time, collector, collection date, acquisition date, origin and acquisition mode were absent from every audited row. These results describe the documentary snapshot rather than the taxonomic composition of the full collection, but they demonstrate that a high rate of populated fields can coexist with limited precision, missing provenance and internal inconsistency.

The originality of the case lies in connecting paleontological research usability to explicit management of uncertainty. Taxonomic and provenance data are not peripheral to museological documentation because they determine whether a fossil can be reliably retrieved, compared, reidentified and placed in geological context. The proposed seven-stage protocol records the broadest defensible rank, determination status, authorship, documentary source, provenance status, specialist referral and revision history. Its operational application to seven inventory cases showed that documentary quality control can distinguish internally coherent entries, review-required conflicts, descriptive labels, unresolved determinations and identifier anomalies without claiming to correct taxonomy. The Darwin Core crosswalk also provides a practical path toward interoperable representation of determinations, uncertainty, attribution and geological context. In a collection housed outside a Biology or Paleontology department, this approach separates curatorial responsibility from taxonomic authority while creating a traceable interface between them. The immediate next step is documentary reconciliation across the inventory, legacy records and public repository. Specialist specimen-based validation remains necessary for biological correction, but the protocol allows that future work to be incorporated without loss of provenance or revision history.