Following the tuna trail: Contrasting global catch estimates from FAO and RFMOs

Following the tuna trail: Contrasting global catch estimates from FAO and RFMOs Bastien Grasset1,2 Emmanuel Chassot3Julien Barde1,2 James Geehan4 Fabio Fiorellato4 Abstract This study compares annual catch statistics from the Global Tuna Atlas (GTA) and FAO FishStat (FS) marine capture datasets, with a focus on tuna and tuna-like species. The analysis first describes the differences between both dataset structures, then applies a harmonized mapping and filtering procedure to enable consistent inter-comparison. At the global scale, total catches from both datasets are highly consistent (differences < 1%), yet this apparent agreement conceals substantial variations at finer levels, particularly by species and fishing fleet. These discrepancies often compensate each other across years, producing an illusion of equivalence in aggregated time series. A regional focus on the Indian Ocean Tuna Commission (IOTC) management area confirms this pattern: while temporal trends are parallel overall, differences emerge for some species such as bigeye tuna and Albacore, often linked to specific fleets. In recent years (post-2014), several species show nearly identical values in both datasets, reflecting cases where one source adopts figures from the other when deemed more reliable. However, differences persist for certain taxa and, in some cases, where the underlying data flows or integration processes differ between the two datasets. This analysis highlights both the complementarity and the limitations of GTA and FS: GTA provides detailed fishing gear and fishing mode, while FS offers finer spatial resolution and broader taxonomic coverage. Understanding these structural and procedural differences is essential for ensuring the comparability and reproducibility of global fisheries statistics. Keywords: Global Tuna Atlas; FishStat; FAO; Fisheries statistics; Data harmonization; Open data; CWP standards; Tuna fisheries 1Institut de Recherche pour le Développement (IRD), 34203 Sète Cedex, France 2MARBEC, University of Montpellier, CNRS, Ifremer, IRD, Sète, France 3Indian Ocean Tuna Commission (IOTC) Secretariat, Victoria, Mahé, Seychelles 4Food and Agriculture Organization of the United Nations (FAO), 00153 Rome, Italy 1 Following the tuna trail: Contrasting global catch estimates from FAO and RFMOs IOTC-2025-WPDCS21-17_Rev1 | DOI: 10.5281/zenodo.17719667 Introduction Reliable and comprehensive catch statistics are fundamental for monitoring fisheries, assessing stocks, and supporting international policy and management decisions. However, uncertainties in catch data remain a major challenge in global fisheries statistics. These uncertainties arise from a range of factors, including incomplete or inconsistent data collection, limited monitoring capacity, reporting errors, differences in national data compilation and submission procedures, and the varying levels of aggregation and validation applied during data processing. At the global scale, fishery catches are generally considered to be widely underestimated (Clarke et al. 2006, Agnew et al. 2009, Pauly & Zeller 2016), although the extent of this underestimation remains difficult to quantify (Yimin Ye 2017). The Food and Agriculture Organization of the United Nations (FAO) and the tuna Regional Fisheries Management Organizations (t-RFMOs) represent two of the main institutional frameworks responsible for compiling and disseminating global and regional fishery statistics. Both rely on data originally collected by national authorities, yet they follow distinct workflows for validation, standardization, and dissemination. Comparing their outputs provides an opportunity to better understand how these processes influence the resulting estimates of catch, and to assess the magnitude and nature of uncertainties embedded in global fishery statistics. In this study, we compare the FAO Global Capture Production Database and the nominal catch dataset compiled through the FIRMS Global Tuna, with a particular focus on tuna and tuna-like species caught in the Indian Ocean. By examining differences between these two datasetseach compiled through independent but interrelated reporting mechanismswe aim to highlight the potential sources of divergence and provide insights into the consistency and reliability of global tuna catch statistics. Materials Both FishStat and the GTA have been developed by FAO to disseminate fisheries data, but they differ greatly in scope and design. Each encompasses multiple datasets and analytical tools, and both contribute to FAO’s global fisheries data governance in complementary ways. The FAO Global Capture Production Database The FAO maintains the Global Capture Production Database (FAO 2025), which provides harmonized statistics on marine and inland capture fisheries production at the global level. The database is compiled annually by the FAO Statistics Division from information submitted by Member States and regional fishery bodies through the National Statistical Questionnaire 1 (NS1) survey. Reported data include annual catch quantities (in tonnes live weight) by country or territory, species (FAO ASFIS list), major fishing area (FAO statistical areas), and production source (marine or inland waters). Following submission, FAO performs quality control, validation, and standardization procedures to ensure internal consistency and comparability across countries and years. As of today, the FAO’s FishStat data, including the global and regional capture production data, are disseminated through a desktop client application (FishStatJ)1as well as through a CSV dataset accessible from the FAO Fisheries and Aquaculture Statistics online portal as a zipped file2. FS constitutes the official global reference for capture fishery production statistics. The version used in this study corresponds to the dataset available from March 2025. The FS dataset is compact, with only seven columns (Table X). It focuses on annual national statistics of capture production. In FS, information on the type of measurement is implicit: all records correspond to nominal catches (“catch”), without discards, and are expressed in metric tonnes (“t”) - the same unit used in the GTA dataset - and in number for a reduced set of species. In GTA, these details are explicitly defined through the columns measurement,measurement_type, and measurement_unit. For some groups, such as tunas and billfishes, FAO integrates best scientific estimates provided by regional tuna commissions. 1Available at: https://www.fao.org/fishery/en/topic/166235/en 2Available at: https://www.fao.org/fishery/static/Data/Capture_2025.1.0.zip 2 Following the tuna trail: Contrasting global catch estimates from FAO and RFMOs IOTC-2025-WPDCS21-17_Rev1 | DOI: 10.5281/zenodo.17719667 The FIRMS Global Tuna Atlas (GTA) The GTA is a global repository of harmonized nominal and geo-referenced catch datasets from tuna and tuna-like fisheries, developed under the auspices of the Fisheries and Resources Monitoring System (FIRMS) of the FAO (FAOFIRMS 2025). The GTA compiles, standardizes, and disseminates public-domain catch data submitted by the five tuna Regional Fisheries Management Organizations (t-RFMOs) the Commission for the Conservation of Southern Bluefin Tuna (CCSBT), the Inter-American Tropical Tuna Commission (IATTC), the International Commission for the Conservation of Atlantic Tunas (ICCAT), the Indian Ocean Tuna Commission (IOTC), and the Western and Central Pacific Fisheries Commission (WCPFC, in collaboration with the Pacific Community, SPC). The GTA was developed through a coordinated process between the FIRMS Secretariat and the t-RFMOs to establish a common data exchange format consistent with the standards of the Coordinating Working Party on Fishery Statistics (CWP) (FAO Coordinating Working Party on Fishery Statistics (CWP) 2025a). This process enables the systematic integration and annual update of harmonized nominal catch data from all t-RFMOs into a single global dataset. The GTA currently disseminates catch data for tuna and tuna-like species, including principal market tunas, billfishes, coastal or neritic tunas, and associated species such as bonitos, mackerels, and pelagic sharks. All GTA datasets are openly accessible on Zenodo with complete metadata and transparent versioning, and are outputs of an R data generation workflow and inputs for Shiny applications for visualization and analysis3. The nominal catch dataset component of the GTA, covering the period 1918-2023, is publicly available from the Zenodo repository4. The full processing workflow is also openly available, with a DOI for the source code5. The GTA is an open, reproducible data system focused on tuna and tuna-like species. The GTA dataset follows the CWP standards. The data structure is made of 16 columns categorizing catches by using multiple dimensions: species, fleet, gear, fishing_mode, area, and time (Table IX). Among these, four temporal columns time_start,year,month, and quarter deliberately provide redundant but complementary time references which facilitate filtering and aggregation at multiple temporal scales. The exchange format is available as JSON file from the Fisheries Data Interoperability Working Group GitHub Repository6. Methods Scope and datasets In this study, we focus only on one specific dataset type shared by both systems, the annual global capture data, to compare their structure, coverage, and consistency. Specifically, we use the datasets FishStat - Global Capture Production. (FAO 2025) and Global Tuna Atlas - Global nominal catches (FAO-FIRMS 2025), corresponding respectively to FAO’s global capture production statistics and the harmonized nominal catch data compiled from t-RFMOs. Only marine capture data were retained (excluding inland records), and the analysis focuses on ISSCAAP groups 36 (Tunas and tuna-like species) and 38 (Sharks, rays, chimaeras). All measurement_value entries represent nominal catches in metric tonnes (t). 3Accessible at: https://tunaatlaspiemapinseetuto.lab.dive.edito.eu/ 4DOI: https://doi.org/10.5281/zenodo.8034730 5DOI: https://zenodo.org/records/15312151 6Available at: https://github.com/fdiwg/fdi-formats/blob/main/cwp_rh_generic_gta_taskI.json 3 Following the tuna trail: Contrasting global catch estimates from FAO and RFMOs IOTC-2025-WPDCS21-17_Rev1 | DOI: 10.5281/zenodo.17719667 Global comparison of available dimensions The GTA dataset explicitly includes several descriptive dimensions such as gear_type, fishing_fleet, fishing_mode, and source_authority, consistent with CWP standards. These fields make it possible to group, filter, or compare data by gear, fleet, or reporting authority. In contrast, FS provides a more aggregated representation: it does not include specific gear or fleet categories, and the reporting entity corresponds to the national authority. However, some missing dimensions can be inferred indirectly - for example, the tRFMO can generally be deduced from the geographic area code (i.e., FAO Major Fishing area). This correspondence is not always unique in the Pacific Ocean, where RFMO boundaries overlap. Table I summarizes the correspondence between key dimensions in the two datasets. Table I: Summary of key structural differences between FS and GTA datasets Dimension FS GTA Key differences Species and group of species 337 65 FS has broader taxonomic coverage; GTA focuses on tuna and tuna-like species. Fishing Fleet 169 (country/territory level) 158 (aggregated categories) FS distinguishes individual territories; GTA merges some under national or RFMO entities. Gear Type Not specified Reported Enables stratification by fishing practice. Fishing Mode Not specified Reported Adds contextual information (e.g. free school vs associated school). Source Authority National authority RFMO or source agency Improves traceability and attribution of data. Area / Region 17 FAO subregions 10 RFMO-based management areas Partial overlap between spatial frameworks (see Table XIII). Measurement Type Nominal Landings (NL) only Nominal Catch + NL (+ few discards) Combined as equivalent for analysis. Temporal Range 1950-2023 1918-2023 Analyses restricted to 1950-2023, the common period covered by both datasets. Taxonomic coverage FishStat includes 337 species, whereas the GTA dataset lists only 65 (Table: I). GTA deliberately focuses on a restricted set of 32 key species, considered the most relevant for tuna and tuna-like fisheries. For the remainder of this analysis, only this 32 species included in both FS and GTA are considered, with specific focus given to the main tuna species in the following sections. (See Appendix XI). Catches of species not present in the GTA, but present in FS, represent 28% of total catches, for over 250 additional species. Fleet coverage Fishing fleet categories are slightly more numerous in FishStat (169) than in GTA (158), reflecting a slightly higher reporting granularity (Table: I). FishStat distinguishes individual countries and territories, while GTA aggregates certain entities under broader national or regional categories. This difference mainly reflects variations in national reporting practices and data compilation rules. Geographic coverage The GTA defines 10 RFMO-based management regions, while FS subdivides the world into 17 FAO Major Fishing Areas and subregions. GTA thus relies on large, RFMO-oriented areas or aggregated FAO regions, whereas FS follows a finer 4 Following the tuna trail: Contrasting global catch estimates from FAO and RFMOs IOTC-2025-WPDCS21-17_Rev1 | DOI: 10.5281/zenodo.17719667 spatial framework defined by FAO statistical divisions. As a result, spatial identifiers between the two datasets only partially overlap. Table XIII, in appendix summarizes the correspondence between GTA management regions and FAO fishing areas by ocean basin, illustrating the partial overlap between the two spatial frameworks. Exact correspondences exist only for the Mediterranean and Black Sea (MD) and the Indian Ocean (IOTC_WEST and IOTC_EAST). In contrast, the Atlantic and Pacific basins are divided into multiple FAO subregions in FS, making direct mapping with GTA’s broader RFMO areas more complex. Antarctic areas, which have no direct equivalent in GTA, include marginal species in FS (e.g. Porbeagle,Rays,Mantas nei), while in GTA, only Southern Bluefin Tuna (SBF) partially overlaps these regions under the CCSBT mandate. Measurement type In the GTA dataset, the variable measurement_type includes both Nominal Catch (NC) and Nominal Landings (NL), with a few occasional records of Dead Discards (DD) and Discarded Live (DL) following CWP standards (FAO Coordinating Working Party on Fishery Statistics (CWP) 2024, 2025b). In contrast, FishStat reports only Nominal Landings (NL). According to the CWP Handbook of Fishery Statistical Standards, Nominal Catch represents “nominal landings plus the component of the catch discarded dead, and post-release mortality of fish discarded alive”, approximating the total biomass removed. Despite this conceptual difference, the sum of Nominal Catch and Nominal Landings in GTA are generally of the same order of magnitude as Nominal Landings in FS. Although Nominal Catch theoretically includes dead discards and postrelease mortalities, it is not possible to quantify how much these components contribute in practice. For this reason, both NC and NL values are combined in the global analysis and compared directly with FS NL data, as a pragmatic approximation pending more detailed metadata on discard estimates. Temporal coverage The time series of FS catch data start in 1950, whereas GTA catch data begin in 1918, thus providing 32 additional years of historical coverage (Table I). Note that dates prior to 1950 are not available for all ocean basins in GTA. Harmonization and mapping process Mapping of fishing fleets Establishing a sound correspondence between the categories fishing_fleet, country, etc. is a complex task: FS offers a much finer level of detail in terms of capture production quantities at country or territorial entity level, and does not aggregate certain entities (e.g., Jersey, or Zanzibar), unlike GTA. In addition, geopolitical developments over the past years have made mapping difficult. We therefore propose a provisional mapping (see Table XIV) for not specific countries/territories, that enables us to analyze certain trends between the data sets, without claiming to be exhaustive for each country. Mapping of geographic areas Given the differences highlighted earlier, spatial comparisons between GTA and FS are conducted at the major oceanbasin level (Atlantic, Indian, and Pacific), rather than by FAO subregion. Antarctic areas, which concern only marginal species (Porbeagle and Rays, stingrays, mantas nei in FS, and Southern bluefin tuna in GTA, absent from FS), were excluded to maintain analytical consistency. The goal of this harmonization is not to reproduce fine-scale geographic details but to identify broad inter-basin patterns in reported catches. Filtering of comparable strata We compared the presence and absence of the 32 retained species across all ocean basins. This comparison reveals that only two species are present in the GTA dataset but absent from FishStat in one or more basins, whereas the opposite situation is much more frequent: 21 species reported by FishStat are not represented in GTA for several 5 Following the tuna trail: Contrasting global catch estimates from FAO and RFMOs IOTC-2025-WPDCS21-17_Rev1 | DOI: 10.5281/zenodo.17719667 Other Common in both datasets Atlantic Ocean Indian Ocean Mediterranean and Black Sea Pacific Ocean Albacore Bigeye tuna Blue marlin Blue shark Devil fish Porbeagle Silky shark Skipjack tuna Swordfish Thresher sharks nei Yellowfin tuna Bigeye thresher Black marlin Bullet tuna Frigate and bullet tunas Frigate tuna Giant manta Great hammerhead Indo−Pacific king mackerel Indo−Pacific sailfish Kawakawa Longfin mako Longtail tuna Narrow−barred Spanish mackerel Oceanic whitetip shark Pelagic thresher Scalloped hammerhead Shortbill spearfish Shortfin mako Smooth hammerhead Streaked seerfish Striped marlin Wahoo Ocean Species Dataset presence Both: FS > GTA Both: GTA > FS FS only GTA only None Figure 1: Presence and absence of the main species and group of species for each ocean 6 Following the tuna trail: Contrasting global catch estimates from FAO and RFMOs IOTC-2025-WPDCS21-17_Rev1 | DOI: 10.5281/zenodo.17719667 basin-species combinations (Figure: 1). To ensure consistent coverage across basins, only species-ocean pairs present in both datasets were retained for the subsequent analyses. However, this assessment should not be interpreted as a direct measure of completeness. Differences in taxonomic mapping, aggregation, or naming conventions between the two datasets, particularly for sharks and other elasmobranchs, can explain apparent absences. In many cases, catches may simply be recorded under broader nei or synonym categories rather than truly missing. As an illustration, a specific case of data aggregation was encountered with Frigate and Bullet tunas, which were reported jointly for many years in FishStat under the combined FAO code ‘FRZ’. We made an exception to retain this aggregated category in our comparison given its relevance for major tuna species and its direct equivalency in the GTA dataset. However, for the ‘Mediterranean and Black Sea’ basin, data for both individual species and the ‘FRZ’ group were removed, as only bullet tuna is represented in GTA dataset for that area. For other broadly aggregated taxa such as “Bonitos nei”, “Tunas nei”, or “Rays, stingrays, mantas nei”, the corresponding catches were excluded to ensure consistency at the species level. However, it is important to note that these excluded categories represent substantial catch volumes for tunas and sharks, as illustrated in Table XII. The goal of this comparison is to illustrate the broader coverage of FishStat and to justify the application of a harmonized filtering by area and species in the subsequent analyses, rather than to infer which dataset is superior. We choose not to include the fishing_fleet dimension in this filtering step, because it is not always possible to verify that fleets are not reported under ‘Other nei’ (Not Elsewhere Included) in one of the datasets. For species, we therefore remove all ‘UNK’ categories to avoid ambiguities. Global comparison of captures In the following subsections we detail the results obtained by comparing the remaining rows of the initial FS and GTA datasets, after applying the previous filters (see explanations in previous section). Overall differences and total values Table II: Total catch (t) for each dataset and relative differences FishStat GTA Difference Difference (in %) 230,241,024 228,485,148 -1,755,876 -0.76 When considering total global captures over the entire time series from 1950 to 2023, the two datasets show remarkably close aggregated values, with less than 1% difference between them. (Table II). Such apparent similarity suggests that, at a broad scale, both FishStat and GTA provide coherent global estimates of total catches over the studied period. However, this global agreement can mask substantial differences when exploring specific dimensions, such as species composition or fishing fleet contributions, which may compensate each other when aggregated. 7 Following the tuna trail: Contrasting global catch estimates from FAO and RFMOs IOTC-2025-WPDCS21-17_Rev1 | DOI: 10.5281/zenodo.17719667 Temporal evolution of global captures 1960 1980 2000 2020 0 2000 4000 6000 Total catches (x 1,000 t) Dataset FishStat GTA Figure 2: Comparison of annual time series of catch (t) of tuna and tuna-like species in the Indian Ocean between for FishStat and GTA datasets for the period 1950-2023 The temporal trends in total catches (Figure: 2) confirm this overall consistency. Both datasets exhibit parallel trajectories, however, this apparent alignment does not imply full equivalence. In some years, FishStat reports higher values, while in others GTA exceeds it, suggesting that small interannual compensations between datasets smooth out when aggregated. These offsetting variations contribute to the impression of overall agreement, even though substantial compositional differences may persist beneath the global totals. 8 Following the tuna trail: Contrasting global catch estimates from FAO and RFMOs IOTC-2025-WPDCS21-17_Rev1 | DOI: 10.5281/zenodo.17719667 Breakdown by species and fleet Table III: Comparison of total catch (t) by taxon between the Global Tuna Atlas and Fishstat database for the 32 retained species and one group of species Loss / Gain Per species Values dataset 1 (FS) Values dataset 2 (GTA) Difference (in %) Difference in value Gain Frigate tuna 1,659,080 3,456,781 108.36 1,797,700 Yellowfin tuna 61,818,850 63,410,344 2.57 1,591,494 Bigeye tuna 19,860,483 21,322,080 7.36 1,461,597 Bullet tuna 323,456 613,264 89.60 289,808 Narrow-barred Spanish mackerel 4,981,372 5,193,418 4.26 212,046 Longtail tuna 3,975,399 4,078,920 2.60 103,521 Shortfin mako 194,601 285,833 46.88 91,232 Wahoo 131,757 181,323 37.62 49,566 Black marlin 656,581 699,705 6.57 43,124 Porbeagle 115,092 152,977 32.92 37,885 Shortbill spearfish 43,198 58,379 35.14 15,181 Smooth hammerhead 3,345 10,158 203.71 6,814 Devil fish 749 6,371 750.76 5,623 Oceanic whitetip shark 17,841 21,925 22.89 4,085 Scalloped hammerhead 5,776 9,760 68.98 3,984 Longfin mako 907 4,839 433.67 3,932 Giant manta 166 2,533 1,424.21 2,367 Pelagic thresher 6,738 7,904 17.31 1,166 Loss Frigate and bullet tunas 3,272,143 0 -100.00 -3,272,143 Skipjack tuna 100,198,153 98,364,215 -1.83 -1,833,938 Swordfish 5,015,175 4,353,926 -13.18 -661,249 Kawakawa 4,288,368 3,856,577 -10.07 -431,791 Indo-Pacific sailfish 1,181,350 869,782 -26.37 -311,567 Blue shark 2,605,833 2,360,481 -9.42 -245,352 Striped marlin 1,004,918 764,201 -23.95 -240,716 Thresher sharks nei 237,338 26,197 -88.96 -211,141 Blue marlin 2,066,180 1,891,178 -8.47 -175,002 Silky shark 329,687 261,903 -20.56 -67,785 Albacore 14,704,002 14,686,461 -0.12 -17,541 Streaked seerfish 14,857 10,511 -29.25 -4,346 Indo-Pacific king mackerel 1,524,961 1,520,887 -0.27 -4,074 Great hammerhead 488 207 -57.61 -281 Bigeye thresher 2,181 2,106 -3.45 -75 A more detailed examination by species reveals significant discrepancies between the two datasets (Table III). The analysis shows marked deficits in GTA for certain tunas and mackerels, with kawakawa (-10.1%) and Indo-Pacific sailfish (-26.4%) displaying substantially lower volumes compared to FishStat. These differences are largely explained by our 9 Following the tuna trail: Contrasting global catch estimates from FAO and RFMOs IOTC-2025-WPDCS21-17_Rev1 | DOI: 10.5281/zenodo.17719667 A) Skipjack tuna 1960 1980 2000 2020 0 200 400 600 Total catches (x 1,000 t) B) Albacore 1960 1980 2000 2020 0 10 20 30 40 50 Total catches (x 1,000 t) C) Bigeye tuna 1960 1980 2000 2020 0 50 100 150 Total catches (x 1,000 t) D) Yellowfin tuna 1960 1980 2000 2020 0 200 400 Total catches (x 1,000 t) E) Swordfish 1960 1980 2000 2020 0 10 20 30 40 Total catches (x 1,000 t) Figure 4: Evolutions of values for the dimension year and differences by ocean for 5 major species between FS (red) and GTA (Blue) 16 Following the tuna trail: Contrasting global catch estimates from FAO and RFMOs IOTC-2025-WPDCS21-17_Rev1 | DOI: 10.5281/zenodo.17719667 Pronounced discrepancies are observed for albacore and bigeye tuna, particularly in the western Indian Ocean, where GTA reports systematically higher values (see Figure: 4). For bigeye tuna, these differences are especially pronounced at the ocean-basin level, suggesting that variations may stem from differences in data integration or national submissions rather than from temporal inconsistency. For yellowfin and skipjack tuna, the two datasets display almost identical trajectories over time, indicating a strong coherence in the reporting of the dominant commercial species. In contrast, swordfish shows moderate discrepancies, mostly during the 1990s and early 2000s, which may correspond to periods of partial data revision or differing treatment of gear-specific catches. Overall, while the general trends remain comparable across datasets, the magnitude and timing of reported catches for Albacore and bigeye highlight potential inconsistencies in how these species are aggregated or reported by ocean basin. Focus on bigeye tuna To better understand the origin of the discrepancies observed at the aggregate level, we focus on bigeye tuna, an illustrative example where the direction of the differences between GTA and FishStat reverses between ocean areas. While not the species with the largest discrepancies, it provides a relevant case study to explore patterns that could similarly be examined for other species. For bigeye tuna, the largest discrepancies occur in the Eastern Indian Ocean, where GTA reports substantially higher total catches than FishStat for several decades (Figure: 5). The fleet breakdown in appendix (Table XVI) shows that these differences are mainly associated with catches attributed to Indonesia, which are markedly lower in FishStat. This difference is mainly associated with the Eastern Indian Ocean which shows significantly high differences in percentage Table VIII). While part of this divergence could stem from the way catches are grouped (for example, through the use of the NEI category in GTA), such reclassification alone cannot explain the overall difference, since it would not affect total basin-level values. In contrast, in the Western Indian Ocean, the two time series show a much closer alignment, with consistent magnitudes and parallel temporal patterns across most years. However, when comparing the two basins, the situation appears more complex: the dataset that reports higher values changes depending on the ocean, and in the Eastern basin, the interannual dynamics also diverge markedly, especially after the early 2000s. As discussed previously, the case of Indonesia remains the most significant: its lower totals in GTA likely reflect historical revisions to Indian Ocean catch data that have been integrated into GTA but are not yet reflected in FishStat. Table VIII: Major differences break down by fishing_fleet_label between FS and GTA datasets, for bigeye tuna catches in Eastern Indian Ocean Loss / Gain Per fishing fleet Values dataset 1 (FS) Values dataset 2 (GTA) Difference (in %) Difference in value Gain Other nei 49,399 193,575 291.86 144,176 Taiwan Province of China 269,273 325,559 20.90 56,286 Republic of Korea 69,094 74,134 7.29 5,039 China 21,532 22,134 2.79 601 United Republic of Tanzania 180 696 285.76 515 Others 98,163 98,216 0.05 53 Loss 17 Following the tuna trail: Contrasting global catch estimates from FAO and RFMOs IOTC-2025-WPDCS21-17_Rev1 | DOI: 10.5281/zenodo.17719667 Table VIII: Major differences break down by fishing_fleet_label between FS and GTA datasets, for bigeye tuna catches in Eastern Indian Ocean Loss / Gain Per fishing fleet Values dataset 1 (FS) Values dataset 2 (GTA) Difference (in %) Difference in value Indonesia 841,626 488,922 -41.91 -352,704 Japan 354,899 352,132 -0.78 -2,767 Italy 323 5 -98.36 -318 Thailand 1,498 1,328 -11.33 -170 Seychelles 6,827 6,757 -1.03 -70 Others 27,964 27,911 -0.19 -53 To better assess the alignment between datasets, a new comparison was performed excluding the Indonesian data, as the associated differences are well understood and explained by historical reporting revisions. These discrepancies are expected to disappear in the next FishStat update; excluding them therefore allows a clearer examination of the remaining differences between the two datasets. Focus on bigeye tuna without Indonesia data When excluding the Indonesian data from the bigeye tuna series, the overall alignment between GTA and FishStat substantially improves, confirming that the observed discrepancy was largely driven by this fleet. Nevertheless, noticeable differences persist in the Western Indian Ocean, suggesting that additional factors, such as spatial aggregation, reporting updates and use of ‘Other nei’ data may still play a role (Figure: 6). The following results also suggest a progressive convergence between the two datasets in recent years. For several major tuna and tuna-like species -such as albacore, bigeye, and to some extent skipjack -the (see Figure: 4)) values reported by FishStat and GTA become nearly identical after 2014, indicating that both datasets may increasingly rely on similar or shared data sources. To verify whether this convergence is systematic or species-specific, the following section focuses on the most recent years of the time series. By comparing the post-2014 period across all major tuna species, we aim to determine whether the observed alignment reflects a broader harmonization of data flows or remains limited to certain taxa or ocean basins. Post-2014 alignment of major species We restrict the analysis to data from 2014 onwards, since this is where the two series visually converge for all major species. On top of that, for all species the differences are lower from this year (See Appendix: Figure: 8) Many similarities are observed between 2014 and 2020. We now investigate, for the species-year combinations with small discrepancies, whether differences still remain at the fishing_fleet_label level. Country-specific behaviour for Indian Ocean Although yellowfin tuna and swordfish do not appear to show a clear convergence between FishStat and GTA in the most recent years, a closer examination reveals that for many countries, the reported values are in fact very similar, sometimes even identical, as they are for the other species (see Figure: 7). At the country level, three broad patterns can be distinguished. 1. Countries with stable agreement: for several reporting States (e.g. Republic of Korea, Mauritius, or Madagascar), the correspondence between the two datasets remains strong and constant over time. 18 Following the tuna trail: Contrasting global catch estimates from FAO and RFMOs IOTC-2025-WPDCS21-17_Rev1 | DOI: 10.5281/zenodo.17719667 1960 1980 2000 2020 0 25 50 75 100 Total catches (x 1,000 t) Dataset FishStat GTA a) Western Indian Ocean 1960 1980 2000 2020 0 20 40 60 Total catches (x 1,000 t) Dataset FishStat GTA b) Eastern Indian Ocean Figure 5: Comparison of Western and Eastern Indian Ocean for bigeye tuna catches 1960 1980 2000 2020 0 25 50 75 100 Total catches (x 1,000 t) Dataset FishStat GTA a) Western Indian Ocean 1960 1980 2000 2020 0 10 20 30 40 Total catches (x 1,000 t) Dataset FishStat GTA b) Eastern Indian Ocean Figure 6: Comparison of Western and Eastern Indian Ocean for bigeye tuna catches, without Indonesian data 19 Following the tuna trail: Contrasting global catch estimates from FAO and RFMOs IOTC-2025-WPDCS21-17_Rev1 | DOI: 10.5281/zenodo.17719667 Vanuatu Yemen Taiwan Province of China Thailand Timor−Leste United Kingdom of Great Britain and Northern Ireland United Republic of Tanzania Republic of Korea Seychelles South Africa Spain Sri Lanka Oman Other nei Pakistan Philippines Portugal Madagascar Malaysia Maldives Mauritius Mozambique Iran (Islamic Republic of) Italy Japan Jordan Kenya Djibouti Egypt France India Indonesia Australia Bangladesh Belize China Comoros 2015 2020 2015 2020 2015 2020 2015 2020 2015 2020 Albacore Bigeye tuna Skipjack tuna Swordfish Yellowfin tuna Albacore Bigeye tuna Skipjack tuna Swordfish Yellowfin tuna Albacore Bigeye tuna Skipjack tuna Swordfish Yellowfin tuna Albacore Bigeye tuna Skipjack tuna Swordfish Yellowfin tuna Albacore Bigeye tuna Skipjack tuna Swordfish Yellowfin tuna Albacore Bigeye tuna Skipjack tuna Swordfish Yellowfin tuna Albacore Bigeye tuna Skipjack tuna Swordfish Yellowfin tuna Albacore Bigeye tuna Skipjack tuna Swordfish Yellowfin tuna Species Category <0.1% <1% >1% Exact Figure 7: Country-specific differences in reported catches (GTA vs FishStat, 2012-2023) 20 Following the tuna trail: Contrasting global catch estimates from FAO and RFMOs IOTC-2025-WPDCS21-17_Rev1 | DOI: 10.5281/zenodo.17719667 2. Countries with persistent differences: some countries, such as South Africa or Jordan, show systematic deviations between datasets, suggesting enduring discrepancies in reporting, conversion, or aggregation practices. 3. Countries with variable alignment: in a few cases (e.g. France, Spain, Great Britain), the relationship between datasets changes from year to year, showing periods of agreement followed by sharp divergences. These alternating patterns are the most challenging to interpret, as they may result from changes in data structuring, aggregation rules, estimation procedures, or even shifts in the data source used by one of the systems. 4. Non-Contracting Parties with data integration lags: The case of Djibouti exemplifies how asynchronous update cycles and data reallocation can create significant temporary discrepancies. While its data is ultimately sourced from FAO by the IOTC, a recent update in FishStat (where some of the catch previously reported as Yellowfin Tuna (YFT) was reallocated to the generic “Tunas nei” (TUN) category) had not yet been reflected in the GTA at the time of analysis. This highlights how taxonomic reclassifications in one dataset can create apparent discrepancies before synchronization occurs. In contrast, data for other non-Contracting Parties like Egypt shows near-perfect alignment, demonstrating that synchronization is maintained when no recent updates have been made to the underlying records in either dataset. Overall, this analysis confirms that while convergence between GTA and FishStat is evident for many country-species combinations, differences remain and are not uniformly distributed. Understanding for each, whether these variations stem from harmonization updates, national resubmissions, or methodological differences in data integration will require a detailed comparison of the underlying reporting flows. Verification example: identical strata To conclude, a final verification was carried out on a single year-species pair to confirm whether identical values between datasets correspond to complete equivalence across all dimensions. We selected albacore in 2015, a representative case where total catches are identical in FishStat and GTA. For this year and species, the data are fully identical between FishStat and GTA, differing only by rounding errors (Appendix: Table XVII). This confirms that in certain strata, both datasets rely on exactly the same source data and transformations, reinforcing the assumption of partial convergence observed in the most recent years. Synthesis and discussion Historical analyses have already highlighted persistent inconsistencies between FAO and RFMO tuna statistics. Garibaldi and Kebe (Garibaldi & Kebe 2005) were the first to document discrepancies between FAO and ICCAT tuna catch statistics in the Mediterranean. These differences were later confirmed at a broader scale by Justel-Rubio et al. (Justel-Rubio et al. 2016), who compared FAO and tuna RFMO datasets globally and showed that such inconsistencies persisted across regions and species despite ongoing harmonization efforts. Although the overall difference between FAO and RFMO datasets was estimated at less than 1% globally, differences exceeding 10% were found for several species or ocean areas. The main causes identified included variations in spatial delineation, flag attribution, the reporting of some fisheries to FAO only or present only in tRFMOs data. These findings mirror the patterns observed in the present study, where discrepancies between FishStat and the Global Tuna Atlas are generally small in aggregate but can reach higher levels when broken down by specific taxa or basins. Together, these analyses reinforce that such divergences largely stem from structural and procedural differences in reporting and harmonization, rather than from contradictory underlying data. It is important to note that the FAO FishStat dataset does not represent a fully independent source from the RFMOs. A substantial part of the data originates from the same regional reporting systems (IOTC, ICCAT, WCPFC, IATTC, CCSBT), complemented by national submissions and FAO adjustments to fill gaps or ensure consistency. The opposite is also true: in some cases, RFMO statistics may directly draw on FishStat estimates for non-reporting members. This interdependence offers a promising pathway for enhancing the GTA. By systematically integrating new FishStat data as soon as it is published, the GTA could achieve more frequent updates and more comprehensive coverage, particularly for non-contracting parties. This approach could help bridge gaps caused by the delayed integration of national data into RFMO databases, or by the inclusion of catches from non-tuna fisheries. Therefore, many differences between 21 Following the tuna trail: Contrasting global catch estimates from FAO and RFMOs IOTC-2025-WPDCS21-17_Rev1 | DOI: 10.5281/zenodo.17719667 the GTA and FishStat reflect not fundamental inconsistencies, but rather different timelines within the same data harmonization process. The FAO performs additional aggregation, validation, and estimation steps that may correct for late reporting or fill missing values, leading to slightly higher totals in some cases. This aligns with findings by Heidrich et al. (2023) ((Heidrich et al. 2023)), who demonstrated that IOTC data under-represent total pelagic catches by about 30 %, suggesting that part of the discrepancies observed between GTA and FishStat could stem from incomplete reporting at the RFMO level. Interestingly, this pattern appears to vary by taxonomic group: for major species, the GTA values are mainly higher than those in FishStat, reflecting the fact that RFMOs like IOTC tend to maintain more up-to-date and comprehensive statistics for their primary target stocks. Conversely, secondary taxa and bycatch groups are often higher in FishStat, possibly due to FAO-level adjustments or reconstructions compensating for the limited coverage of these species in RFMO datasets. This may also reflect the inclusion of catches from non-tuna fisheries reported under broader categories. Overall, discrepancies between FishStat and RFMO-based products such as the GTA have tended to decrease since 2014, following FAO’s efforts to enhance alignment with RFMO data. However, historical differences persist, as countries seldom revise older submissions when updating their national reports to tRFMO. Residual mismatches are also partly explained by differences in species or fleet mappings between FAO and RFMO classification systems, and by the use of approximate or overlapping spatial areas that are not handled in a consistent way across datasets. In addition, the conceptual distinction between nominal landing in FishStat and nominal catches in GTA may introduce minor biases when comparing aggregated totals, as the two variables are not strictly equivalent. This inconsistency cannot be fully resolved at present, but it highlights the need for continued clarification and harmonization of definitions and reporting practices across global tuna datasets. Acknowledgments This work has received funding from the European Union’s Horizon Europe research and innovation programme under the Blue-Cloud 2026 project (Grant agreement No 101094227). How to cite this document Grasset, B.,Chassot, E.,Barde, J.,Geehan, J.,Fiorellato, F. (2025). Following the tuna trail: Contrasting global catch estimates from FAO and RFMOs. IOTC-2025-WPDCS21-17_Rev1.\ DOI: https://doi.org/10.5281/zenodo.17719667 22 Following the tuna trail: Contrasting global catch estimates from FAO and RFMOs IOTC-2025-WPDCS21-17_Rev1 | DOI: 10.5281/zenodo.17719667 Bibliography Agnew DJ, Pearce J, Pramod G, Peatman T, Watson R, Beddington JR, Pitcher TJ (2009) Estimating the worldwide extent of illegal fishing. PLoS ONE. doi:10.1371/journal.pone.0004570 Clarke SC, McAllister MK, Milner-Gulland EJ, Kirkwood GP, Michielsens CGJ, Agnew DJ, Pikitch EK, Nakano H, Shivji MS (2006) Global estimates of shark catches using trade records from commercial markets. Ecology Letters. doi:10.1111/j.1461-0248.2006.00968.x FAO (2025) Global capture production. FAO Coordinating Working Party on Fishery Statistics (CWP) (2024) CWP global code list - catch concepts (cl_catch_concepts.csv). FAO Coordinating Working Party on Fishery Statistics (CWP) (2025b) Catch and landings - CWP handbook of fishery statistical standards. FAO Coordinating Working Party on Fishery Statistics (CWP) (2025a) CWP handbook of fishery statistical standards. FAO-FIRMS (2025) Global tuna atlas - global nominal catches (1950-2023). Garibaldi L, Kebe P (2005) Discrepancies between the FAO and ICCAT databases for tuna catches in the mediterranean. Collect Vol Sci Pap ICCAT Heidrich KN, Meeuwig JJ, Zeller D (2023) Reconstructing past fisheries catches for large pelagic species in the indian ocean. Frontiers in Marine Science 10:1177872. doi:10.3389/fmars.2023.1177872 Indian Ocean Tuna Commission (IOTC) (2013) Resolution 13/03 on the recording of catch and effort data by fishing vessels in the IOTC area of competence. Justel-Rubio A, Garibaldi L, Hampton J, Maunder M (2016) A comparative study of annual tuna catches from two different sources: FAO global capture database vs tuna RFMOs statistical databases (2000–2014). Technical Report, ISSF Technical Report 2016-15. International Seafood Sustainability Foundation (ISSF), Washington, D.C., USA Pauly D, Zeller D (2016) Catch reconstructions reveal that global marine fisheries catches are higher than reported and declining. Nature communications. doi:10.1038/ncomms10244 Yimin Ye MB Manuel Barange (2017) FAO’s statistical databases and the sustainability of fisheries and aquaculture: Comments on pauly and zeller (2017). Marine Policy. doi:10.1016/j.marpol.2017.03.013 23 Following the tuna trail: Contrasting global catch estimates from FAO and RFMOs IOTC-2025-WPDCS21-17_Rev1 | DOI: 10.5281/zenodo.17719667 List of Figures 1 Presence and absence of the main species and group of species for each ocean . . . . . . . . . . . . 6 2 Comparison of annual time series of catch (t) of tuna and tuna-like species in the Indian Ocean between for FishStat and GTA datasets for the period 1950-2023 . . . . . . . . . . . . . . . . . . . . . 8 3 Comparison of the annual time series of catch (t) of tuna and tuna-like species in the Indian Ocean between for FishStat and GTA datasets for the period 1950-2023 . . . . . . . . . . . . . . . . . . . . 13 4 Evolutions of values for the dimension year and differences by ocean for 5 major species between FS (red)andGTA(Blue)............................................ 16 5 Comparison of Western and Eastern Indian Ocean for bigeye tuna catches . . . . . . . . . . . . . . . 19 6 Comparison of Western and Eastern Indian Ocean for bigeye tuna catches, without Indonesian data . 19 7 Country-specific differences in reported catches (GTA vs FishStat, 2012-2023) . . . . . . . . . . . . . 20 8 Comparison of species-to-species differences between GTA and FishStat datasets . . . . . . . . . . . 33 List of Tables I Summary of key structural differences between FS and GTA datasets . . . . . . . . . . . . . . . . . . 4 II Total catch (t) for each dataset and relative differences . . . . . . . . . . . . . . . . . . . . . . . . . 7 III Comparison of total catch (t) by taxon between the Global Tuna Atlas and Fishstat database for the 32 retained species and one group of species . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 IV Comparison of total catch (t) by Area between the Global Tuna Atlas and Fishstat database for a set ofselectedtaxa.............................................. 10 V Comparison of total catch (t) by fishing fleet between the Global Tuna Atlas and Fishstat database for asetofselectedtaxa ........................................... 10 VI Comparison of total catch (t) by species between the Global Tuna Atlas and Fishstat database for a set ofselectedtaxa.............................................. 14 VII Total captures (t) for each dataset and relative differences, for datasets filtered on major species . . . 15 VIII Major differences break down by fishing_fleet_label between FS and GTA datasets, for bigeye tuna catchesinEasternIndianOcean ..................................... 17 IX Extract of the Global Tuna Atlas dataset, with redundant temporal columns (year, month, quarter) omitted .................................................. 26 X FirstlinesoftheFSdataset ........................................ 27 XII Sample of aggregated species group removed for the comparative analysis between GTA and FS . . . 27 XI Species retained for the comparative analysis between FishStat and the Global Tuna Atlas (n = 32) . . 28 XIII Geographic correspondence between GTA management areas and FishStat (FAO) subregions . . . . 29 XV Number of dimensions for each dataset filtered on FAO areas 51 and 57, after mapping of fishing_fleet forFS ................................................... 29 XIV Mapping from FishStat to GTA fishing_fleet_label . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30 XVI Major differences break down by fishing_fleet between FS and GTA datasets, filtered on speciesoceanscommonspairs .......................................... 31 XVII Major differences break down by fishing_fleet between FishStat and GTA for albacore tuna data catchesforyear2015 ........................................... 32 XVIII Major differences break down by fishing_fleet between FishStat and GTA for bigeye tuna in western indianocean................................................ 32 XIX List of species codes, common names, and scientific names (IOTC focus). . . . . . . . . . . . . . . . . 34 24 Following the tuna trail: Contrasting global catch estimates from FAO and RFMOs IOTC-2025-WPDCS21-17_Rev1 | DOI: 10.5281/zenodo.17719667 Appendix 25 Following the tuna trail: Contrasting global catch estimates from FAO and RFMOs IOTC-2025-WPDCS21-17_Rev1 | DOI: 10.5281/zenodo.17719667 Table XVII: Major differences break down by fishing_fleet between FishStat and GTA for albacore tuna data catches for year 2015 Dimension Loss / Gain Precision Values dataset 1 Values dataset 2 Difference (in %) Difference in value fishing-fleetlabel Gain China 1,843 1,843 0.01 0 fishing-fleetlabel Loss South Africa 13 12 -12.31 -2 Table XVIII: Major differences break down by fishing_fleet between FishStat and GTA for bigeye tuna in western indian ocean Dimension Loss / Gain Precision Values dataset 1 Values dataset 2 Difference (in %) Difference in value fishing-fleetlabel Gain Japan 394,632 426,884 8.17 32,253 Republic of Korea 341,054 370,883 8.75 29,829 Other nei 194,355 217,995 12.16 23,640 Taiwan Province of China 855,139 869,303 1.66 14,164 Spain 396,205 398,271 0.52 2,066 Mozambique 4,592 6,172 34.39 1,579 Maldives 34,340 35,070 2.13 731 China 88,911 89,155 0.28 245 South Africa 3,901 4,142 6.19 241 Seychelles 255,162 255,330 0.07 169 Others 308,992 309,286 0.10 294 fishing-fleetlabel Loss Italy 7,812 3,040 -61.09 -4,772 32 Following the tuna trail: Contrasting global catch estimates from FAO and RFMOs IOTC-2025-WPDCS21-17_Rev1 | DOI: 10.5281/zenodo.17719667 Table XVIII: Major differences break down by fishing_fleet between FishStat and GTA for bigeye tuna in western indian ocean Dimension Loss / Gain Precision Values dataset 1 Values dataset 2 Difference (in %) Difference in value United Republic of Tanzania 9,858 6,760 -31.43 -3,098 Iran (Islamic Republic of) 33,747 31,688 -6.10 -2,059 India 19,955 19,781 -0.88 -175 Kenya 1,834 1,810 -1.34 -25 Madagascar 2,200 2,197 -0.17 -4 Philippines 16,187 16,187 0.00 0 0.00% 0.25% 0.50% 0.75% 1.00% 2000 2005 2010 2015 2020 Year Relative part (%) of year Relative difference <0.1% <1% >1% Exact Figure 8: Comparison of species-to-species differences between GTA and FishStat datasets 33 Following the tuna trail: Contrasting global catch estimates from FAO and RFMOs IOTC-2025-WPDCS21-17_Rev1 | DOI: 10.5281/zenodo.17719667 Table XIX: List of species codes, common names, and scientific names (IOTC focus). Code Common_name Scientific_name ALB Albacore Thunnus alalunga BET Bigeye tuna Thunnus obesus BLM Black marlin Istiompax indica BLT Bullet tuna Auxis rochei BUM Blue marlin Makaira nigricans COM Narrow-barred Spanish mackerel Scomberomorus commerson FRI Frigate tuna Auxis thazard GUT Indo-Pacific king mackerel Scomberomorus guttatus KAW Kawakawa Euthynnus affinis LOT Little tunny Euthynnus alletteratus MLS Striped marlin Kajikia audax SBF Southern bluefin tuna Thunnus maccoyii SFA Indo-Pacific sailfish Istiophorus platypterus SKJ Skipjack tuna Katsuwonus pelamis SWO Swordfish Xiphias gladius YFT Yellowfin tuna Thunnus albacares FRZ Frigate and bullet tuna Auxis thazard and Auxis rochei Albacore Bigeye tuna Black marlin Blue marlin Blue shark Bullet tuna Devil fish Frigate tuna Giant manta Great hammerhead Indo−Pacific king mackerel Indo−Pacific sailfish Kawakawa Longfin mako Longtail tuna Narrow−barred Spanish mackerel Oceanic whitetip shark Pelagic thresher Porbeagle Scalloped hammerhead Shortbill spearfish Shortfin mako Silky shark Skipjack tuna Smooth hammerhead Streaked seerfish Striped marlin Swordfish Thresher sharks nei Wahoo Yellowfin tuna 2000 2005 2010 2015 2020 Year Species Relative difference <0.1% <1% >1% Exact Dominant difference category by species a year ## [[1]] 34 Following the tuna trail: Contrasting global catch estimates from FAO and RFMOs IOTC-2025-WPDCS21-17_Rev1 | DOI: 10.5281/zenodo.17719667 t 1990 2000 2010 2020 0 50 100 150 year Values Dataset FishStat−Djibouti GTA−Djibouti 35