A Large-Scale AIS Dataset from Finnish Water
Debayan Bhattacharya, Ikram Ul Haq, Carlos Pichardo Vicencio, Sebastien Lafond
cs.LG
Quick verdict
Dataset paper releasing 20 months of AIS data from Finnish waters and lakes; regionally valuable but marred by internal count inconsistencies.
Summary
This paper introduces a freely available Automatic Identification System (AIS) dataset covering Finnish waters — the Baltic Sea and, unusually, Finnish lakes — over 20 months from April 2021 to December 2022. Data were collected via Fintraffic's public MQTT WebSocket API and released on Zenodo (DOI 10.5281/zenodo.8112336) as daily location and metadata CSV files whose fields and sentinel values are documented in detail. The abstract reports 229 million data points, while Section 3 states 2,293,129,345 — an unreconciled order-of-magnitude discrepancy. Alongside the dataset, the authors present a Web of Science literature review of 35 pre-2023 papers spanning nine maritime ML application domains, a survey of existing free and paid AIS datasets, ship-type distributions by unique MMSI and IMO numbers (cargo and tankers dominate at 51.5% and 22.3% of MMSIs), lake-specific statistics from a bounding-box subset, and seasonal Folium heatmaps showing frozen lakes in winter versus heavy summer traffic. Limitations: purely descriptive statistics, no data-quality audit (drop rates, duplicates, spoofing), inconsistent counts (35 vs 33 reviewed papers; percentages summing over 100%), broken table cross-references, and no downstream ML benchmark demonstrating the dataset's utility.
Review complete
Assessment at a glance
4
risk flags identified
● high 1
● medium 2
● low 1
5 strengths · 4 weaknesses · 5 author questions
AI text signal = Binoculars zero-shot detector score over 20 sampled passages (Fast-DetectGPT co-reported) — absence does not prove human authorship, and a flag does not imply misconduct.
Levels reflect verification complexity — not a quality score. Evolve never rates your paper.
Key claims under review
- The paper curates and categorises existing AIS datasets and releases a new free AIS dataset from Finnish waters collected via Fintraffic's public MQTT WebSocket API.
- The dataset spans 20 months (April 2021-December 2022) and contains 229 million data points per the abstract, or 2,293,129,345 per Section 3.
- It is uniquely distinguished by including AIS data from Finnish lakes, which prior free datasets do not cover.
- Detailed ship-type distributions (Tables 4-6) and seasonal traffic visualisations (Figs 4-7) make the resource a gateway for diverse maritime machine-learning applications.
Strengths
Genuine regional gap filled with a free release
20 months (Apr 2021-Dec 2022) of AIS covering the Baltic Sea plus Finnish lakes — coverage absent from the free datasets tabulated in Table 3 (US Coast Guard, Danish waters) — released with a DOI on Zenodo [40] as daily location/metadata CSV pairs.
Field-level documentation faithful to the AIS specification
Sections 3.1-3.2 enumerate all 12 location and 14 metadata fields with correct ITU sentinel values (sog=1023 unavailable, heading=511, timestamp 60-63 status codes, '@'-filled text fields), sparing downstream users classic AIS decoding errors.
Usable entry-point literature survey
Seasonal visualisations substantiate the frozen-water differentiator
Ship-type statistics at three granularities
Weaknesses
Headline dataset size is internally contradictory
Literature-review accounting does not add up
No data-quality audit and no ML benchmark
Broken cross-references and editorial sloppiness
Full Review Report
$1.99 · one-timeUnlock the complete preliminary analysis Evolve produced for this paper.
Report preview
Preliminary Review: A Large-Scale AIS Dataset from Finnish Water
The remaining sections cover Methodology Assessment, Statistical Evidence, Results & Reproducibility, Limitations & Risks, and Verdict & Recommendations.
- All 5 strengths & 4 weaknesses, analyzed in detail
- 4 risk flags with severity, section & reviewer questions
- Methods & statistical evidence assessment
- 5 questions a peer reviewer would ask
- Full six-dimension evaluation report & recommendations
One-time purchase · this paper only
One-time purchase, this paper only. AI-generated preliminary analysis — not peer review.
Frequently asked questions
Where can I download the Finnish AIS dataset from this paper?
The authors state it is freely available on Zenodo under DOI 10.5281/zenodo.8112336, organised as two CSV files per day: a location file (position reports with MMSI, latitude/longitude, speed over ground, course, heading, navigational status, etc.) and a metadata file (voyage data such as destination, IMO number, ship type, draught, ETA, vessel dimensions and name). The underlying feed is Fintraffic's public digitraffic.fi MQTT API.
How big is the dataset - 229 million or 2.29 billion data points?
The paper is inconsistent: the abstract and introduction say 229 million data points, while Section 3 states 2,293,129,345 (about 2.29 billion). This factor-of-ten discrepancy is never reconciled in the text, so users should verify the true size against the actual Zenodo download.
What makes this AIS dataset different from other free AIS datasets?
Its regional scope: 20 months (April 2021-December 2022) over the Baltic Sea including Finnish lakes, where seasonal freezing reshapes traffic patterns. The free datasets listed in the paper cover US waters (US Coast Guard Navigation Center) and Danish waters only, while comparable global sources (MarineTraffic, FleetMon, VesselTracker) are paid or restricted. The release also keeps raw AIS field values with documented sentinel codes rather than cleaned aggregates.
Does the paper include any machine-learning experiments on the dataset?
No. The paper is a dataset description: a literature review of AIS applications, ship-type count tables, and Folium heatmap visualisations. It contains no models, baselines, or benchmarks; the cs.LG categorisation reflects intended downstream use rather than experiments performed in the paper.
About this review
This is an AI-generated preliminary review by Evolve.Pub, produced from the openly available text of the paper. It is not peer review and does not represent the authors' views. The canonical record of the paper is at the publisher link above.
- Figures 1-7 appear only as captions in the extraction; the imagery itself is absent, so heatmap legibility and the plots behind the country/year statistics could not be visually assessed.
- Table 6's row labelled '136' with one ship is interpreted as an unlabelled regional ship-type code (100-199 are reserved for regional use per Section 3.2); a garbled PDF table extraction cannot be fully ruled out.
- The Zenodo record [40] was not accessed; availability, file sizes, licence and completeness of the daily CSVs rest on the paper's claims.
Go deeper on this paper
Wrote a paper? Stress-test it before reviewers do.
Upload your manuscript for the same structured scrutiny — reasoning, statistics, evidence, figures, and reproducibility.