IBM–Yamagata AI survey confirms 303 new Nazca geoglyphs in Peru

An IBM–Yamagata deep-learning survey guided fieldwork that confirmed 303 new Nazca geoglyphs, published in PNAS (Sep 2024).

23ⓘ

AI system

System

The geoglyph classifier developed by IBM and Yamagata University, with IBM PAIRS for imagery. [1]

Model

ResNet50 with a two-layer classifier, adapted from ImageNet training. [1]

Release

The study does not give a release or checkpoint identifier. [1]

Provider

IBM Research and Yamagata University, with the German Aerospace Center. [1]

Role

The classifier estimated where relief-type geoglyphs might appear in aerial images. [1]

Configuration

The model analyzed 112 × 112-pixel images at 5-meter intervals. Candidate selection used probability threshold 0.55 and at least two adjoining image patches. [1]

Inputs and outputs

The classifier analyzed aerial photographs and returned a map of probabilities and candidate locations. [1]

Autonomy

Archaeologists screened the candidates and visited selected locations to confirm the geoglyphs. [1]

Workflow

High-resolution aerial imagery of the whole survey area. A detector trained with limited labeled examples of relief-type geoglyphs. Archaeologist screening of candidates followed by ground verification.

Human oversight

Not assessed

Relation to AI

The model proposed candidate locations that directed fieldwork. 178 of the 303 confirmed geoglyphs were individually suggested by the AI. Of the remaining 125, 66 were found as part of an AI-discovered group. The other 59 were found without any help from AI or in screening outside the AI-suggested boxes. Confirmation was by ground survey, not by the model.

AI contribution

The model proposed candidate locations that directed fieldwork. 178 of the 303 confirmed geoglyphs were individually suggested by the AI. Of the remaining 125, 66 were found as part of an AI-discovered group. The other 59 were found without any help from AI or in screening outside the AI-suggested boxes. Confirmation was by ground survey, not by the model.

Benefits

Outcome identifiers

Not assessed

Benefit

303 previously unknown figurative relief-type geoglyphs, almost doubling the known total

Category

Improved Knowledge Discovery

Status

Not assessed

Beneficiary groups

Organizations. Society.

Beneficiaries

Not assessed

Reach

Not assessed

Distribution and access

Not assessed

Duration

Not assessed

Expected?

Yes

Costs and harms

Not assessed

Measurement

Studies

Not assessed

Study design

Not assessed

Measurement

Not assessed

Setting and sample

Not assessed

Measures

303 previously unknown figurative relief-type geoglyphs, almost doubling the known total

Comparison

the discovery rate with remotely sensed high-resolution imagery but without AI (2004 to 2020)

Results

Reference: 18.7/y from 2004 to 2020. Result: 303 new figurative geoglyphs during the 2022/23 season of field work (a 16-fold acceleration).

Additional value from AI

Not assessed

Limitations

Not assessed

Denominators

Not assessed

Materials for replication

Not assessed

Money

Kind of value

Not assessed

Amount

Not assessed

Currency and prices

Not assessed

Period and unit

Not assessed

Realization

Not assessed

Recipients and payers

Not assessed

Gross or net

Not assessed

Included costs

Not assessed

Comparison

Not assessed

Method

Not assessed

Return and payback

Not assessed

People and places

Participants

Developer: IBM Research (IBM T.J. Watson Research Center) with Yamagata University and DLR. Operator: Yamagata University (Institute of Nasca field team).

Jurisdictions

Benefit: Nazca, Peru. Developer: United States. Operator: Japan.

Ownership and control

Not assessed

Dependencies

Not assessed

Cross-border connections

Not assessed

Geopolitical implications

Not assessed

Follow-up

Conditions for replication

Targets must be visible at the imagery resolution (about 10 cm). Each candidate must be confirmed on the ground before being counted.

Replication attempts

Not assessed

Later findings

14 August 2025. The same Yamagata-IBM team announced 248 further AI-assisted geoglyph finds at Expo 2025 Osaka, bringing the total to 893, of which 781 came from AI and aerial-image analysis.

Availability

Not assessed

Institutional decision

Not assessed

Grounds

Not assessed

Action and result

Not assessed

Success criterion

Not assessed

Evidence

Source numbers

AIBD-0019-source-0001–0005

Citations

Numbered references below.

Language

Not assessed

Source relationships

Developer, Independent

Original evidence

The original paper and supplement describe the classifier, image processing, and archaeological fieldwork. [1]

Access and versions

The original article and its supplementary methods were accessible. [1]

Supported findings

Not assessed

Evidence strength

Not assessed

Missing evidence

Not assessed

Grades

Assessed benefits

Organizations. Society.

Depth

Organizations: 2. Society: 3.

Reasons

Organizations: Yamagata University's Institute of Nasca reports a sixteen-fold increase in discovery rate, achieving in six months what DLR says previously took decades. Society: The known corpus of figurative geoglyphs at a UNESCO World Heritage site nearly doubled, with 303 confirmed sites added to Peru's archaeological record.

Observed or intended

Not assessed

Rubric

Not assessed

Assessor and date

Not assessed

Alternative assessments

Not assessed

Fill

Independently reported

Entry details

Kind

Not assessed

Field

Archaeology. Academic research with industry partner.

Purpose

Built to flag candidate geoglyph locations in aerial imagery so that archaeologists could prioritize ground survey; the discoveries were that purpose.

Dates

Date basis unassessed: 2022 (precision unassessed). Updated: 25 October 2026.

Stage

Not assessed

Version

SHA-256: 199358dd8dd5…

References

Other identifiers

Not assessed

Category mappings

Not assessed

Brief

Not assessed

Revisions

7 revisions. Latest: 25 October 2026.

Sources

  1. Peer-reviewed study Developer

    Sakai, M., Sakurai, A., Lu, S., Olano, J., Albrecht, C. M., Hamann, H. F. and Freitag, M. (2024). AI-accelerated Nazca survey nearly doubles the number of known figurative geoglyphs and sheds light on their purpose. PNAS 121(40), e2407652121 (PMC11459208).

    Describes imagery, candidate generation, human review, and field confirmation. Results; Materials and Methods, Data Preparation

    Original research or documentation from the system’s researchers. This supports identity and methods, not independent verification of benefits.

    Original research or documentation from the system’s researchers. This supports identity and methods, not independent verification of benefits. Names ResNet50 and its classifier and documents threshold settings. This is the original paper’s supplement. Details of the artificial neural network, pp. 2–3; Postprocessing, pp. 3–4

    Original research or documentation from the system’s researchers. This supports identity and methods, not independent verification of benefits.

  2. Press release Developer

    Yamagata University (2024). AI-accelerated Nazca survey nearly doubles the number of known figurative geoglyphs and sheds light on their purpose. University news.

  3. Press release Developer

    German Aerospace Center DLR, Earth Observation Center (2024). Accelerated Discovery of Nasca Geoglyphs Guided by AI.

  4. News report Independent author

    Archaeology Magazine, Archaeological Institute of America (2024). More Than 300 Geoglyphs Discovered in Peru.

  5. News report Independent author

    del Río ML. More than 200 geoglyphs discovered in Nazca Desert with help of AI. The Art Newspaper, 14 August 2025.

Collections: Cross-border cases.