Digitizing old boring logs: from scanned PDFs to structured data

What does digitizing a boring log produce?#

Digitizing a boring log produces structured records rather than a searchable PDF: one header record per boring and one row per depth interval, each linked back to the page it came from. OCR alone makes text searchable, but it does not tell software which number is a depth and which is a blow count.

The output should let an engineer filter borings by location, find every interval classified as lean clay, or plot SPT N values against depth across a site. That is the practical test of whether the work was worth doing.

Step checklist: from shelf to database#

The step checklist below works for a firm digitizing a few projects or a whole archive. Run a pilot on a small set of projects from different eras and drillers before committing to the full backlog, because template changes over the years are where most surprises hide.

• 1Inventory: list projects, number of borings, formats (paper, scan, gINT, report PDF) and where each sits.
• 2Triage scans: grade each file as clean typed, poor typed, handwritten or mixed, and rescan the worst pages.
• 3Define the schema: choose AGS, DIGGS or an internal schema, and write field definitions and units.
• 4Extract header data: project, boring ID, coordinates, ground surface elevation, dates, driller, rig and hammer type.
• 5Extract interval data: top and bottom depth, sample type and number, blow counts, SPT N, recovery, USCS symbol and description.
• 6Capture groundwater: depth encountered, depth at completion and any later readings with dates.
• 7Attach lab results: moisture content, Atterberg limits, gradation and other tests by sample.
• 8Check quality: compare a sample of extracted values to source pages and run automated checks on every log.
• 9Store and control access: load to the database with source links, project metadata and client restrictions.

Which fields should you extract?#

The fields you should extract are the ones engineers use to compare borings across a site, plus the metadata that makes them trustworthy. The table lists the core set and the problem each one usually causes during extraction.

How should you handle poor scans and handwritten logs?#

Poor scans and handwritten logs should go down a separate path rather than through the same automated extraction as clean typed logs. Faded carbon copies, skewed pages and handwritten field logs generate the errors that later undermine trust in the whole dataset.

Rescan originals where paper still exists, in grayscale and at a higher resolution than the office copier default, and straighten pages before extraction. Where both the field log and the final typed log exist, extract from the typed log and keep the field log as a linked reference. Flag handwritten-only borings so reviewers know to check them more closely.

Manual entry, OCR templates or AI extraction?#

Manual entry, OCR templates and AI document extraction each fit different parts of an archive, and most firms end up combining them by file type.

If an outside service or software vendor processes the logs, read its terms on data retention and model training, and check client contracts for confidentiality before any project files leave the firm.

Which quality checks catch the most errors?#

The quality checks that catch the most errors are simple rules run on every log, combined with a person comparing a sample against the source pages.

Track error rates by source type. If handwritten logs fail far more often than typed ones, change the path for those files instead of reviewing everything at the same depth.

• 1Depth continuity: intervals should not overlap or leave unexplained gaps.
• 2N value arithmetic: SPT N should equal the sum of the last two increments where the log records all three.
• 3Classification consistency: the USCS symbol should match key words in the description.
• 4Groundwater sanity: recorded groundwater depth should fall within the boring depth.
• 5Total depth: the bottom of the last interval should match the stated boring depth.
• 6Location: borings should plot within the project site boundary.

Illustrative: a geotechnical firm digitizes its archive#

Illustrative: a fictional geotechnical firm with offices in neighboring states holds many years of boring logs: early ones typed on paper, later ones in gINT, and many only as appendices to PDF reports. The data manager inventories the archive and finds several log templates and a large handwritten set in old field files.

The firm adopts DIGGS as its target schema, uses AI extraction for typed logs, keys handwritten logs by hand and runs depth and N value checks on every boring. Proposal teams now pull nearby borings for desk studies quickly instead of hunting through report PDFs, and the principal has an accurate inventory if a licensing fit check ever comes up.

Rights, confidentiality and how SourceX approaches digitized logs#

Digitized boring logs carry the same rights questions as the original reports. Client contracts may require confidentiality, owner-furnished borings usually belong to the owner, and public agency projects can carry their own terms, so store client and contract restrictions alongside each project record.

SourceX reviews digitized geotechnical records through the SourceX five-step transaction: Supply, Rights, Preparation, Approval and Delivery. The fit check needs only metadata, such as record types, years covered and formats. Under the SourceX Enterprise Data Value Framework, data cleanliness and domain expertise raise value while preparation cost reduces net value, which is one more reason a consistent schema pays off.