Regional Groundwater Database for Arequipa and Surroundings in Southern Peru from Unstructured Sources Using an Integrated Optical Character Recognition and Large Language Model Workflow

A groundbreaking regional groundwater database for Arequipa, southern Peru, has been created by integrating optical character recognition (OCR) and large language models to extract data from 3675 unstructured sources. The dataset includes 4775 georeferenced records spanning 1966–2025, covering parameters like depth to water, well geometry, hydraulic conductivity, transmissivity, storage coefficient, lithology, and test type. A rigorous three-phase human-in-the-loop audit corrected 4.1% of records, primarily unit misassignments. Independent validation confirmed the dataset's fidelity, with transmissivity values aligning closely with calculated products (94% within a factor of two) and matching official reports for extreme values. However, the database highlights a critical hazard: mixing seven measurement supports without filtering biases regional estimates downward by ~48%. The dataset is distributed under CC BY 4.0 on HydroShare, with explicit guidance to filter test types before aggregation.

Venegas Quiñones, H. L., Guillen, M., Garcia-Chevesich, P. A., González, E., Ticona, J., Díaz, J., Zea, J., Alejo, F., & McCray, J. E. (2026). Regional Groundwater Database for Arequipa and Surroundings in Southern Peru from Unstructured Sources Using an Integrated Optical Character Recognition and Large Language Model Workflow. https://doi.org/10.5194/essd-2026-667