Design preview
Powered by PyHydroGeophysX

The planner turns this request into a structured plan. In this preview the plan is fixed sample content.

Analysis plan

Each value states where it came from.

Schema v1.2 · validated

Water content

Volumetric water content (m³/m³)

Uncertainty

One standard deviation of water content, on the same axes.

1σ (m³/m³)

About this preview

A non-functional design preview, made for a proposal figure. It holds sample data and performs no computation.

PyHydroGeophysX is an open-source Python package that connects near-surface geophysics with hydrology. It reads field data from common instruments; runs forward models and inversions for electrical resistivity, seismic refraction, ground-penetrating radar, electromagnetic and magnetotelluric surveys; and converts the results to water content through petrophysical relationships, carrying the uncertainty through each step. It also turns the output of hydrological models such as MODFLOW and ParFlow into the geophysical data they predict.

Today the package is used from Python, from a desktop studio and from a web app. A run can keep a provenance record: the input files with their hashes, the sequence of calls, the warnings raised and the quality of the fit.

UnEARTHeD would add a catalog of archived surveys with curated metadata, requests stated in plain language, and deposit of each product with its record. A language model serves as the planner. It turns the request into structured output, decoded against a plan schema and limited to registered tools. The validated plan then runs deterministically in a pinned environment, with no model in the loop.

Chen, Niu and Wu (2026), PyHydroGeophysX, SoftwareX 34, 102710 ↗ Source code: github.com/geohang/PyHydroGeophysX ↗ Documentation ↗

Plan schema v1.2

JSON Schema, draft 2020-12. The planner's output is produced by constrained decoding against this schema, so every plan it returns conforms: only registered tools, an origin for every parameter, citations for every inferred value, and uncertainty that is always propagated.