Mükremin Nergiz

Tools and research

Detectability

Pre-survey detectability, modelled and reproducible

Before a survey, the practical question is rarely "what will the anomaly look like?" but "will we see the target at all, and how dense must the stations be?". Detectability answers that in seconds, shows the physics behind the answer and gives a script that anyone can rerun.

Open the tool Method and validation Source code

What it does

  1. You describe the target (sphere, tunnel, shaft, room, wall or dyke, fault), its depth and size, what it is made of and the host rock, and the survey: instrument, field noise and location.
  2. It models the gravity and magnetic anomaly along a profile across the target, with the Earth's main field at the site and date from IGRF-14.
  3. It compares the anomaly with the effective noise and gives a verdict for each method: clearly detectable, marginal or not detectable.
  4. It recommends the method, the station spacing and the profile length, lists every assumption, and returns a Python script that reproduces the result.

An example

A 20 m air-filled cave whose top is 15 m deep in limestone, surveyed near Sivas in 2026:

−114 µGalgravity anomaly; S/N 10, clearly detectable
0.46 nTmagnetic anomaly; below noise, not detectable
≤ 5 m · ≥ 153 mstation spacing and profile length for gravity

Limestone is almost non-magnetic, so the cave has little magnetic contrast; its density contrast, however, makes it a clear gravity target. The tool says so, and shows why.

How it is validated

Every model is checked against a result computed in a different way, and every check is an automated test with a stated tolerance that runs on every change:

  • analytical limits: point mass, Bouguer slab, half-width rules;
  • numerical integration of point dipoles against the closed-form magnetic models;
  • Poisson's relation between the independent gravity and magnetic codes, including dipping dykes;
  • the main field against the British Geological Survey's IGRF-14 service at five sites, to 1 nT.

The Validation page lists each check with its largest error and tolerance, and shows each comparison as a figure with the difference beneath it.

Open science

The Python library is open source under the BSD-3-Clause licence, with a citation file and the author's ORCID; the code and its automated tests are on GitHub. The website, the HTTP API and the planned ChatGPT and Claude apps call the same code, so they give the same numbers.

Cite

Nergiz, M. (2026). Detectability: Pre-survey detectability,
modelled and reproducible (Version 0.1.1) [Computer software].
Zenodo. https://doi.org/10.5281/zenodo.23931815

Author: Mükremin Nergiz, Geophysical Engineer & Software Developer, Independent Researcher, Sivas, Türkiye. ORCID 0009-0001-6789-2774.