How satellites measure land recovery on reclaimed well sites

Reclamation is, at its core, a question about vegetation and soil: has the disturbed ground returned to a state comparable to the surrounding, undisturbed land? Multispectral satellites answer that question by measuring far more than the eye can see — capturing reflectance in visible, near-infrared, and shortwave-infrared bands that respond directly to plant health, moisture, and bare soil.

The indices that matter

A handful of well-established spectral indices translate raw reflectance into interpretable signals of recovery:

• NDVI (Normalized Difference Vegetation Index) — the workhorse measure of green, photosynthesizing vegetation.
• EVI and SAVI — vegetation indices that correct for soil background and atmospheric effects, useful on sparsely vegetated or early-recovery sites.
• NDWI / NDMI — water and moisture indices that track surface water and vegetation water content.
• BSI (Bare Soil Index) — highlights exposed soil, which is exactly what a disturbed, un-recovered lease looks like from orbit.

From indices to a recovery score

A single index on a single day is noisy — clouds, season, and soil all interfere. A robust recovery score does three things: it composites several indices so no single signal dominates, it compares the site against a baseline (its own pre-disturbance history and the surrounding reference area), and it looks at the trajectory over time rather than one snapshot. The result is a 0–100 score that says how far the land has come back, with a confidence level attached.

Why rules-based, not black-box

It is tempting to throw a machine-learning model at the problem, but reclamation decisions are audited. A score that has to survive a regulatory redetermination or anchor a credit memo needs to be explainable: every number should trace back to an observation and a documented rule. That is why Meridian uses an interpretable, rules-based composite of vegetation, moisture, and bare-soil indices rather than an opaque model — a machine-learning layer is most useful later, for anomaly detection, once there are years of longitudinal data.

Why free, frequent imagery changes the economics

Sentinel-2 provides 10-metre multispectral imagery on roughly a five-day revisit, at no per-scene cost. That is enough resolution for site-level vegetation analysis and frequent enough to build a real time series — which means portfolio-wide monitoring is possible from day one, with no per-site data bill standing between a question and its answer. Higher-resolution commercial imagery can be layered in where a specific site warrants it.