Your portfolio companies file their PAI numbers once a year, usually through a template their sustainability team fills in from internal records. That's the disclosure. It is not independent evidence. When a mining subsidiary reports no material impact on a biodiversity-sensitive area, or a midstream asset reports flaring "in line with regulatory permits," you're taking their word for it, and their word is exactly what SFDR asks you to look past.
A quarterly satellite pass over the site itself can't verify everything in Annex I Table 1. But it can verify more than most analysts assume, and it's worth being precise about where the line actually falls.
What a site screen can corroborate
Medium-resolution multispectral imagery, the 3-10 m class that revisits a site every few months rather than every few years, picks up physical change on the ground. That's its whole job. It can tell you:
- Whether tree cover inside or adjacent to a concession boundary has thinned or disappeared since the last pass, which speaks directly to PAI 7, activities negatively affecting biodiversity-sensitive areas.
- Whether a tailings dam's footprint has expanded, a useful physical-risk check even though tailings management itself sits across several PAI categories rather than one clean line item.
- Whether a flare stack shows an active thermal signature on a given pass. That's a presence/absence read, and it gives you a real check against a "minimal flaring" disclosure even without a volume figure attached.
- Whether infrastructure, clearing, or access roads have crept into a buffer zone the company's own disclosure says is undisturbed.
All four of those are things a change log built from repeat imagery can show you in black and white: here's the boundary in Q1, here's the same boundary in Q4, here's what moved.
What it can't give you
This is the part worth saying plainly, because overclaiming here is how a screening tool becomes a liability in your own file. A site screen does not measure tonnes of CO2e, water discharge chemistry, hazardous waste ratios, gender pay gaps, board composition, or whether a company has a functioning grievance mechanism under the UN Guiding Principles. None of that is visible from orbit, and no imagery product should tell you otherwise. Flaring detection is a proxy for fossil-fuel-related activity, not a substitute for the GHG intensity figure in the company's own filing. Treat it as a flag that a disclosure needs a closer look, not as the number itself.
The indicators that line up best with this kind of screen are the ones with a physical footprint: PAI 7 most directly, with tailings and encroachment feeding into your broader physical-risk assessment even where they don't map to a single PAI line. The social and governance indicators in the table stay entirely dependent on self-reported data, because there's nothing to photograph.
Using it as a corroboration layer, not a replacement
The practical workflow most stewardship teams land on is: take the PAI statement as filed, then run the sites against independent change detection before the annual re-engagement cycle. If the imagery and the disclosure agree, you've got a cleaner audit trail for your own PAI statement to the regulator. If they don't, you've got a specific, dated, boundary-level discrepancy to raise with the portfolio company rather than a vague suspicion.
That's the gap ESG Screening is built for: a quarterly map-layer screen of each holding's site, flagging deforestation, tailings-dam footprint change, flaring, and encroachment, delivered as a per-holding risk map and change log you can sit next to the company's own PAI submission.
If your PAI statement currently rests entirely on what portfolio companies choose to tell you, it might be worth seeing what a quarterly look at the actual site adds to the file.
