Reconstructing a Greenhouse Event from Telemetry and LiDAR

Environmental telemetry and a spatial capture can describe the same greenhouse without describing the same moment.

Reconstruction is the work of establishing which observations belong to the event under review, how they relate, and where the available evidence stops.

The useful starting point is an operator's question and a bounded interval. Name the facility and crop area, identify the available sources, and state what the reconstruction is intended to help someone inspect.

Then preserve each input before normalizing it. An observation needs its source identity, original time, applicable location, units or spatial frame, and known quality limits. Ingestion time should remain distinct from measurement time.

Unit conversions, timezone normalization, filtering, and coordinate transformations may be necessary. They should produce traceable derived records while keeping the original evidence available. A reviewer needs to know which transformation placed an observation in a scene.

That is the foundation described in the sensor-provenance article.

Build the timeline before treating the scene as a common state. Determine when each observation was captured, over what duration, and with what clock information. Being recorded on the same day or assigned to the same bed does not establish simultaneity.

Next, evaluate the relationship required by the intended use. Comparing separate captures is different from fusing them into an estimate of one instant. Spatial registration also requires support for coordinate relationships and calibration; matching timestamps alone is insufficient.

Streamline's GH1 / Bed 4 reconstruction kept spatial observations separate when their timing and available calibration evidence did not justify a synchronized fusion. The capture-gap note documents that particular result.

Refusal should preserve a useful output: the original observations, the proposed comparison, the missing support, and the narrower inspection still possible. An unsuccessful fusion need not make the underlying records inaccessible.

The resulting replay should let a reviewer move in both directions—from source records into the representation and from a represented element back to its evidence. Observed geometry, derived placement, inferred state, and unavailable information should remain distinguishable.

Validation then checks those relationships. Does source identity survive transformation? Are times preserved? Can a reviewer recover the observation behind a displayed element? Does insufficient evidence prevent the unsupported operation?

Those checks establish behavior within the tested reconstruction. They do not validate sensor accuracy, crop diagnosis, live reliability, or equipment control.

The current Streamline OS boundary is recorded and read-only. A next decision-support workflow would need to establish whether the reconstructed evidence can support a useful recommendation under appropriate comparison conditions.

The practical deliverable is an event someone can investigate: which observations are relevant, which connections are supported, and what additional evidence would be required.

A reconstruction is successful when it makes the event easier to inspect without making the evidence appear more complete than it is.