Harmonizer For Emulator Inputs

The website centers the quantum emulator, but the emulator still needs good inputs. In environmental data science, those inputs often begin as mismatched geospatial layers: biodiversity observations, climate grids, remote sensing products, land-cover rasters, field sites, access costs, or planning units.

The harmonizer is the preparation layer that turns those sources into comparable data. In this repository, its role is to support the move from environmental synthesis to binary optimization.

raw environmental layers -> harmonized layers -> decision table -> QUBO

Why Harmonization Still Matters

The emulator does not operate directly on raw rasters or shapefiles. It operates on a decision table where each row is a candidate site or planning unit and each column describes something relevant to the decision.

Harmonization helps create that table by making datasets comparable:

Input mismatch	Harmonizer task	Why the emulator cares
Different coordinate systems	Reproject	Site features must refer to the same locations
Different raster resolutions	Resample	Candidate sites need comparable environmental values
Different study boundaries	Clip	The decision problem needs a shared study area
Vector and raster formats	Convert or summarize	Site rows need consistent feature columns
Different ecological meanings	Document and derive variables	QUBO rewards and penalties must be interpretable

From Layers To Decision Columns

A monitoring-site table might include:

• species_richness
• climate_refugia_score
• habitat_connectivity
• cost
• mean_temp
• annual_precip
• elevation
• region
• lat
• lon

Those columns can come from field observations, raster extraction, vector overlays, or summaries of geospatial data cubes. Once they are in a clean table, the optimizer can normalize them and build a QUBO.

Preserved Harmonizer Example

The Colorado fire-risk example remains in the repository as a reference for AI-assisted geospatial preparation:

python3 examples/colorado_fire_risk/colorado_harmonization.py

It shows how datasets can be downloaded, reprojected, clipped, resampled, and visualized. In the quantum-emulator storyline, that work is not the endpoint. It is what prepares environmental layers for a later data-to-decision workflow.

Scientific Caution

Harmonization choices are scientific choices. Resampling, aggregation, normalization, and clipping affect what the later optimization problem means.

Before building a QUBO, check:

• whether the variables are comparable
• whether categorical data used nearest-neighbor resampling
• whether continuous data used an appropriate interpolation method
• whether the study boundary matches the decision question
• whether cost and biological value are measured or estimated consistently

The emulator can only explore the decision space it is given. Good harmonization is how that space becomes meaningful.