Reading FRED's answers
FRED's recommendations come back with several parts. This page walks through each part, what it means, and when to look at which piece.
#Anatomy of an answer
A typical FRED response has four components:
- The headline: a one-sentence recommendation with the most important numbers.
- The map: a visualization of the recommended facility location, contributing clusters, and the relevant spatial context.
- The supply curve: a chart showing how much biomass is available at what cost, often broken out by year for multi-year queries.
- The justification: a written paragraph explaining how FRED arrived at the recommendation and what the key tradeoffs were.
#The headline number
An example headline:
Reading this:
- Site near Burney, CA: FRED's recommended facility coordinates, resolved to the nearest named place.
- 42.3k BDT/yr: total annual sustainable biomass supply available to this facility. That's bone-dry tons per year.
- $87.40/BDT: levelized cost per ton, weighted across all contributing clusters. Includes harvest cost and transport.
- 5.8x fire-risk leverage: ratio of fire reduction to cost premium versus the cost-only optimum. A 5.8x leverage means each 1 percent of extra cost buys roughly 5.8 percent of fire reduction.
- alpha = 0.05: the fire weighting FRED recommended for this query, at the knee of the Pareto curve.
#The map
The map shows three things by default:
- The recommended facility location, marked with a breathing pin.
- The contributing clusters, color-coded by their contribution. Darker colors mean larger contribution.
- The procurement radius, drawn as a soft boundary around the facility.
Toggle the map controls in the top right to add overlays. Fire-risk heatmap and transport circuity zones are the most commonly useful. The H3 hex aggregation makes regional patterns more visible than the per-cluster view at large radii.
#The supply curve
For single-year queries, the supply curve shows clusters ranked from cheapest to most expensive on the x-axis with cumulative annual supply on the y-axis. The slope flattens out where supply gets expensive; the inflection point is usually a good indicator of where additional supply stops being economic.
For multi-year queries, the supply curve is replaced by a year-by- year stacked chart. The bars often show a "sawtooth" pattern; this is a real signal, not a glitch. It reflects genuine terrain heterogeneity: in any given year, the available clusters may cluster into easier or harder terrain depending on what's left.
#The justification
The written paragraph below the recommendation explains the reasoning. A typical justification covers:
- Why FRED picked this location over alternatives nearby.
- What the dominant cost drivers are (terrain, transport, or fire weighting).
- Any constraints that mattered, like coverage gaps or supply ceilings.
- What to consider for follow-up analysis.
Read the justification before acting on the headline. The headline is a useful summary, but the justification often surfaces caveats that change how you interpret the number.
#When to trust the answer
FRED's predictions are validated against the source data they approximate. The harvest cost surrogate has R² = 0.997 against held- out FRCS runs with mean absolute error under $2.10 per BDT. The transport circuity model is built from 445,000 real road network routes. Within California, within standard prescription regimes, on typical haul distances, FRED's recommendations are dependable.
Be more cautious when the query pushes the edges:
- Very remote terrain or very long hauls. The transport model interpolates from sampled routes; areas with few nearby samples have wider uncertainty.
- Unusual harvesting systems. If a query forces a helicopter operation on flat ground, FRED's feasibility classifier will catch it, but the cost prediction for marginal cases is noisier.
- Very high fire alpha. At alpha > 0.5, the model is operating in a regime where structural ceilings dominate. The predictions are still correct but the result is largely shaped by where biomass and high-fire-risk areas coincidentally overlap, which is sparse.
- Edge of the year horizon. Year-10 projections have more accumulated uncertainty than year-1. The depletion model assumes prescriptions stay constant; real procurement decisions may alter that.