Frequently asked questions

Answers to questions FRED's users actually ask.

#General

#Who built FRED?

FRED is a research project developed by Aunsh Bandivadekar with collaborators. The underlying research is documented in a peer-reviewed thesis and several published papers.

#Is FRED free to use?

Yes, for research, policy analysis, and California biomass procurement evaluation. FRED is a research tool maintained by the project team. Commercial use at scale requires no fee but does require coordination; email contact@biofred.us.

#How does FRED handle costs internally?

FRED uses Claude as its reasoning backbone, which has per-query compute cost. For users of the chat interface, this is transparent. The query cost scales with conversation length and complexity but is generally a fraction of a cent per turn.

#Coverage and scope

#Which states does FRED cover?

Currently California only. The cluster dataset, fire raster, and transport model are all California-specific. Oregon and Washington coverage is on the roadmap pending equivalent regional data.

#Does FRED cover agricultural residues?

No. FRED's supply is forest biomass only. Orchard residues, dairy waste, crop residues, and other agricultural waste streams are not in the cluster catalog.

#Does FRED distinguish private versus public land?

FRED knows the ownership type of each cluster but does not filter on it by default. If you want only private-land supply, ask: "find biomass on private land near X." Most users find that ownership composition matters less than terrain and transport in the cost analysis.

#Accuracy and trust

#How accurate are FRED's cost predictions?

The harvest cost surrogate validates against FRCS with R² = 0.997 and mean absolute error under $2.10 per BDT. For typical California cluster configurations within the C-BREC catalog, FRED's predictions are dependable to within a few dollars per BDT. See Harvest cost surrogate for the full validation details.

#Why does FRED's cost differ from a contractor's quote?

FRED reports engineering cost. Contractor quotes include stumpage payments, contract margin, overhead, and operational reserves that FRED does not model. A reasonable rule of thumb is that delivered contract pricing runs 15-25 percent higher than FRED's engineering cost. The difference is not a modeling error; it's the contractor's business overhead.

#Does FRED produce confidence intervals?

Not as standard output. The surrogate's R² and MAE statistics give a global error estimate, but per-prediction intervals are not exposed in the chat interface. If you need formal uncertainty quantification for a specific procurement plan, email the team.

#Using FRED

#How long do queries take?

Typical query time is 2 to 5 seconds for first-time queries and under 1 second for follow-ups that hit the cache. Complex multi- year projections in large radii can take up to 7 seconds.

#How long does FRED remember a conversation?

Within a session, FRED remembers everything you've said. Sessions time out after 24 hours of inactivity, after which the conversation is cleared. To revisit a prior analysis later, copy the recommendation or use the export option.

#Can I export results?

Yes. FRED's chat interface includes export options for the recommendation card, supply curve, map snapshot, and cluster list. Exports are available in CSV, JSON, and PDF formats. For programmatic access to FRED's underlying data, contact the team.

#Research and citation

#How do I cite FRED in a paper?

See Citing FRED for the recommended citations. Cite the relevant prediction layer paper for specific methodology, or the umbrella thesis when referencing FRED as a system.

#Where are the underlying papers?

The harvest cost surrogate, transport circuity model, and CS-RAG architecture each have their own paper. Drafts and preprints are linked from the FRED landing page in the Research section.

#Can I reproduce FRED's results outside the chat interface?

Yes. The harvest cost surrogate is available as a standalone XGBoost model. The IDW transport raster is available as a GeoTIFF. The USDA FSim burn probability data is publicly available from USDA. Email contact@biofred.us to request the artifacts for reproducibility.

TIP
If your question isn't answered here, just ask FRED. Many questions about FRED's behavior, methodology, and outputs can be answered directly by the agent itself.