SAR Multi-Crop Acreage Estimator
Round 1 of a SAR-based agricultural-intelligence challenge: estimate the cultivated hectares of Rice, Cotton, Maize, Bajra and Groundnut per village from multi-temporal X-band SAR, scored by MSE. A complete, dependency-free SAR → acreage pipeline with a documented real-data seam.
At a glance
The pipeline
SAR preprocessing
Calibration (dB ↔ linear power), multi-temporal Lee speckle filtering, and temporal stacking across the Kharif season.
Agricultural extent
Cropland is separated from water and built-up by its high multi-temporal backscatter variability (bare → canopy → harvest).
Linear unmixing
Backscatter mixes in linear power weighted by crop area — so acreage is recovered by physically-matched ridge least squares (closed-form, fast).
Exact submission
Emits ID,Rice_ha,Cotton_ha,Maize_ha,Bajra_ha,Groundnut_ha, non-negative and area-capped, scored by MSE.
Screenshot
Deliverables
Overview
What it is and how to run it
Outcome Contract
Scope, honest constraints, success criteria
Verification Report
12/12 checks + benchmark
Proof Report
Status (PROOF_INCOMPLETE), evidence lineage, disclosed seams
Executive Evidence
Answers the 10 IRS_AUDITOR questions with lineage
Limitations
Synthetic/seam/unverified disclosures
Auditor Objections
Pre-written objections + evidence responses
Reproduce
Exact commands to reproduce every number
Verify
How verification runs + what each check asserts
User Guide
Synthetic demo, real data, explorer
Run & Deploy
Local run + raster→zonal seam
Release Notes
v1.0.0 summary
Honest scope
The official competition SAR tiles, village boundaries and labels are not present in this environment, so no leaderboard MSE has been produced. The pipeline is verified end-to-end on a physically-motivated synthetic Kharif benchmark (R² —, ). The raster→zonal-statistics stage is documented as a seam; everything downstream — ingestion, modeling, submission — runs live. See the certification.
Delivery metrics
Tokens, elapsed time, and cost for producing this outcome. Basis: Estimated — reproducible model over this outcome’s published artifacts. metrics.json
Cost and tokens are estimates derived deterministically from published artifacts and representative list pricing; actual billing may differ. Model basis: claude-sonnet-class (representative).