CropHist — Project Status

Overall Progress

US History

100%

Canada

85%

Argentina

85%

Brazil

50%

EU Raster

EU Vector

Prediction

Platform

100%

Timeline

Feb 25

Self-serve API + Stripe billing launched

Feb 26

Interactive map + search + viewport stats

Feb 27

MCP server published (npm + Lambda)

Mar 3

CDL confidence layer (2017-2023)

Mar 14

First organic signup (Fort Worth ISD)

Mar 24

International expansion started (CA/AR/BR)

Mar 27

CA/AR crop-coded tiles building on LUMI

Mar 28

Brazil: transfer to S3, DynamoDB, frontend

Apr

EU pipeline (Copernicus CLMS + EuroCrops)

Apr-May

Crop prediction model v1

NEW Goal 3: Crop Prediction Model

For every field, predict next season's crop using historical sequences + spatial context. Aggregate to county/state/country level for area change estimation.

Features (input per field)

Own crop history: last 3, 5, 7 years

Neighbor crops: 100m, 500m, 1km, 5km radius

Historical crop prices (USDA NASS / CBOT)

County-level planting trends

Field size, location (lat/lng, county)

Model Pipeline

Baseline: last-year repeat (~70% acc for corn/soy)

LightGBM: tabular features per field

Spatial: neighbor crop distribution features

Sequence: LSTM/Transformer on crop timeseries

Ensemble: best of above, per-crop calibration

Output

Per-field: predicted crop + probability

Per-county: predicted area by crop + YoY change

Per-state/country: aggregate forecasts

Confidence intervals on all estimates

Compare vs USDA prospective plantings report

Deployment

API: /v1/predict/{field_id}

API: /v1/forecast?county=X&year=Y

Map layer: fields colored by predicted crop

Dashboard: county/state forecast charts

Annual retraining after CDL release

Full Goal Hierarchy

CropHist.com

✅ 1. US Crop History Done

1.1 Field Boundaries (29.1M, CDL 2008-2025)

1.2 History Pipeline (shards + index + DynamoDB)

1.3 Self-Serve API (6 endpoints, demo + auth)

1.4 Interactive Map (MapLibre + CDL raster + PMTiles)

1.5 Stripe Billing (Free/Pro/Enterprise, live)

1.6 SEO (49 state + 3,109 county pages)

1.7 CDL Confidence (2017-2023)

1.8 MCP Server (npm + Lambda, 9 tools)

🔄 2. International Expansion 45%

2.0 Foundation (taxonomy, config, API routing) Done

2.1 Canada — boundaries, zonal, history done; crop tiles building 85%

2.2 Argentina — same as Canada 85%

2.3 Brazil — 40yr zonal + history complete on LUMI; needs S3 transfer 50%

2.4 EU Raster (Copernicus CLMS) Todo

2.5 EU Vector (EuroCrops IACS) Todo

🔮 3. Crop Prediction Model New

3.1 Data Prep (feature engineering, train/val split)

3.2 Models (baseline → LightGBM → spatial → sequence)

3.3 Area Estimation (county/state/country aggregation)

3.4 API + Frontend (predict endpoint, forecast dashboard)

3.5 Validation + Deployment (backtest, Lambda, retraining)

✅ 4. Platform Done

Alarms, Analytics, Admin, Rate Limiting, SES

CropHist.com — Project Status