Jeff Bradshaw, Founder, Adaptris. NYC 2017
summary
Data from farm management systems, tractors, drones, satellites and sensors is ingested into the open source HPCC Systems pipeline, cleaned with ML entity linking and dimension reduction, and turned into dashboards and recommendations for agronomists. The dataset spans ~570 million farms and ~25 million tractors producing 30 MB per tractor per day; average canola yield is about 3,700 kg/ha with ~1,500 kg/ha typical variance. Analysis of 2016 finds a wet spring and warm winter plus timely pesticide application correlate with higher yields, while hybrids do not show clear advantage; results are actionable but correlation not causation.
AI-written summary · about
chapters (24)
- 0:00Intro
- 0:23Probe Agrokor
- 1:16Why Agriculture
- 2:29Why Adaptris
- 3:14Weather
- 3:51Data Collection
- 5:42Overview of Approach
- 6:06HPCC
- 7:11Seed Rape
- 8:04Yields
- 8:22Equipment
- 9:04Yield Variance
- 9:33Cost Consistency
- 10:00Variety Choice
- 10:11North vs South
- 10:37Hybrids
- 11:06Data realization
- 11:33London bus
- 11:45Big red bus
- 11:58Graphing
- 12:25Varieties
- 12:41What causes yield variation
- 14:41Answers
- 15:37Key takeaways
related talks
26mAI, big data, digital and future of agriculture
31mBridging IoT and Machine Learning with Python for Sustainable Agriculture
04mAI and big ag data for food security
38mBetter Farming and Better Eating Thanks to Software: Tech Serving Agriculture
28mHarnessing Machine Learning to Improve Agricultural Resilience in Africa
07mAI Infrastructure Transforming Modern Agriculture
19mPlant Factory: Sensor, Data, Machine Learning
21mUnderstanding World food economy with satellite