Tweets Yield Crop Progress

Traditionally, USDA Crop Progress reports are generated by sending some 4,000 surveys out to Extension and USDA reporters every week during the crop season. What if there were another way?

By Marilyn Cummins | Photos By Jamie Cole

Sam Zipper, a postdoctoral fellow at the University of Victoria in British Columbia, Canada, was curious to know if Twitter data could be a useful source for such information, so he set up his own trial run to see if he could determine weekly U.S. crop planting progress solely based on tweets during the 2017 planting season.

He designed a system called AgroStream to automatically search, filter and store tweets about agricultural planting daily. He kept all relevant ones with known locations, then narrowed further to only tweets with hashtags #corn, #soy or #soybean(s) used together with #plant17 or similar.

Zipper built a model to calculate planting progress during the 20 weeks in 2017 between Feb. 27 and July 16 based on the content and frequency of that set of tweets, then compared his results to weekly USDA National Agricultural Statistics Service reports. In the January 2018 issue of Agronomy Journal, he reported that Twitter data accurately captured planting progress across most of the 10 state-crop combinations he studied.

He says he’s not suggesting Twitter can replace traditional crop monitoring. He does say mining Twitter data can give researchers answers “very quickly compared to survey-based approaches,” and already has a colleague who’s using his method to search tweets for how farmers perceive and respond to drought