In order to make these forecasts, we need to know where crops are, which is why we’re pioneering new techniques for crop mapping.
For a long time, the best way to map crops was to actually walk (or drive) through fields, tagging areas as corn or soy (for example). This may work on a local basis, but relying on “ground truthing” for the world’s more than 300 million hectares would be impractical (Spock would even say…illogical).
Our data scientists use years of images from NASA and ESA to develop algorithms that can recognize “field boundaries” around the world. These field boundaries are then recorded in our database, allowing us to identify characteristics (or “features”) for each field, like reflectance, crop stage, and crop health.
Each of the fields is made up of pixels (like on your screen). There is always uncertainty associated with remotely sensed images, and we reduce that uncertainty by building a model for each pixel. We take the best pixels from each image in the time series and merge them together to get a harmonized record of many years, which helps us detect anomalies in each field. This helps us predict, in-season, which crop is going to be prominent - something that hasn’t been done on a large scale before!
In the past, we’ve had to rely on a hodgepodge of sources to know which crop has been planted where. At the end of the season, the United States Department of Agriculture (USDA) generates a comprehensive report with detailed crop map information for the United States… but farmers and agribusinesses can’t wait that long! Also… what about the rest of the world??
Leading agribusinesses, commodities funds, and possibly even the Borg have already assimilated this into their systems, what could YOU do with a global crop map built from space?
This is the future for TellusLabs, we hope you are as excited as we are!
Live long and prosper!
The team at TellusLabs