AI for Yield Prediction in Corn under Soil Moisture Stress
This is an executive summary of Grain Data Solutions’ blog post.
Application of AI for yield prediction is an important aspect of artificial intelligence in agriculture. AI techniques are able to analyze several attributes in crop production to predict yield.
Soil moisture is one of the most important factors affecting health of crops. Water stress significantly affects crop yield. Although soil moisture can be quantified as a value (percentage or index), this value has a temporal (time) attribute, and cannot accurately be expressed by single value such as average.
In research labs where all parameters could be controlled, role of moisture stress can be evaluated as an individual parameter. In real farming, moisture stress is not a stand alone parameter, and it affects plants in combination with other temporal attributes such as humidity, temperature, wind speed, and sun radiation. Plants of similar input history (moisture stress, humidity, temperature, etc.) output alike. The challenge for farmers is that even if they have collected their input/output data over various farming seasons from different locations of a farm, predicting an output cannot readily be inferred from the high dimensional inputs that are spread over time.
Grain Data Solutions deals with such projects using an AI technique called DTW, which can find farms of similar history even if there are many inputs measured over time.
Applying this method on yield of corn farms affected by moisture stress over multiple farming seasons, we could predict yield with error less than 5% in most cases.
Our approach could predict yields where number of farming experiments where limited and variation was high in yield from one to another farming season.
