You are here

Better Way to Predict Corn Yields

By Gene Johnston
This question may surface in your head as you harvest your corn this fall: Is there a better way to predict corn yields during the growing season?

Haishun Yang thinks so. The University of Nebraska agronomy professor is a computer modeling expert who is using a program to crunch numbers into useful data for farmers.

The program he’s developed, called Hybrid-Maize, is a computer application that lets you upload all the pertinent weather (and some agronomic) conditions for your farm. At any point in the growing season, it will take your rainfall, soil moisture, daily temperature readings, planting date, and plant population, and give you an estimated corn yield. Yang says the Hybrid-Maize software has a very friendly graphic interface and is easy to use.

Yang says the model relies on a combination of real-time weather data and historical weather data, the single biggest influencer of corn yields. At whatever point in the season you run the yield estimate, it first uses this season’s up-to-date weather, then it uses each of the previous years in the historical data to finish the rest of the season.

For instance, if you run the program on June 15, it takes conditions up to that point and assumes the rest of the season could look like each of the previous years from that point forward.

As the season progresses, the yield prediction gains accuracy as current weather is updated. 

“If you have 20 years of historical data, the program will calculate 20 possible outcomes from best to worst and then give you an average,” says Yang. “As the summer goes on, it gets closer to predicting the actual final yield, as there is less time to deviate from the average.”

In 2014, Yang’s group tested the Hybrid-Maize model with Extension collaborators from 25 locations across the Corn Belt (Kansas to Ohio). They entered their local conditions from various official reports, and the model estimated their local yields.

If enough of those local collaborators (including farmers) are strung together, Yang says, the data could be used to estimate yield on a regional, state, or national basis throughout the growing season. The more participants and the more historical data that are included, the more accurate it becomes at predicting final yield.

“When the Hybrid-Maize model simulates corn yield of a particular field, it assumes that the field is uniform in terms of crop management and soil properties,” says Yang. “Such a field can represent an area where conditions are similar, including weather, crop management, and soil properties.”

The good news is that most of the data needed is available in public records such as National Weather Service (NWS) data or USDA digital soil maps. Other pertinent things, including planting date, plant population, and hybrid maturity of the growing crop, are also included as inputs. “All of this data is out there,” says Yang.

Surprisingly, he says, the NWS data isn’t as easy to access as you might think. Some of its stations are not integrated into the national network, so it’s not automatic that you can get and upload its data.

Limitations
Of course, yield estimations ahead of harvest have limitations. The Hybrid-Maize program can’t account for excessive nutrient deficiencies, hail damage, weed infestations, or crop diseases. It assumes those issues will conform to historical averages.

“In a year when a crop disease is unusually severe, this model will have a hard time picking up on it, because it doesn’t incorporate ground observation,” says Yang.

He points to the 2012 crop year as an illustration of this. That year grew increasingly dryer as the summer wore on. Each biweekly yield forecast assumed average or normal weather from that point on, but it only got drier. The full extent of yield damage wasn’t known until harvest.

In 2014, the Hybrid-Maize model was very accurate at an early date in predicting near-record yields.

“It is noteworthy that in our first 2014 yield forecast on July 20, we anticipated with a high level of probability above-average dryland corn at 11 of the 14 locations that ultimately reached above-average yields,” points out Yang.

Some cornfields in 2014 suffered from late-season nitrogen shortages due to excess rainfall, he continues. The Hybrid-Maize model did not account for that, and in a few cases, it predicted higher yields than actually materialized.

Yang says many people have shown interest in his computer model, starting with farmers. Also, insurance companies, the seed industry, food companies, and the biofuels industry have great interest in early forecasts of crop yields.

“The more you know, the better,” says Yang. A similar model for soybeans is in the beginning stages.

In the future, he hopes to add inputs that will make the Hybrid-Maize model better, such as local management issues like the amount of no-till and water conservation techniques.

“We wish we knew more about water stress on the crop at different stages of production,” he adds.

USDA crop reporting officials have been reporting in-season yield estimates for years. Their estimates involve more local sampling of crop conditions on the ground than does the Hybrid-Maize.

Both systems are good, says Yang, but a combination may prove to be the best. Hybrid-Maize is more efficient at pulling data together from various sources, he says, because it doesn’t require people to go out and do the scouting.

Read more about