Predicting yield loss in maize fields and developing decision support for post‐emergence herbicide applications

Summary This work was initiated to integrate an image analysis system and a prediction equation to support decisions for post‐emergence herbicide applications under field conditions. Data were collected from 1999 to 2001 in 32 commercial fields to obtain weed cover data at the three to four leaf stage of maize ( Zea mays L.), and crop yield at maturity. Relative crop yield was predicted using a non‐linear sigmoidal equation with relative weed cover as the predictor variable ( P < 0.0001; R 2 = 0.39). The decision procedure consists of using the equation within the limits of a yield loss threshold that represents the loss one is willing to tolerate. The tolerance threshold (TT) allows determination of a weed threshold (WT). The procedure considers the variability around the prediction equation by setting the WT at the intersection between the lower 95% confidence interval of the prediction line and the TT. It also considers the variability around the weed cover estimate. For a given field, the decision is made by comparing the average weed cover corrected for sampling error, to the WT. We tested the performance of the decision procedure and found it could lead to a saving of 25% of herbicide use. We also computed a probability table showing the chances of getting relative yield above or below the TT. We suggest using the probability table in combination with the decision procedure to manage risks. The proposed approach does not offer a set ‘yes’ or ‘no’ answer but rather provides a framework to support decisions by producers who ultimately must manage the risks.