Abstract Grain sorghum Sorghum bicolor (L.) Moench response to stand reduction following hail events may have changed over time due to improvements in crop genetics and management. Field trials were imposed from 2017 to 2020 in College Station, TX, and Manhattan, KS, to evaluate the impact of stand reduction on sorghum yield and yield components. A split‐plot randomized block experimental design with four replications was used to evaluate stand reductions (0%, 25%, 50%, 75%, and 90%) and timings (GS2 = six‐leaf stage, GS5 = late boot, and GS7 = soft dough, where GS is growth stage) at two seeding rates (subplots). In Texas, grain yield averaged 6.13 Mg ha − 1 without stand loss, and high seeding rates yielded 9.2% more than low rates. Yield losses increased with later growth stages and higher stand reduction, reaching 86.2% yield loss at 90% stand loss during GS7. In Kansas, mean yield was 7.98 Mg ha − 1 without stand reduction. Yield was maintained up to 50% stand loss at GS2 and 25% at GS5, but declined at any level during GS7, with 87% yield loss at 90% stand reduction. Early‐stand loss (GS2) allowed yield compensation through tillering to maintain panicle density. GS5 relied more on increased grain number per panicle, and GS7 showed sharp declines with any loss. Grain size contributed less to yield compensation but increased slightly with severe reductions. Machine learning models (gradient boosting and deep neural networks) outperformed traditional linear and nonlinear regressions for predicting yield loss, with gradient boosting achieving the highest accuracy (root mean square error: 10%–11%).
Nielsen et al. (Fri,) studied this question.