A Bayesian latent‐scale model for quantifying turfgrass seasonality from turf quality ratings in the National Turfgrass Evaluation Program

Seasonal variation in turfgrass quality is important for cultivar evaluation in the National Turfgrass Evaluation Program (NTEP). However, quantifying these patterns directly from visual ratings is difficult, because these ratings are ordinal, subjective, and collected repeatedly from spatially structured field trials. We developed a Bayesian latent‐scale model to estimate entry‐specific seasonal patterns from NTEP turf quality ratings while accounting for rater‐specific scale use and within‐trial spatial heterogeneity. The model combines item response theory with Gaussian processes to place ratings on a common latent quality scale and estimate smooth seasonal trajectories for each entry. We applied the model to data collected from the 2017 NTEP Kentucky bluegrass ( P. pratensis L.) trial in Adelphia, NJ, and evaluated it using model comparison, blocked held‐out‐date validation, and synthetic parameter recovery. Models that included entry‐specific seasonality improved predictive performance relative to matched spatial‐only models, indicating that the seasonal structure explained important variation in the rating data. The fitted trajectories summarized differences among entries in seasonal timing, amplitude, and average latent turf quality, providing a more informative view of cultivar performance than a single static summary or direct aggregation of raw ratings. The rater‐threshold formulation described differences in category use among raters, although it was not the best predictive rating specification for this dataset. Overall, this analysis demonstrates how Bayesian latent‐scale modeling can quantify turfgrass seasonality from ordinal NTEP ratings, while broader validation across additional locations, years, species, and traits remains needed.