Improved temperature and photoperiod response functions to better link genotype to phenotype prediction for soybean phenology

Abstract

Gene-to-phenotype modeling requires robust mechanistic frameworks, yet widely used soybean phenology models rely on older datasets with limited genotypic representation and do not explicitly account for twilight effects. We quantified soybean phenology responses to temperature (11–35 °C) and photoperiod (12–16 h) using 20 modern cultivars spanning maturity groups (MG) 0–7, evaluated across 12 controlled-environment experiments and 13 field experiments in the USA. Our objectives were to (1) reconstruct temperature and photoperiod response functions for three developmental phases (emergence – flowering, flowering – beginning seed filling, and beginning seed filling – physiological maturity), (2) determine whether these responses vary among cultivars, and (3) identify the twilight value that maximizes prediction accuracy and evaluate an improved soybean phenology module. Temperature responses were best described by a curvilinear function, with optimum temperature declining from vegetative to reproductive development (30.4 to 22.1 °C). Photoperiod responses followed a bilinear model, with photoperiod sensitivity and minimum phase duration varying systematically with maturity group, whereas no photoperiod response was detected during the seed filling period. Prediction accuracy across independent field experiments was maximized using a twilight angle of −6°, with an uncalibrated multiplicative model achieving a root mean squared error (RMSE) <10 days and an index of agreement (d-index) >0.96. Our results show that variation in photoperiod response is driven primarily by photoperiod sensitivity rather than the critical photoperiod threshold, improving cultivar parameterization, reducing equifinality, and providing a stronger physiological foundation for improving crop model prediction accuracy across diverse genotypes and environments.