Comparative Analysis of Conventional and Non-linear Growing Degree Day Methods for Wheat Yield Prediction in Punjab, India

Abstract

Aims: To evaluate the accuracy and reliability of three Growing Degree Day (GDD) calculation methods—linear (Method 1), improved linear (Method 2), and cosine-wave (Method 3)—for predicting wheat phenological stages and yield in Punjab, India, to optimize agricultural practices in semi-arid regions. Study Design: Comparative field-based experimental study. Place and Duration of Study: Research Farms of Punjab Agricultural University, Ludhiana; Research Station, Faridkot, and Research Station, ballowal Saunkhri. Methodology: Field experiments involved the wheat variety Unnat PBW 550, planted on November 15th at three locations. Weather data from Agrometeorological Observatories were used to calculate GDD using three methods. Phenological stages were observed, and GDD was computed for each method. Model performance was assessed using the coefficient of variation (CV), Willmott’s refined index of agreement (dr), and logistic modeling of dry matter accumulation (R² and RMSE). Results: GDD for maturity at Ballowal Saunkhri was 1880.8 ± 34 °C (Method 1), 1710.65 ± 34 °C (Method 2), and 1618 ± 38 °C (Method 3). Method 3 showed the lowest CV (3.5% for tillering) and highest dr (0.98 in 2020 at Ballowal Saunkhri). Logistic modeling indicated Method 3’s superior accuracy (R² = 0.989, RMSE = 0.039 at Ballowal Saunkhri) compared to Method 1 (R² = 0.982, RMSE = 0.047) and Method 2 (R² = 0.970, RMSE = 0.062). Conclusion: The cosine-wave method (Method 3) offers superior precision for GDD estimation, enhancing wheat yield predictions and resource management in semi-arid regions. Further validation across diverse crops and climates is recommended.