Calibration and Validation of the APSIM Model for Sugarcane Yield Simulation and Climate Change Impact Assessment in Coastal Andhra Pradesh, India
Ch. Apparao *
Dr. NTR College of Agricultural Engineering, Bapatla, India.
A. Mani
ANGRAU, Lam, Guntur, India.
K. Krupavathi
Dr. NTR College of Agricultural Engineering, Bapatla, India.
S. Prathibha Sree
RARS, ANGRAU, Lam, Guntur, India.
A. Ashok Kumar
Dr. NTR College of Agricultural Engineering, Bapatla, India.
*Author to whom correspondence should be addressed.
Abstract
Sugarcane (Saccharum officinarum L.) is a strategically important agro-industrial crop, and its productivity is increasingly threatened by climate variability and climate change. This study calibrated and validated the Agricultural Production Systems Simulator (APSIM) for sugarcane yield simulation in coastal Andhra Pradesh, India, and projected future cane yields for 25 years (2026–2050) under representative climate-change scenarios. The APSIM-Sugarcane module was calibrated using field observations from 2021 to 2023 and independently validated against 2024 data. The model performance was quantified using the Root Mean Square Error (RMSE), Mean Absolute Error (MAE), coefficient of determination (R²), and Nash–Sutcliffe Efficiency (NSE). During calibration, the model showed a systematic underestimation tendency (mean bias = −2.73 t ha⁻¹), whereas validation against 2024 data revealed overestimation (+5.96 t ha⁻¹, +8.07%). The overall model performance metrics were RMSE = 3.59 t ha⁻¹, MAE = 3.07 t ha⁻¹, R² = 0.748, and NSE = 0.136, confirming moderate predictive skill with acceptable absolute accuracy (~5.2% of the mean observed yield). Annual future projections (2026–2050) indicated substantial inter-annual variability (range: 74.08–112.12 t ha⁻¹), with a declining decadal mean trajectory from 103.11 t ha⁻¹ (2026–2030) to 95.73 t ha⁻¹ (2031–2040), and 90.83 t ha⁻¹ (2041–2050). The projected absolute peak was 112.12 t ha⁻¹ in 2028, whereas the minimum was 74.08 t ha⁻¹ in 2049. Future projections were generated using CMIP6 multi-model ensemble data under the SSP2-4.5 scenario through a delta-change downscaling approach applied to the 1991–2024 baseline climate record. However, projection uncertainty may arise from climate model variability, limited calibration years, and static management assumptions used in the simulations. These trends reflect the progressive shift from CO₂ fertilization dominance in the near term to increasing thermal stress in the later decades. This study provides a transparent, data-driven, and quantitative basis for climate-informed sugarcane production planning in coastal Andhra Pradesh.
Keywords: Sugarcane yield simulation, model calibration and validation, climate change impact, inter-annual variability, CO₂ fertilization