Modeling nutrient stress in paddy rice: Calibration practicalities, data needs and performance trade-offs for AquaCrop and Decision Support System for Agrotechnology Transfer models

Abstract

Importance of the work: Often, the application of crop simulation models for nutrient
management in paddy rice is constrained by the need for demanding calibration and
intensive data requirements.
Objectives: To compare the calibration methods, data requirements and performance
of the AquaCrop and Decision Support System for Agrotechnology Transfer (DSSAT)
models in simulating the effects of nitrogen on RD43 rice.
Materials and Methods: A pot experiment was conducted using two planting methods
(wet direct sowing and transplanting) and three nitrogen levels: recommended (N1),
approximately one-half-recommended (N2) and zero-application (N3). Data from this
experiment were used to assess model performance in simulating biomass, yield and crop
evapotranspiration for RD43 rice.
Results: Nitrogen availability was a key determinant of observed biomass and yield,
with yields diminishing as the N application was reduced from N1 to N3. Although both
models simulated yield accurately under high fertility (N1), their predictions diverged as
fertility declined. Under moderate (N2) and severe (N3) stress, AquaCrop substantially
overestimated yield due to calibration limitations. Conversely, DSSAT produced
a more graduated response across the N gradient but was more demanding to calibrate.
Both models consistently underestimated total crop evapotranspiration.
Main finding: AquaCrop was suitable for general assessment, yet its simplicity
constrained precision under severe nutrient deficiency. DSSAT produced a more
graduated response at the cost of a more demanding calibration, revealing the necessity to
refine paddy-specific biogeochemistry representations in models.