Application of CERES-Sorghum : Sensitivity Analysis.

Understanding the grain sorghum models will provide useful insights about the crop’s functioning and the interactions between the crop and their environments. While modeling sorghum, the questions we need to ask are (1) how well does the model under consideration represent the underlying physics of crop growth and yield? (2) What confidence can one have on the model’s numerical results to be correct? (3) How far can the calculated results be extrapolated? (4) How can the predictability and/or extrapolation limits be extended and/or improved? The objective of this work is to attempt answer the first and forth questions using sensitivity (SA) and uncertainty analyses (UA).

The developed methodology is demonstrated using decision support system for the Agro-transfer technology (DSSAT) embedded with CERES-sorghum for a location in Kansas (Manhattan), a leading state for grain sorghum production in USA. OAT (One at a time) method of local SA is carried out on genetic (five), climatic (four), soil (six) and agronomic (four) input parameters. In OAT method, each input parameter is perturbed at a time. The results of SA were analyzed using mathematical (Sensitivity index; SI) and graphical approaches for combinations of input parameters and six model response variables (yield, biomass, leaf area index, leaf number at maturity, days to anthesis and maturity). To the best of our knowledge no studies have done such an elaborate SA for CERES-sorghum.

Results revealed that depending on the response variable, the ranking of sensitive parameters/variables varied. For example, temperature is the most sensitive factor for all response variables except leaf number. For leaf number, two genetic parameters are the most sensitive. A small decrease in temperature (1-2⁰C) increases the yield by 7-9 %, whereas for more extreme change in temperature (±4⁰C), the decrease was about 30%. In general, based on the SA carried out using graphical and mathematical approaches, the underlying physical crop growth and yield are captured by the model. Although sorghum yields are often found to be constrained by water, surprisingly, the model is not found to be very sensitive to rainfall. We intend to explore this in detail in future.