Sugarcane for water-limited environments: enhanced capability of the APSIM sugarcane model for assessing traits for transpiration efficiency and root water supply

Abstract Genetic variation in traits affecting transpiration efficiency has been reported in sugarcane, but the impact of this variation on yield in a range of production environments needs to be estimated for assessing the priority and selection weightings to apply to these traits in crop improvement programs. A modelling approach may be useful and even necessary for reasonable assessment of these traits across production environments with different and temporally variable levels of water availability. Earlier theoretical modelling using the Agricultural Production Systems Simulator (APSIM)-Sugar module, found that traits affecting transpiration efficiency (TE) and water uptake by roots were important for improving sugarcane grown with highly variable rainfall. However, there were limitations of APSIM-Sugar in accommodating key physiological mechanisms, and this led to a revision of the APSIM-Sugar module described in this paper. Four key features were added to enable genetic variation in TE traits and root water supply (RWS) known to exist, to be modelled and assessed for predicted impact. These features were 1) the response of TE to water stress, 2) the midday flattening of hourly transpiration when plants are stressed, 3) conductance limits to hourly transpiration, which can apply even without stress and 4) the separation of soil hydraulic conductivity ( k) and root length density ( l) rather than the use of combined kl for determining RWS. A dataset of 182 observations of above ground biomass from 13 field experiments of sugarcane were used to check firstly that the new sugarcane module did not affect the simulation results when all the new features were disabled (model stability), secondly to check that the new features did not greatly reduce model performance, and thirdly, to determine the response or sensitivity of yields to the new features. Variation in parameter settings for the new features were based on the best evidence available for genetic variation in these traits, and were set before any testing against observations were made. With these features enabled in partial factorial combinations, the accuracy of the simulation of observed biomass, changed only to a minor extent compared to when no features were enabled. Separating k and l had the most consistent effect on improving model performance. Three additional published experiments with varying levels of imposed water stress were also simulated, with and without the new features. The simulation of biomass yield from two of the experiments was remarkably accurate regardless of which features were used in the simulation. Dry stalk yields reported for the third experiment were simulated accurately when no features were enabled and when midday flattening of transpiration was enabled and separate k and l enabled, one at a time. When a limit was placed on hourly transpiration, simulated TE and yield increased substantially when water was limiting but not in well-watered conditions; where yield was reduced. The new APSIM-Sugar features address the limitations of the original module (developed in 1999) for assessing water use related traits, including TE component traits and root growth. Effects of conductance on TE can now be simulated by limiting maximum hourly transpiration at the leaf level or k and specific root length at the root level. The new sugar module will also allow for better discrimination between sugarcane cultivars, now that vigour traits such leaf area development and radiation use efficiency are linked to root water uptake.