Liu, Jian2022-02-272022-02-272021https://hdl.handle.net/10182/14639Lucerne is an ideal plant for east coast sheep and beef farmers to integrate on-farm to cope with a dry and drying climate. Expanded use of lucerne on-farm leads to on-farm questions around feed supply, environmental impacts and farm resilience. Many of these questions can only be answered with process-based models. These have been helping researchers, policy makers and farmers to make informed decisions about farm practices worldwide. However, the current lucerne model in the Agricultural production system simulator next generation (APSIMX) is not designed for simulating lucerne responses to dryland conditions. Hence, this study aimed to incorporate previous knowledge obtained from dryland experiments from Lincoln University into the APSIMX-Lucerne mode. New equations were introduced to the model to constraint the growth and development processes including leaf area expansion rate, radiation use efficiency and phyllochron under water-limited conditions. Secondly, this study investigated an alternative approach for increasing the efficiency and reliabilities of model parameter estimation. The reproducibility of the study was addressed as the third objective by adapting data science concepts and state-of-art tools. A literature review of lucerne responding to water stress (water deficiencies) laid out the physiological knowledge to design the mechanism of the APSIMX-Lucerne model. The thesis initially documents previous experiments, the APSIMX framework and data science tools. The conventional approach for APSIMX model development was conducted to gain experience and understanding of the structure and operation of the APSIMX-Lucerne model, verify the implementation of the lucerne model in the APSIMX framework and identify major issues with current model implementation to guide subsequent improvements. An alternative approach was applied via the R programme and a workflow manager to implement an optimisation procedure for estimating nine water-related parameters in a simple APSIMX water balance model. The optimised parameter values were later transferred into the APSIMX-Lucerne model to evaluate the model performance compared with the conventional approach. Negligible improvement (1% normalised root mean square error reduction) was gained in profile soil water content prediction for dryland trial by the alternative approach although it achieved full reproducibility and quantifiable resource expenses. These results might be caused by multiple contributors. First, both approaches demonstrated that the demand-related parameters in the model were inadequate to impose the correct waster stress effects on lucerne. More specifically, the model failed to constrain above ground variables at water supply limited conditions while the model extracted inadequate water from soil at water demand limited conditions. Furthermore, the model had no mechanism to represent water stress effects on lucerne height. More investigation is necessary to implement the relationship between water stress levels and lucerne height since it is lacking in the current model. Moreover, root distribution patterns differ in different soil types, whereas this study assumed an exponential decay distribution for both stony and non-stony soils, which may not represent reality in deep soils. Lastly, lucerne showed different phenological development in the supply limited conditions in comparison with temperature-driven development under non-limiting conditions. Therefore, in-silico approaches, such as Bayesian inference or specialised root structure mechanistic models, might be required to assist the understanding in the full picture of lucerne growth and development under dryland conditions.endrylandwater stresslucernesoil water balancetranspirationwater extraction patternsoil characteristicsAgricultural production system simulator next generation (APSIMX)water deficitAPSIMX model (Agricultural production system simulator)ThesisANZSRC::30 Agricultural, veterinary and food scienceshttps://creativecommons.org/licenses/by/4.0/Attribution 4.0 International