Modelling variation in the physiology of Bambara Groundnut (Vigna subterranea (L.) Verdc. )

The aim of this PhD project was to construct a model based on physiological and socio-economic factors related to the growth, development and yield of bambara groundnut landraces in relation to their environment. The model (BamGro) is an adaptation of the PALM (Matthews, 2005) model for a leguminous crop. It is a sink-orientated model, i.e. the number of available sinks (pods) determines the final production. The model is a stand-alone computer program written in Delphi 6 (Borland®). It uses climate data, landrace specific parameters and physiological relationships and runs on a daily time-step to determine the biomass production and yield of a landrace in a specific environment. The parameters of the model have been determined with experiments in the field (Swaziland) and glasshouses (TCRU, University of Nottingham). Large differences between glasshouse data and field data in leaf appearance rate and consequently leaf area development were found. In this study the leaf appearance rate was typically three times higher in the field, than in the glasshouse for the same landrace. When the relation between leaf area per plant and leaf number per plant is observed, there is no difference between the UK and Swaziland. The differences between the field and the glasshouse are therefore likely to be the result of an effect of environment on the leaf initiation. These differences meant that the model had to be developed with two different parameter sets, one for the landraces used in the field and one for the landraces used in the glasshouse. BamGro is capable of describing differences between landraces, and the influence of both drought and photoperiod are simulated using a simplified approach, and these aspects can be improved when sufficient high quality data becomes available. BamGro has been validated against three independent sets of data. BamGro achieves an excellent fit between observed and predicted data for leaf area index and pod yield, but underestimates the total above ground biomass by 50% in the TCRU glasshouses (2003 season). For the Swaziland ‘Malkerns’ field site (2002-2003 season) BamGro predicts the total above ground biomass excellently. BamGro achieves a good fit between observed and predicted pod yield data, but underestimates the leaf area index. For the Swaziland ‘Luve’ field site (2002-2003 season) the predictions are poor, with the model underestimating the total above ground biomass, leaf area index and pod yield for most landraces. BamGro is most sensitive to its crop parameters. BamGro seems not to be sensitive to changes in seasonal rainfall or initial soil moisture content. The unavailability of data on soil water relations and incomplete agronomic data sets meant that the water routines of the model could not be validated against field data from Namibia and Botswana. Three potential uses for BamGro have been presented.