According to the framework presented in Chapter 2, assessing the scope to increase production of feed-crop livestock systems requires to benchmark both feed crop production and livestock production (Fig. 7.1). Crop growth models based on concepts of production ecology are widely used to assess the scope to increase crop
production (Bouman et al., 1996, Jones et al., 2003, Keating et al., 2003, Van Ittersum et al., 2003), and can thus be readily applied to feed crops (x-axis Fig. 7.1). Literature review showed that the current livestock models were developed for other purposes than assessing the scope to increase production generically (Appendix 1A). Although many models contain aspects of concepts of production ecology, a generic model to assess the scope to increase livestock production (y-axis Fig. 7.1) was not available at the start of this research. Chapter 3, therefore, described the model LiGAPS-Beef, which aims to simulate potential and feed-limited production of beef cattle in different beef production systems under different agro-ecological conditions. This model combines existing models and concepts on thermoregulation (McGovern and Bruce, 2000, Turnpenny et al., 2000a), feed intake and feed digestion (Chilibroste et al., 1997), and energy and protein utilisation (NRC, 2000, CSIRO, 2007). The novelty of LiGAPS-Beef thus lies in the fact that it combines existing models which were never combined before. This combination provided new ways to visualise the most constraining factors for livestock production on a daily basis in Chapter 3 (Fig. 3.5, Supplementary Information Fig. S12-S31) and Chapter 4 (Fig. 4.4). Such graphs clearly illustrate which factor constrains livestock production at what moment, and provide opportunities to identify effective improvement options. LiGAPS-Beef was developed with the purpose to estimate potential and feed-limited production of farming systems with beef cattle in different agro-ecological environments. In Chapter 4, the model was tested by simulating live weight gain in beef production systems in Australia, Uruguay, and the Netherlands. Evaluation of LiGAPS-Beef at animal level showed that live weight gain was predicted fairly well (mean absolute error = 15.4% of measured average daily gain). Together with the evaluation of sub-models in Chapter 3, the results of Chapter 4 provide confidence that LiGAPS-Beef is suited for its purpose.
7.3.2 Data availability and data accuracy for model evaluation
The performance of LiGAPS-Beef was evaluated for three different beef production systems (Chapter 4). Evaluating the model for more systems may further advance insight in its validity domain. Model evaluation is, however, hampered by a significant lack of experimental data. Firstly, experimental data are abundant for specific life phases of individual animals, but evaluation of LiGAPS-Beef requires preferably data over entire life spans of all animals within herds or flocks. Such data are scarce, since long-term experiments with multiple animals are costly and time-consuming. As a result, livestock production at herd and farm level is generally not measured in experiments (Morel et al., 2016).
additions to evaluate the predictions of LiGAPS-Beef with regard to edible beef production will be measuring the carcass percentage and the percentage of edible beef in carcasses.
Thirdly, LiGAPS-Beef is a dynamic model requiring daily inputs of weather, feed quality, and feed quantity. The accuracy of model output is expected to increase with an increasing accuracy of measured input data, and with smaller time steps. Measured weather data are freely accessible in on-line repositories for several regions in the world (AGBOM, 2016, NIWA, 2016). Generated or intrapolated weather data are available also for several regions (Agri4Cast, 2013), although these are inferior to measured weather data. Availability of weather data was generally not a bottleneck for model evaluation in this thesis, but it might be when simulating beef production systems in countries where weather data are hardly available or accessible. Experimental data about feed quality and the available feed quantity were much more scarce than weather data during model evaluation. The feed quality of feed types was often not measured in experiments. If absent, feed quality was assumed to correspond with the default values for feed quality given in feed tables (Jarrige, 1989, Chilibroste et al., 1997, Kolver, 2000). The quality of some feed types, such as grasses, is known to vary significantly among grass species, grass cultivars, geographical locations, and seasons (Smith et al., 1998). In addition, grass quality is affected by management and nitrogen fertilisation (Hoekstra et al., 2007). The accuracy of the output of LiGAPS-Beef is likely to decrease if input data for forage quality are inaccurate.
Inaccurate data for crude protein content are not likely to affect beef production, as protein is not among the main constraining factors for growth in Chapters 3 and 4. Inaccurate data for the metabolisable energy content, however, do affect beef production. Sensitivity analysis in Chapter 4 showed that a 10% change in metabolisable energy content (conversion from digestible to metabolisable energy) resulted in a larger change of feed efficiency of ¾ Brahman × ¼ Shorthorn cattle in Australia (14%) and of Hereford cattle in Uruguay (12%) under feed quality limitation. These results suggest that some errors in feed quality result in even larger errors in the estimates of feed efficiency at herd level, and highlight the importance of accurate feed quality data.
Despite the difficulties in model evaluation, the silver lining is that simulations with LiGAPS-Beef allow more targeted measurements in livestock systems. The sensitivity analysis performed in Chapter 4 identified the most influential parameters. With regard to energy and protein utilisation, future experiments may measure and calculate energy requirements for maintenance, the conversion from digestible to metabolisable energy, protein absorption, and protein accretion efficiency. With regard to thermoregulation, the sweating capacity, body area, and heat transfer between the body core and skin may be determined more precisely. Measuring the
genetic potential for growth and calculating parameters of the Gompertz curve is also key to ensure model accuracy. Using measurements this way, simulation and experimentation can reinforce each other.