The evolution and evaluation of dairy cattle models for predicting milk production: an agricultural model intercomparison and improvement project (AgMIP) for livestock

Abstract

The contemporary concern about anthropogenic release of greenhouse gas (GHG) into the environment and the contribution of livestock to this phenomenon have sparked animal scientists’ interest in predicting methane (CH4) emissions by ruminants. We contend that improving the adequacy of mathematical nutrition model estimates of production of meat and milk is a sine qua non condition to reliably determine ruminants’ worldwide contribution to GHG. Focusing on milk production, we address six basic nutrition models or feeding standards (mostly empirical systems) and five complex nutrition models (mostly mechanistic systems), describe their key characteristics, and highlight their similarities and differences. We also present derivative systems. We compiled a database of milk production information from 37 published studies from six regions of the world, totalling 173 data points: 19 for Africa, 45 for Asia, 16 for Europe, 12 for Latin America, 44 for North America and 37 for Oceania. Four models were used to predict milk production in lactating dairy cows, and the adequacy of their predictions was measured against the observed milk production from our database. Even though these mathematical nutrition models shared similar assumptions and calculations, they have different conceptual and structural foundations inherent to their intended purposes. A direct comparison among these models was further complicated by the different models requiring unique inputs that are very often not available, and the low reliability of the inputs prevents an unbiased assessment of the model predictions. Very few studies have collected the necessary information to run more mechanistic systems, and users have to rely on standard information to populate many model inputs. Study effect was a critical source of variation that limited our ability to conclusively evaluate the models’ applicability under different scenarios of production around the world. Only after study variation was removed from the database did the adequacy of the model predictions of milk production improved, but deficiencies still existed. On the basis of these analyses, we conclude that not all models were suitable for predicting milk production and that simpler systems might be more resilient to variations in studies and production conditions around the world. Improving the predictability of milk production by mathematical nutrition models is a prerequisite to further development of systems that can effectively and correctly estimate the contribution of ruminants to GHG emissions and their true share of the global warming event.

Invited review: Nitrogen in ruminant nutrition: A review of measurement techniques

“…Current protein evaluation systems estimate rumen N balance, which is indicative of N surplus or shortage, to sustain MPS in the rumen (reviewed by Dijkstra et al, 1998, andTedeschi et al, 2014). As discussed earlier, estimates of rumen availability and degradability of dietary N and OM are obtained by external washing procedures and in situ incubations of feeds in the rumen.…”

Modeling European ruminant production systems: Facing the challenges of climate change

“…For modelers of ruminant production systems, the complexity of farm-scale interactions creates a major challenge for the scaling up of 'animal' and 'field' scale modeling to the national, regional and global levels most relevant for policy makers. A range of modeling approaches has been applied to European ruminant livestock systems and their various components (Box 1) with a number of technical reviews providing comprehensive comparisons of a range of models, for example (Holzworth et al, 2015;Snow et al, 2014;Tedeschi et al, 2014).…”

The Contribution of Mathematical Modeling to Understanding Dynamic Aspects of Rumen Metabolism

“…Such an approach is fundamentally different from the approaches used in the aforementioned static models, which may represent multiple microbial classes but lack a representation of dynamic interactions, making the final modeling outcome for the whole rumen the arithmetic summation of the outcomes of the representations for the individual functional classes. Recently, other modeling efforts have been undertaken to improve the representation of interactions between functional classes of micro-organisms (review Tedeschi et al, 2014). For example, a new version of the rumen submodel in CNCPS was constructed that includes more detailed aspects of growth and activity of Holotrichs and Entodiniomorphid protozoa (Higgs, 2014).…”

“…Whether the extra detail introduced by such a consensus outweighs the benefit of an empirical approach based on in vivo measurements in the target animal (Bannink et al, 2006b; Ghimire et al, 2014) remains unclear. Other aspects that have been incorporated in the various rumen models reported in literature include effective fiber, peptides, distribution of particles, fats and fatty acids, and sulfate (review Tedeschi et al, 2014), which have their own merit when aiming to explain rumen response to related factors.…”