Comparison of Molly and Karoline models to predict methane production in growing and dairy cattle

Numerous empirical and mechanistic models predicting methane (CH₄) production are available. The aim of this work was to evaluate the Molly cow model and the Nordic cow model Karoline in predicting CH₄ production in cattle using a data set consisting of 267 treatment means from 55 respiration chamber studies. The dietary and animal characteristics used for the model evaluation represent the range of diets fed to dairy and growing cattle. Feedlot diets and diets containing additives mitigating CH₄ production were not included in the data set. The relationships between observed and predicted CH₄ (pCH₄) were assessed by regression analysis using fixed and mixed model analysis. Residual analysis was conducted to evaluate which dietary factors were related to prediction errors. The fixed model analysis showed that the Molly predictions were related to the observed data (± standard error) as CH₄ (g/d) = 0.94 (±0.022) × pCH₄ (g/d) + 31 (±6.9) [root mean squared prediction error (RMSPE) = 45.0 g/d (14.9% of observed mean), concordance correlation coefficient (CCC) = 0.925]. The corresponding equation for the Karoline model was CH₄ (g/d) = CH₄ (g/d) = 0.98 (±0.019) × pCH₄ (g/d) + 7.0 (±6.0) [RMSPE = 35.0 g/d (11.6%), CCC = 0.953]. Proportions of mean squared prediction error attributable to mean and linear bias and random error were 10.6, 2.2, and 87.2% for the Molly model, and 1.3, 0.3, and 98.6% for the Karoline model, respectively. Mean and linear bias were significant for the Molly model but not for the Karoline model. With the mixed model regression analysis RMSPE adjusted for random study effects were 10.9 and 7.9% for the Molly model and the Karoline model, respectively. The residuals of CH₄ predictions were more strongly related to factors associated with CH₄ production (feeding level, digestibility, fat concentrations) with the Molly model compared with the Karoline model. Especially large mean (underprediction) and linear bias (overprediction of low digestibility diets relative to high digestibility diets) contributed to the prediction error of CH₄ yield with the Molly model. It was concluded that both models could be used for prediction of CH₄ production in cattle, but Karoline was more accurate and precise based on smaller RMSPE, mean bias, and slope bias, and greater CCC. The importance of accurate input data of key variables affecting diet digestibility is emphasized.

Estimating in vitro ruminal ammonia-N using multiple linear models and artificial neural networks based on the CNCPS nitrogenous fractions of cattle rations with low concentrate/roughage ratios

The objectives of this study were to investigate the relationship between the in vitro ruminal ammonia nitrogen (NH₃ -N) concentration and the Cornell Net Carbohydrate and Protein System (CNCPS) N-fractions of feeds for cattle and further compare the performance of developing multiple linear regression (MLR) and artificial neural network (ANN) models in estimating the NH₃ -N concentration in rumen fermentation. Two data sets were established, of which the training data set containing forty-five rations for cattle with concentrate/roughage ratios of 50:50, 40:60, 30:70, 20:80 and 10:90 used for developing models and the test data set containing ten other rations with the same concentrate/roughage ratios with the training data set were used for validating of models. The NH₃ -N concentrations of feed samples were measured using an in vitro incubation technique. The CNCPS N-fractions (g), for example PB₁ (rapidly degraded true protein), PB₂ (neutral detergent soluble nitrogen), PB₃ (acid detergent soluble nitrogen) of rations, were calculated based on chemical analysis. Statistical analysis indicated that the NH₃ -N concentration (mg) was significantly correlated with the CNCPS N-fractions (g) PB₁ , PB₂ and PB₃ in a multiple linear pattern: NH₃ -N = (130.70±33.80) PB₁  + (155.83±17.89) PB₂ - (85.44±37.69) PB₃  + (42.43±1.05), R²  = 0.77, p < 0.0001, n = 45. The results indicated that both MLR and ANN models were suitable for predicting in vitro NH₃ -N concentration of rations using CNCPS N-fractions PB₁ , PB₂ , and PB₃ as independent variables while the neural network model showed better performance in terms of greater r² , CCC and lower RMSPE between the observed and predicted values.

A revised representation of ruminal pH and digestive reparameterization of the Molly cow model

Ruminal pH is a critical factor to regulate nutrient degradation and fermentation. However, it has been poorly predicted in the Molly cow model, and recent improvements in the representation of nitrogen cycling across the rumen wall altered some of the modeled responses to feed nutrients, resulting in some model bias. The objectives of this study were to further improve the representation of pH and to refit parameters related to ruminal metabolism and nutrient digestion in the model to resolve this bias, and to use the improved model to estimate nitrogen and energy fluxes with varying rumen-degradable protein (RDP; 40 vs. 60%) and ruminally degraded starch (RDSt; 50 vs. 75%). A meta data set containing 284 peer reviewed studies with 1,223 treatment means was used to derive parameter estimates for ruminal metabolism and nutrient digestions. Refitting the parameters significantly improved the accuracy and precision of the model predictions for ruminal nutrient outflow [acid detergent fiber (ADF), neutral detergent fiber (NDF), total N, microbial N, nonammonia N, and nonammonia nonmicrobial N], ammonia and blood urea concentrations, and fecal nutrient outflow (protein, ADF, and NDF). The prediction error for body weight was decreased from 19.3 to 6.2% with decreased mean bias (from 76.0 to 11.5%) and slope bias (from 17.2 to 7.7%), primarily due to improved representations of ruminal dry matter and liquid pool size. Adding ammonia concentration as a driver to the pH equation increased the precision of predicted ruminal pH and, thereby, the precision of predicted volatile fatty acid (VFA) concentrations, due to improved representation of pH regulation of VFA production rates. Although minor mean and slope bias were observed for ruminal pH and VFA concentrations, the concordance correlation coefficients indicated that much of the observed variation in these variables remains unexplained. Overall, the biological functions of nutrient degradation and digestion appear to be represented without bias. Simulated results indicated that decreasing RDP and RDSt proportions in an isonitrogenous and isocaloric diet can slightly improve N efficiency, and increasing RDSt proportions can increase energy efficiency.

Molly reborn in C++ and R

The dairy cow model 'Molly' is a mixed discrete event-continuous system model that simulates feeding, metabolism and lactation of dairy cows. Decades of model development have resulted in a valuable tool in dairy science. Due to the deprecation of the ACSL (Advanced Continuous Simulation Language) programming language, Molly has been translated into C++. This paper describes the translation process and discusses the advantages of the new implementation, one of which is the ability to run Molly within RStudio, a popular integrated development environment (IDE) for data science.

Use of Mechanistic Nutrition Models to Identify Sustainable Food Animal Production

To feed people in the coming decades, an increase in sustainable animal food production is required. The efficiency of the global food production system is dependent on the knowledge and improvement of its submodels, such as food animal production. Scientists use statistical models to interpret their data, but models are also used to understand systems and to integrate their components. However, empirical models cannot explain systems. Mechanistic models yield insight into the mechanism and provide guidance regarding the exploration of the system. This review offers an overview of models, from simple empirical to more mechanistic models. We demonstrate their applications to amino acid transport, mass balance, whole-tissue metabolism, digestion and absorption, growth curves, lactation, and nutrient excretion. These mechanistic models need to be integrated into a full model using big data from sensors, which represents a new challenge. Soon, training in quantitative and computer science skills will be required to develop, test, and maintain advanced food system models.

A redefinition of the modeled responses of mammary glands to once-daily milking

Milking cows once daily is a management tool that has been implemented to improve physical and financial results of seasonal pasture-based dairy farms. The Molly cow model integrates physiology and metabolism of dairy cattle; however, milk production during short-term changes in milking frequency (e.g., 1× milking) is not well represented. The model includes a representation of variable rates of cell quiescence and death. However, the rate constants governing cell death and the return of quiescent to active cells are not affected by milking frequency. An empirical assessment of the problem was conducted, and it was hypothesized that changing the current representation of the rate of cell death in response to short-term 1× milking would more accurately represent active and quiescent cells and improve predictions of milk production. An extra senescent cell flux was added to account for cell loss during periods of 1× milking. Additional changes included a gradual decline in the rate of 1× stimulated senescence during 1× milking, and a structural change in cell cycling between active and quiescent cells during and after short-term 1× milking. Data used for parameter estimation were obtained from 5 studies where 1× milking or different feeding strategies were tested. Parameter estimates of cell loss indicated that 1× milking would affect a small proportion of quiescent cells to cause extra cell death. This added cell senescence was influenced by the length of 1× milking such that cell senescence peaked on d 1 of 1× milking and decayed from that point. The new structure in the model includes a variable rate of cell death in response to 1× milking and a gradual rate of return of quiescent cells back to the active pool in response to switching to 2× milking after short-term 1× milking. Root mean square errors, mean bias, and slope bias declined by at least 50% for predictions of energy-corrected milk yield and fat percent. The model showed quantitative agreement with production data from short-term 1× milking. The accuracy of predictions was improved and the error was reduced by implementing modifications in the model in response to changes in milking frequency.

A revised representation of urea and ammonia nitrogen recycling and use in the Molly cow model

Accurately predicting nitrogen (N) digestion, absorption, and metabolism will allow formulation of diets that more closely match true animal needs from a broad range of feeds, thereby allowing efficiency of N utilization and profit to be maximized. The objectives of this study were to advance representations of N recycling between blood and the gut and urinary N excretion in the Molly cow model. The current work includes enhancements (1) representing ammonia passage to the small intestine; (2) deriving parameters defining urea synthesis and ruminal urea entry rates; (3) adding representations of intestinal urea entry, microbial protein synthesis in the hindgut, and fecal urea-N excretion; and (4) altering existing urinary N excretion equations to scale with body weight and adding purine derivatives as a component of urinary N excretion. After the modifications, prediction errors for ruminal outflows of total N, microbial N, and nonammonia, nonmicrobial N were 29.8, 32.3, and 26.2% of the respective observed mean values. Prediction errors of each were approximately 7 percentage units lower than the corresponding values before model modifications and fitting due primarily to decreased slope bias. The revised model predicted ruminal ammonia and blood urea concentrations with substantially decreased overall error and reductions in slope and mean bias. Prediction errors for gut urea-N entry were decreased from 70.5 to 26.7%, which was also a substantial improvement. Adding purine derivatives to urinary N predictions improved the accuracy of predictions of urinary N output. However, urinary urea-N excretion remains poorly predicted with 69.0% prediction errors, due mostly to overestimated urea-N entry rates. Adding representations of undigested microbial nucleic acids, microbial protein synthesized in the hindgut, and urea-N excretion in feces decreased prediction errors for fecal N excretion from 21.1 to 17.1%. The revised model predicts that urea-N entry into blood accounts for approximately 64% of dietary N intake, of which 64% is recycled to the gut lumen. Between 48 and 67% of the urea recycled to the gut flows into the rumen largely depending on diet, which accounts for 29 to 54% of total ruminal ammonia production, and 65 to 76% of this ammonia-N is captured in microbial protein, which represents 17% of N intake. Based on model simulations, feeding a diet with moderately low crude protein and high rumen-undegradable protein could increase apparent ruminal N efficiency by 20%.

Assessing the production and economic benefits from preventing cows grazing on wet soils in New Zealand

BACKGROUND

Intensive grazing by cattle on wet pasture can have a negative effect on soil physical quality and future pasture production. On a North Otago dairy farm in New Zealand, experimental plots were monitored for four years to assess whether preventing cow grazing of wet pastures during the milking season would improve soil structure and pasture production compared with unrestricted access to pastures. The DairyNZ Whole Farm Model was used to scale up results to a farm system level and ascertain the cost benefit of deferred grazing management.

RESULTS

Soils under deferred grazing management had significantly higher total porosity, yet no significant improvement in macroporosity (values ranging between 0.112 and 0.146 m(3)  m(-3) ). Annual pasture production did not differ between the control and deferred grazing treatments, averaging 17.0 ± 3.8 and 17.9 ± 4.1 t DM ha(-1) year(-1) respectively (P > 0.05). Furthermore, whole farm modelling indicated that farm operating profit was reduced by NZ$1683 ha(-1) year(-1) (four-year average) under deferred grazing management.

CONCLUSION

Deferring dairy cow grazing from wet Pallic soils in North Otago was effective in improving soil structure (measured as total soil porosity), yet did not lead to a significant increase in pasture production. Whole farm modelling indicated no economic benefit of removing cows from wet soils during the milking season. © 2016 Society of Chemical Industry.

TRIENNIAL LACTATION SYMPOSIUM: Systems biology of regulatory mechanisms of nutrient metabolism in lactation

A major role of the dairy cow is to convert low-quality plant materials into high-quality protein and other nutrients for humans. We must select and manage cows with the goal of having animals of the greatest efficiency matched to their environment. We have increased efficiency tremendously over the years, yet the variation in productive and reproductive efficiency among animals is still large. In part, this is because of a lack of full integration of genetic, nutritional, and reproductive biology into management decisions. However, integration across these disciplines is increasing as the biological research findings show specific control points at which genetics, nutrition, and reproduction interact. An ordered systems biology approach that focuses on why and how cells regulate energy and N use and on how and why organs interact through endocrine and neurocrine mechanisms will speed improvements in efficiency. More sophisticated dairy managers will demand better information to improve the efficiency of their animals. Using genetic improvement and animal management to improve milk productive and reproductive efficiency requires a deeper understanding of metabolic processes throughout the life cycle. Using existing metabolic models, we can design experiments specifically to integrate data from global transcriptional profiling into models that describe nutrient use in farm animals. A systems modeling approach can help focus our research to make faster and larger advances in efficiency and determine how this knowledge can be applied on the farms.

Screening for diets that reduce urinary nitrogen excretion and methane emissions while maintaining or increasing production by dairy cows

Farmers face complex decisions at the time to feed animals, trying to achieve production goals while contemplating social and environmental constraints. Our purpose was to facilitate such decision making for pastoral dairy farmers, aiming to reduce urinary N (UN) and methane emissions (CH4), while maintaining or increasing milk production (MP). There is a number of feeds the farmers can choose from and combine. We used 50 feeds (forages and grains) combined systematically in different proportions producing 11,526 binary diets. Diets were screened, using an a posteriori approach and a Pareto front (PF) analysis of model (Molly) outputs. The objective was to identify combinations with the best possible compromise (i.e. frontier) between UN, CH4, and MP. Using high MP and low UN as objective functions, PF included 10, 14, 12 and 50 diets, for non-lactating, early-, mid- and late-lactation periods, with cereals and beets featuring strongly. Using the same objective functions, but including ryegrass as dietary base PF included 2, 4, 8 and 4 diets for those periods. Therefore, from a wide range of diets, farmers could choose from few feeds combined into binary diets to reduce UN while maintaining or increasing MP. If the intention is maintaining pasture-based systems, there are fewer suitable options. Reducing UN will simply require dilution of N supplied by pasture by supplementing low N conserved forages. The results also evidence the risk of pollution swapping, reaching the frontier means arriving at a point where trade-off decisions need to be made. Any further reduction in UN implies an increment in CH4, or reduction in CH4 emissions increases UN. There is no perfect diet to optimize all objectives simultaneously; but if the current diet is not in the frontier some options can offset pollution swapping. The choice is with the farmers and conditioned by their context.

Evaluation of predictions of volatile fatty acid production rates by the Molly cow model

Predicting ruminal volatile fatty acid (VFA) production is important, as VFA are an energy source to the animal, affect nutrient partitioning, and dictate methane production. The VFA production submodel in the Molly cow model was evaluated using data from 8 publications that reported VFA production rates for cattle. Evaluations were conducted with ruminal water balance predictions enabled and the ruminal VFA stoichiometry coefficients set to "mixed" for all diets, or "mixed" when forage represented between 20 and 80% of the diet, "concentrate" when <20% forage, or "forage" when >80% forage. Prediction errors were relatively insensitive to changes in VFA coefficients by diet type. Root mean square prediction errors (RMSPE) were 63, 63, and 49% for acetate, propionate, and butyrate production rates, respectively. A large proportion of the error was slope bias for acetate and butyrate, and a modest proportion for propionate. Because interconversions between acetate and propionate represent approximately 15% of the variation in net production rates, lack of such consideration in the model may contribute to the substantial model prediction errors. The potential of using thermodynamic equations to predict interconversions was assessed using observed ruminal pH and VFA concentrations from 2 studies and assuming constant hydrogen pressure and concentrations of CO₂, H₂O, adenosine diphosphate, ATP, and inorganic P. Rate constants for conversion of acetate to propionate and propionate to acetate were derived independently from the control treatments and used to predict the fluxes for the other treatment. The observed changes in VFA concentrations and pH explained the observed changes in conversion of acetate to propionate, but overpredicted the change in the propionate to acetate flux in one study. When applied to the other study, the equations predicted the increase in propionate to acetate flux, but failed to predict the observed reduction in acetate to propionate flux. The inability to predict responses accurately may be due to a lack of data for controlling factors other than pH and VFA concentrations.

Revised digestive parameter estimates for the Molly cow model

The Molly cow model represents nutrient digestion and metabolism based on a mechanistic representation of the key biological elements. Digestive parameters were derived ad hoc from literature observations or were assumed. Preliminary work determined that several of these parameters did not represent the true relationships. The current work was undertaken to derive ruminal and postruminal digestive parameters and to use a meta-approach to assess the effects of interactions among nutrients and identify areas of model weakness. Model predictions were compared with a database of literature observations containing 233 treatment means. Mean square prediction errors were assessed to characterize model performance. Ruminal pH prediction equations had substantial mean bias, which caused problems in fiber digestion and microbial growth predictions. The pH prediction equation was reparameterized simultaneously with the several ruminal and postruminal digestion parameters, resulting in more realistic parameter estimates for ruminal fiber digestion, and moderate reductions in prediction errors for pH, neutral detergent fiber, acid detergent fiber, and microbial N outflow from the rumen; and postruminal digestion of neutral detergent fiber, acid detergent fiber, and protein. Prediction errors are still large for ruminal ammonia and outflow of starch from the rumen. The gain in microbial efficiency associated with fat feeding was found to be more than twice the original estimate, but in contrast to prior assumptions, fat feeding did not exert negative effects on fiber and protein degradation in the rumen. Microbial responses to ruminal ammonia concentrations were half saturated at 0.2mM versus the original estimate of 1.2mM. Residuals analyses indicated that additional progress could be made in predicting microbial N outflow, volatile fatty acid production and concentrations, and cycling of N between blood and the rumen. These additional corrections should lead to an even more robust representation of the effects of dietary nutrients on ruminal metabolism and nutrient absorption, of animal performance, and the environmental impact of dairy production.

The relationship between milk production and farm-gate nitrogen surplus for the Waikato region, New Zealand

As the scope and scale of New Zealand (NZ) dairy farming increases, farmers and the industry are being challenged by Government and the New Zealand public to address growing environmental concerns. Dairying has come under increasing scrutiny from local authorities tasked with sustainable resource management. Despite recent efforts of farmers and industry to improve resource use efficiency, there is increasing likelihood of further regulatory constraints on water use and nutrient management. This study uses available data on farm-gate nitrogen (N) surpluses and milk production from the Waikato, New Zealand's largest dairying region, together with a farm scale modeling exercise, to provide a perspective on the current situation compared to dairy farms in Europe. It also aims to provide relevant guidelines for N surpluses and efficiencies under NZ conditions. Waikato dairy farms compare favorably with farms in Europe in terms of N use efficiency expressed as L milk/kg farm-gate N surplus. Achievable and realistic good practice objectives for Waikato dairy farmers could be 15,000 L milk/ha (1200 kg milk fat plus protein/ha) with a farm-gate N surplus of 100 kg/ha giving an eco-efficiency (L milk/kg N surplus) of 150, and long-term average nitrate leaching losses of approximately 25-30 kg/ha/yr. This can be achieved by increasing the N conversion efficiency through lower replacement rates (16 versus 22%), lower stocked (< 3 cows/ha) high genetic merit cows (30 L milk/day at peak) milked for longer (277 versus 240 days), feeding effluent-irrigated, home-grown, low-protein supplements to cows on high-protein, grass-clover pastures to dilute N concentration in the diet, removing some of the urinary N from the paddocks during critical times by standing cows on a loafing pad for part of the day, and through lower N fertilizer rates (50-70 kg/ha/yr compared to the norm of 170-200 kg/ha/yr) and using a nitrification inhibitor and gibberellins to boost pasture growth and the former to reduce N leaching.

Technical note: a simple model to estimate changes in dietary composition of strip-grazed cattle during progressive pasture defoliations

Methodological problems occur in measuring herbage intake and diet quality during short-term (4-24h) progressive defoliations by grazing. Several models were developed to describe pasture component selection by grazing ruminants, particularly sheep. These models contain empirical coefficients to determine preferences that require laborious and data-demanding calibration. The objective was to develop a simple and practical model of changes in diet composition (green:dead) of pastures strip-grazed by dairy cows. The model was based on 3 premises when cows are strip-grazed in relatively homogeneous swards: 1) cows eat dead material only when green leaf and uncontaminated material have been removed; 2) dead material increases toward the bottom of the sward canopy; and 3) cows progressively defoliate pasture in layers. The main simplification in this model was assuming a linear decrease of green mass from the top to the bottom of the sward canopy. Thus, the proportion of green mass in the stratum eaten depended on the proportion of green in the entire sward canopy and its vertical profile. The model offers a simple solution to estimate changes in dietary compositions in pastures strip-grazed by dairy cattle during progressive pasture defoliations. It uses 2 inputs, the green mass proportion of the total herbage mass and the proportion of total herbage mass eaten during grazing. This can be optionally complemented with inputs of herbage chemical composition. The main outputs of the model are the proportions of green and dead herbage mass in the diet. For example, if the green proportion in the sward was 0.5 and the proportion of herbage mass eaten was 0.5, then the diet would be 0.75 green:0.25 dead; assuming 0.8 and 0.4 digestibility for green and dead material, respectively, the diet digestibility would be 0.7.