modelling biochemical rumen functions with special emphasis on methanogenesis

8/11/2019 Modelling Biochemical Rumen Functions With Special Emphasis on Methanogenesis

1/49

ModellingModelling

biochemical rumen functionsbiochemical rumen functions

with special emphasis onwith special emphasis onmethanogenesismethanogenesis

DDD

Dr Jan DijkstraDr Jan DijkstraAnimal Nutrition GroupAnimal Nutrition Group

Wageningen UniversityWageningen University


2/49

Overview

Principles of mathematicalPrinciples of mathematical modellingmodelling

Empirical models of methane productionEmpirical models of methane production

Mechanistic models of methane productionMechanistic models of methane production

J. Dijkstra

Modelling methane emissions


3/49

Prediction ~ mathematical models

A model is anA model is an equation or set of equationsequation or set of equations

thatthatrepresent therepresent the behaviour of a systembehaviour of a system

France and Thornley (1984)France and Thornley (1984)

Prediction requires use of modelsPrediction requires use of models

A model can be viewed as an idea, hypothesis orA model can be viewed as an idea, hypothesis orrelation expressed in mathematicsrelation expressed in mathematics

Symbiosis between experimentation andSymbiosis between experimentation andmodellingmodelling

J. Dijkstra



4/49

Model classification

dynamic OR staticdynamic OR static

deterministic OR stochasticdeterministic OR stochastic

mechanistic OR empiricalmechanistic OR empirical

To put categories into a more familiar context, a modelTo put categories into a more familiar context, a modelbased on:based on:

regression analysisregression analysis

static, stochastic, empiricalstatic, stochastic, empirical

linear programminglinear programming static, deterministic, empiricalstatic, deterministic, empirical

differentialdifferential eqnseqns

dynamic, deterministic,dynamic, deterministic,

mechanisticmechanistic J. DijkstraModelling methane emissions


5/49

Levels of organization

1

1

-

-

+

i

i

i

2

Description of levelLevel

Herd / flock

Animal

Organ / tissue

Cell

i

Dijkstra & France

(2005)

J. Dijkstra



6/49

Properties of hierarchical systems

Each level has its own concepts and languageEach level has its own concepts and language

Each level is an integration of items from lowerEach level is an integration of items from lowerlevelslevels

Successful operation of a level requires lowerSuccessful operation of a level requires lower

levels to function properly, but not necessarilylevels to function properly, but not necessarily viceviceversaversa

LevelLevel Description of levelDescription of level

i + 1i + 1

ii

i - 1i - 1

i - 2i - 2

Herd / flockHerd / flock

AnimalAnimalOrgan / tissueOrgan / tissue

CellCell

J. Dijkstra



7/49

Contributions of modelling

Models make best use of (precious) dataModels make best use of (precious) data

Models provide a convenient data summary,Models provide a convenient data summary,

useful for interpolation and cautious extrapolationuseful for interpolation and cautious extrapolation

Models provide quantitative description andModels provide quantitative description and

understanding of biological problemsunderstanding of biological problems

ModellingModelling

provides strategic and tactical supportprovides strategic and tactical support

to researchto research programmesprogrammes

ModellingModelling

allows exploration of possibleallows exploration of possible

outcomes when data are not availableoutcomes when data are not available

J. Dijkstra



8/49

Overview




J. Dijkstra



9/49

Livestock greenhouse gases

manure /

fertilizer

CO2

N2

O

CH4

deforestatio

n

enteric

fermentation

FAO (2006)

GWP*-100 yr:

CO2 1

CH4 25

N2

O 298*Global Warming

Potential

J. Dijkstra



10/49

Livestock greenhouse gases

manure /

fertilizer

CO2N2

O

CH4

deforestatio

n

enteric

fermentation

GWP*-20 yr:

CO2 1

CH4 72

N2

O 289*Global Warming

Potential

J. Dijkstra


FAO (2006)


11/49

Rumen methanogenesis

Fermentative microFermentative micro--organisms utilize dietaryorganisms utilize dietary

organic matter to produce VFA plus gases (e.g.,organic matter to produce VFA plus gases (e.g.,

COCO22

, H, H22

))

amount of Hamount of H22

depends on type of VFAdepends on type of VFA

MethanogensMethanogens

reduce COreduce CO22

to CHto CH44

using Husing H22

,,keeping Hkeeping H22

partial pressure in rumen lowpartial pressure in rumen low

HH22

is used up as it is producedis used up as it is produced

CHCH44

production:production:

22

12% of GE intake12% of GE intake

1010

35 g/kg DM intake35 g/kg DM intake J. DijkstraModelling methane emissions


12/49

Organic

matterMicro-

organismsVF

A

feed

intake

Rumen

2. Microbial

growth

(efficiency)

3. Type VFA

f (substrate type, pH, )

Gut

outflow

absorption

Factors involved in rumen methanogenesis

1. Chemical composition & degradation characteristics

H2 CH4

CO2

J. Dijkstra



13/49

Empirical models of methane production

> 30 empirical models available> 30 empirical models available

inventoryinventory

mitigation strategymitigation strategy

Independent variables include live weight, milkIndependent variables include live weight, milk

production, feed intake, dietary components,production, feed intake, dietary components,

digestibilitydigestibility

Applied in models of greenhouse gas emissionsApplied in models of greenhouse gas emissionsin whole farm settingin whole farm setting

J. Dijkstra



14/49

Assessment of accuracy of empirical models

9 methane equations applied in 8 whole farm9 methane equations applied in 8 whole farm

modelsmodels

169 observations from 9 studies169 observations from 9 studies

Ellis et al. (2010) Global Change Biology

MeanMean SDSD MinMin MaxMaxFeed intake (kg DM/d)Feed intake (kg DM/d) 19.619.6 4.04.0 11.211.2 32.032.0

Milk production (kg/d)Milk production (kg/d) 30.330.3 9.059.05 8.88.8 49.449.4

Methane production (g/d)Methane production (g/d) 371371 77.177.1 117117 698698

J. Dijkstra



15/49

Empirical models of methane production - 1

274 (Europe) or 323 (N274 (Europe) or 323 (N--America)America)

IPCC (1997) Tier IIPCC (1997) Tier I

0.06 x GE intake / 55.650.06 x GE intake / 55.65

IPCC (1997) Tier IIIPCC (1997) Tier II

10 + 4.9 x milk yield + 1.5 x LW10 + 4.9 x milk yield + 1.5 x LW0.750.75KirchgeKirchgenerner

et al. (1995)et al. (1995) -- 11

137 + 10 x milk yield137 + 10 x milk yield

CorreCorre

(2002)(2002)

CHCH44

production (g/d) estimated as:production (g/d) estimated as:

J. Dijkstra



16/49

Empirical models of methane production - 2

(45(45

0.02 x DMI0.02 x DMI22

1.8 x C18:21.8 x C18:2

84 x C84 x C20) x20) x

DMIDMIGigerGiger--ReverdinReverdin

et al. (2003)et al. (2003)

[1.3 + 0.11 x[1.3 + 0.11 x DigDigmm

++ MnMn

x (2.4x (2.4

0.05 x0.05 x DigDigmm

)] x)] x

GEIGEIBlaxterBlaxter

&& ClappertonClapperton

(1965)(1965)

63 + 79 x CF + 10 x NFE + 26 x CP63 + 79 x CF + 10 x NFE + 26 x CP

212 x FAT212 x FATKirchgeKirchgenerner

et al. (1995)et al. (1995) -- 22

(3 + 0.5 x NSC + 1.7 x HC + 2.7 x CEL)/55.65(3 + 0.5 x NSC + 1.7 x HC + 2.7 x CEL)/55.65Moe & Tyrrell (1979)Moe & Tyrrell (1979)

CHCH44

production (g/d) estimated as:production (g/d) estimated as:

20 x concentrate + 22 x maize20 x concentrate + 22 x maize silsil

+ 27 x grass+ 27 x grass

((silsil)) SchilsSchils

et al. (2006)et al. (2006)J. Dijkstra



17/49

Evaluation of empirical methane models

PredPred

CHCH44(g/d)(g/d)

RMSPE(%)

CCC

IPCC (1997) Tier IIPCC (1997) Tier I 304304 27.6 0.01

IPCC (1997) Tier IIIPCC (1997) Tier II 399399 20.9 0.49

CorreCorre

(2002)(2002) 440440 34.2 0.13

KirchgeKirchgenerner

et al. (1995)et al. (1995)

11404404 29.5 0.29

GigerGiger--ReverdinReverdin

et al. (2003)et al. (2003) 230230 52.5 0.12

BlaxterBlaxter


(1965)(1965) 332332 21.2 0.27

Moe & Tyrrell (1979)Moe & Tyrrell (1979) 391391 20.2 0.46KirchgeKirchgenerner

et al. (1995)et al. (1995)

22 345345 20.9 0.22

SchilsSchils

et al. (2006)et al. (2006) 483483 39.5 0.25

Ellis et al. (2010)

Observed CH4371 g/d

J. Dijkstra



18/49


PredPred

CHCH44(g/d)(g/d)

RMSPERMSPE(%)(%)

CCC

IPCC (1997) Tier IIPCC (1997) Tier I 304304 27.627.6 0.01

IPCC (1997) Tier IIIPCC (1997) Tier II 399399 20.920.9 0.49

CorreCorre

(2002)(2002) 440440 34.234.2 0.13


et al. (1995)et al. (1995)

11404404

29.529.5 0.29


et al. (2003)et al. (2003) 230230 52.552.5 0.12

BlaxterBlaxter


(1965)(1965) 332332 21.221.2 0.27

Moe & Tyrrell (1979)Moe & Tyrrell (1979) 391391 20.220.2 0.46KirchgeKirchgenerner

et al. (1995)et al. (1995)

22 345345 20.920.9 0.22

SchilsSchils

et al. (2006)et al. (2006) 483483 39.539.5 0.25

Ellis et al. (2010)

J. Dijkstra



19/49


PredPred

CHCH44(g/d)(g/d)

RMSPERMSPE(%)(%)

CCCCCC

IPCC (1997) Tier IIPCC (1997) Tier I 304304 27.627.6 0.010.01

IPCC (1997) Tier IIIPCC (1997) Tier II 399399 20.920.9 0.490.49

CorreCorre

(2002)(2002) 440440 34.234.2 0.130.13


et al. (1995)et al. (1995)

11 404404 29.529.5 0.290.29


et al. (2003)et al. (2003) 230230 52.552.5 0.120.12

BlaxterBlaxter


(1965)(1965) 332332 21.221.2 0.270.27

Moe & Tyrrell (1979)Moe & Tyrrell (1979) 391391 20.220.2 0.460.46KirchgeKirchgenerner

et al. (1995)et al. (1995)

22 345345 20.920.9 0.220.22

SchilsSchils

et al. (2006)et al. (2006) 483483 39.539.5 0.250.25

Ellis et al. (2010)

J. Dijkstra



20/49

Evaluation of empirical methane modelsEllis et al. (2010)

Model:

Giger-Reverdin

et al. (2003)0 100 200 300 400 500 600 700Observed CH

4(g/d)

0

100

200

300

400

500

600

700

Predic

tedCH4

(g

/d)

y = x

J. Dijkstra



21/49

Evaluation of empirical methane modelsEllis et al. (2010)

Model:

Moe & Tyrrell

(1979)

0 100 200 300 400 500 600 700

Observed CH4(g/d)

0

100

200

300

400

500

600

700

Predic

tedCH4

(g

/d)

y = x

J. Dijkstra



22/49

Conclusions empirical methane models

Predictions poor; significant bias and deviation ofPredictions poor; significant bias and deviation ofregression slope from unityregression slope from unity

Simple, generalized models performed worseSimple, generalized models performed worse

than models based on diet compositionthan models based on diet composition

Substantial errors into inventories of whole farmSubstantial errors into inventories of whole farmgreenhouse gas emissions are likelygreenhouse gas emissions are likely

Low prediction error may lead to incorrectLow prediction error may lead to incorrect

mitigation recommendationsmitigation recommendations

J. Dijkstra



23/49

Overview




J. Dijkstra



24/49

Rate : state formalism

State variables:State variables: QQ11,,

QQ22,,

..,.., QQnn

Change of state variables with timeChange of state variables with time tt::

ddQQ11/d/dt = ft = f11 ((QQ11, Q, Q22,, ,, QQnn;;

PP ))

ddQQ22/d/dt = ft = f22 ((QQ11, Q, Q22,, ,, QQnn;;

PP ))

. .. .

. .. .

ddQQnn

/d/dtt = f= fnn

((QQ11

, Q, Q22

,, ,, QQnn

;;

PP ))

Differential equations based on law of massDifferential equations based on law of mass

conservation, 1conservation, 1stst

law of thermodynamics, etclaw of thermodynamics, etc

J. Dijkstra



25/49

Organic

matterMicro-

organismsVF

A

feed

intake

Rumen

2. Microbial

growth

(efficiency)

3. Type VFA

f (substrate type, pH, )

Gut

outflow

absorption

Factors involved in rumen methanogenesis

1. Chemical composition & degradation characteristics

H2 CH4

CO2

J. Dijkstra



26/49

Mechanistic rumen module

smallsmall

intestineintestineNDFNDFNDF

feed intakefeed intake rumenrumen

gut wallgut wall

NDFmicrobmicrob

starchstarchstarchstarch

VFAVFA

Ac Pr Bu

starch

LCFA LCFA

protein proteinproteinprotein

NH3NHNH33

amino

acids

aminoamino

acidsacids

NH3urea

protein

sugars

F-AA +pept

glucoseglucoseglucose

NDF

glucose

amino acid

NH3urea

microb

Dijkstra et al. (1992)

Modifications:

Dijkstra (1994)

Mills et al. (2001)

Dijkstra et al. (2002)Kebreab

et al.(2004)

Bannink

et al.(2006)

In use in:

Netherlands

UK

AustraliaBrazil

Canada

USA

and many moreJ. Dijkstra



27/49

Efficiency of microbial growth

Substrate is used for

non-growth purposes (maintenance)

growth purposes

Yield is related to fractional

growth rate

(Pirt, 1965)

1 / Y = M /

+ 1 / Ymax

Y = growth yield

M = maintenance requirement

Ymax

= maximum yield

= fractional growth rate 0.00 0.10 0.20 0.30Fractional growth rate (/h)

0.00

0.20

0.40

0.60

Growthyield(g/g)

J. Dijkstra



28/49

Microbial metabolism: N source

Yield affected by availability of preformed molecules

0 500 100090

110

130

150

per unit OM

per unit CHO

Bacterialproteinyield

(gprotein/kg

degrOMor

CHO)

Trypticase (mg/l)

Russell & Sniffen (1984)

J. Dijkstra



29/49

Energy and N synchrony: energy spillingK. aerogenes; Neijssel

& Tempest

(1976)

0.00 0.25 0.50 0.750

5

10

15

20

Dilution rate (/h)

glycerol limited

ammonia limited

Micro

bialyield

(gDM

/molATP)

J. Dijkstra


Uglycerol

in energy spilling

= vmax

Qmicrobes

/(1+[ammonia]/Jammonia

)


30/49

Significance of VFA molar proportion

CC66

HH1212

OO66

+2H+2H22

OO

2C2C22

HH44

OO22

(acetate) + 2CO(acetate) + 2CO22

++ 4H4H22

CC66

HH1212

OO66

++2H2H22

2C2C33

HH66

OO22

(propionate) + 2H(propionate) + 2H22

OO

CC66

HH1212

OO66

CC44

HH88

OO22

(butyrate) + 2CO(butyrate) + 2CO22

++ 2H2H22

COCO22

++ 4H4H22

CHCH44

+ 2H+ 2H22

OO

acetateacetate

70%70%

propionatepropionate

15%15%butyratebutyrate

15%15%

CHCH44

0.39 mol/mol VFA0.39 mol/mol VFA

60%60%

25%25%15%15%

0.31 mol/mol VFA0.31 mol/mol VFAJ. Dijkstra



31/49

Mechanistic model methane production

(Dijkstra et

al. 1992; Mills et al. 2001; Bannink et al. 2006)

Acetate

H2

Propionate

Butyrate

Lipidhydrogenation

Valerate

Microbial growthwith ammonia

Microbial growthwith amino acids

H2 source H2 sink

Fermentation

Feed input

Large IntestinalModel

Small intestinaldigestion

Rumen ModelFermentation

FermentationMethane

CO2

+ 4H2

CH4

+2H2

O

EXCESS

Methane

module

Methane

module

J. Dijkstra



32/49

VFA stoichiometry

Type of VFA formed related to substrateType of VFA formed related to substratefermented, rumen pH, roughagefermented, rumen pH, roughage vsvs

concentrateconcentrate

J. Dijkstra


0.0

0.5

1.0

1.5

molacetate/molsubstrate

sugars

sta

rch

cellulo

se

hemi-

cellu

lose

roughage concentrate

0.0

0.5

1.0

1.5

molpropionate/molsubstrate

sugars

sta

rch

cellulo

se

hemi-

cellu

lose

roughage concentrate

Murphy et al.

(1984)

acetate propionate


33/49

Analysissubstrate type

roughage

vs.

concentratesefficiencymicrobialgrowth

kineticsVFA-absorption

pH

In vivo datarumendigestionlactating cows

VFA stoichiometry

and rumen pHBannink

et al. (2008)

J. Dijkstra


Sc: sugars

St: starch


34/49

Methane formation from various substrates

J. Dijkstra


Starch Cell wall Protein0.00

0.20

0.40

0.60

0.80

1.00

1.20

Relativeme

thaneformation

pH 5.5

pH 6.0

pH 6.5 CH4

from

soluble

carbohydrates

set at unity

Bannink

et al. (2008)


35/49

Mechanistic models of methane formation

Prediction based on description of the rumen inPrediction based on description of the rumen in

terms of components and associated processesterms of components and associated processes

Mechanistic models superior predictive powerMechanistic models superior predictive power

BenchaarBenchaar

et al. (1998); Mills et al. (2001);et al. (1998); Mills et al. (2001); KebreabKebreab

et al. (2006)et al. (2006)

Allows evaluation of dietary mitigation optionsAllows evaluation of dietary mitigation options

Inventories under Kyoto protocolInventories under Kyoto protocol

J. Dijkstra



36/49

Example: Netherlands, database 1990 -

2008

Milk production and composition dataMilk production and composition data

Feed intake dataFeed intake data

feed categories: fresh grass, grass silage, maizefeed categories: fresh grass, grass silage, maizesilage, wet byproducts, concentratessilage, wet byproducts, concentrates

concentrate chemical composition from centralconcentrate chemical composition from central

databasedatabase

roughage chemical composition from Laboratory forroughage chemical composition from Laboratory for

Soil and Crop TestingSoil and Crop Testing

Dietary changes in 1990

2008

more maize silage and less fresh grass

crude protein content decreased

starch and sugar content increased J. DijkstraModelling methane emissions


37/49

DM intake and fat corrected milk (FCM)

J. Dijkstra


1990 1995 2000 2005 2010

Year

12

14

16

18

20

22

24

FCMorDMI(kg/d)

FCM production

DM intake


38/49

J. Dijkstra


Simulated methane production

1990 1995 2000 2005 2010

Year

100

105

110

115

120

125

130

Methaneproduction(kg

/cow/yr)

15

16

17

18

19

20

Methane(g/kgDMorg

/kgFCM)

Methane (kg/cow/yr)

Methane (g/kg DM)

Methane (g/kg FCM)


39/49

Simulated methane production

J. Dijkstra


1990 1995 2000 2005 2010

Year

100

105

110

115

120

125

130

Methaneproduction(kg

/cow/yr)

15

16

17

18

19

20

Methane(g/kgDMorg

/kgFCM)

Methane (kg/cow/yr)

Methane (g/kg DM)

Methane (g/kg FCM)


40/49


41/49

Methane Conversion Factor (MCF)

Trend methaneTrend methaneproductionproduction

(kg/cow/yr):(kg/cow/yr):

IPCC 1.38IPCC 1.38

model 1.05model 1.05 1990 1995 2000 2005 2010

Year

5.7

5.8

5.9

6.0

6.1

6.2

6.3

M

CF(%G

E)

Mechanistic model

IPCC Tier 2 (1997)

J. Dijkstra



42/49

Conclusions mechanistic methane models

Continued interaction between mechanisticContinued interaction between mechanisticmodellingmodelling

and experimentation allows fasterand experimentation allows faster

progressprogress

combine expertise in traditional fermentation parameters,combine expertise in traditional fermentation parameters,

molecular microbiological tools, and mathematicsmolecular microbiological tools, and mathematics

Mechanistic methane models enable predictionMechanistic methane models enable prediction

based on understanding of the systembased on understanding of the system

J. Dijkstra



43/49

Acknowledgements

AndreAndre BanninkBannink

LelystadLelystad, the Netherlands, the Netherlands

Jennifer EllisJennifer Ellis

UnivUniv

Guelph, CanadaGuelph, Canada

James FranceJames France

UnivUniv

GuelpGuelp, Canada, Canada

ErmiasErmias

KebreabKebreab

UnivUniv

Davis, USADavis, USA

SecundinoSecundino

LopezLopez

UnivUniv

Leon, SpainLeon, Spain

J. Dijkstra



44/49

Acknowledgements

AndreAndre BanninkBannink

LelystadLelystad, the Netherlands, the Netherlands

Jennifer EllisJennifer Ellis

UnivUniv

Guelph, CanadaGuelph, Canada

James FranceJames France

UnivUniv

GuelpGuelp, Canada, Canada

ErmiasErmias

KebreabKebreab

UnivUniv

Davis, USADavis, USA

SecundinoSecundino

LopezLopez

UnivUniv

Leon, SpainLeon, Spain

J. Dijkstra



45/49

Practical solutions: model results

0

510

15

20

25

mean NL doubled

maize

silage

early cut

grass

silage

reduced

fertiliser

+2% fat

Metha

ne(g/kg)

per kg feed

per kg milk

J. Dijkstra



46/49


0

510

15

20

25

mean NL doubled

maize

silage

early cut

grass

silage

reduced

fertiliser

+2% fat

Metha

ne(g/kg)

per kg feed

per kg milk

-4%

-5%

J. Dijkstra



47/49


48/49


0

510

15

20

25

mean NL doubled

maize

silage

early cut

grass

silage

reduced

fertiliser

Metha

ne(g/kg)

per kg feed

per kg milk

-3%

-6%

+8%

+11%-4%

-5%

J. Dijkstra



49/49


0

510

15

20

25

mean NL doubled

maize

silage

early cut

grass

silage

reduced

fertiliser

+2% fat

Metha

ne(g/kg)

per kg feed

per kg milk

-3%

-6%

+8%

+11%-2%

-6%

-4%

-5%

J. Dijkstra

modelling biochemical rumen functions with special emphasis on methanogenesis

Documents