Modelling of feeding behaviour, rumen load and the kinetics of digestion and passage of digesta in domestic and wild ruminants.
Date
2021
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Roughage intake is affected by a collection of factors that include feeding behaviour, and the
weight of rumen digesta which is a function of digesta clearance from the rumen as governed
by the rates of degradation and passage. Accurate prediction of intake depends on the ability to
predict these factors. In literature, there are few models, if any, that can be used to estimate the
weight of rumen digesta load, simulate feeding behaviour, predict passage rates of solid and
liquid digesta, and degradability in the rumen of ruminants inhabiting environments with
different diet qualities.The five main objectives of this study were to (1) investigate effects of
diet and roughage quality on feeding behaviour; and to determine the main factors affecting
and developing Random Forest models to estimate (2) time spent on diurnal feeding
behaviours, (3) digestion of feeds in the rumen, (4) weight rumen digesta load and (5) rate of
passage of digesta in the rumen.
The effects of diet and roughage quality on dry matter intake, duration and number of
daytime and night-time eating bouts, and ruminating activities in small ruminants were
investigated. In Exp 1 and 2, roughage quality was improved by urea treatment of veld hay,
while diet quality was improved by supplementing with Lucerne hay (Exp 3), sunflower meal
and lespedeza (Exp 4), fish meal (Exp 5a), and sunflower meal (Exp 5b). Daytime (06:00 to
18:00 h) and night-time (18:00 to 06:00 h) feeding behaviour activities were recorded.
Roughage quality affected rumination index in Exp 1, but not in Exp 2, 3, and 5. Time spent
eating and ruminating was affected by roughage quality (Exp 1, 3, and 4), period of day (all
experiments) and their interaction (Exp 1). Period of day affected the duration of rumination
sessions (Exp 1, 2, and 3); diet quality or roughage quality affected the duration of eating bouts
(Exp 3) and rumination sessions (Exp 1 and 2). roughage quality had a significant effect on the
duration eating sessions in Exp 3 only, whilst period of day affected this same behaviour in
Exp 2 and 3.
To ascertain the influence of the period of the day, ambient temperature, climatic region,
and ruminant feeding type on daytime and night-time feeding behaviour of ruminants a dataset
was collected from studies that measured feeding behaviour. Studies that qualified for inclusion
into the dataset should have (1) reported times spent eating (TSE), ruminating (TSR) and idling,
number and duration of ruminating and eating sessions during a 12h day and 12h night period,
and 24 h period (2) measured body weights of animals used, and (3) stated feeds or proportion
of feeds in diets fed to or consumed by the animals. Diet properties, animal and environmental
factors affecting feeding behaviour were identified in the studies. A mixed effects and
regression models captured the influence and response to the period of the day, ambient
temperature, climatic region and ruminant feeding type of feeding behaviour. During the day,
time spent ruminating and chewing became longer in large ruminants than at night. Predictions
showed that time spent eating during the day and at night are expected to decline with an
increase in ambient temperature, while times spent ruminating during the day will increase.
Grazers and intermediate feeders spent more time eating during the day than at night, while
browsers spent more time eating at night than during the day.
The influence on 24 h diurnal feeding behaviour patterns of ruminants in response to
ambient temperature and ruminant feeding type were ascertained. Feeding behaviours scaled
allometrically with body weight for all ruminant feeding types, except for TSE by browsers
and intermediate feeders, and TSR by grazers. Times spent eating and TSR become shorter in
large compared to small ruminants. Time spent ruminating became shorter in large browsers,
while large intermediate feeders spent more TSR than their smaller counterparts. Browsers had
less TSE, highest DEB and lowest number of eating bouts compared to grazers and
intermediate feeders. Trends from this study showed that TSE, DEB, and idling are projected
to increase with ambient temperature, while TSR is likely to decrease.
Models to predict TSE and TSR for grazing and browsing ruminants were developed. A
dataset was created from studies that reported TSE and TSR, number of eating (NEB) and
ruminating bouts (NRB), and the duration of ruminating (DRB) and eating bouts (DEB) over
a 24h period. Factors affecting feeding behaviour were identified from each study and grouped
into (1) diet properties, (2) animal and (3) environmental factors. These factors were used as
input variables for the prediction of feeding behaviour. The dataset was randomly divided into
two subsets: 70% for model training and 30% for model testing. Developed models accounted
for 95% (TSE), 90% (TSR), 93% (DEB), 93% (DRB), 78% (NEB) and 90% (NRB) of the
variation in prediction of feeding behaviour. The models attained 87% (TSE), 62% (TSR), 93%
(DEB), 83% (DRB), 82% (NEB) and 77% (NRB) precision in prediction during testing using
an independent dataset. This study developed good simulation models for feeding behaviour
of ruminants.
The consequences of increases in ambient temperature and effect of climate type on
digestibility of forages by ruminants using meta-analysis in relation to global warming were
evaluated. A dataset on nylon bag degradability parameters bearing the chemical composition
of roughages, grains, leaves, stems, fruits, concentrates, and diets given to animals, climate
type, and ambient temperature were compiled. Data were analysed using mixed model
regression and simple linear regression methodologies. Negative correlations between ambient
temperature and degradability parameters were observed. Potential degradability was highest
for studies carried out in cold and temperate climates compared to tropical and arid climates.
A 1 °C increase in ambient temperature decreased PD by 0.39% (roughages), 0.76%
(concentrates), and 2.41% (mixed diets), with an overall decrease of 0.55% for all feed types.
The “b” fraction decreased by 0.1% (roughages), 1.1% (concentrates), 2.27% (mixed diets),
and 0.35% (all feed types) for every 1 °C increase in ambient temperature. Increasing ambient
temperature by 1 °C increased the neutral detergent fibre content of feeds by 0.4%. A test of
slopes showed that the predicted decrease in rumen digestibility of feeds with ambient
temperature would be most severe in tropical and arid regions compared to cold and temperate
regions.
An evaluation and prediction of the nutritive and feeding value of underutilised forages
that have a potential of being ruminant feeds was done. Underutilised forage legumes,
leaves/trigs of forage trees and shrubs (non-leguminous), commonly used grass forages and
concentrates were collected from various regions. The nylon bag method was used to determine
degradability of the underutilised forage legumes, leaves/trigs of forage trees and shrubs (nonleguminous)
in the rumen. A step-wise regression procedure was used to develop regression
equations to predict degradability of forages in the rumen. Of the underutilised forages, the
crude protein content tended to be double for Brassica oleracea var. acephala compared to
Colophospermum mopane leaves and pods. Forage grasses (62.9±34 g/kgDM) tended to have
very low crude protein contents compared to legumes (137.6±69 g/kgDM) and concentrates
(177±39.9 g/kgDM). Underutilised Brassica oleracea var. acephala (305 g/kgDM) tended to
have higher crude protein levels compared to commonly used protein sources (cotton seed cake
= 222 g/kgDM). The regression model for predicting the soluble fraction accounted for 59%
and 71% of the variation in model development and validation of predictions, respectively. The
regression model for predicting the potential degradability accounted for 65% and 24% of the
variation in model development and validation, respectively.
A dataset to enable prediction of degradation parameters in the rumen were collected from
studies that (1) reported values for in-sacco degradability parameters viz. soluble fraction (a),
slowly degradable fraction (b), potential degradability (PD) and rate of degradation (c) of
roughages, grains, leaves, stems, fruits and concentrate formulations, and (2) stated the diets
given to animals fed at ad-libitum. Two datasets were collated, one on studies that used the
time-lag model and another on studies that used the no-time lag model in computing
degradation parameters. Factors that affect degradability were identified in each of these
studies and categorised into (i) diet properties (ii) feed sample properties (iii) ruminant feeding
type and (iv) environmental factors. These factors were used as input variables to enable
prediction of degradability. Each dataset was randomly divided into two subsets: 70% for
training and 30% for testing. The no time-lag models attained 88% (“a”), 93% (“b”), 76% (“c”)
and 90% (“PD”) precision in prediction during training and 58% (“a”), 52% (“b”), 48% (“c”)
and 53% (“PD”) precision in testing. Time lag models accounted for 91% (“a”), 84% (“b”),
79% (“c”), 91% (“PD”) and 87% (lag) of the variation in prediction during training and 64%
(“a”), 57% (“b”), 29% (“c”), 52% (“PD”) and 59% (lag) precision in testing. Both sets of
models predicted “a”, “b”, PD, and lag with appreciable precision, but models for the prediction
of the rate of degradation require improvement.
The influence of liquid passage rates on solid digesta passage rates and the possibilities of
simultaneous prediction of solid and liquid passage rates in ruminants was examined. Artificial
neural networks were used to develop models of solid and solid plus liquid passage rates.
Studies that reported fractional passage rates, class and body mass of ruminants were included
in the dataset. Factors affecting the rate of passage were identified from each study and grouped
into (i) diet properties, (ii) animal, (iii) feed particle properties and (iv) environmental factors.
Animal and feed factors that affect the rate of passage were identified in studies and used as
input variables to estimate rate of passage in the rumen. The database was composed of
observations of domestic and wild ruminants of variable body mass (1.5 to 1238 kg) from 74
(solid using predicted liquid passage rate) and 31 (solid using observed liquid passage rate)
studies. Observations were randomly divided into 2 data subsets: 75% for training and 25% for
validation. Developed models accounted for 76 and 77% of the variation in prediction of solid
passage rates using predicted and observed liquid passage rate as inputs, respectively.
Simultaneous prediction accounted for 83 and 89% of the variation of solid and liquid passage
rates, respectively. On validation using an independent dataset, these models attained 45%
(solid using predicted liquid), 66% (solid using observed liquid), 50% (solid predicted with
liquid) and 69% (liquid predicted with solid) of precision in predicting passage rates.
Simultaneous prediction of solid and liquid passage rate yielded better predictions (+7%)
compared to independent predictions of solid passage rate.
Scaling relationships of rumen digesta load with body weight and the influence on
ruminant digesta load in response to climatic region and ruminant feeding type were evaluated.
A dataset on rumen digesta load (RDL) parameters bearing body weights of ruminants,
proximate chemical composition of feeds and diets fed to or eaten by ruminants and climate
type was created. Data were analysed using a linear regression and mixed model regression
methodology. Grazers and intermediate feeders had hypoallometric scales of RDL with BW,
while the scale was hyperallometric for browsers. Wet and liquid RDL of grazers and browsers
scaled isometrically with BW. Intercepts of scaling relationships of RDL and BW were highest
for intermediate feeders and lowest for browsers. For all RDL, body mass and animal
production were both influential covariates. Ruminant species and ruminant feeding type
(p<0.05) influenced all measures of RDL and was highest in grazers and lowest in browsers.
The response of RDL to increases in ambient temperature where more linear than they were
quadratic. Liquid and dry rumen digesta load were predicted to decrease in proportion by 0.02
(p<0.0001) for every 1°C increase in ambient temperature.
Models to estimate the weight of rumen digesta in ruminants were developed. A dataset
was created from studies that (1) measured either the rumen dry matter load (RDML), rumen
wet matter load (RWML) or rumen liquid matter load (RLML) by complete evacuation of the
rumen through fistulas or after slaughtering, (2) reported body weights of animals and (3) stated
the diets fed to or eaten by the animals. Factors affecting rumen digesta load were identified
from each study and included animal (ruminant feeding type, body weight, degree of maturity,
animal production level, days in lactation and pregnancy), diet composition (dry matter, neutral
detergent fibre, crude protein, starch and ash content), management (grazing or fed-indoors)
and environmental (climate type and ambient temperature) factors. These factors were used as
input variables in predicting rumen digesta load. The dataset was divided into 2 subsets: 70%
for model training and 30% for testing. The models accounted for 81% (scaled RDML) and
90% (unscaled RDML) of the variation in prediction of RDML. On testing, the models attained
59% (scaled RDML) and 84% (unscaled RDML) precision in prediction. Models attained high
precision in prediction of RWML (R2 = 0.94) and RLML (R2 = 0.94) during training and testing
of RWML (R2 = 0.85) and RLML (R2 = 0.88) using an independent dataset. In conclusion, the
models gave good predictions of the weight of rumen digesta load. However, there is a need to
correct for the effect of time delay from the point when feeding stops till when rumen digesta
load is measured; this is quite cardinal in regressing in time to the exact rumen digesta load
when the animal stopped eating.
In summary, results from this study showed that increases in ambient temperature will
decrease rumen digestibility of forages and these will be more pronounced in arid regions.
Small-sized ruminants adapted their feeding behaviour and rumen digesta load better than
large-size ruminants. This implies that local breeds which are generally small in size can be
better utilised to mitigate climate change by farmers in arid regions. High accuracy in
prediction of feeding behaviour, rumen degradability, passage rate of digesta in the rumen and
rumen digesta load would enable better prediction of dry matter intake by ruminants.
Description
Doctoral Degree. University of KwaZulu-Natal, Pietermaritzburg.