As described
previously,1 (link) there is significant uncertainty
(and probably high variability between locations) in the values of
ABM parameters related to hydrolysis and microbial kinetics. New estimates
were developed here using only measurements from section C. For parameter
estimation, slurry production (inferred from slurry heights during
batches) and slurry temperatures were fixed to measured values from
section C. The composition of the produced slurry was calculated based
on the fresh feces and urine composition, assuming that urine and
feces were excreted in a ratio of 3:1 w/w %.28 (link) Degradable VS was calculated as 70% of VS,29 (link),30 (link) and a factor of 1.54 gCOD gVSd–1 was used for
conversion to COD.1 (link) The ratio between qmax,opt for the three methanogen populations
(m0:m1:m2) was fixed at 1:2.4:3.73 during optimization. Optimization
was performed with a quasi-Newton method31 (link) using the optim() function in base R (stats package, v4.2.1) (R
core team, 2022) and specifying method argument as “L-BFGS-B”
with parameter boundaries. Optimization minimized the absolute difference
between measured and ABM-calculated daily methane emission rates (g
d–1) and concentrations of VFA (gCOD kg–1slurry). The methane emission rate and VFA concentration
were equally weighted by centering and scaling to a mean of 0 and
standard deviation of 1 in measurements. Period 4 was excluded from
the optimization and validation due to uncertainty in the washing
procedure and a 1 month delay in delivery of piglets for the fourth
batch period.
The optimized parameter values were used for validation
of the model. Sections WF, SF, and ST during periods 1–3 were
used as validation datasets. Slurry temperature was not measured in
sections SF and ST, and instead the slurry temperature from section
WF was used as the model input. Temperature data from section WF rather
than from section C was used because the retained slurry mass in section
WF better represented slurry masses in sections ST and SF. Hence,
slurry temperatures were expected to respond similarly to heat transfer
from the air and surroundings. The slurry production rate in ST was
set to the rate in the control section since slurry production measurements
were systematically underestimated in the ST section.
previously,1 (link) there is significant uncertainty
(and probably high variability between locations) in the values of
ABM parameters related to hydrolysis and microbial kinetics. New estimates
were developed here using only measurements from section C. For parameter
estimation, slurry production (inferred from slurry heights during
batches) and slurry temperatures were fixed to measured values from
section C. The composition of the produced slurry was calculated based
on the fresh feces and urine composition, assuming that urine and
feces were excreted in a ratio of 3:1 w/w %.28 (link) Degradable VS was calculated as 70% of VS,29 (link),30 (link) and a factor of 1.54 gCOD gVSd–1 was used for
conversion to COD.1 (link) The ratio between qmax,opt for the three methanogen populations
(m0:m1:m2) was fixed at 1:2.4:3.73 during optimization. Optimization
was performed with a quasi-Newton method31 (link) using the optim() function in base R (stats package, v4.2.1) (R
core team, 2022) and specifying method argument as “L-BFGS-B”
with parameter boundaries. Optimization minimized the absolute difference
between measured and ABM-calculated daily methane emission rates (g
d–1) and concentrations of VFA (gCOD kg–1slurry). The methane emission rate and VFA concentration
were equally weighted by centering and scaling to a mean of 0 and
standard deviation of 1 in measurements. Period 4 was excluded from
the optimization and validation due to uncertainty in the washing
procedure and a 1 month delay in delivery of piglets for the fourth
batch period.
The optimized parameter values were used for validation
of the model. Sections WF, SF, and ST during periods 1–3 were
used as validation datasets. Slurry temperature was not measured in
sections SF and ST, and instead the slurry temperature from section
WF was used as the model input. Temperature data from section WF rather
than from section C was used because the retained slurry mass in section
WF better represented slurry masses in sections ST and SF. Hence,
slurry temperatures were expected to respond similarly to heat transfer
from the air and surroundings. The slurry production rate in ST was
set to the rate in the control section since slurry production measurements
were systematically underestimated in the ST section.
