Refinement, evaluation of ALFAM2 model and development of parameter set 3.
Sasha D. Hafner. Contact information here: https://au.dk/sasha.hafner@bce.
Work on ALFAM2 model, all in R.
Check of measurements for censoring low, late emission rates.
ALFAM2 database data in data-emission/data.
Data are downloaded from https://github.com/sashahafner/ALFAM2-data.
Saved data files are excluded from Git in .gitignore, so contributors will have to run scripts/main.R to download data locally.
See data-emission/data/data_version.txt for current data version.
Summary of latest data.
R functions used in the other subdirectories. Some of these are from https://github.com/sashahafner/jumbled/, others were written for this (or earlier) ALFAM2 work.
Development and evaluation of parameter set 3. The largest component, with several further sub-directories for different results. Some of these are described here.
main.R runs everything, but it would take a day or more to run, mainly because of the bootstrap and cross-validation parts.
Some parts are described here.
The subset.R and prep.R scripts get measurement data ready.
par_est.R does parameter estimation using the optim() function.
There are a lot of choices that must be made about how this should take place.
The settings are passed to either the resCalc() function (for a single response) or resCalcComb() for using 2 responses simultaneously in the objective function.
These two functions are defined in the functions directory.
Output from the optim() calls are saved in a list called mods (as in the parameter sets each represent a unique "model").
This list includes parameter estimates, objective function value, convergence information, and whatever else optim() returns.
In extract_pars.R the parameters are extracted for export and plotting.
calc_emis.R calculates emission using all the parameter sets in mods.
To work with parameter estimation results without re-running everything, use the saved workspace.
See commented out call near top of pars/scripts/main.R.
These files are excluded via .gitignore, so are only helpful if the analysis had been carried out previously.
Plots of cumulative emission curves for all field plots used for parameter estimation.
Measurements are blue, parameter set 3 calculated values in red.
Plot titles have identification key (pmid) and some other information.
Has plots of residuals (all as ALFAM2 value minus measured value) in 168 h cumulative emission versus some predictor variables: DM, pH, and average air temperature. These show that parameter set 3 (ps3 in names) addresses some problems that were present in sets 1 and 2 (ps1, ps2). Plots are included for all parameter sets.
Bivariate plots of measured vs. calculated 168 h cumulative emission for different parameter sets.
Exploration of effect of new sink with r5 parameter.
Sensitivity of model to predictor variables, including uncertainty from bootstrap analysis.
Sensitivity with predictors on some common scale.
Sensitivity table like Table 4 in the 2019 paper, with emission predictions for all application methods x multiple scenarios defined by weather and slurry characteristics.
A test of the closed-form solution by comparison to a numerical approach.
Evaluation of alfam2() evaluation on multiple machines.
Evaluation of model structure with respect to pH effects.
Example to show importance of parameter correlation in estimating uncertainty.
| Paper section | Content | Script source | Output source |
|---|---|---|---|
| 2.5 | Number of plots outside of emis. limits | pars/scripts/prep.R | pars/logs/prep.txt |
| 2.5 | Number of plots with missing values | pars/scripts subset_summ.Rmd | pars/logs/subset_summ.md |
| 3.1 | Fig. 2 | data-summary/scripts/plot | data-summary/plots/emis_summ.png |
| 3.2 | Results on bootstrap par. correlation | pars/scripts/boot_cor.R | pars/output/boot_correlation.cav |
| 3.3 | Fig. 3 | pars/scripts/plot_scatter_emis_ps3.R | pars/plots-scatter/e168_comp_ps3.pdf |
| 3.3 | Table 2 | pars/scripts/fit_summ.R | pars/output/fit_168.csv, pars/output/fit_cv.csv |
| Supplement | Table S-1 | pars/scripts/counts.R | pars/output/d1_counts.csv |
| Supplement | Table S-2 | pars/scripts/counts.R | pars/output/d1_incorp_counts.csv |
| Supplement | Table S-3 | pars/scripts/fit_summ.R | pars/output/fit_168.csv |
| Supplement | Table S-4 | pars/scripts/fit_summ.R | pars/output/fit_table_comb.csv |
| Supplement | Table S-6 | speed/scripts/speed_comp_ubuntu.R | speed/output/times_ubuntu.csv |
| Supplement | Fig. S-1 | pars/scripts/plot_long_resids.R | pars/plots-resids/long_resids.png |
| Supplement | Fig. S-2 | pars/scripts/plot_long_resids.R | pars/plots-resids/long_resids_box.png |
| Supplement | Fig. S-3 | pars/scripts/plot_pars.R | pars/plots-pars/pars_sel.png |
| Supplement | Fig. S-4 | pars/scripts/plot_cv.R | pars/plots-cross-val/cross_val_error.png |
| Supplement | Fig. S-5 | pars/scripts/plot_cv.R | pars/plots-cross-val/cross_val_rel_error.png |
| Supplement | Fig. S-6 | sens/scripts/plot.R | sens/plots/sens5_pig.png |
| Supplement | Fig. S-7 | pars/scripts/plot_resids.R | pars/plots-resids/resids_emis_DM_ps3.png |
| Supplement | Fig. S-8 | pars/scripts/plot_resids.R | pars/plots-resids/resids_emis_temp_ps3.png |
| Supplement | Fig. S-9 | sens-tab/scripts/plot_curves.R | sens-tab/plots/sens_curves.png |
| Supplement | Fig. S-10 | sens-tab/scripts/plot_curves.R | sens-tab/plots/sens_comp_1col_ALFAM.png |
| Supplement | Fig. S-11 | pars/scripts/plot_scatter_emis.R | pars/plots-scatter/e168_comp_loocv-202.png |
| Supplement | Fig. S-12 | pars/scripts/plot_scatter_emis.R | pars/plots-scatter/e168_comp_loocv-214.png |
| Supplement | Fig. S-13 through S-26 | pars/scripts/plot_scatter_emis_ps3.R | pars/plots-scatter/e168_comp_loocv-*.png |
| ------------- | ------- | ------------- | ------------- |