fingR is a comprehensive package designed to support Sediment Source
Fingerprinting studies. It provides essentials tools including: dataset
characterisation, tracer selection from analysed properties through the
Three-step method, model source contributions modelling with the
Bayesian Mixing Model (BMM), and assessment of modelling predictions
prediction through the use of virtual mixtures, supporting BMM and
MixSIAR models.
The fingR package is available in this
Github repository and
archived on Zenodo.
- Installation
- Usage
- Future updates
- Getting help
- Citation
- References
#install.packages(devtools)
library(devtools)
# Install the lastest version from GitHub
devtools::install_github("https://github.com/tchalauxclergue/fingR/releases/tag/2.1.1", ref = "master", force = T)
# Alternatively, from the downloaded .tar.gz file
devtools::install_local("path_to_file/fingR_2.1.1.tar.gz", repos = NULL) # 'path_to_file' should be modified accordingly to your working environmentTo illustrate the usage of the package, we are using the database of the sediment core sampled in the Mano Dam reservoir (Fukushima, Japan) and associated soil samples. The 38 sediment core layer are used as target, and 68 soil samples as potential sources. The potential source include three classes: undecontaminated cropland (n = 24), remediated cropland (n = 22), forest (n = 24), and subsoil (mainly granite saprolite; n = 24).
All samples were sieved to 63 microns and analysed for organic matter, elemental geochemistry and diffuse reflectance spectrocolourimetry for sediment source fingerprinting.
The dataset, along with detailed measurement protocols, is available for download on Zenodo at Chalaux-Clergue et al., 2024 (Version 2).
library(fingR)
# Get the dir to data and metadata files within the R package
data.dr <- system.file("extdata", "TCC_MDD_20210608_data_ChalauxClergue_et_al_v240319.csv", package = "fingR")
metadata.dr <- system.file("extdata", "TCC_MDD_20210608_metadata_ChalauxClergue_et_al_v240319.csv", package = "fingR")
# Load the csv files of data and metadata - replace the dir with your file direction
db.data <- read.csv(data.dr, sep = ";", fileEncoding = "latin1", na = "")
db.metadata <- read.csv(metadata.dr, sep = ";", fileEncoding = "latin1", na = "")Verify the different samples classes
table(db.metadata$Class_decontamination)
#>
#> Forest Remediated Subsoil Target
#> 24 10 10 38
#> Undecontaminated
#> 24We join the metadata (general information) and the data (analyses) so that all the information is on a single dataframe. Both dataframes are joined by common variables, here IGSN and Sample_name. In addition, only the analyses performed on the sample fraction below 63 microns are kept.
library(dplyr)
# Create a single dataframe with metadata and data information
database <- dplyr::left_join(db.metadata, db.data, by = join_by(IGSN, Sample_name)) %>% # Joining metadata and data data frame
dplyr::filter(Sample_size == "< 63 µm") %>% # select sample fraction on which analyses were performed
dplyr::filter(Class_decontamination != "Remediated") # to simplify the example remediated cropland are removedtable(database$Class_decontamination)
#>
#> Forest Subsoil Target Undecontaminated
#> 24 10 38 24Among the analysed properties, 31 properties from organic matter and elemental geochemistry analyses were selected as potential tracers. Together with the properties, their measurement uncertainties are selected.
# colnames(database)
# Select the names/colnames of the properties
prop.values <- database %>% dplyr::select(TOC_PrC, TN_PrC,# organic matter properties
EDXRF_Al_mg.kg.1:EDXRF_Zr_mg.kg.1) %>% names # elemental geochemistry
# Select the names/colnames of the property measurement uncertainties/errors
prop.uncertainties <- database %>% dplyr::select(TOC_SD, TN_SD, # organic matter
EDXRF_Al_RMSE:EDXRF_Zr_RMSE) %>% names # elemental geochemistry
names(prop.uncertainties) <- prop.values # Add property names to property uncertainty for easier selectionprop.values
#> [1] "TOC_PrC" "TN_PrC" "EDXRF_Al_mg.kg.1" "EDXRF_Ca_mg.kg.1"
#> [5] "EDXRF_Co_mg.kg.1" "EDXRF_Cr_mg.kg.1" "EDXRF_Cu_mg.kg.1" "EDXRF_Fe_mg.kg.1"
#> [9] "EDXRF_K_mg.kg.1" "EDXRF_Mg_mg.kg.1" "EDXRF_Mn_mg.kg.1" "EDXRF_Ni_mg.kg.1"
#> [13] "EDXRF_Pb_mg.kg.1" "EDXRF_Rb_mg.kg.1" "EDXRF_Si_mg.kg.1" "EDXRF_Sr_mg.kg.1"
#> [17] "EDXRF_Ti_mg.kg.1" "EDXRF_Zn_mg.kg.1" "EDXRF_Zr_mg.kg.1"unname(prop.uncertainties)
#> [1] "TOC_SD" "TN_SD" "EDXRF_Al_RMSE" "EDXRF_Ca_RMSE"
#> [5] "EDXRF_Co_RMSE" "EDXRF_Cr_RMSE" "EDXRF_Cu_RMSE" "EDXRF_Fe_RMSE"
#> [9] "EDXRF_K_RMSE" "EDXRF_Mg_RMSE" "EDXRF_Mn_RMSE" "EDXRF_Ni_RMSE"
#> [13] "EDXRF_Pb_RMSE" "EDXRF_Rb_RMSE" "EDXRF_Si_RMSE" "EDXRF_Sr_RMSE"
#> [17] "EDXRF_Ti_RMSE" "EDXRF_Zn_RMSE" "EDXRF_Zr_RMSE"First, we use data.watcher to check that the selected properties meet
the quality criteria, particularly in terms of their measurement
uncertainty. Several criteria are evaluated (e.g. presence of some
negative values or high uncertainty) and presented as indicators to
consider the use of a property.
library(fingR)
fingR::data.watcher(data = database, properties = prop.values, prop.uncer = prop.uncertainties)
#>
#> Following column(s) contain(s) some negative values: EDXRF_Cr_mg.kg.1.
#> Following column(s) have a measurement uncertainty that makes some values to be virtually impossible: EDXRF_Co_mg.kg.1, EDXRF_Cr_mg.kg.1, EDXRF_Cu_mg.kg.1, EDXRF_Ni_mg.kg.1.
#> Following column(s) have a relative measurement uncertainty above 5% (up to - number): EDXRF_Co_mg.kg.1 (max:753% - n:26), EDXRF_Cr_mg.kg.1 (max:211% - n:38), EDXRF_Ni_mg.kg.1 (max:105% - n:96), EDXRF_Cu_mg.kg.1 (max:103% - n:52), EDXRF_Rb_mg.kg.1 (max:89% - n:93), TN_PrC (max:45% - n:91), EDXRF_Pb_mg.kg.1 (max:38% - n:91), EDXRF_Zn_mg.kg.1 (max:34% - n:96), EDXRF_Sr_mg.kg.1 (max:15% - n:46), TOC_PrC (max:14% - n:95), EDXRF_Zr_mg.kg.1 (max:7% - n:2).According to data.watcher results: Co, Cr, Cu, Ni, and Rb have too
high measurement uncertainty and in addition Cr has some negative values
among the samples. These properties will be removed from following
study.
# Remove Co, Cr, Cu, Ni and Rb from the vector of properties
prop.values <- prop.values[!prop.values %in% c("EDXRF_Co_mg.kg.1", "EDXRF_Cr_mg.kg.1", "EDXRF_Cu_mg.kg.1", "EDXRF_Ni_mg.kg.1", "EDXRF_Rb_mg.kg.1")]
# Keep uncertainties associated to the new vector of properties
prop.uncertainties <- prop.uncertainties[prop.values]prop.values
#> [1] "TOC_PrC" "TN_PrC" "EDXRF_Al_mg.kg.1" "EDXRF_Ca_mg.kg.1"
#> [5] "EDXRF_Fe_mg.kg.1" "EDXRF_K_mg.kg.1" "EDXRF_Mg_mg.kg.1" "EDXRF_Mn_mg.kg.1"
#> [9] "EDXRF_Pb_mg.kg.1" "EDXRF_Si_mg.kg.1" "EDXRF_Sr_mg.kg.1" "EDXRF_Ti_mg.kg.1"
#> [13] "EDXRF_Zn_mg.kg.1" "EDXRF_Zr_mg.kg.1"In the three-step method, the conservative behaviour is assessed by range tests (RT), also known as bracket tests. To be considered to have a conservative behaviour, all target samples values should lye within the range of the potential source classes. The range of the potential source classes is defined as the highest and lowest source class value of a certain criterion.
Various criteria for range tests are documented in the literature, including minimum-maximum (MM), minimum-maximum plus/minus 10% (MMe) -to account for measurement error- , boxplot whiskers -as threshold to identify extreme values-, boxplot hinge -50% of the population-, mean, mean plus/minus one standard deviation (mean.sd) and median. The mean and mean.sd criteria are performed on log-transformed values, assuming a Normal distribution of the samples.
By default, the function applies all these criteria, though their effectiveness in identifying conservative characteristics may vary. Among these, the mean.sd criterion is mathematically the most robust.
The range.test function returns a list containing two data frames -
results.df: A summary overview of the range test results. -
results.RT: Detailed results for each target sample’s range test for
each property. The result of the range test in detailled as: TRUE for
samples within the range, low for sample values lower than the range,
and high for sample values higher than the range.
rt.results <- fingR::range.tests(data = database, # Dataset containing source and mixture information
class = "Class_decontamination", # Column containing the classification or grouping of sources and mixtures
mixture = "Target", # Identifier for mixtures within the class variable
properties = prop.values, # Properties to be tested for conservativeness
sample.id = "Sample_name", # Identifier for individual samples
criteria = c("mean.sd") # Criteria for conducting range tests (options: "MM", "MMe", "whiskers", "hinge", "mean", "mean.sd", "median", or "all")
# MM.error = c(0.1), # Optional: Set the minimum-maximum plus/minus error as 10%
# save.dir = dir.example, # Optional: Directory path for saving the results
# note = "example" # Optional: Additional note to append to the file name
)rt.results$results.RT$EDXRF_Pb_mg.kg.1[1:5,]
#> Sample_name Property n_source RT_mean.sd
#> 1 ManoDd_2106_00-01 EDXRF_Pb_mg.kg.1 58 TRUE
#> 2 ManoDd_2106_01-02 EDXRF_Pb_mg.kg.1 58 TRUE
#> 3 ManoDd_2106_02-03 EDXRF_Pb_mg.kg.1 58 high
#> 4 ManoDd_2106_03-04 EDXRF_Pb_mg.kg.1 58 high
#> 5 ManoDd_2106_04-05 EDXRF_Pb_mg.kg.1 58 TRUErt.results$results.df[1:5]
#> Property n_source n_mixture NAs RT_mean.sd_single
#> 1 TOC_PrC 58 38 0 TRUE
#> 2 TN_PrC 58 38 0 TRUE
#> 3 EDXRF_Al_mg.kg.1 58 38 0 TRUE
#> 4 EDXRF_Ca_mg.kg.1 58 38 0 FALSE
#> 5 EDXRF_Fe_mg.kg.1 58 38 0 FALSE
#> 6 EDXRF_K_mg.kg.1 58 38 0 FALSE
#> 7 EDXRF_Mg_mg.kg.1 58 38 0 FALSE
#> 8 EDXRF_Mn_mg.kg.1 58 38 0 FALSE
#> 9 EDXRF_Pb_mg.kg.1 58 38 0 FALSE
#> 10 EDXRF_Si_mg.kg.1 58 38 0 FALSE
#> 11 EDXRF_Sr_mg.kg.1 58 38 0 FALSE
#> 12 EDXRF_Ti_mg.kg.1 58 38 0 TRUE
#> 13 EDXRF_Zn_mg.kg.1 58 38 0 FALSE
#> 14 EDXRF_Zr_mg.kg.1 58 38 0 FALSEThe is.conservative function returns a list of vector of conservative
properties based on the results of range tests. If multiple criteria are
used, a vector is generated for each criterion.
prop.cons <- fingR::is.conservative(data = rt.results$results.df, # Data frame containing the results of range tests, typically generated by fingR::range.tests
# property = "Property", # Optional: Column containing the names of properties being tested for conservativeness
# test.format = "RT", # Optional: Indicates the common pattern in column test names (default: "RT")
# position = 2, # Optional: Position of the test name in the column name (default: 2)
# separator = "_", # Optional: Character used to split test names in the column (default: "_")
# note = "example" # Optional: Additional note to append to the file name
)prop.cons
#> $mean.sd
#> [1] "TOC_PrC" "TN_PrC" "EDXRF_Al_mg.kg.1" "EDXRF_Ti_mg.kg.1"Inthe three-step method, the capacity of a property to discriminate among source groups is commonly assessed using a Kruskal-Wallis H-test. The discriminant.test function arguments are very similar to range.tests. As an alternative Kolmogov-Smirnov two-samples tests can be used. It provides more detailled results as source groups are compared to each other.
KS.results <- fingR::discriminant.test(data = database, # Dataset containing source and mixture information
class = "Class_decontamination", # Column containing the classification or grouping of sources and mixtures
mixture = "Target", # Identifier for mixtures within the class variable
test = "KS", # Type of test performed, Kruskal-Wallis (KW) or Kolmogorov-smirnov (KS)
properties = prop.values, # Properties to be tested for conservativeness
p.level = .01, # Optional: p-value significance level (default = 0.05)
# save.discrim.tests = T, # Optional: If two-samples tests should be saved
# save.dir = dir.example, # Optional: Directory path for saving the results
# note = "example" # Optional: Additional note to append to the file name
)KS.results[1:5,]
#> Property n.diff.groups Kolmogorov.Smirnov_discriminant
#> 1 TOC_PrC 3 TRUE
#> 2 TN_PrC 3 TRUE
#> 3 EDXRF_Al_mg.kg.1 3 TRUE
#> 4 EDXRF_Ca_mg.kg.1 1 TRUE
#> 5 EDXRF_Fe_mg.kg.1 0 FALSEProperties that get a Kruskal-Wallis p-value bellow 0.05 (p.value = 0.05), are selected as discriminant properties. The function is.discriminant list them. The function automatically recognise data.frame produced by discriminant.test but it is possible to set it for other data.frame format.
prop.discrim <- fingR::is.discriminant(KS.results, # data.frame from discriminant.test or any df with the same organisation.
# property = "Property", # Optional: Column containing the names of properties being tested for conservativeness
# test.format = "Kruskal.Wallis_p.value", # Optional: Indicates the common pattern in column test names (default: "RT")
# position = 1, # Optional: Position of the test name in the column name (default: 1)
# separator = "_", # Optional: Character used to split test names in the column (default: "_")
# p.level = 0.05, # Optional: p-value significance level (default = 0.05)
# note = "example" # Optional: Additional note to append to the file name
)prop.discrim
#> $Kolmogorov.Smirnov
#> [1] "TOC_PrC" "TN_PrC" "EDXRF_Al_mg.kg.1" "EDXRF_Ca_mg.kg.1"
#> [5] "EDXRF_K_mg.kg.1" "EDXRF_Pb_mg.kg.1" "EDXRF_Si_mg.kg.1" "EDXRF_Sr_mg.kg.1"
#> [9] "EDXRF_Zn_mg.kg.1" "EDXRF_Zr_mg.kg.1"Tracers are conservative and discriminant properties.
tracers <- fingR::selected.tracers(cons = prop.cons, # character vector of conservative properties
discrim = prop.discrim) # character Vector of discriminant propertiestracers
#> $mean.sd_Kolmogorov.Smirnov
#> [1] "TOC_PrC" "TN_PrC" "EDXRF_Al_mg.kg.1"Tracer selection are labelled by selected.tracers accordingly to the
range test criteria (e.g. mean.sd, hinge…) and discriminant test
(i.e. Kruskal.Wallis or Kolmogorov.Smirnov). However, sometimes this
label is to long for file labelling therefore, you may replace it
accondingly.
names(tracers) <- "msd_KS" # replace tracers names with the new nametracers
#> $msd_KS
#> [1] "TOC_PrC" "TN_PrC" "EDXRF_Al_mg.kg.1"The conventional three-step method apply a DFA forward stepwise selection on the selected tracers. This DFA stepwise selection aims to retain tracers that maximize source discrimination. However, this step has faced criticism. Observing the results for a large selection of tracers can be insightful. However, it is not useful for small selection of tracers, as it is the case here.
tracers.SW <- fingR::stepwise.selection(data = database, # Dataset containing source and mixture information
class = "Class_decontamination", # Column containing the classification or grouping of source and mixtures
tracers = tracers$msd_KS, # Character vector containing tracers to consider
target = "Target" # Identifier for target samples within the "class" column
# save.dir = dir.example, # Optional: Directory path for saving the results
# note = "example" # Optional: Additional note to append to the file name
)tracers.SW
#> [1] "EDXRF_Al_mg.kg.1" "TOC_PrC" "TN_PrC"The DFA stepwise selection did not removed any of the selected tracers. If the DFA selects different tracers, examining the modelling results for both sets can provide useful insights.
Both tracer selections could joint like following:
# Joining two tracers vector in a list
all.tracers <- list("msd_KS" = tracers$msd_KS, "msd_KS_DFA" = tracers.SW)To evaluate the accuracy of un-mixing models, virtual mixtures are used.
These virtual mixtures, serving as target samples with known
contributions, allow for the calculation of modelling accuracy metrics.
The VM.contrib.generator generate virtual mixture contributions from
the min to the max contribution set with a specified step.
Contribution could be set as percentage (min = 0, max = 100) or as a
ratios (min = 0, max = 1). Smaller step result in a higher number of
virtual mixtures, such as 231 virtual mixtures for a 5% step and 5151
virtual mixtures for a 1% step. Alternatively, virtual mixtures can be
generated within VM.builder.
# Generate virtual mixture source contributions
VM.contrib <- fingR::VM.contrib.generator(n.sources = 3, # Number of source levels
min = 0, # Minimum contribution (here percentage)
max = 100, # Maximum contribution (here percentage)
step = 5, # Step between two contribution levels (here percentage)
sources.class = c("Forest", "Subsoil", "Undecontaminated"), # Optional: Classification of sources
save.dir = dir.example, # Optional: Directory path for saving the results
# note = "example" # Optional: Additional note to append to the file name
# VM.name = "Sample_name", # Optional: Name of the column containing virtual mixture labels
# fileEncoding = "latin1", # Optional: File encoding, important if special character are used in source levels
# return = TRUE, # Optional: Whether the function should return the result
# save = TRUE # Optional: Whether the function should save the result
)VM.contrib[1:5,]
#> Sample_name Forest Subsoil Undecontaminated
#> 1 VM-001 0 0 100
#> 2 VM-002 0 5 95
#> 3 VM-003 0 10 90
#> 4 VM-004 0 15 85
#> 5 VM-005 0 20 80Next, virtual mixture properties are calculated as simple proportional
mixture of source signature (i.e. mean values). This approach is a
simple mass balance approach. The VM.builder function saves and
returns a list containing three data.frame objects: one with the
$property values, the other with the $uncertainty values (with
corresponding labels when given in $uncertainty if not simply “_SD”
is added at the end of the tracer label), and the last one $full where
property and uncertainty were join.
To run un-mixing models, source and target information should be within the same data frame. Source informations are added at the end of all the data.frame created.
VM <- VM.builder(data = database, # Dataset containing source samples
material = "Material", # Column indicating the difference between source and target
source.name = "Source", # Identifier for source samples within the material column
class = "Class_decontamination", # Column containing the classification or grouping of sources and mixtures
tracers = tracers$msd_KS, # Character vector containing tracers to consider
uncertainty = unname(prop.uncertainties[tracers$msd_KS]), # Character vector containing tracers uncertainty labels
contributions = VM.contrib, # Virtual mixture contributions
VM.name = "Sample_name", # Column with virtual mixture labels in the 'contribution' (i.e. VM.contribution)
add.sources = TRUE, # Add source information at the end of the VM data frames
save.dir = dir.example, # Optional: Directory path for saving the results
# note = "example" # Optional: Additional note to append to the file name
)VM$full[1:5,]
#> Sample_name Class_decontamination TOC_PrC TN_PrC EDXRF_Al_mg.kg.1 TOC_SD
#> 1 VM-001 Virtual Mixture 5.16 0.42 84858.53 4.75
#> 2 VM-002 Virtual Mixture 4.97 0.40 86004.71 4.75
#> 3 VM-003 Virtual Mixture 4.78 0.39 87150.90 4.75
#> 4 VM-004 Virtual Mixture 4.60 0.38 88297.08 4.75
#> 5 VM-005 Virtual Mixture 4.41 0.36 89443.26 4.75
#> TN_SD EDXRF_Al_RMSE
#> 1 0.28 17840.72
#> 2 0.28 17840.72
#> 3 0.28 17840.72
#> 4 0.28 17840.72
#> 5 0.28 17840.72Here an example of sets to generate virtual mixture with the
VM.builder function without previously running the
VM.contrib.generator function.
VM <- VM.builder(data = database, # Dataset containing source samples
material = "Material", # Column indicating the difference between source and target
source.name = "Source", # Identifier for source samples within the material column
class = "Class_decontamination", # Column containing the classification or grouping of sources and mixtures
tracers = tracers$msd_KS, # Character vector containing tracers to consider
uncertainty = unname(prop.uncertainties[tracers$msd_KS]), # Character vector containing tracers uncertainty labels
VM.range = c(0, 100), # Minimum and maximum contribution (here percentage)
VM.step = 5, # Step between two contribution levels (here percentage)
VM.name = "Sample_name", # Column with virtual mixture labels in the 'contribution' (i.e. VM.contribution)
add.sources = TRUE, # Add source information at the end of the VM data frames
save.dir = dir.example, # Optional: Directory path for saving the results
# note = "example" # Optional: Additional note to append to the file name
)Create a folder where all modelling results will be saved
# Create new folder to save tracer modelling results
dir.create(file.path(dir.example, "Modelling/"), showWarnings = FALSE)
dir.modelling <- paste0(dir.example, "Modelling/")Create a folder specific from BMM modelling results.
# Create new folder to save BMM modelling results
dir.create(file.path(dir.modelling, "BMM/"), showWarnings = FALSE)
dir.mod.BMM <- paste0(dir.modelling, "BMM/")Run BMM models for actual sediment samples (mix) and virtual mixtures (VM). The BMM model performs a Bayesian un-mixing with a Monte-Carlo chain, the prediction is corrected using the sum of squared relative error of each tracer. Without isotopic ratio within the tracers, there is no need to take any precautions when setting up the model.
# Run BMM model for sediment samples
BMM.mix <- fingR::run.BMM(data = database, # Dataset containing source and target samples
class = "Class_decontamination", # Column containing the classification or grouping of sources and mixtures
mixture = "Target", # Column name identifying the target samples
sample.id = "Sample_name", # Column name for sample identifiers
tracers = tracers$msd_KS, # Character vector containing tracers to consider
uncertainty = unname(prop.uncertainties[tracers$msd_KS]), # Optional: Character vector containing uncertainty of the tracers
n.iter = 30, # Number of iterations for the model (30 for test version - 2500 or 5000 iterations are recommended) 'prop.uncertainties'
save.dir = dir.mod.BMM, # Optional: Directory path for saving the results - 'BMM_previsions.CSV'
#note = "example" # Optional: Additional note to append to the file name
)# Run BMM model for virtual mixtures
BMM.VM <- fingR::run.BMM(data = VM$full, # Dataset containing source and target samples
class = "Class_decontamination", # Column containing the classification or grouping of sources and mixtures
mixture = "Virtual Mixture", # Column name identifying the target samples
sample.id = "Sample_name", # Column name for sample identifiers
tracers = tracers$msd_KS, # Character vector containing tracers to consider
uncertainty = unname(prop.uncertainties[tracers$msd_KS]), # Optional: Character vector containing uncertainty of the tracers
n.iter = 30, # Number of iterations for the model (30 for test version - 2500 or 5000 iterations are recommended)
save.dir = dir.mod.BMM, # Optional: Directory path for saving the results - 'BMM_previsions_VM.CSV'
note = "VM" # Optional: Additional note to append to the file name
)When dealing with isotopic ratios, which are non-linear properties,
errors should be calculated considering relative property content (see
Laceby et al. (2015) for further
details). For example, the delta 13C ratio indicates the isotopic ratio
of 12C to 13C in organic matter, the run.BMM function should be
configured in this way:
# Run BMM model for sediment samples
BMM.iso <- fingR::run.BMM(data = database, # Dataset containing source and target samples
class = "Class_decontamination", # Column containing the classification or grouping of sources and mixtures
mixture = "Target", # Column name identifying the target samples
sample.id = "Sample_name", # Column name for sample identifiers
tracers = tracers$msd_KS, # Character vector containing tracers to consider
uncertainty = unname(prop.uncertainties[tracers$msd_KS]), # Optional: Character vector containing uncertainty of the tracers
isotope.ratio = c("d13C_PrM"), # Optional: Character vector containing isotopic ratios
isotope.prop = c("TOC_PrC"), # Optional: Character vector containing isotopic ratios respective properties
isotopes.unc = c("d13C_SD"), # Optional: Character vecotr containing uncertainty of the isotopic ratios
n.iter = 30, # Number of iterations for the model (30 for test version - 2500 or 5000 iterations are recommended) 'prop.uncertainties'
save.dir = dir.mod.BMM, # Optional: Directory path for saving the results - 'BMM_previsions.CSV'
#note = "example" # Optional: Additional note to append to the file name
)After running the models, we extract the prediction information from the
iteration previsions. The BMM.summary function provides a summary of
the predictions, including the mean, standard deviation, and various
quantiles (2.5, 5, 25, 50, 75, 95, 97.5%) for each mixture (sediment
sample or virtual mixture). From this summary, the BMM.pred function
extracts the ‘Median’ and/or ‘Mean’ for each mixture. Finally, the
ensure.total function ensures that the total predicted contribution
from all sources sums to 1 or 100%.
# For sediment samples
## Summarise BMM model previsions
BMM.summary.mix <- fingR::BMM.summary(pred = BMM.mix, # Predicted contributions from BMM
#sample.id = "mix.names", # Column name for sample identifier
#source = "source", # Column name for source identifier
#value = "value", # Column name for prediction value identifier
save.dir = dir.mod.BMM, # Optional: Directory path for saving the results
#note = "example" # Optional: Additional note to append to the file name
)
## Extracts the median value of the previsions
BMM.preds.mix <- fingR::BMM.pred(data = BMM.summary.mix, # Summary statistics of the predicted contribution by BMM, data from fingR::BMM.summary.mix
stats = "Median", # The summary statistics for source contribution, Could be Mean or Median
#sample.id = "mix.names", # Column name for sample identifier
#source = "source", # Column name for prediction value identifier
save.dir = dir.mod.BMM, # Optional: Directory path for saving the results
#note = "example" # Optional: Additional note to append to the file name
)
## Ensure that the total predicted contribution sums to 1 or 100%
BMM.preds.mixE <- fingR::ensure.total(data = BMM.preds.mix, # Predicted source contribution for each sample, data from fingR::BMM.pre
sample.name = "mix.names", # Column name for sample identifier
path = dir.mod.BMM, # Optional: Directory path for saving the results
#note = "example" # Optional: Additional note to append to the file name
)BMM.preds.mixE[1:5,]
#> mix.names Median_Forest Median_Subsoil Median_Undecontaminated total
#> 1 ManoDd_2106_00-01 0.001 0.711 0.288 1
#> 2 ManoDd_2106_01-02 0.001 0.939 0.060 1
#> 3 ManoDd_2106_02-03 0.001 0.687 0.312 1
#> 4 ManoDd_2106_03-04 0.097 0.902 0.001 1
#> 5 ManoDd_2106_04-05 0.001 0.614 0.385 1Same code for virtual mixtures:
# For virtual mixtures
## Summarise BMM model previsions
BMM.summary.VM <- fingR::BMM.summary(pred = BMM.VM, # Predicted contributions from BMM
#sample.id = "mix.names", # Column name for sample identifier
#source = "source", # Column name for source identifier
#value = "value", # Column name for prediction value identifier
save.dir = dir.mod.BMM, # Optional: Directory path for saving the results
note = "VM" # Optional: Additional note to append to the file name
)
## Extracts the median value of the previsions
BMM.preds.VM <- fingR::BMM.pred(data = BMM.summary.VM, # Summary statistics of the predicted contribution by BMM, data from fingR::BMM.summary.mix
stats = "Median", # The summary statistics for source contribution, Could be Mean or Median
#sample.id = "mix.names", # Column name for sample identifier
#source = "source", # Column name for prediction value identifier
save.dir = dir.mod.BMM, # Optional: Directory path for saving the results
note = "VM" # Optional: Additional note to append to the file name
)
## Ensure that the total predicted contribution sums to 1 or 100%
BMM.preds.VME <- fingR::ensure.total(data = BMM.preds.VM, # Predicted source contribution for each sample, data from fingR::BMM.pre
sample.name = "mix.names", # Column name for sample identifier
path = dir.mod.BMM, # Optional: Directory path for saving the results
note = "VM" # Optional: Additional note to append to the file name
)BMM.preds.VME[1:5,]
#> mix.names Median_Forest Median_Subsoil Median_Undecontaminated total
#> 1 VM-001 0.120 0.879 0.001 1
#> 2 VM-002 0.001 0.943 0.056 1
#> 3 VM-003 0.026 0.973 0.001 1
#> 4 VM-004 0.012 0.987 0.001 1
#> 5 VM-005 0.001 0.998 0.001 1The modelling accuracy of BMM model is evaluate with the virtual mixtures. These virtual mixtures, serving as target samples with known contributions (VM.contrib), allow for the calculation of modelling accuracy metrics based on their prediction.
The eval.groups function calculates several common modelling accuracy
metrics: ME, RMSE, squared Pearson’s correlation coefficient (r2), and
Nash-Sutcliffe Modelling Efficiency Coefficient (NSE).
BMM.stats <- fingR::eval.groups(df.obs = VM.contrib, # Theoretical contribution
df.pred = BMM.preds.VME %>% dplyr::select(-total), # Predicted contribution (remove the $total column from ensured data.frame)
by = c("Sample_name" = "mix.names"), # Column where mixtures labels are specified (for `dplyr::left_join` function)
path = dir.mod.BMM, # Optional: Directory path for saving the results
#note = "example" # Optional: Additional note to append to the file name
)BMM.stats
#> Type Source ME RMSE r2 NSE
#> 1 Median Forest -0.16 0.24 0.58 0.13
#> 2 Median Subsoil 0.42 0.49 0.33 -2.77
#> 3 Median Undecontaminated -0.26 0.39 0.00 -1.34The CRPS functions calculate the continuous ranking probability score
and returns a list contraining two data.frame objects; one with the
$samples CRPS values per source class group (saved as CRPS.csv), the
other is $mean with the mean of the CRPS per source class groups
(saved as CRPS_mean.csv).
# Calculate prediction CRPS values
BMM.CRPS <- fingR::CRPS(obs = VM.contrib, # Observed contributions
prev = read.csv(paste0(dir.mod.BMM, "BMM_prevision_VM.csv")), # Predicted prevision from BMM saved by `rum.BMM()`
source.groups = c("Forest", "Subsoil", "Undecontaminated"), # Source class groups
mean.cal = TRUE, # Calculate mean CRPS per source class group
save.dir = dir.mod.BMM, # Optional: Directory path for saving the results
#note = "example" # Optional: Additional note to append to the file name
)
#> Le chargement a nécessité le package : scoringRulesBMM.CRPS$samples[1:6,]
#> Sample_name Forest Subsoil Undecontaminated
#> 1 VM-001 0.0680 0.4323 0.6693
#> 2 VM-002 0.0546 0.3434 0.5180
#> 3 VM-003 0.0464 0.3368 0.5116
#> 4 VM-004 0.0663 0.3151 0.5692
#> 5 VM-005 0.0508 0.3873 0.6046
#> 6 VM-006 0.0613 0.2569 0.4796BMM.CRPS$mean
#> Source CRPS.mean
#> 1 Forest 0.1371
#> 2 Subsoil 0.1682
#> 3 Undecontaminated 0.1863The interval.width functions calculate two prediction interval width:
The W50 contains 50% of the prevision (Q75-Q25) and the W95 contains
95% of the prevision (Q97.5-Q2.5). It returns a list contraining two
data.frame objects; one with the $samples prediction interval width
values per source class group (saved as Interval_width.csv), the other
is $mean with the mean of the prediction interval width per source
class groups (saved as Interval_width_mean.csv).
# Calculate prediction interval width (W95, W50)
BMM.predWidth <- fingR::interval.width(path.to.prev = paste0(dir.mod.BMM, "BMM_prevision_VM.csv"), # Path to prediction file
mean.cal = TRUE, # Calculate mean of interval width per source group
save = TRUE, # Save the results at the same location of the path.to.prev
#note = "exemple" # Optional: Additional note to append to the file name
)BMM.predWidth$samples[1:6,]
#> mix.names source W50 W95
#> 1 VM-001 Forest 0.242 0.794
#> 2 VM-001 Subsoil 0.584 0.997
#> 3 VM-001 Undecontaminated 0.332 0.845
#> 4 VM-002 Forest 0.322 0.595
#> 5 VM-002 Subsoil 0.613 0.997
#> 6 VM-002 Undecontaminated 0.498 0.897BMM.predWidth$mean
#> # A tibble: 3 × 3
#> Source W50.mean W95.mean
#> <chr> <dbl> <dbl>
#> 1 Forest 0.352 0.873
#> 2 Subsoil 0.586 0.973
#> 3 Undecontaminated 0.41 0.927The ESP function calculates the Encompassed Sample Prediction (ESP).
The ESP is a newly introduced statistics in Chalaux-Clergue et al
(under review) and was created to assess the transferability of the
statistics calculated on virtual mixtures to actual sediment samples.
The ESP was calculated as the percentage of actual samples for which the
predicted contributions remained within the lowest and the highest
predicted contributions obtained for the virtual mixtures. When
expressed as a percentage, ESP ranges from 0 to 100%, the latter
providing an optimal value. Values close to 100% indicate a higher
transferability of modelling evaluation statistics calculated on virtual
mixture to actual sediment samples.
sources.lvl <- c("Forest", "Subsoil", "Undecontaminated")
# Calculate encompassed sample predictions (ESP)
BMM.ESP <- fingR::ESP(obs = BMM.preds.VM, # Virtual mixtures predicted contributions
pred = BMM.preds.mixE, # Actual sediment samples predicted contributions
sources = paste0("Median_", sources.lvl), # Sources labels in prediction objects
count = "Both" # Count 'Number' and 'Percentage'
)BMM.ESP
#> Source ESP.Number ESP.Percentage
#> Median_Forest Forest 37 97
#> Median_Subsoil Subsoil 38 100
#> Median_Undecontaminated Undecontaminated 18 47Modelling accuracy statistics could be interpreted the following way: “Higher values of W50 indicate a wider distribution, which is related to a higher uncertainty. The sign of the ME indicates the direction of the bias, i.e. an overestimation or underestimation (positive or negative value, respectively). As ME is affected by cancellation, a ME of zero can also reflect a balanced distribution of predictions around the 1 : 1 line. Although this is not a bias, it does not mean that the model outputs are devoid of errors. The RMSE is a measure of the accuracy and allows us to calculate prediction errors of different models for a particular dataset. RMSE is always positive, and its ideal value is zero, which indicates a perfect fit to the data. As RMSE depends on the squared error, it is sensitive to outliers. The r2 describes how linear the prediction is. The NSE indicates the magnitude of variance explained by the model, i.e. how well the predictions match with the observations. A negative RMSE indicates that the mean of the measured values provides a better predictor than the model. The joint use of r2 and NSE allows for a better appreciation of the distribution shape of predictions and thus facilitates the understanding of the nature of model prediction errors. The CRPS evaluates both the accuracy and sharpness (i.e. precision) of a distribution of predicted continuous values from a probabilistic model for each sample (Matheson and Winkler, 1976). The CRPS is minimised when the observed value corresponds to a high probability value in the distribution of model outputs.” (Chalaux-Clergue et al, 2024).
The MixSIAR is an R package designed to create and run Bayesian mixing
models. This package is widely used in the sediment source
fingerprinting community to predict source contribution. To explore more
about MixSIAR, including detailed tutorials, examples, and technical
documentation, please visit the official MixSIAR
website. Additionally,
the source code and further resources can be found on the MixSIAR
GitHub page.
According to MixSIAR guide, installation should follow these steps:
install.packages("MixSIAR", dependencies=TRUE)You can install the GitHub version
#install.packages(remotes)
remotes::install_github("brianstock/MixSIAR", dependencies=T)Create a folder specific from BMM modelling results.
# Create new folder to save BMM modelling results
dir.create(file.path(dir.modelling, "MixSIAR/"), showWarnings = FALSE)
dir.mod.MixSIAR <- paste0(dir.modelling, "MixSIAR/")To MixSIAR models require data in a specific format to load the
information of mixtures and sources samples. The data.for.MixSIAR
function generates csv files that conform to the format required by
MixSIAR loading functions. The function generates three files:
MixSIAR_mix.csv containing mixtures information, MixSIAR_sources
containing the mean and standard deviation (sd) of the source classes,
and MixSIAR_discrimination which is a matrix of zero as there is no
throphic information in sediment source fingerprinting studies.
Of note, if several selection of tracers were obtained from the tracer
selection different files should be created. Use the note argument to
differentiate them.
fingR::data.for.MixSIAR(data = database, # Dataset containing source samples
class = "Class_decontamination", # Column containing the classification or grouping of sources and mixtures
target = "Target", # Identifier for mixture samples within the class column
tracers = tracers$msd_KS, # Character vector containing tracers to consider
sample.name = "Sample_name", # Column containing sample names in data
save.dir = dir.mod.MixSIAR, # Directory path for saving the files
# note = "exemple", # Optional: Additional note to append to the file name
# fileEncoding = "latin1", # Optional: File encoding, important if special character are used in source levels
# show.data = FALSE, # Optional: Return generated files in R
)fingR::data.for.MixSIAR(data = VM$full, # Dataset containing source samples
class = "Class_decontamination", # Column containing the classification or grouping of sources and mixtures
target = "Virtual Mixture", # Identifier for mixture samples within the class column
tracers = tracers$msd_KS, # Character vector containing tracers to consider
sample.name = "Sample_name", # Column containing sample names in data
save.dir = dir.mod.MixSIAR, # Directory path for saving the files
note = "VM", # Optional: Additional note to append to the file name
# fileEncoding = "latin1", # Optional: File encoding, important if special character are used in source levels
# show.data = FALSE, # Optional: Return generated files in R
)Load mixture, source and discrimination data for sediment samples.
library(MixSIAR)
# Load sediment samples data
MSIAR.mix <- MixSIAR::load_mix_data(filename = paste0(dir.mod.MixSIAR, "MixSIAR_mix.csv"), # File containing real samples data
iso_names = tracers$msd_KS, # Names of tracers
factors = c("Sample_name"), # Columns used to differentiate samples
fac_random = FALSE, # Indicates if the factor is a random effect
cont_effects = NULL # Continuous effect column not specified
)
# Load source data
MSIAR.source <- MixSIAR::load_source_data(filename = paste0(dir.mod.MixSIAR, "MixSIAR_sources.csv"), # File containing source data
source_factors = NULL, # No source factors specified
conc_dep = FALSE, # Concentration dependence not considered
data_type = "means", # Type of data provided is means
mix = MSIAR.mix # Actual samples mixtures
)
# Load discrimination data
MSIAR.discr <- MixSIAR::load_discr_data(filename = paste0(dir.mod.MixSIAR, "MixSIAR_discrimination.csv"), # File containing discrimination data
mix = MSIAR.mix) # Actual samples mixturesLoad mixture, source and discrimination data for virtual mixtures.
library(MixSIAR)
# Load virtual mixtures data
MSIAR.VM <- MixSIAR::load_mix_data(filename = paste0(dir.mod.MixSIAR, "MixSIAR_mix_VM.csv"), # File containing virtual mixtures data
iso_names = tracers$msd_KS, # Names of tracers
factors = c("Sample_name"), # Columns used to differentiate samples
fac_random = FALSE, # Indicates if the factor is a random effect
cont_effects = NULL) # Continuous effect column not specified
# Load source data
MSIAR.source.VM <- MixSIAR::load_source_data(filename = paste0(dir.mod.MixSIAR, "MixSIAR_sources_VM.csv"), # File containing source data
source_factors = NULL, # No source factors specified
conc_dep = FALSE, # Concentration dependence not considered
data_type = "means", # Type of data provided is means
mix = MSIAR.VM # Actual samples mixtures
)
# Load discrimination data
MSIAR.discr.VM <- MixSIAR::load_discr_data(filename = paste0(dir.mod.MixSIAR, "MixSIAR_discrimination_VM.csv"), # File containing discrimination data
mix = MSIAR.VM) # Actual samples mixturesWrite the JAGS file, which define model structure. The model will be
saved as model_file (“MixSIAR_model.txt” is default).
# Write JAGS model file for actual samples
MixSIAR::write_JAGS_model(filename = paste0(dir.mod.MixSIAR, "MixSIAR_model.txt"), # File path and name to write the JAGS model
resid_err = FALSE, # Whether to include residual error in the model
process_err = TRUE, # Whether to include process error in the model
mix = MSIAR.mix, # Actual samples mixtures dataset
source = MSIAR.source # Source dataset
)# Write JAGS model file for virtual mixtures
MixSIAR::write_JAGS_model(filename = paste0(dir.mod.MixSIAR, "MixSIAR_model_VM.txt"), # File path and name to write the JAGS model
resid_err = FALSE, # Whether to include residual error in the model
process_err = TRUE, # Whether to include process error in the model
mix = MSIAR.VM, # Virtual mixtures dataset
source = MSIAR.source.VM # Source dataset loaded with virtual mixture mix
)When running MixSIAR model you should choose one of the MCMC run
option.
Here run is set to “test” as it is an example.
# note if "Error: .onload ... 'rgags' -> it's because R version is too old need at least R.2.2
# Run MixSIAR model for sediment samples
jags.mix <- MixSIAR::run_model(run = "test", # Type of run (e.g. "test", "long"...)
mix = MSIAR.mix, # Sediment samples dataset
source = MSIAR.source, # Source dataset
discr = MSIAR.discr, # Discrimination dataset
model_filename = paste0(dir.mod.MixSIAR, "MixSIAR_model.txt") # File path to the JAGS model
)
#> Compiling model graph
#> Resolving undeclared variables
#> Allocating nodes
#> Graph information:
#> Observed stochastic nodes: 114
#> Unobserved stochastic nodes: 93
#> Total graph size: 2838
#>
#> Initializing model# note if "Error: .onload ... 'rgags' -> it's because R version is too old need at least R.2.2
# Run MixSIAR model for Virtual mixtures
jags.VM <- MixSIAR::run_model(run = "test", # Type of run (e.g. "test", "long", "very long"...)
mix = MSIAR.VM, # Virtual mixtures dataset
source = MSIAR.source.VM, # Source dataset loaded with virtual mixture mix
discr = MSIAR.discr.VM, # Discrimination dataset
model_filename = paste0(dir.mod.MixSIAR, "MixSIAR_model_VM.txt") # File path to the JAGS model
)
#> Compiling model graph
#> Resolving undeclared variables
#> Allocating nodes
#> Graph information:
#> Observed stochastic nodes: 693
#> Unobserved stochastic nodes: 479
#> Total graph size: 16541
#>
#> Initializing modelAfter running the models, we extract the prediction information from the
MixSIAR model predictions. The MixSIAR.summary function provides a
summary of the predictions, including the mean, standard deviation, and
various quantiles (2.5, 5, 25, 50, 75, 95, 97.5%) for each mixture
(sediment sample or virtual mixture). From this summary, the
MixSIAR.pred function extracts the ‘Median’ and/or ‘Mean’ for each
mixture. Finally, the ensure.total function ensures that the total
predicted contribution from all sources sums to 1 or 100%.
## Summarise MixSIAR model previsions
MixSIAR.summary.mix <- fingR::JAGS.summary(jags.1 = jags.mix, # Data from the MixSIAR model `MixSIAR::run_model()`
mix = MSIAR.mix, # Sediment dataset
sources = MSIAR.source, # Source dataset
path = dir.mod.MixSIAR, # Directory path for saving the files
#note = "example", # Optional: Additional note to append to the file name
save_pred = TRUE # Optional: Save the MixSIAR modelling predictions (heavy files)
)
## Extracts the median value of the previsions
MixSIAR.preds.mix <- fingR::JAGS.pred(path = paste0(dir.mod.MixSIAR, "contrib.csv"), # location of files generated by `JAGS.summary`
stats = "Median", # Summary statistics to calculate (Median or Mean)
save = TRUE, # If the result should be saved
#note = "example" # Optional: Additional note to append to the file name
)
## Ensure that the total predicted contribution sums to 1 or 100%
MixSIAR.preds.mixE <- fingR::ensure.total(data = MixSIAR.preds.mix, # Predicted source contribution for each sample, data from fingR::BMM.pre
sample.name = "sample", # Column name for sample identifier
path = dir.mod.MixSIAR, # Optional: Directory path for saving the results
#note = "example" # Optional: Additional note to append to the file name
)MixSIAR.preds.mixE[1:5,]
#> sample Median_Forest Median_Subsoil Median_Undecontaminated total
#> 1 ManoDd_2106_00-01 0.191 0.031 0.778 0.999
#> 2 ManoDd_2106_01-02 0.070 0.032 0.898 0.999
#> 3 ManoDd_2106_02-03 0.072 0.044 0.884 0.999
#> 4 ManoDd_2106_03-04 0.073 0.063 0.864 0.998
#> 5 ManoDd_2106_04-05 0.068 0.027 0.905 1.000Same code for virtual mixtures:
## Summarise MixSIAR model previsions
MixSIAR.summary.VM <- fingR::JAGS.summary(jags.1 = jags.VM, # Data from the MixSIAR model `MixSIAR::run_model()`
mix = MSIAR.VM, # Virtual mixtures dataset
sources = MSIAR.source.VM, # Source dataset loaded with virtual mixture mix
path = dir.mod.MixSIAR, # Directory path for saving the files
note = "VM", # Optional: Additional note to append to the file name
save_pred = TRUE # Optional: Save the MixSIAR modelling predictions (heavy files)
)
## Extracts the median value of the previsions
MixSIAR.preds.VM <- fingR::JAGS.pred(path = paste0(dir.mod.MixSIAR, "contrib_VM.csv"), # location of files generated by `JAGS.summary`
stats = "Median", # Summary statistics to calculate (Median or Mean)
save = TRUE, # If the result should be saved
note = "VM" # Optional: Additional note to append to the file name
)
## Ensure that the total predicted contribution sums to 1 or 100%
MixSIAR.preds.VME <- fingR::ensure.total(data = MixSIAR.preds.VM, # Predicted source contribution for each sample, data from fingR::BMM.pre
sample.name = "sample", # Column name for sample identifier
path = dir.mod.MixSIAR, # Optional: Directory path for saving the results
note = "VM" # Optional: Additional note to append to the file name
)MixSIAR.preds.VME[1:5,]
#> sample Median_Forest Median_Subsoil Median_Undecontaminated total
#> 1 VM-001 0.150 0.398 0.451 0.999
#> 2 VM-002 0.069 0.418 0.513 1.000
#> 3 VM-003 0.084 0.452 0.464 1.000
#> 4 VM-004 0.081 0.480 0.439 1.000
#> 5 VM-005 0.076 0.510 0.414 1.000The modelling accuracy of MixSIAR model is evaluate with the virtual mixtures. These virtual mixtures, serving as target samples with known contributions (VM.contrib), allow for the calculation of modelling accuracy metrics based on their prediction.
The eval.groups function calculates several common modelling accuracy
metrics: ME, RMSE, squared Pearson’s correlation coefficient (r2), and
Nash-Sutcliff Modelling Efficiency Coefficient (NSE).
MixSIAR.stats <- fingR::eval.groups(df.obs = VM.contrib, # Theoretical contribution
df.pred = MixSIAR.preds.VME %>% dplyr::select(-total), # Predicted contribution (remove the $total column from ensured data.frame)
by = c("Sample_name" = "sample"), # Column where mixtures labels are specified (for `dplyr::left_join` function)
path = dir.mod.MixSIAR, # Optional: Directory path for saving the results
#note = "example" # Optional: Additional note to append to the file name
)MixSIAR.stats
#> Type Source ME RMSE r2 NSE
#> 1 Median Forest -0.20 0.25 0.73 -0.01
#> 2 Median Subsoil 0.08 0.14 0.79 0.69
#> 3 Median Undecontaminated 0.12 0.28 0.08 -0.27The CRPS functions calculate the continuous ranking probability score
and returns a list contraining two data.frame objects; one with the
$samples CRPS values per source class group (saved as CRPS.csv), the
other is $mean with the mean of the CRPS per source class groups
(saved as CRPS_mean.csv).
# Calculate prediction CRPS values
MixSIAR.CRPS <- fingR::CRPS(obs = VM.contrib, # Observed contributions
prev = read.csv(paste0(dir.mod.MixSIAR, "MixSIAR_prevision_VM.csv")), # Predicted prevision from MixSIAR saved by `JAGS.summary()`
source.groups = c("Forest", "Subsoil", "Undecontaminated"), # Source class groups
mean.cal = TRUE, # Calculate mean CRPS per source class group
save.dir = dir.mod.MixSIAR, # Optional: Directory path for saving the results
#note = "example" # Optional: Additional note to append to the file name
)MixSIAR.CRPS$samples[1:6,]
#> Sample_name Forest Subsoil Undecontaminated
#> 1 VM-001 0.0838 0.3950 0.4923
#> 2 VM-002 0.0500 0.2945 0.3661
#> 3 VM-003 0.0570 0.2867 0.3633
#> 4 VM-004 0.0572 0.2645 0.3391
#> 5 VM-005 0.0519 0.2555 0.3261
#> 6 VM-006 0.0455 0.2404 0.3042MixSIAR.CRPS$mean
#> Source CRPS.mean
#> 1 Forest 0.1520
#> 2 Subsoil 0.0801
#> 3 Undecontaminated 0.1715The interval.width functions calculate two prediction interval width:
The W50 contains 50% of the prevision (Q75-Q25) and the W95 contains
95% of the prevision (Q97.5-Q2.5). It returns a list contraining two
data.frame objects; one with the $samples prediction interval width
values per source class group (saved as Interval_width.csv), the other
is $mean with the mean of the prediction interval width per source
class groups (saved as Interval_width_mean.csv).
# Calculate prediction interval width (W95, W50)
MixSIAR.predWidth <- fingR::interval.width(path.to.prev = paste0(dir.mod.MixSIAR, "MixSIAR_prevision_VM.csv"), # Predicted prevision from MixSIAR saved by `JAGS.summary()`
mean.cal = TRUE, # Calculate mean of interval width per source group
save = TRUE, # Save the results at the same location of the path.to.prev
#note = "exemple" # Optional: Additional note to append to the file name
)BMM.predWidth$samples[1:6,]
#> mix.names source W50 W95
#> 1 VM-001 Forest 0.242 0.794
#> 2 VM-001 Subsoil 0.584 0.997
#> 3 VM-001 Undecontaminated 0.332 0.845
#> 4 VM-002 Forest 0.322 0.595
#> 5 VM-002 Subsoil 0.613 0.997
#> 6 VM-002 Undecontaminated 0.498 0.897BMM.predWidth$mean
#> # A tibble: 3 × 3
#> Source W50.mean W95.mean
#> <chr> <dbl> <dbl>
#> 1 Forest 0.352 0.873
#> 2 Subsoil 0.586 0.973
#> 3 Undecontaminated 0.41 0.927The ESP function calculates the Encompassed Sample Prediction (ESP).
The ESP is a newly introduced statistics in Chalaux-Clergue et al
(under review) and was created to assess the transferability of the
statistics calculated on virtual mixtures to actual sediment samples.
The ESP was calculated as the percentage of actual samples for which the
predicted contributions remained within the lowest and the highest
predicted contributions obtained for the virtual mixtures. When
expressed as a percentage, ESP ranges from 0 to 100%, the latter
providing an optimal value. Values close to 100% indicate a higher
transferability of modelling evaluation statistics calculated on virtual
mixture to actual sediment samples.
sources.lvl <- c("Forest", "Subsoil", "Undecontaminated")
# Calculate encompassed sample predictions (ESP)
MixSIAR.ESP <- fingR::ESP(obs = MixSIAR.preds.VM, # Virtual mixtures predicted contributions
pred = MixSIAR.preds.mixE, # Actual sediment samples predicted contributions
sources = paste0("Median_", sources.lvl), # Sources labels in prediction objects
count = "Both" # Count 'Number' and 'Percentage'
)MixSIAR.ESP
#> Source ESP.Number ESP.Percentage
#> Median_Forest Forest 35 92
#> Median_Subsoil Subsoil 0 0
#> Median_Undecontaminated Undecontaminated 9 24Modelling accuracy statistics could be interpreted the following way: “Higher values of W50 indicate a wider distribution, which is related to a higher uncertainty. The sign of the ME indicates the direction of the bias, i.e. an overestimation or underestimation (positive or negative value, respectively). As ME is affected by cancellation, a ME of zero can also reflect a balanced distribution of predictions around the 1 : 1 line. Although this is not a bias, it does not mean that the model outputs are devoid of errors. The RMSE is a measure of the accuracy and allows us to calculate prediction errors of different models for a particular dataset. RMSE is always positive, and its ideal value is zero, which indicates a perfect fit to the data. As RMSE depends on the squared error, it is sensitive to outliers. The r2 describes how linear the prediction is. The NSE indicates the magnitude of variance explained by the model, i.e. how well the predictions match with the observations. A negative RMSE indicates that the mean of the measured values provides a better predictor than the model. The joint use of r2 and NSE allows for a better appreciation of the distribution shape of predictions and thus facilitates the understanding of the nature of model prediction errors. The CRPS evaluates both the accuracy and sharpness (i.e. precision) of a distribution of predicted continuous values from a probabilistic model for each sample (Matheson and Winkler, 1976). The CRPS is minimised when the observed value corresponds to a high probability value in the distribution of model outputs.” (Chalaux-Clergue et al, 2024).
Upcoming updates will introduce graphical support functions such as Bayesian prediction density plots, prediction vs. observation plots, and ternary diagrams.
If you encounter a clear bug, please file and issue or send an email to Thomas Chalaux-Clergue and Rémi Bizeul.
To cite this packages:
utils::citation(package = "fingR")
#> To cite the 'fingR' package in publications please use:
#>
#> Chalaux-Clergue, T. and Bizeul, R (2024). fingR: A package to support
#> sediment source fingerprinting studies, Zenodo [Package]:
#> https://doi.org/10.5281/zenodo.8293595, Github [Package]:
#> https://github.com/tchalauxclergue/fingR, Version = 2.0.0.
#>
#> Une entrée BibTeX pour les utilisateurs LaTeX est
#>
#> @Manual{,
#> title = {fingR: A package to support sediment source fingerprinting studies},
#> author = {{Chalaux-Clergue} and {Thomas} and {Bizeul} and {Rémi}},
#> year = {2024},
#> month = {6},
#> note = {R package version 2.0.0},
#> doi = {https://doi.org/10.5281/zenodo.8293595},
#> url = {https://github.com/tchalauxclergue/fingR},
#> }-
Chalaux-Clergue, T., Bizeul, R., Foucher, A., & Evrard, O. (2024a). An unified template for sediment source fingerprinting databases (24.03.01) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.10725787.
-
Chalaux-Clergue, T., Evrard, O., Durand, R., Caumon, A., Hayashi, S., Tsuji, H., Huon, S., Vaury, V., Wakiyama, Y., Nakao, A., Laceby, J. P., & Onda, Y. (2024b). Organic matter, geochemical, visible spectrocolorimetric properties, radiocesium properties, and grain size of potential source material, target sediment core layers and laboratory mixtures for conducting sediment fingerprinting approaches in the Mano Dam Reservoir (Hayama Lake) catchment, Fukushima Prefecture, Japan (Version 2) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.10836974.
-
Chalaux-Clergue, T., Bizeul, R., Batista, P. V. G., Martinez-Carreras, N., Laceby, J. P., Evrard, P. (2024c). Sensitivity of source sediment fingerprinting to tracer selection. SOIL, 10(1), 109-138. https://doi.org/10.5194/soil-10-109-2024.
-
Chalaux-Clergue, T., & Bizeul, R. (2024d). fingR: A support for sediment source fingerprinting studies (All version). Zenodo. https://doi.org/10.5281/zenodo.8293595. Github. https://github.com/tchalauxclergue/fingR.
-
Laceby JP, Olley J, Pietsch TJ, Sheldon F, Bunn SE. Identifying subsoil sediment sources with carbon and nitrogen stable isotope ratios. Hydrological Processes. 15 avr 2015;29(8):1956‑71. https://doi.org/10.1002/hyp.10311
-
Stock, B. C., Jackson, A. L., Ward, E. J., Parnell, A. C., Phillips, D. L., & Semmens, B. X. (2018). Analyzing mixing systems using a new generation of Bayesian tracer mixing models. PeerJ, 6, e5096. https://doi.org/10.7717/peerj.5096.
-
Stock, B. C., Jackson, A. L., Ward, E. J., Parnell, A. C., Phillips, D. L. (2020). MixSIAR: Bayesian Mixing Models in R (Version 3.1.12). Zenodo. https://doi.org/10.5281/zenodo.594910. Github. https://github.com/brianstock/MixSIAR/tree/3.1.11