MSLT

Git page for the multiple scale line transect (MSLT) R-package mslt.
The mslt package uses line transect data to estimate the rate of a thinned spatial point process, which can be used to estimate the abundance of ,e.g., marine mammals. For a detailed description of the model, see https://doi.org/10.1080/01621459.2025.2566422.

Installation

The model is installed by typing

remotes::install_github("NorskRegnesentral/MSLT/MSLT")

A quick example

Here is a quick example of how to use the mslt package. For the full R code to run an example, see the section "A runnable example" blow.

Data

The length data must be in the format of a data frame with the following columns: Description of each columns are provided below

code            time         lon     lat       vessel  Port Stbd   dist     size     perpDist
  0   2019-01-14 10:56:10   -64.886 -56.934    KPH    EMB  <NA>   0.000    NA       NA
  1   2019-01-14 10:59:45   -64.881 -56.938    KPH    EMB  <NA>   0.568    NA       NA
  1   2019-01-14 10:59:45   -64.876 -56.942    KPH    EMB  <NA>   0.568    NA       NA
  1   2019-01-14 11:01:04   -64.871 -56.946    KPH    EMB  <NA>   0.451    NA       NA
  2   2019-01-14 11:02:56   -64.866 -56.950    KPH    EMB  <NA>   0.581    2        0.221
  1   2019-01-14 11:06:01   -64.861 -56.954    KPH    EMB  <NA>   0.506    NA       NA

Here is a description of each column in the line transect survey data:

code Description of what the row represents: 0) start of a transect. 1) no fin whale was detected. 2) a fin whale group was detected.

time Time of recording (POSIXct).

lon Longitude of this point on the transect (numeric).

lat Latitude of this point on the transect (numeric).

vessel Name of the vessel

Port Observer on the port side (string); NA if no observer is present.

Stbd Observer on the starboard side (string); NA if no observer is present.

dist Distance in kilometers with sampling effort from the previous row (numeric); defined as zero when the row represents the start of a transect.

size Number of animals in the detected group (numeric); NA if no group is detected.

perpDist Perpendicular distance in kilometers to the detected whale group (numeric).

Set up configurations and run model

Set up configurations, see ?defConf for details:

conf = defConf(...)

Set up the data and the parameters:

data = setData(surveyData,predAreaUTM,conf)
par = setPar(data,conf)

Estimating the model:

run = fitMSLT(data,par,conf)

A runnable example

This example reproduces the main results in the JASA 2025 paper: "Spatial Variation on Multiple Scales in Line Transect Data; the Case of Antarctic Fin Whales"

The paper can be found here: https://doi.org/10.1080/01621459.2025.2566422

Load the R-package:

library(mslt)

Read Data

Read the line transect data and the survey area. The data is stored in folder papers/jasa2025/data

setwd(tempdir())
files = c("lineTransectData.rds", "surveyArea.rds")
url <- "https://raw.githubusercontent.com/NorskRegnesentral/MSLT/main/papers/JASA2025/data/"
d <- lapply(files, function(f)download.file(paste(url,f,sep=""), f))
surveyData <- readRDS("lineTransectData.rds")
predAreaUTM <- readRDS("surveyArea.rds")

Set configurations for the model:

conf = defConf(detectionTrunc = 3.5, #Truncation limit 3.5 km in our case, see Eq. (20)
               g_function = 2, #Hazard detection function, see eq. (20)
               cutoff = 8, #Maximum distance between SPDE mesh nodes along transects, see Section 2.2 and S3.1.
               max.edge = c(20,80), #Maximum distance between SPDE mesh nodes in survey area and outer area, see Section 2.2  and S3.1.
               cellsize = 50, #Distance between spatial integration points, see Eq (19)
               matern_intensity = 1, #Activate SPDE, see Eq (2)
               mmpp = 1, #Activate MMPP , see Eq (2)
               applyPodSize = 1, #Activate group size model, see section 2.4.
               covDetection = "vessel" #Use vessel as a covariate in Eq. (2)
)

Set up the data and the parameters:

data = setData(surveyData,predAreaUTM,conf)
par = setPar(data,conf)

Estimating the model:

run = fitMSLT(data,par,conf)

Inspect Results

We can inspect the estimated model parameters reported in Table 1 in https://doi.org/10.1080/01621459.2025.2566422:

run$rep

We can inspect the derived model parameters reported in Table 2 by calling partable_derived(run). However, we have not yet performed the bias-correction to obtain $\hat{N}_{\text{BC}}$.

Bias-correction

Bias correction can be performed like this:

index = list(run$obj$env$ADreportIndex()$logAbundance) #Only do bias-correction for estimated abundance
sdrep <- TMB::sdreport(run$obj,
                               bias.correct = TRUE,
                               bias.correct.control = list(split = index,sd = TRUE),
                               getReportCovariance = FALSE)

We can inspect the bias-corrected abundance estimate reported in Table 2 in https://doi.org/10.1080/01621459.2025.2566422:

partable_derived(run,sdrep)

Estimate simpler submodels

The different submodels can be estimated by modifying the conf object accordingly:

SPDE: To estimate the SPDE-only model, set conf$mmpp = 0 and conf$matern_intensity = 1.

MMPP: To estimate the MMPP-only model, set conf$mmpp = 1 and conf$matern_intensity = 0.

HPP: To estimate the HPP model, set both conf$mmpp = 0 and conf$matern_intensity = 0.

Only number of groups: To estimate only the number of groups, set conf$applyPodSize = 0.

Practical convenient details

All estimated parameters are included in the object run$pl. That object has exactly the same format as the initial parameters par. Similarly, the standard deviations of the estimated parameters are stored in run$plSd. Note that the run object includes everything needed to reproduce and inspect the results. For example, the SPDE mesh is located here: attributes(run$data)$mesh.