Skip to contents

3b. Modeling with a GAM

Source: vignettes/V3b_GAM.Rmd
V3b_GAM.Rmd

Summary

Theory

TemporalModelR operates across two complementary dimensions to support temporally explicit modeling:

  • E-space - (Environmental Space) represents a location in terms of its environmental conditions, independent of geography. A species’ niche in E-space is its tolerance for a given set of environmental variables.
  • G-space - (Geographic Space) represents a location in terms of where it physically sits on the landscape. Every point in G-space has a corresponding location in E-space.

Because environmental conditions change over time, the same point in G-space may move through E-space across years, decades, or seasons. Traditional, temporally-static workflows assume both G-space and E-space to be stable in time. TemporalModelR assumes only that the niche (i.e., the species’ tolerance in E-space) is stable across the study period, while the E-space coordinates of any G-space location may change over time. Species observations are therefore matched to E-space data correct in both space and time, the niche is modeled in time-independent E-space. This is done in the Preprocessing temporally explicit data vignette.

By modeling the niche in time-independent E-space using time-independent training data, we can generate a static prediction of a species’ environmental tolerances. We can then project these static E-space predictions back onto G-space at explicit time periods, resulting in dynamic, temporally-explicit ENM predictions of species distributions.


Overview

build_temporal_gam() fits one binomial generalized additive model per cross-validation fold as was created during the Preprocessing temporally explicit data vignette. Each fold’s model is trained on all data outside the fold and evaluated on the held-out fold. The user supplies the right-hand side of the model formula directly with smooth terms via s(), te(), or ti(). See Smooth term syntax. The link function defaults to logit but can be set to probit, complementary log-log, or cauchit. A threshold is selected on the training data and applied to the continuous predictions to produce binary suitability output for downstream summarization.

Unlike GLMs, GAMs let environmental responses take flexible, data-driven shapes through smooth terms while still producing interpretable per-variable effect curves. This is useful when the species’ response to a predictor is potentially nonlinear or unimodal, but a user still wants to prioritize interpretability over fitting very complex relationships (contrasting random forests).

This vignette runs the seasonal workflow using the bundled tmr_partition and tmr_absences objects, which are pre-built outputs of the partitioning and pseudoabsence steps produced by the Preprocessing temporally explicit data vignette using the same call patterns shown there. The dataset itself is described in About the Example Dataset. This is the same dataset as is used in each modeling vignette GLM, GAM, Random Forest or Hypervolume.

The mgcv package is a hard dependency of build_temporal_gam() and must be installed before running:

library(TemporalModelR)
library(terra)
#> terra 1.9.34

data(tmr_partition, package = "TemporalModelR")

data(tmr_absences,  package = "TemporalModelR")


Fitting a temporal GAM

The minimal call needs the partition, the pseudoabsences, a formula with smooth terms, and the time columns. We fit an additive model with thin plate regression spline smooths on each of the three predictors and let the function select per-fold thresholds via the True Skill Statistic (TSS), though this threshold may also be set manually or using prevalence. See Threshold selection.

create_plot = TRUE allows the function to generate visual response curves for each per-fold model result. On the below graphs, response curves are shown per-fold, and the selected threshold is indicated with a horizontal dashed line.

gam_out <- build_temporal_gam(
  partition_result     = tmr_partition,
  pseudoabsence_result = tmr_absences,
  model_formula        = ~ s(elevation) + s(forest_cover) + s(prseas),
  link                 = "logit",
  gam_params           = list(method = "REML"),
  threshold_method     = "tss",
  output_dir           = file.path(tempdir(), "GAM_Models"),
  create_plot          = TRUE,
  time_cols            = c("year", "season"),
  verbose              = FALSE
)

The smoothed response curves show how the model captures any nonlinearity in each predictor’s effect on suitability without us having to specify polynomial terms. The same create_plot = TRUE function also visualizes time independent model assessment metrics relevant to the GAM (above). See E-space performance.

class(gam_out)
#> [1] "TemporalGAM" "list"

names(gam_out)
#>  [1] "models"             "thresholds"         "threshold_method"  
#>  [4] "model_formula"      "link"               "model_vars"        
#>  [7] "fold_training_data" "fold_test_metrics"  "output_dir"        
#> [10] "model_type"         "plots"

The returned object is a list of class "TemporalGAM" containing the following objects:

  • $models - named list of gam objects, one per fold (fold1, fold2, …). Each is a standard mgcv::gam fit and accepts R’s standard summary(), coef(), and predict() calls, as well as mgcv specific tools like plot.gam().
  • $thresholds - named numeric vector of probability thresholds used to binarize predictions for each fold.
  • $threshold_method - character or numeric value recording how the thresholds were chosen ("tss", "prevalence", or a fixed numeric value).
  • $model_formula - the right hand side formula supplied to the function, including smooth wrappers like s() or te().
  • $link - the link function used ("logit", "probit", "cloglog", or "cauchit") supplied to the function.
  • $model_vars - character vector of the predictor variable names extracted from the formula. Smooth wrappers are unwrapped to the base variable name. Used by generate_spatiotemporal_predictions() to know which raster layers to assemble at projection time.
  • $fold_training_data - named list of the training data frames used for each fold, with presences and pseudoabsences combined and environmental columns attached.
  • $fold_test_metrics - data frame of per-fold E-space metrics computed on the held-out test set. See E-space performance.
  • $output_dir - path to the directory where the saved RDS of fitted models and the fold test metrics CSV were written.
  • $model_type - "gam", used by generate_spatiotemporal_predictions() to choose the correct projection logic.
  • $plots - list of recorded base-R plots produced when create_plot = TRUE (smoothed response curves and the fold metrics bar chart).


Smooth term syntax

The response variable presence is added automatically to each formula, so only the right-hand side needs to be supplied. Standard mgcv syntax applies, including s() for univariate smooths, te() for tensor product smooths between predictors on different scales, and ti() for tensor interaction smooths with main effects removed. Parametric terms can be mixed in alongside smooths. These examples that all work:

### Univariate smooths
~ s(elevation) + s(forest_cover) + s(prseas)

### Mix smooth and parametric terms
~ s(elevation) + forest_cover + prseas

### Tensor product smooth between forest cover and precipitation
~ s(elevation) + te(forest_cover, prseas)

### Tensor interaction with main effects also kept
~ s(elevation) + s(forest_cover) + s(prseas) + ti(forest_cover, prseas)

### Constrained basis dimension and explicit basis type
~ s(elevation, k = 5, bs = "tp") + s(forest_cover, k = 5) + s(prseas, k = 5)
  • s() fits a univariate smooth. The default basis is thin plate regression splines (bs = "tp") and the basis dimension k is chosen automatically by mgcv unless specified.
  • te() fits a tensor product smooth, useful when two predictors interact and may be on different scales.
  • ti() is similar but with main effects removed, useful when including main effects separately in the formula.
  • Constraining k (e.g. k = 5) is sometimes necessary when sample sizes are small. The default basis dimension can otherwise be larger than the data supports and produce a warning.

Predictor names are extracted with all.vars(), so smooth wrappers like s(elevation) are correctly unwrapped to the base variable name for response curve plotting and prediction.

If you choose a link other than logit, set it via the link argument.

Anything that would normally be passed to mgcv::gam() can go in gam_params. See ?mgcv::gam for the full list.


Threshold selection

GAMs produce probabilities on the 0–1 scale. The rest of the TemporalModelR pipeline operates on binary suitability rasters, so a probability threshold must be chosen to convert probabilities into 0/1 predictions. threshold_method supports three options:

  • "tss" (default) - selects the threshold that maximizes True Skill Statistic (sensitivity + specificity − 1) on the training data. See E-space performance. This balances commission and omission.
  • "prevalence" - sets the threshold equal to the prevalence (proportion of presences) in the training data for that fold.
  • A numeric value between 0 and 1 - used as a fixed threshold for all folds, e.g. threshold_method = 0.4.

The chosen thresholds for our four folds:

gam_out$thresholds
#> fold1 fold2 fold3 fold4 
#>  0.34  0.30  0.27  0.31

These are per-fold values because each fold trains on a different subset of data and may end up with different probability distributions. Using a single global threshold across folds risks underclassifying suitability in some folds.


E-space performance

Each fold’s held-out test set provides a set of presence and pseudoabsence points the model has never seen based on the folds defined in the Preprocessing temporally explicit data vignette. We compare predicted vs observed at those points and compute confusion-matrix metrics. Because these are evaluated in E-space (the species’ environmental tolerance, independent of geography or time), they give a stable picture of how well the model is preforming.

gam_out$fold_test_metrics
#>   Fold Threshold Testing_TP Testing_FN Testing_TN Testing_FP Sensitivity
#> 1    1      0.34         32          5         38         36      0.8649
#> 2    2      0.30         35          2         44         30      0.9459
#> 3    3      0.27         42          1         26         60      0.9767
#> 4    4      0.31         32          1         35         31      0.9697
#>   Specificity    TSS  Kappa    AUC
#> 1      0.5135 0.3784 0.3128 0.6907
#> 2      0.5946 0.5405 0.4545 0.8338
#> 3      0.3023 0.2791 0.2078 0.6160
#> 4      0.5303 0.5000 0.4074 0.7098

Columns:

  • Fold - fold identifier matching the folds in tmr_partition$points_sf$fold.
  • Threshold - the per-fold probability threshold used to binarize predictions.
  • Testing_TP - count of test-set presence points correctly classified as suitable (true positives).
  • Testing_FN - count of test-set presence points incorrectly classified as unsuitable (false negatives).
  • Testing_TN - count of test-set pseudoabsence points correctly classified as unsuitable (true negatives).
  • Testing_FP - count of test-set pseudoabsence points incorrectly classified as suitable (false positives).
  • Sensitivity - proportion of test-set presences correctly classified, Testing_TP / (Testing_TP + Testing_FN). Values close to 1 mean the model rarely misses presences.
  • Specificity - proportion of test-set pseudoabsences correctly classified, Testing_TN / (Testing_TN + Testing_FP). Values close to 1 mean the model rarely flags pseudoabsence points as suitable.
  • TSS - True Skill Statistic, Sensitivity + Specificity - 1. Ranges from -1 to 1, with 0 indicating performance no better than random and values above ~0.4 considered useful for most applications.
  • Kappa - Cohen’s kappa statistic adjusting overall classification accuracy for the accuracy expected by chance. Ranges from -1 to 1 with similar interpretation to TSS.
  • AUC - area under the ROC curve, computed on continuous probabilities rather than binarized predictions, so it is threshold-independent. Useful for comparing the models overall performance independent of threshold.

The full ROC curve and each above metric are also graphed as an output of create_plot = T in the Fitting a temporal GAM section.

E-space metrics are robust to imbalanced sample sizes across time, because they pool across the time series. They are a good metric for assessing overall model fit. Time-specific (G-space) metrics can also be assessed later when we project the model to specific G-space and time combinations.


Projecting predictions

generate_spatiotemporal_predictions() projects each fold’s model onto a stack of environmental rasters matching each requested time step, producing one fold-vote raster per time step. The same call works for all four model types; model_type in the model object tells the function which projection logic to use.

For this example we project across all 15 years and 4 seasons, producing 60 prediction layers total, each summarizing votes across the four folds:

scaled_dir <- system.file("extdata/rasters_scaled", package = "TemporalModelR")

time_steps <- expand.grid(
  year             = 1:15,
  season           = c("Spring", "Summer", "Autumn", "Winter"),
  stringsAsFactors = FALSE
)

preds <- generate_spatiotemporal_predictions(
  partition_result     = tmr_partition,
  model_result         = gam_out,
  pseudoabsence_result = tmr_absences,
  raster_dir           = scaled_dir,
  variable_patterns    = c(
    "elevation"    = "elevation",
    "forest_cover" = "forest_cover_YEAR",
    "prseas"       = "prseas_YEAR_SEASON"
  ),
  time_cols            = c("year", "season"),
  time_steps           = time_steps,
  output_dir           = file.path(tempdir(), "GAM_Predictions"),
  overwrite            = TRUE,
  verbose              = FALSE
)

We can now visualize the 60 prediction rasters as a year-by-season grid. Each cell shows the number of folds (0 to 4) that classified that pixel as suitable.

terra::plot() limits each call to 16 layers, so we render the stack in four blocks of four years each.

pred_stack <- terra::rast(preds$prediction_files)

pred_names    <- basename(preds$prediction_files)
pred_seasons  <- sub(".*_(Spring|Summer|Autumn|Winter)\\.tif$", "\\1", pred_names)
pred_years    <- as.numeric(sub(".*_(\\d+)_(Spring|Summer|Autumn|Winter)\\.tif$",
                                "\\1", pred_names))
season_levels <- c("Spring", "Summer", "Autumn", "Winter")
stack_order   <- order(pred_years, match(pred_seasons, season_levels))

pred_stack <- pred_stack[[stack_order]]
ordered_years   <- pred_years[stack_order]
ordered_seasons <- pred_seasons[stack_order]
names(pred_stack) <- paste0("Y", ordered_years, "_", ordered_seasons)

block1 <- which(ordered_years %in%  1:4)
block2 <- which(ordered_years %in%  5:8)
block3 <- which(ordered_years %in%  9:12)
block4 <- which(ordered_years %in% 13:15)
terra::plot(pred_stack[[block1]], nr = 4, nc = 4,
            mar = c(1.0, 1.0, 1.5, 3.0), legend = FALSE)

terra::plot(pred_stack[[block2]], nr = 4, nc = 4,
            mar = c(1.0, 1.0, 1.5, 3.0), legend = FALSE)

terra::plot(pred_stack[[block3]], nr = 4, nc = 4,
            mar = c(1.0, 1.0, 1.5, 3.0), legend = FALSE)

terra::plot(pred_stack[[block4]], nr = 3, nc = 4,
            mar = c(1.0, 1.0, 1.5, 3.0), legend = FALSE)

Visualizing our predictions across each season and year, we see that there is generally high agreement among models. These rasters represent consensus votes among each of our four folds, with yellow pixels having strong positive consensus among all folds (all four identify the pixel as suitable) and blue pixels having low consensus among folds (few to no folds identify the given pixel as suitable). We also see that the models correctly visually show one of the main temporal trends in the data: loss of habitat through deforestation starting around year 6.

There is lower consensus among models during the summer predictions. Some folds are correctly picking up on the limitations imposed to occupancy based on precipitation variability for this example species, and are correctly predicting constrained suitability during the summer. Others (like fold 2) see no substantial influence of precipitation on occupancy and therefor predict similar pixels as suitable across all years. Similarly to the GLM method in the previous vignette, this is in part due to the lack of low precipitation value background points in our dataset which would correctly identify the gap here and this methods reliance on that background data to make inference. We can explore if other modeling methods are able to fix this hurdle in other vignettes. Alternatively, we could use an alternative method for generating absence points which we describe more in the the Preprocessing vignette.

Note that here we show that time_steps can handle making predictions from compound variables or variables at different temporal scales so long as their temporal scales are nested. For example here “elevation” has no temporal value and is considered to be static across all time steps. “forest_cover” is measured annually, but is considered to be static across all seasons within a year for the purposes of predictions. “preseason” is measured both by year and season, so resulting seasonal predictions reflect that. However if precipitation was only measured based on aggregate seasons but had no associated year, predictions would fail. Predictions can also be made where all variables share the same time step- for example annual forest cover, annual temperature, and annual precipitation.

Additionally, a plain vector like time_steps = 1:15 produces one prediction per value of the first time column; the other time columns are filled in with every unique value present in the occurrence data. So a vector input with time_cols = c("year", "season") would produce one prediction per year-season combination observed in the data. A data frame like the expand.grid() above gives explicit control: any combination of values can be requested, including ones not present in the occurrence data, as long as the matching rasters exist (Such as Summer or Winter predictions).


G-space performance

Once predictions are projected, per-time-step metrics become available. These are computed by overlaying the held-out test points (presences plus, optionally, pseudoabsences) on the prediction raster for each time step they fall in and counting correct vs incorrect classifications.

head(preds$timestep_metrics)
#>   Fold Pct_Suitable N_Pres TP FN Sensitivity         CBP N_Abs TN FP
#> 1    1       0.2800      2  2  0         1.0 0.078400000     4  2  2
#> 2    2       0.3067      2  2  0         1.0 0.094044444     4  2  2
#> 3    3       0.3089      5  5  0         1.0 0.002811975    10  2  8
#> 4    4       0.2933      0 NA NA          NA          NA    NA NA NA
#> 5    1       0.2733      0 NA NA          NA          NA    NA NA NA
#> 6    2       0.2933      2  1  1         0.5 0.414577778     4  2  2
#>   Specificity TSS year season
#> 1         0.5 0.5    1 Spring
#> 2         0.5 0.5    1 Spring
#> 3         0.2 0.2    1 Spring
#> 4          NA  NA    1 Spring
#> 5          NA  NA    2 Spring
#> 6         0.5 0.0    2 Spring

Columns:

  • Fold - fold identifier; one row per fold per time step.
  • Pct_Suitable - proportion of the study area predicted suitable in that time step.
  • N_Pres - number of held-out presence test points falling in that time step for that fold.
  • TP - true positives, presences correctly classified as suitable in that time step.
  • FN - false negatives, presences incorrectly classified as unsuitable in that time step.
  • Sensitivity - TP / (TP + FN) for that fold-time step.
  • CBP - cumulative binomial probability. Under the null that each test point lands in suitable area at the rate Pct_Suitable, this is the probability of observing exactly TP correctly classified points by chance. Small values indicate predictions are better than random.
  • N_Abs - number of held-out pseudoabsence test points falling in that time step. NA for presence-only models.
  • TN - true negatives, pseudoabsences correctly classified as unsuitable.
  • FP - false positives, pseudoabsences incorrectly classified as suitable.
  • Specificity - TN / (TN + FP) for that fold-time step.
  • TSS - Sensitivity + Specificity - 1 for that fold-time step.
  • year and season - the values of time_cols for that row, identifying which time step the metrics belong to.

G-space metrics are powerful in time steps with substantial sample sizes but lose meaning when few records exist (as with out example). A fold with only one presence in a given year can score sensitivity 0 or 1 with no real information content. Likewise, the ability to earn a significant G-space CBP score explicitly depends on sample size. Report E-space (from $fold_test_metrics) and G-space (from $timestep_metrics) together for a complete view of model performance.

The overall summary gives a per-fold summary across all time steps:

preds$overall_summary
#>   Fold N_Timesteps Mean_Pct_Suitable Total_TP Total_FN Overall_Sensitivity
#> 1    1          60            0.1705       32        5              0.8649
#> 2    2          60            0.2415       35        2              0.9459
#> 3    3          60            0.2247       42        1              0.9767
#> 4    4          60            0.1855       32        1              0.9697
#>    Overall_CBP Total_TN Total_FP Overall_Specificity Overall_TSS
#> 1 4.439470e-20       38       36              0.5135      0.3784
#> 2 9.708071e-20       44       30              0.5946      0.5405
#> 3 1.953461e-26       26       60              0.3023      0.2791
#> 4 1.035253e-22       35       31              0.5303      0.5000

Rapid visual assessment of these metrics can be done using the function plot_model_assessment() which generates simple summary plots for each.

plot_model_assessment(
  predictions         = preds,
  time_column         = c("year", "season"),
  secondary_time_mode = "combine",
  model_result        = gam_out
)
#> Loaded fold_test_metrics for 4 fold(s).

#>          Metric    Mean     SD
#>    Pct_Suitable  0.2055 0.0898
#>     Sensitivity  0.9306 0.2136
#>     Specificity  0.4916 0.2757
#>  CBP < 0.05 (%) 39.7000     NA

This produces per-fold time series of percent suitable, sensitivity, specificity, CBP, TP/FN, and TN/FP. When model_result is also supplied, overall sensitivity and specificity from $fold_test_metrics are added as reference lines. When compound time steps are used (such as year and season), you must choose how they are visualized. Choosing secondary_time_mode = "combine" will roll them into one continuous x axis. secondary_time_mode = "facet" produces stacked plots as seen below, where the first time_col is displayed as the x axis, and a different plot is made for each secondary variable (season here). By default, the threshold for which CBP is identified as significant is 0.05, but this may also be adjusted manually.

plot_model_assessment(
  predictions         = preds,
  time_column         = c("year", "season"),
  secondary_time_mode = "facet",
  model_result        = gam_out,
  cbp_threshold       = 0.001
)
#> Loaded fold_test_metrics for 4 fold(s).

#> Facet mode: produced 4 stacked plot(s) across 4 season value(s).

In both visualizations, we can clearly see the constrictions in suitable area during the summer in the G-space predictions per time step graphs.


Next steps

Now that predictions have been generated, you can assess the model and see if it is preforming satisfactory enough for what your goals are. If this is the case, you can preform post-processing analyses to try to gain additional inference about the spatiotemporal patterns of change in the study region See the Post-processing predictions vignette.

For comparison with other algorithms applied to the same dataset: