Changes in version 2.0.3 (2026-05-03)                  

Patch release with substantive vignette improvements and several factual
corrections to the worked example. No changes to analytical behaviour or
to any function signature.

  - The vignette now live-renders all worked-example figures from the
    bundled Code Lake dataset; the previous build showed only static
    code blocks. Figure chunks are gated on the availability of ggplot2,
    patchwork, and ggtext (all in Suggests) and skip cleanly if any is
    missing.
  - Two diagnostic figures, char_plot_raw() (Fig. 1) and
    char_plot_thresh_diag() (Fig. 2), are now included in the vignette
    alongside the analytical figures (Figs. 3, 5, 6, 7, 8).
  - A new bundled parameter file CO_compensated_charParams.csv (with
    companion CO_compensated_charData.csv) is shipped in
    inst/validation/. It is identical to the standard CO_charParams.csv
    except that the working threshold percentile (threshValues[4]) is
    lowered from 0.99 to 0.95 to compensate for language-induced drift
    in the Gaussian mixture model (GMM) noise estimation. The vignette
    uses this file for the worked example. With this compensation, R
    identifies 50 peaks for Code Lake (close to the published MATLAB
    v2.0 result of 48 peaks) and reproduces the significant decrease in
    fire-return intervals from Zone 1 to Zone 2 reported in Higuera et
    al. (2009). The strict 1-to-1 reference configuration remains
    available as CO_charParams.csv.
  - The vignette's Comparison with MATLAB v2.0 section gained a
    Threshold column in the validation table and a new "CO
    (compensated)" row; the narrative below the table was rewritten in
    terms of compensation for GMM drift rather than ad-hoc tuning.
  - Vignette factual corrections: Code Lake is in Alaska (not Colorado);
    the smoothing-method label for Code Lake in the validation table is
    Method 4 (moving median), not Method 1 (lowess); the discussion of
    smoothing-related differences was adjusted to identify Method 2
    (robust lowess) as the only method that diverges between R and
    MATLAB.
  - Fixed a stale system.file() path in the vignette (extdata →
    validation) that would have returned an empty string if the chunk
    had been evaluated.

                 Changes in version 2.0.2 (2026-04-28)                  

Patch release addressing the third round of CRAN reviewer feedback, plus
a related rendering fix discovered during local testing. No changes to
analytical behaviour.

  - Title field shortened to "Peak Detection and Fire History from
    Sediment-Charcoal Records" (62 characters) to satisfy the CRAN
    convention of titles under 65 characters.
  - char_write_results() and char_plot_all() no longer default out_dir
    to the working directory. out_dir is now required for
    char_write_results() and required when save = TRUE for
    char_plot_all(). This brings the package into compliance with the
    CRAN policy against writing to the user's home filespace by default.
    Users should pass an explicit path; tempdir() is acceptable for a
    transient location.
  - Vignette updated to write its example output to tempdir() instead of
    a Results/ directory in the user's working directory, with a note
    that real users would substitute their own path.
  - Fixed axis-label rendering in char_plot_peaks(),
    char_plot_fire_history() (peak-magnitude panel), and
    char_plot_zones() (CDF and box-plot panels). These labels were still
    using expression(paste(...)) (plotmath) syntax and rendered as raw
    text under the ggtext::element_markdown() axis-title theme
    introduced in 2.0.1. They now use the same conditional HTML-tag /
    plain-text pattern as the FRI and fire-frequency labels, so super-
    and subscripts render correctly when ggtext is available.

                        Changes in version 2.0.1                        

Patch release addressing CRAN reviewer feedback on the initial
submission. No changes to analytical behaviour.

  - DESCRIPTION: condensed to a single paragraph and removed the
    paragraph separators that were rendering as double periods in CRAN
    metadata.
  - char_parameters() is now exported. Its help page previously
    contained an example for an unexported function.
  - Replaced \dontrun{} wrappers in all examples with \donttest{} (or
    unwrapped entirely, where the example runs in under 5 seconds). All
    examples now use the bundled validation dataset via system.file()
    and write any output to tempdir().

                        Changes in version 2.0.0                        

First R implementation of CharAnalysis, a direct translation of
CharAnalysis v2.0 (MATLAB). Analytical outputs are validated against
four benchmark datasets; user testing is ongoing. The package is in the
experimental lifecycle stage: the API may change as user feedback is
incorporated.

Please report issues at https://github.com/phiguera/CharAnalysis/issues.

New features relative to MATLAB v2.0

  - Full R package with devtools::install_github() installation.
  - CharAnalysis() returns a named list with S3 class "CharAnalysis";
    print(), summary(), and plot() methods are provided.
  - Eight publication-quality ggplot2 figures (char_plot_peaks(),
    char_plot_fire_history(), char_plot_cumulative(), char_plot_fri(),
    char_plot_zones(), char_plot_raw(), char_plot_thresh_diag(),
    char_plot_sni()); save all to PDF with char_plot_all().
  - char_write_results() writes the 33-column output matrix to CSV with
    column headers and numeric format matching the MATLAB output
    exactly.

Known differences from MATLAB v2.0

Robust lowess background (smoothing method 2). MATLAB's Curve Fitting
Toolbox smooth(..., 'rlowess') and the R char_lowess() port handle NaN
gaps differently inside the bisquare robustness iteration, producing
slightly different C_background series for records with missing values.
For gap-free records the difference is negligible (< 0.001); for records
with NaN gaps the maximum absolute difference across validated datasets
is 0.267 (Chickaree Lake). Smoothing method 1 (plain lowess) is
unaffected and agrees to within floating-point noise.

GMM peak counts. The Gaussian mixture model EM algorithm accumulates
floating-point arithmetic differently in R and MATLAB, causing slightly
different threshold values in some local windows. Peak counts differ by
10–20% across validated datasets, with the direction varying by dataset.

Figures 9 and 10. The threshold-sensitivity detail plot (Fig. 9) and
multi-site comparison plot (Fig. 10) from the MATLAB interface are not
implemented in this R package.

Smoothed FRI column (col 23). The R package computes smoothed
fire-return intervals in this column; MATLAB v2.0 stores NaN.

Validation summary

| Dataset | Site | charBkg max|diff| | Peaks R | Peaks MATLAB |
|---------|------|-------------------|---------|-------------| | CO |
Code Lake, AK | ~5 × 10^-6^ | 39 | 48 | | CH10 | Chickaree Lake, CO
| 0.267 | 59 | 50 | | SI17 | Silver Lake, CO | 0.130 | 25 | 31 | | RA07
| Raven Lake, AK | < 0.001 | 15 | 17 |

Full validation details are in
inst/z_Validation_report_R_vs_MATLAB_V_2.0.md.

Citation

If you use CharAnalysis in published research, please cite:

Higuera, P.E., L.B. Brubaker, P.M. Anderson, F.S. Hu, and T.A. Brown.
2009. Vegetation mediated the impacts of postglacial climate change on
fire regimes in the south-central Brooks Range, Alaska. Ecological
Monographs 79:201–219. https://doi.org/10.1890/07-2019.1

If you used Version 2.0 specifically, please also cite the software:

Higuera, P.E. 2026. CharAnalysis: Diagnostic and analytical tools for
peak analysis in sediment-charcoal records (Version 2.0). Zenodo.
https://doi.org/10.5281/zenodo.19304064