BSync

Bayesian Synchronisation for Proxy Time‑Series

BSync is an open‑source R implementation of the Bayesian Synchronisation framework described in our forthcoming paper Bayesian Synchronization of Proxy Paleorecords with Reference Chronologies. The method aligns an undated (or imprecisely dated) proxy record to one or more well‑dated reference records while fully propagating alignment uncertainty. BSync supports single‑target, double‑target and target‑with‑age‑uncertainty workflows.

---

Table of contents

1. Overview
2. Key features
3. Installation
4. Quick start
5. Function reference
6. Reproducing the paper figures
7. Citing BSync
8. Contributing
9. Licence
10. Contact

---

Overview

Synchronising palaeo‑environmental records is essential for multi‑archive climate reconstructions, yet traditional "tie‑point" approaches provide little or no statistical measure of the alignment uncertainty. BSync tackles this by casting synchronisation as a Bayesian inverse problem:

• A flexible link function (inverse‑gamma process) deforms the input age/depth scale.
• Prior information on sedimentation behaviour and memory is encoded through interpretable hyper‑parameters.
• The likelihood couples the rescaled input and target proxy signals using a heavy‑tailed t distribution or a kernel density estimator when the target carries its own age uncertainty.
• The full posterior is sampled with the t‑walk MCMC algorithm, enabling credible intervals for the alignment, accumulation rates and derived chronologies.

Status of the accompanying manuscript A detailed description of the model, benchmarking and application case‑studies is in preparation and will be submitted for peer review in late‑2025. A pre‑print will be posted on ArXiv as soon as it is ready. Early adopters are encouraged to cite this repository and to contact the author for an advance copy.

---

Key features

• Three alignment modes – single target, double target (optimal mixture of two chronology‑consistent targets) and target with reported age uncertainty.
• Objective yet configurable – default priors work out‑of‑the‑box, but every hyper‑parameter is exposed for expert tuning.
• Heavy‑tailed error model to reduce the impact of local noise and proxy‑specific variance.
• Automatic convergence monitoring through log‑objective traces and ESS diagnostics.
• Publication‑ready plots of proxy alignment, age‑depth models and parameter posteriors.

---

Installation

BSync is written in R (>= 4.2) with small C++ extensions via Rcpp. You will need the following libraries:

install.packages(c("Rcpp", "KernSmooth", "coda"))

Clone the repository and source the main script:

git clone https://github.com/maquinolopez/BSync.git
cd BSync

---

Quick start

# Load the core routine
source("Bsynchv4.R")  # script will load the helpers automatically

# files shall be in a folder name "input target"

# Run Bayesian synchronisation (defaults are usually fine)
Bsynch(input,           # name of the input record
       target,          # name of the target record
       burn      = 5e4,
       iterations = 3e5,
       thinning   = 100)

See examples/ for fully‑worked scripts that reproduce selected figures from the manuscript.

---

Function reference

The main entry point is Bsynch()

Argument	Type	Default	Description
input_record	data.frame	required	Proxy record to be aligned
target_record	data.frame	required	Reference proxy record
folder	character	getwd()	Output directory for MCMC chains, plots & logs
burn	integer	5e4	Burn‑in iterations
iterations	integer	3e5	Total iterations
thinning	integer	100	Keep every n‑th sample
sigma.prior	numeric	1	Prior SD of proxy residuals
n_sections	integer	50	Number of inverse‑gamma sections
shape_acc, mean_acc	numeric	1.5, 50	Hyper‑parameters of accumulation‑rate prior
strength_mem, mean_mem	numeric	10, 0.5	Memory prior parameters
sd_shape, sd_scale	numeric	1.5, 0.01	Hyper‑parameters of the SD prior
savefig	logical	TRUE	Save alignment plot as PDF

Full documentation is available in the function header.

---

Reproducing the paper figures

Scripts in paper/ replicate every figure and table from the upcoming manuscript, including:

• Iberian Margin case‑study (benthic & planktic δ¹⁸O).
• Synthetic‐core benchmarks for single‑, double‑ and uncertainty‑aware targets.
• Comparison with BIGMACS.

Running make paper will execute all workflows and save outputs in results/ (≈ 4 h on a modern laptop).

---

Citing BSync

If you use BSync in your research, please cite both the software and the manuscript (once available):

@software{aquino2025_bsync,
  author  = {Marco A. Aquino‑López},
  title   = {BSync: Bayesian Synchronisation of Proxy Palaeo‑records},
  year    = {2025},
  url     = {https://github.com/maquinolopez/BSync}
}

@unpublished{aquino2025_manuscript,
  author = {Aquino‑López, M. A. and Osman, M. and Muschitiello, F.},
  title  = {Bayesian Synchronization of Proxy Paleorecords with Reference Chronologies},
  note   = {Manuscript in preparation, planned submission 2025}
}

---

Contributing

Pull requests are welcome! Please open an issue first to discuss major changes. By contributing you agree to license your work under the terms below.

---

Licence

BSync is released under the MIT License – see LICENSE for details.

---

Contact

Questions, suggestions or requests for an early copy of the manuscript? Drop me a line:

Marco A. Aquino‑López Centro de Investigación en Matemáticas (CIMAT), Mexico ✉️ aquino@cimat.mx

---

Happy synchronising!