Documentation
The complete SedSAT3 User’s Manual provides step-by-step guidance for every tool in the program, from importing raw data through final source apportionment and uncertainty analysis. It also includes a mathematical appendix describing the theoretical basis of each modeling method.
User’s Manual
View the SedSAT3 User’s Manual (PDF) →
The manual is maintained alongside its source files in a dedicated repository, so it is always kept current with the latest version of the software:
User’s Manual repository on GitHub →
What’s inside
1. Raw data preparation
How to format your Excel workbook (one tab per source group plus a target tab),
import it into SedSAT3, assign constituent types (element, isotope, particle size,
organic carbon, or exclude), and include or exclude individual samples. Projects are
saved in a readable JSON format with a .cmb extension.
2. Deterministic fingerprinting tools
The full deterministic workflow, in order: outlier analysis, organic matter and particle size correction, bracketing analysis, stepwise discriminant function analysis for tracer selection, and Levenberg–Marquardt maximum likelihood fingerprinting — available as both single-sample and batch tools.
3. Bayesian sediment fingerprinting
Bayesian chemical mass balance analysis using MCMC, which produces a full posterior distribution for each source contribution rather than a single point estimate. Covers chain settings, burn-in, convergence diagnostics, credible intervals, and the batch version that writes results to text files.
4. Other statistical tools
Supporting analyses for exploring your data and deciding which elements and samples to include: correlation matrices, analysis of variance, auto-select elements, two-way and one-vs-the-rest DFA, distribution fitting, discriminant power, error analysis, Kolmogorov–Smirnov tests, optimal Box–Cox parameters, source verification, and genetic algorithm estimation.
5. Mathematical basis
The theory behind the software: the mass balance formulation (including stable isotopes), maximum likelihood estimation (with source composition treated either deterministically or as unknown), and Bayesian inference.
A typical workflow
- Import your Excel data and assign constituent properties.
- Run outlier analysis and exclude any problematic samples.
- Apply organic matter and particle size corrections if appropriate.
- Run a bracketing analysis for your target sample and exclude unbracketed elements.
- Use stepwise discriminant function analysis to select effective tracers.
- Estimate source contributions with Levenberg–Marquardt, genetic algorithm, or Bayesian methods.
- Quantify uncertainty with Bayesian credible intervals or bootstrap error analysis.
New to the tool? Download the sample dataset and follow along with the manual, or watch the instructional videos.