Temporal Monitoring and Source Change Detection

The Quick Detection module, Temporal Monitoring tab, analyses a series of dated isotopic measurements to detect statistically significant changes in source or dominant process, and to quantify isotopic drift over time.

The Problem This Tab Solves

A site manager monitors a river downstream from an industrial zone. Dissolved antimony analyses show stable concentrations, but isotopic signatures progressively drift from +0.76‰ in March to +0.85‰ in September, before returning in December. Concentrations do not change, but the isotopic source does.

This type of drift is invisible to conventional chemical analyses. It can signal a process change at an upstream facility, the activation of a new drainage zone, a seasonally variable contribution from a second deposit, or a change in soil oxidation conditions. Isotopic temporal monitoring is the only analytical tool that makes this change detectable and quantified.

The same reasoning applies in an industrial context. An automotive battery manufacturer monitors monthly the isotopic signature of incoming lead (²⁰⁷Pb/²⁰⁶Pb ratio) across its refined lead deliveries. The chemical composition of batches is compliant at each receipt, but the isotopic ratio drifts from 0.8521 in January to 0.8674 in July, before returning to 0.8538 in November. This amplitude, significant at 2.3σ, reveals that a supplier introduced lead from a different provenance for six months, likely sourced from second-hand recycled batteries rather than certified primary lead, without this being detected by certificates of analysis or chemical controls. In a RoHS or REACH compliance context, this silent substitution constitutes a non-conformity traceable by isotopy.

Expected Data

Temporal monitoring works with two data sources:

Option 1: Import a CSV File

The CSV file must contain at minimum three columns:

The separator can be a comma or semicolon. The accepted date format is YYYY-MM-DD. Uncertainty is at 2σ. If the uncertainty column is absent, IsoFind uses a default value of 0.05‰.

Option 2: Load from the Database

If a sample has been measured multiple times on different dates in IsoFind (each measurement recorded with its collection_date field), it can be selected directly from the interface. An element and isotopic ratio selector allows choosing the tracer to analyse when the sample has multiple tracers available. IsoFind automatically loads the time series and displays a preview chart.

Step-by-Step Workflow

1. Choose the Data Source Switch to the Import CSV or From Database tab. For CSV import, drop the file or click to select it. For the database, type the name of the sampling site and select the sample from the results.
2. Check the Preview After loading, the preview chart displays the time series with uncertainty bars on each measurement. Verify that dates are correctly interpreted and that values are consistent.
3. Set the Sensitivity and Window The sensitivity determines the sigma threshold for breakpoint detection: High (1.5σ) detects small shifts, Medium (2.0σ) is the standard setting, Low (3.0σ) only flags major changes. The comparison window defines the number of measurements before/after each point for calculating moving averages (minimum 2, recommended 3).
4. Run the Analysis and Interpret Click Analyse Series. The result chart displays detected breakpoints in red, the trend line in purple dashes, and descriptive statistics (mean, standard deviation, range, R²).

The Breakpoint Detection Algorithm

IsoFind uses a CUSUM (Cumulative Sum) algorithm on a sliding window. For each point i in the series, it calculates the difference between the mean of the N preceding measurements and the mean of the N following measurements, normalised by the pooled uncertainty. A breakpoint is flagged when this difference exceeds the configured sigma threshold.

Breakpoints too close together (distance less than the window) are deduplicated, retaining only the most significant. The global trend is calculated by OLS linear regression, with R² as an indicator of drift linearity.

Interpreting the Results

Practical Case: Seasonal Monitoring of a Sampling Station

Station WAT-OR-001 (Rio Desaguadero, Bolivia) was measured 5 times between March 2024 and March 2025. The IsoFind test dataset shows:

• T1 (March 2024): +0.764‰, start of wet season
• T2 (June 2024): +0.798‰, transition to dry season
• T3 (September 2024): +0.851‰, advanced dry season, peak
• T4 (December 2024): +0.729‰, return of rains
• T5 (March 2025): +0.680‰, wet season

Analysis results (High sensitivity, window 2): one positive breakpoint detected between T2 and T3 (+0.053‰, significant at 1.9σ), global trend slightly decreasing (slope = −0.018‰/measurement, R² = 0.42). The geochemical interpretation is consistent with a dry-season concentration effect (less dilution) that increases the relative contribution of an isotopically heavier source (AMD or oxidation process), offset by the return of precipitation.