Data & Figures

For Everyone

Browse visuals showing when models agree, where they differ, and how confidence changes over time.

For Researchers

Each figure includes dataset membership, metric definition, time span, and provenance links. Time units: kyr BP (Present=1950).

Exploratory Visuals

Some panels are exploratory diagnostics. We keep visuals that add clarity and drop those that don’t. The data, not a narrative, drives what stays.

Dataset Scope Context

Why amplitudes differ: 2206 uses full-span normalization; CALS10k spans 0–10 ka. This affects relative amplitudes. Use --annotate-scope (see README) to surface span context without changing underlying stats. Differences also reflect inversion regularization, temporal coverage, and baseline choices.

Reading Coverage

Tiers partition time: A (0–10 ka, four datasets), B (10–100 ka, two), C (100–2000 ka, PADM2M only). Use tiers to interpret divergence and agreement: Tier C often lacks peers, not necessarily agreement.

Figure Gallery

Divergence Multiview Panels – Reading Guide

Each divergence card bundles four lenses on a single reconstruction (strength C, change rate C′):

  1. Raw dual-axis: Smoothed C and C′ (5 ka window) share time, separate y-axes.
  2. Global z-scores: Standardized smoothed series reveal relative departures.
  3. Percent anomalies: Deviation from median (or rolling median for short spans).
  4. Δz divergence: Difference z(C′)−z(C) highlighting rate vs magnitude mismatch.
For comparative clarity, some figures restrict spans. The 2206 divergence panel shown here is truncated to 0–10 ka so its variability can be read directly against CALS10k on identical time bounds. See README section “Divergence Multiview Scaling” for command-line parameters.

Downloads

Methodology · Reproduce · Interactive · Divergence Guide