WeatherXBiodiversity — does Soroye 2020’s bumble bee–climate claim replicate?

What this site is¶

A FORRT (Framework for Open and Reproducible Research Training) replication study of Soroye, Newbold & Kerr (2020, Science).

The original paper claims that the frequency with which local temperatures exceed species-specific historical thermal tolerances — not just mean warming — predicts local extirpation of bumble bee species across North America and Europe. We test whether that claim:

1. Reproduces when the authors’ R analysis is re-implemented in Python, on the same underlying data, and

2. Replicates when the same Python pipeline is applied to independent open occurrence data for the Iberian Peninsula.

Both tests come back Validated, High confidence, with the Iberian coefficient ~3× larger in magnitude than the continental mean.

Headline result¶

The mechanism Soroye describes is real, and on the warm-edge subregion of Iberia it is stronger — exactly what one would predict if species reaching their thermal limit are most vulnerable to thermal-exposure events.

The full FORRT chain¶

Nine nanopublications, all on Science Live:

                        Soroye et al. 2020
                              ▲
              ┌───────────────┴───────────────┐
              │                               │
    Quote-with-comment  ───►  AIDA  ───►  FORRT Claim
                                              │
                                              │ (tested by)
                                              │
                          ┌───────────────────┴───────────────────┐
                          ▼                                       ▼
                Phase 2 Replication Study           Phase 3 Replication Study
                    (Robustness)                       (Regional replicability)
                          │                                       │
                          ▼                                       ▼
                 Phase 2 Outcome                          Phase 3 Outcome
                  Validated / High                          Validated / High
                          │                                       │
                          ▼                                       ▼
              CiTO confirms Soroye 2020              CiTO confirms Soroye 2020
                          │                                       │
                          └─────────────► Wikidata  ◄─────────────┘

• The claim level is paper-rooted: a verbatim quote from Soroye’s Discussion is wrapped with a personal comment and an AIDA-format claim.

• Each Phase has its own Replication Study + Outcome — Phase 2 swaps the analysis software (R → Python) while keeping data fixed; Phase 3 swaps the data (continental → Iberian GBIF) while keeping software fixed. Together they isolate two different senses of “the claim holds up.”

• CiTO citations wire each Outcome back to the original paper DOI as typed citations (cito:confirms). This makes the replication machine-readable in the open scholarly graph and makes it eligible for downstream import into Wikidata / Scholia via dedicated pipelines such as Egon Willighagen’s; inclusion there is gated by Wikidata’s own notability criteria and requires the import to be run.

The next pages walk through each layer.

Quick links¶

• The Claim — Quote, AIDA, FORRT Claim

• Phase 2 — Reproduction on Soroye’s data

• Phase 3 — Iberian replication

• Methods — Python port, grid, TEI/PEI, GLMM

• Discussion — Why is Iberia stronger? What’s next?

• References

Reuse and citation¶

This work is meant to be reused — to project bumble bee extirpation risk onto future climate, to flag candidate climate refugia, or to extend the mechanism to other thermally-sensitive insect taxa. The validated pipeline is archived at Zenodo and the GBIF download at GBIF.org User (2026).

If you build on this, please cite both the original paper and this repository.

Context¶

Prepared for the BioHackathon Europe 2026 project WeatherXBiodiversity, which packages cross-domain biodiversity × climate data as ARC / RO-Crate FAIR Digital Objects.