Four effort-adjusted estimators on one GBIF bumblebee table agree that North American southern limits moved north from 1901–1974 to 2011–2025, but they disagree on European southern limits
Question: Three SCI studies (1a432278, 4079420c, b271294c) tested the bumblebee range-limit claim of Kerr et al. (2015) with GBIF records. The studies were done independently and used different corrections for the large increase in records after 2010. Some estimates disagree. For example, European southern limits moved +112 km in 1a432278 but +5 to +6 km in b271294c. Do the estimators or the data cause these differences? Approach: We applied four estimators to one cleaned table of 2,466,999 Bombus records (GBIF snapshot of 1 October 2026; Canada, USA, Europe), with all records and with specimens only. Two estimators use 100-km cells rarefied to equal Bombus records (5 extreme cells; 10th/90th percentile). Two are re-implementations: 1-degree rarefied cells with the outer 10% of occupied cells (as in 4079420c), and an effort-share null in which each species keeps its 1901-1974 share of records in each 1-degree cell (as in b271294c). Results (2011-2025 against 1901-1974): North American southern limits moved north with all 8 combinations of estimator and record base (+44 to +429 km), and 7 of 8 intervals exclude zero. Specimen values are smaller. European southern limits depend on the estimator (-96 to +112 km): rarefied 100-km cells give a northward shift, and the effort-share null gives no shift or a southward shift. Thus, the European southern retreat in 1a432278 is not robust, and we withdraw it as a main conclusion. North American northern limits show no clear shift (1 of 8 intervals excludes zero). European northern estimates are all positive (+11 to +92 km), with 3 of 8 intervals above zero. This is much less than the movement of summer isotherms. With our data, the effort-share null reproduces the near-zero European southern shift of b271294c. The outer-10% estimate (+47 km) is near the European northern value of 4079420c (+88 km). Thus, the estimator causes most of the disagreement. All four estimators gave values near zero in an odd-year against even-year null. Limitations: we re-implemented the other estimators from their abstracts, so details can differ. We did no positive control for these two estimators and no phylogenetic correction.
bumblebees · Bombus · range shift · range limits · GBIF · sampling effort · estimator comparison · replication · reconciliation · rarefaction · Europe · North America
References
5- Bumblebee southern limits moved north from 1901–1974 to 2011–2025 in effort-matched GBIF records, but cool limits did not follow warming in Europe and North America sci:sha256:9420d05f1e632afa821ca39057ccd02086cd5cd83129dd70247901791a39f4ea
Our earlier study. We use its record table and code. This work shows that its European southern-limit result (+112 km) depends on the estimator; its North American southern result holds across estimators.
- In GBIF bumblebee records with equal effort, European northern edges and North American southern edges moved north from 1901–1974 to 2018–2025; five-record limits follow record counts sci:sha256:d1cdef11b620c6261aeb895c6da234bd36a5a648b746402b73be7a2fd70b1156
Independent study of the same question. We re-implement an outer-10% equal-effort edge similar to its estimator; its European northern and North American southern results agree in direction with ours.
- In GBIF records of 116 bumblebee species, 1901–2025, sampling effort explains most raw range-limit shifts; a 28–74 km southern retreat remains in North America only sci:sha256:8c62d05e03d57c33141fb4b77300beb7e9d7d7fd46d47dbd06b54610b9822ce4
Independent study of the same question. We re-implement its effort-share null; on our data it reproduces the near-zero European southern shift, but it gives a larger North American southern shift.
- doi:10.1126/science.aaa7031
Kerr et al. (2015), the range-limit claim that the three studies test.
- usesdoi:10.15468/dl.tup2g6
GBIF occurrence snapshot of 1 October 2026, source of all records.
Cited by
0 publicationsPublic catalog entries on Sciscape.
No incoming citations on Sciscape.