ETH Zurich researcher advancing ML weather forecast evaluation and spectral diagnostics of AI prediction models
2
papers here
2
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
5
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-09-20) and cover the author's entire public record, not just the papers listed here. Reproducibility signal comes from Evolve's own reviews.
Research profile
Output—
The output axis is not yet scored: this snapshot lacks a reliable total for this author record's publication volume and venue mix, so production cannot be computed. It would measure aggregate works per year and their distribution across venues and publication types.
Impact—
Impact data is insufficient here: citation totals and downstream uptake are unavailable for this record, so the axis stays unscored. Once populated it would measure citations, h-index trajectory, and influence on weather-ML evaluation practice.
Focus0/100
The focus score is 0, which contrasts with the raw field distribution: every fielded record sits in just two adjacent labels — atmospheric science (AI/ML weather prediction) at 0.5 and machine learning for weather forecasting at 0.5. Both reviewed papers target forecast evaluation, so the evidence points to a tightly concentrated agenda; the 0 more likely reflects how the metric handles the split dual labeling than genuinely scattered interests.
Collaboration46/100
Collaboration scores 46 on a compact base: 9 distinct coauthors appear in the recent window, with Jiakai Chen, Simon Driscoll, Sebastian Schemm and Fanny Lehmann among recurring partners. In both reviewed 2026 papers Oskarsson sits middle or last rather than corresponding author, indicating a contributing specialist role inside a small ETH-anchored group rather than solo work or consortium leadership.
Activity75/100
Activity scores 75, backed by the countsByYear series: 6 works in 2023, 4 in 2024, 5 in 2025 and 3 already in 2026 — 12 works in the last three years per the recent3yWorks figure. After sparse earlier entries in 2014, 2018, 2020 and 2022, the record shows a clear acceleration into a sustained, high-tempo ML-weather research phase.
Reproducibility78/100
Reproducibility scores 78, computed from two reviewed papers (reproMean 78, reproPapers 2). The learned-critic study rates high readiness, its classifier two-sample test on ERA5 reanalysis being straightforward to rerun, while the four-model kinetic-energy-cascade study rates medium owing to its single-case spectral design. An arXiv-first release pattern supports verification, though a two-paper base keeps the estimate thin.
Research summary
Joel Oskarsson is a researcher at ETH Zurich working at the interface of machine learning and atmospheric science; his OpenAlex field profile splits evenly between ML-based weather forecasting and AI-driven atmospheric science. His recent record shows clear momentum: 12 works in the last three years, with yearly counts of 6 (2023), 4 (2024), 5 (2025) and 3 (2026 to date). Two 2026 arXiv papers reviewed by Evolve define his current agenda — evaluating probabilistic ML weather models. One introduces a learned atmospheric critic that adapts where fixed forecast metrics fail, built on classifier two-sample testing and f-divergence estimation against ERA5 reanalysis; the other examines the butterfly effect and kinetic energy cascade across four ML weather models, finding realistic spectral magnitudes but flawed cross-scale energy transfer. Evolve rates the first high reproducibility readiness and the second medium, yielding a reproducibility mean of 78 across two papers. Citation-based output and impact aggregates are not yet available in this snapshot (h-index 5, i10-index 2 under partial coverage), so this profile should be read as an active record of early visibility, tightly concentrated on forecast evaluation methodology.
Collaboration pattern
The collaboration footprint is small and stable: 9 distinct coauthors in the recent window, with Jiakai Chen, Simon Driscoll, Sebastian Schemm and Fanny Lehmann recurring. Both reviewed 2026 papers place Oskarsson as middle or last author, not corresponding, consistent with a specialist contributor role inside a compact ETH Zurich group. The source coauthor list also contains Oskarsson's own name, an artifact of the underlying record.
Research keywords
Highlights
A learned atmospheric critic exposes blind spots in fixed forecast metrics
Four-model spectral study: realistic KE magnitudes, wrong cross-scale transfer
Timeline
- 2014Earliest indexed record in the OpenAlex countsByYear series (1 work) — the visible starting point of this author record.
- 2023Most productive year in the record with 6 indexed works, marking the shift into a sustained ML-weather research phase.
- 20244 indexed works keep output steady, within a recent window spanning 9 distinct coauthors.
- 20255 indexed works confirm a multi-year run of continuous, active output.
- 20263 works indexed so far, including both Evolve-reviewed arXiv papers on learned forecast criticism and kinetic energy cascades.
Papers in the Evolve catalog
Frequent co-authors
About this profile
This profile is AI-structured by Evolve.Pub from public bibliographic data (OpenAlex) and Evolve's own paper reviews. It is not written or endorsed by the researcher unless the profile is claimed. Radar dimensions are computed from real bibliometrics with fixed formulas — raw figures are shown alongside; missing dimensions mean insufficient data, never zero. Errors? Claim the profile to correct or hide it.
- Bibliometrics derive from OpenAlex only; the record is name-resolved, and topic clusters mixing weather, traffic and hydrology may include works by other authors sharing this name.
- Output and impact axes are unscored because citation aggregates are not yet available; h-index (5) and i10-index (2) reflect partial coverage.
- Quality and reproducibility assessments rest on only two Evolve-reviewed 2026 arXiv preprints, not peer-reviewed venue records.
- The frequent-coauthor list includes the researcher's own name, a source-data artifact.