Climate scientist · Berlin
Independent climate scientist. I quantify basis risk in parametric climate covers — the gap between what a weather index pays and the loss actually suffered — with wind and solar drought in Europe as the working case. Before that: data assimilation, paleoclimate reconstruction and neural downscaling, in peer-reviewed work since 2014.
Working write-ups with numbers — the analysis that does not fit in a paper and should not wait for one.
“Just re-calibrate the trigger” is the standard answer to basis risk. On ten years of European wind data, re-tuning does help, up to a point: the fairest available setting still misses roughly 6.7% of whole droughts and leaves about 19% of drought-hours unpaid.
Parametric covers pay on a weather number, not on your loss — and the gap between the two has a name almost nobody measures. What basis risk is, why a 0.98 correlation doesn't protect you, and why the quiet failure mode costs more than the dramatic one.
In July 2026 the Mediterranean averaged 27.0 °C at the surface — 2.4 °C above the 1982–2011 norm, the warmest July in the 45-year satellite record. Four datasets built by rival methods in three countries agree on why: the basin is warming at ~0.36 °C per decade, twice the global-ocean rate.
The same drought was worth almost nothing, then worth US$8m — and nothing about the weather changed. An independent re-analysis on public data, asking which half of a parametric payout was actually doing the work.
Static maps of Germany's wind resource, drawn from public reanalysis. Dated exhibits, re-drawn when the next full year lands — never a live feed.
100 = the best five percent of German cells in this dataset; the score weighs energy yield (0.5), firmness of the bad days (0.3) and persistence of calm (0.2). Drawn on the German Weather Service's regional reanalysis COSMO-REA6 at six kilometres, 9,534 cells.
COSMO-REA6 publishes daily means of 100 m wind through a generic, uncalibrated power curve, so the energy term under-counts gusty days; the index is scaled to this dataset's own best cells and compares in pattern, not in points, with the ERA5 version. A then-versus-now version on ERA5 (1981–1990 against 2015–2025) was built and then checked against 118 long DWD station records: the stations do not confirm the reanalysis change (rank correlation −0.10, same sign in 49 % of stations), so that map stays unpublished. What is published is the check.
A drought day sits below the cell's own 10th-percentile daily wind inside a spell of at least three days — the same convention as the Opinion #001 series. Drawn on the German Weather Service's regional reanalysis COSMO-REA6 at six kilometres, which resolves the calm pockets of the Alps, the Black Forest and the Bavarian Forest that a 25-kilometre reanalysis smears away.
COSMO-REA6 publishes daily means and ends in August 2019; the then-versus-now comparison therefore stays on ERA5.
Data: ERA5 (Copernicus Climate Change Service, CC-BY-4.0), pulled July 2026; COSMO-REA6 regional reanalysis (Deutscher Wetterdienst open data, free use with attribution), pulled October 2026; borders: Natural Earth (public domain). Static exhibits, dated; re-drawn when the next full year lands.
Disclaimer — published in a personal capacity, under my own name. This is independent work: produced entirely on my own time and my own equipment, from public data (ERA5 reanalysis; ENTSO‑E generation), using no resources, data, or working time of any employer, and outside my assigned duties. It carries no institutional affiliation and is neither connected with nor endorsed by any employer or institution; the views are mine alone. It is not financial, actuarial, or legal advice, and nothing here is an offer of services.
Every peer-reviewed article and book chapter, newest first. Titles and journal names as published, in English. The current citation record lives on Google Scholar; preprints, abstracts, datasets and software stay on ORCID.
Work that is not a paper. A book can carry an argument to a reader who would never open a journal, and it can be held to the same standard on the way — every number computed rather than recalled.
Download the PDF — 23 MBZenodo record
Mazandaran is the province I am from: a strip of Iran between the Caspian, which is the largest lake on earth, and the Alborz, which carries a 5,600-metre volcano. A European reader tends to picture Iran as desert. The book answers that with arithmetic rather than protest — Ramsar takes 1,238 mm of rain a year on 89 rain days, against Hamburg's 771 mm on 188; two thirds of the province lies above 500 m; and the forest on its slopes held beech, oak, elm and hornbeam through the last glacial, while Europe was steppe and tundra.
Every number printed on a plate was computed from open data by the pipeline published with the book, and stored with the method that produced it. Every claim that is not a number resolves to a source with a retrieval date. Ninety acceptance checks run on each build, and the one that still fails is named on the last page instead of hidden. Where the data cannot answer, the book says so: all four rain gauges in the province sit at sea level, so above 500 m the rainfall map is a model no gauge has checked, and the plate says that.
Text, maps and design CC BY 4.0; the code MIT; the two photographs under the Unsplash License and not sub-licensable. Free to read, translate and reprint, with credit.
Download the chapter PDF — open accessChapter at Springer
Central Asia warms faster than the planet. The chapter sets out why — a continental interior far from any ocean, with mountains that amplify the warming with height — and what the record already shows: surface air temperature about 1.2 °C above the pre-industrial baseline, and under the high-emission pathway SSP5-8.5 a rise approaching 6 °C by the end of the century. The projections rest on 442 CMIP6 simulations, downscaled to the region with a neural network.
The second half is about what that heat does to water. The glaciers of the Tien Shan have lost about 27 % of their mass since 1961, and 22 million people downstream depend on the meltwater they buffer; roughly 80 million live on the balance between glacial melt and river flow. Extremes that were rare are no longer: multi-year droughts have become seven times more frequent, and in 2024 rapid snowmelt on saturated soil flooded Kazakhstan and displaced 118,000 people. The chapter is open access, as is the whole volume.
Open access under CC BY 4.0 — free to read, share and adapt, with credit. Chapter 2 of a volume on climate, water and human security in Central Asia, edited by I. Didovets, V. Krysanova, F. F. Hattermann and A. Gafurov.
Email is the surest way to reach me — I read everything, and I answer most things within a week.