remove data/hydro_capacities.csv with untraceable source

[fneum]

This is kind of a step backward as it uses crude fill values, but I am skeptical regarding the data file that is currently being used.

• The source is untraceable.
• The storage capacities seem dubiously high (for most countries the existing storage capacity values only reach <20% of it). I don't trust it. Better to be conservative.
• The powerplantmatching has ok coverage of storage capacities (~40% or reservoirs weighted by power capacity, many of the larger hydro reservoirs, best coverage in Norway)
• Use 10% quantile of storage capacities as fill-value. Is this a good number? For all of Europe, 40 hours.
• Use 24 hours of storage duration as the backstop.
• Obviously, it would be better if we had all the data.

Distribution of storage durations (h) in powerplantmatching

[koen-vg]

I guess your digging for the source also ended around here?

I had a brief look at the two cited publications but in a few minutes I couldn't find the promised hydro numbers.

I quickly checked the numbers for the three countries with the largest reservoir capacities (Norway, Sweden, Spain), and they seem to be very accurate, however. (Just looking at E_store[TWh].) Did you check different countries for which the results were clearly wrong?

I think this PR is reasonable if we know that we want to throw this unexplainable file out. But it would be a shame if that happened to significantly worsen the data quality for the largest hydro countries. What's the difference, more or less?

On the other hand we all know that hydro is notoriously difficult; even with the correct storage capacities we don't automatically get correct results anyway, and lowering reservoir capacities might actually make the final results more accurate since the model tends to use reservoir capacities more aggressively than is realistic.

[fneum]

Thanks @koen-vg, @lisazeyen and I decided to put this on hold for the moment.

The motivation was indeed to remove data without a source.

We have asked some more hydro-experts we know, whether they can help.

[fneum]

Potential alternative source: https://www.mdpi.com/1996-1073/10/11/1841

[coroa]

Regarding the original file, it was shared with us directly by Alexander Kies who worked at FIAS during the nascent days of PyPSA, i also went once more over the linked reports and while this is work that Alex contributed to the data is not in any way part of it that i can see (A. Kies report for RESTORE 2050 ). I guess the story is that this is actually the dataset that ISI used for their study (since their scenario b is explicitly about using the hydro storage capacities to buffer the variability) and shared it with Alex for his renewable hydro work on the RESTORE 2050, who shared it with us. But none of them actually reproduced it in their papers.

If there is any way, i would also suggest to remove it (and i was considering to do so myself before)!

I am unsure how well Matteo De Felice's work at JRC compares to what powerplantmatching is having currently, but that might also be a good source: https://github.com/energy-modelling-toolkit/hydro-power-database .