I am making an argument in support of setting a minimum transitional period where both ERA5 and ERA6 will be published, in the hope that someone with authority over this will see it. I am working to structure a project finance deal for a new wind farm where wind resource volatility / risk will be outsourced to the insurance / derivative market via a long-term hedge on wind speeds using ERA5 data. The wind hedge is being imbedded into the deal’s debt sizing, with precise wind speed coverage levels varying year over year based on the project’s expected cash flows. This structure of incorporating a long-term (8+ years) wind hedge into a project financing has never been done before; if successful, it could have a significant impact on how wind power projects are financed in future. The most material barrier to closing this or future similar financings is the inability to guarantee that a single wind speed data set will be available for the full duration of the deal (8-10 years). The contemplated wind hedge structure is covering very remote wind speeds (P95+); ensuring consistency between ERA5 and ERA6 this far out on the distribution becomes very difficult because of how few data points exist at this end of the curve. Confidence that wind speed coverage levels can be locked in at financial close using one consistent and stable data set / methodology will be a boon to the financing of wind power projects. Even after the shift to ERA6, data shift risk would remain for these structures given an inability to guarantee when ERA 6 will be eliminated in favor of ERA 7.
Thank you Michael, this is an interesting and important point. Rest assure that I have taken notice of it. What is not entirely clear to me is whether you suggest for us to keep producing two operational reanalyses in parallel or whether instead you are suggesting to keep both data stream available once one of the two is updated. If it is the former, I am not sure we would be able to afford it as these systems tend to be computationally expensive to run. If it is the latter, I do have good news for you as we do expect to maintain the access to ERA5 for a relatively long period even after ERA6 had been rolled out operationally. This is not dissimilar to what we did when we passed from ERA-Interim to ERA5. Incidentally we are planning to do the same for the future transition between ERA6 and ERA7.
Thank you Carlo. Unfortunately it is the former - we need one consistent data source & methodology from the closing of the deal through the duration of the wind speed hedge tenor to ensure that the amount of risk transfer and the target coverage levels remain consistent. Is it possible to continue providing new ERA5 data for just one or two grid points after the transition to ERA6? Are you able to run the computations for individual locations without running them for the full planet / data set? We would be able to pay for such a service to ensure consistent data methodology for our target location.
Hello Michael, this is a very understandable situation, there may be alternatives and work-arounds for this. Feel free to get in touch. All the best from Denmark. Rémi, C2Wind. Rémi Gandoin - Vejle, Region of Southern Denmark, Denmark | Professional Profile | LinkedIn
Hi Michael,
no, unfortunately it is not possible to generate the time series for ERA5 for just a point without running the whole system. Could the use of some sort of emulator (possibly one based on ML/AI ) , trained over the period that ERA5 and ERA6 will have in common, be part of the solution to your problem? If so such a possibility could be explored further.
Thanks Carlo. Unfortunately it is more of a statistical problem, whereby the sample set of only 40-80 historical years is insufficient to create a distribution smooth / complete enough to where we can keep risk transfer consistent using percentiles. Are you able to share an estimate for the annual computational cost of running ERA5?
Hi Carlo, it looks like the ERA5 dataset is currently updated daily. For the purposes of using ERA5 based wind speed hedges to support the financing of wind farms, we would only need updates on a quarterly basis. Let me know if you think such a reduction in update frequency would reduce the computational costs to keep ERA5 going after the release of ERA6.
Hello Michael,
ERA5 and ERA5T will continue to be produced for some time even after ERA6 and ERA6T are released to the public in late 2026. There won’t be a brutal end to ERA5 and we will make plans to give users ample time to migrate over to using ERA6.
We are aware that maintaining the production of ERA5T is a strong user requirement by many users, including yourself. Like any other user requirement we receive, it will need to be assessed in terms of priority, capabilities, resources, etc before an outcome decision is made.
To record your user requirement as per our user requirement procedure, we shall now raise a support portal ticket on your behalf (outside of the forum) and continue the conversation via the ticket to ensure relevant parties are kept informed.
Again, feel re-assured that the production of ERA5T will not end as soon as, all going well, ERA6 starts being made available towards the end of 2026.
Regards,
ECMWF Support
Thank you Anabelle. Unfortunately time to transition to ERA6 is not my concern. We are working to incorporate very precise long-term wind speed hedges into financial contracts to support wind farm financings. The level of precision necessary is such that any change in data methodology / data set over the course of the hedge’s tenor can cause a unacceptably large shift to the hedge’s coverage levels compared to what is required by the financing. We could quickly change our processes to utilize ERA6 once it comes out, but the change in historical data would cause either the target coverage levels or the amount of risk transfer to materially change. We could theoretically re-do the hedge at this time to correct for the changes, but this would require a re-pricing of the hedge which is unacceptable for non-recourse financings utilized by wind farms. I will continue this discussion via the ticket you created; thanks much for doing so.
I would like to fully second the request made by Michael Kalnas. Indeed, in the German energy context, we are currently looking for a stable set of weather data exactly for this purpose: Writing financial contracts based on long-term stable definitions of weather data.
Please do let me know if there are any more concrete plans in regard to long-term maintenance and how long you plan to support each version.