Dear ISIMIP modelers,
We would like to share more figures that illustrate the extent of the discontinuity problem. Please find those plots of regional and mean annual cycles for all climate models and bias-adjusted climate variables at [1]. The averages shown there were calculated over all years of a climate scenario (color) and all land grid cells belonging to a Giorgi region [2] (panel).
We have also been thinking about how to solve the issue. As explained in the previous email, the root of the problem is that the bias adjustment was done month by month. We think that an application using a running window will generate continuous annual cycles. The 31 day-wide window would be moved over the course of the year in steps of n days (n = 1, or n = 3 to save computational time). All data within the window would be used to estimate the parameters of the transfer function. This function would then be used for a bias adjustment of only the central n days of the window.
We will now implement and test this approach. If it turns out to work as expected, we will apply the updated method and replace all ISIMIP3b bias-adjusted climate input data. In this case, we will need to decide whether we should still accept impact model outputs based on the old climate inputs to be uploaded to the ISIMIP data portal. We will discuss this with the sector coordinators. So if you have an opinion on this, please let your sector coordinator know.
Dams and reservoirs:
The ISIMIP3 dams data sets have been replaced, now featuring the following time periods:
- ISIMIP3a: reservoirs-dams_1850_2020.xlsx includes the dams and reservoirs until 2020 instead of 2025; additionally, the reservoir areas from KSU have been updated.
- ISIMIP3b: reservoirs-dams_1850_2015.xlsx is the same file as in ISIMIP3a, but cropped until 2015.
Best wishes,
the ISIMIP data team at PIK
[1] https://www.isimip.org/files/plots/inputdata/ISIMIP3BASDv2.4.1/annual_cycle/png_united/
