Forcing datasets for land surface models

Below is the copy of an email from David Lawrence sent to the ILMF mailing list.
It discusses the various forcing datasets for land models and their shortcomings. It also asks for input from all modelling centres on datasets they are using.

This issue will be discussed at the Land Working Group meetings.

Thanks to everyone who commented on my original question on this (see end of message for the original email). The ILMF Steering Committee met a few weeks ago and we talked about this issue, the lack of a global long-term historical, bias-corrected, and regularly updated forcing dataset for land models, which is a general problem and also a specific problem for CMIP7 land-hist experiment and possibly/probably also for TRENDY. We decided that we needed to get some more information in advance of our next discussion on this towards the end of April.

  1. We would like to get a better sense of what datasets are out there and in use by different modeling groups. This spreadsheet has two Sheets. The first one is a listing of known existing global historical forcing datasets that Danica Lombardozzi and I have started to populate. Please add to this list if you know of other datasets. The second Sheet is a list for modeling centers to put in what datasets you are currently using (for example, I entered for CLM where we are mainly using CRUJRA and GSWP3 … which produce vastly different results!).

  2. It seems that there is some momentum towards a switch to ERA5 as the underlying reanalysis product (with bias correction). The issue with this reanalysis product, as has been noted, is that it starts in year 1940. Previously, the datasets that we used for CMIP6 all extended back to 1901 through one of the long-term historical reanalysis products. So, I wanted to know how spinning up for 1940 climate era conditions would impact long-term carbon simulations so we ran with CLM two simulations with the CRUJRA forcing dataset, one using the climate forcing from the years 1901-1920 for spinup and for the pre-1901 period and one using the years 1940-1949 for the spinup and the pre-1940 period. The global carbon cycle trends from these two experiments are plotted here. Though there is a bit of an offset in soil carbon, the overall trends, especially during the modern period, look quite similar, which implies that we may be able to live with a dataset that only extends back to 1940 without impacting results too much.

  3. I’ve also been in communication with Hyungjun Kim, the creator of the very nice GSWP3 dataset. He is exploring whether or not he could use JRA-3Q, which I understand is the successor to JRA55, to form the basis of a product using similar methodology to how GSWP3 was generated. JRA-3Q goes back to 1948. This seems like another possible alternative to ERA5 for a dataset that could be regularly updated.

So, thanks again to anyone who commented previously. It would be really helpful if a representative from different modeling centers could fill in the second sheet linked above and also for anyone to comment further on this issue. We are trying, in the absence of a funded group stepping forward to fill the void, to work towards a collaborative plan to produce a dataset that could be of wide use for the land model research community so all help / advice about how to get there is very much appreciated.