Update on ISI-MIP bias correction method, data etc.

Posted by Lila Warszawski on April 24, 2012

With this circular, we are sending you some important information on the following topics:

1) Changes in bias correction method, replacement of input data
Status of precipitation and snowfall data
New email archive on website
New tools and auxiliary files available on website
Timeline for data upload

1) Changes in bias correction method, replacement of input data

What’s the problem?

In pre-processing the ISI-MIP climate input data, we have so far been using an existing bias correction method (used e.g. in the WaterMIP project), which we believe is well suited particularly for studies focusing on periods of relatively stable climate. However, we have recently become aware of some drawbacks of this method when it comes to future climate projections.

Most importantly, projected long-term trends (relative precipitation and absolute temperature change) are not preserved by the current bias-correction algorithm in all cases. In particular, monthly global mean temperatures calculated using future daily temperature values can be rendered incompatible with the uncorrected global monthly means. Since a principal aim of ISI-MIP is the quantification of uncertainties at different levels of global warming, it is essential that the time series of global mean temperature change is consistent. Therefore, we are currently revising the bias correction method. We will provide corrected temperature data, that preserves the monthly temperature averages, except for an offset. This unfortunately means that we will have to replace the temperature data.

In addition, we are working on a modified version of the precipitation code to ensure that the relative change in precipitation is preserved.

What to expect?

At the moment we are testing the revised bias correction method, and are putting all our efforts towards completing the revision of the data as quickly as possible.


In the best case we will provide a completely new data set for all variables, that preserves the absolute or relative trend of the original GCM data.


At the very least we will provide revised temperature data, that preserves the trend, while the other variables are still corrected based on the WaterMIP method.


In any case we will provide a complete set of bias corrected historical and RCP8.5 climate data for the HadGEM2-ES model by 1 May, so that you can start model production runs by that date; and we will of course continue processing additional data as fast as we can after that date.

We will of course notify you of any changes in data availability, and kindly ask for your patience.

How to proceed?

If you have already done model runs using the currently available input data, this work remains useful. To minimise delay, it will be extremely helpful if you have already tested and prepared your model and routines, so that once the revised data is available, production runs can start immediately. 

We will provide you with a detailed explanation of the revised bias correction method as soon as possible. We would also like to thank everybody who pointed out to us specific problems with the currently available input data! We are confident that we can address these problems with a revised bias correction method, and that the revised data will be much improved.

We are genuinely sorry for the inconvenience that this procedure is causing. After considering all aspects of the problem, we came to the conclusion that only by revising the data we can ensure that the research objectives in ISI-MIP can be met and high-quality scientific results achieved. We sincerely hope that you understand our motivation, and we apologize for the additional work that this decision may cause you.

2) Status of precipitation and snowfall data

As reported in the two previous circulars, unrealistically high monthly-mean precipitation (pr) and snowfall (prsn) values were found in some input data files. We have identified the source of the problem and resolved it. The revised data set (see above) will contain corrected pr and prsn data, and we will notify you when it is available.

3) New email archive on website

To make sure that all the information and updates we are communicating to you are permanently accessible independent of people's email clients, we have set up an email archive on the website which will be frequently updated. Visit www.isimip.org, then click "E-mail archive".

4a) Regridding tools available on website for HadGEM2-ES grid under "Input data"

As indicated in the simulation protocol, ISI-MIP input data is provided on a 0.5 x 0.5 degree grid, and ISI-MIP participants are asked to run their models at 0.5 x 0.5 degree horizontal resolution where possible. Nonetheless, some participants will need to run at a coarser resolution in order to be able to cover at least the minimal scenario set-up.

The JULES model team has kindly provided some tools that can be used to interpolate the data to the 1.875 x 1.25 degree grid of the HadGEM2-ES climate model. You may use these if you choose to regrid to this resolution. Visit the ISI-MIP website www.isimip.org and click "Input data", then "Regridding to coarser resolution" or simply follow the link:


4b) WFD lat-lon file available on website under "Input data"

The WATCH Forcing data files come without latitude-longitude information. On the website www.isimip.org under "Input data" you will find a NetCDF file that contains that information and can be used to map the WATCH data to the latitude-longitude grid.

4c) Land-cover data on website under "Input data"

Harmonizing land cover and land-use data is not intended within the ISI-MIP Fast Track. However, if you need data for your model runs and do not have your own set-up you may use the data listed on www.isimip.org. Go to "Input data", then click "Land-cover and topographic data" . For references please see the simulation protocol (section 4.3).

5) Timeline for data upload

Because of the delay caused by the revision of the input data (see 1)), we need to relax the May 1st deadline for submitting a first output data set for testing. We now simply ask you to provide us with a first piece of output data for testing purposes as soon as you have completed your first model production run. A web platform will be made available where you can upload output data, and we will inform you about the details as soon as they are sorted out.

Please do not hesitate to contact us if you have any questions or comments.