Created by on - Updated on 04/27/2017 22:53

Climate Comparisons in the MERRA Atlas

The MERRA Atlas provides monthly, annual, seasonal and climate comparisons among several of the latest atmospheric reanalyses (including ERA-Interim, CFSR, and JRA25) and various global observation datasets.

Reanalysis Comparisons on the WRIT webpages

The NOAA-CIRES Web-based Reanalysis Intercomparison Tools (WRIT) has a prototype plotting/comparison page that is available for testing and commenting by users and others. To obtain a login to go to Create New Account.

Observational Studies that examine reanalysis datasets and compare them and observations are described here.

Inter-reanalysis studies that intercompare reanalyses are described here.

WCRP Task Team for the Intercomparison of ReAnalyses (TIRA) - A group of reanalysis developers and users charged to develop a broad reanalysis intercomparison project.


SPARC Reanalysis Intercomparison Project 

A coordinated activity under the WCRP/SPARC focusing on reanalysis output in the stratosphere, upper troposphere, and lower mesosphere. 


Tables: Overview Comparison

ClimateDataGuide Summary and Comparison

Ana4MIPS Data Description


While NCEP/NCAR is using a much older model for the first guess, and an older data assimilation system, it is still widely used around the world. The main paper on the NCEP/NCAR reanalysis, Kalnay et al. (1996)<0437:TNYRP>2.0.CO;2. , specifically lists the expected quality of surface-related variables. For the case of southern africa, the recent paper Zhang et al (2012) examines precipitation. You can make quick comparisons of several surface variables using the Web-based reanalysis intercomparison tool or other tools listed at The other question you need to ask is what observations are included. The 2m air temperature from stations is indirectly assimilated in ERA-Interim, while it is not in NCEP/NCAR. Please let us know of other questions, or if anything can be clarified. best wishes,

ERA-Interim (like ERA-40 and the JRA reanalyses, but unlike the NCEP/NCAR reanalysis) used the 2m temperature observations to produce its 2m temperature values. Any comparisons thus have to be carried out particularly carefully. We did this recently for the Arctic (Simmons & Poli, doi: 10.1002/qj.2422), though for instantaneous temperatures not daily means. This paper and a couple of Simmons et al. papers it refers to provide details on how we analyse screen-level observations. The ERA-Interim system (which is more than nine years old) is rather limited in the feedback information it saves from its 2m temperature analysis. This will be remedied in the replacement for ERA-Interim that is currently being prepared for production.


Tue, 08/19/2014 - 12:00

I am no expert on reanalyses, so I thought I would ask this community this question.

I have been comparing 20cr to ncep/ncar reanalysis. I have computed DJF mslp anomalies wrt 1981-2010 over part of the North Atlantic (40W-5E, 45-70N). Not only do the two agree very well in the overlap period but there is very little spread in the 20cr ensemble in values going back to 1900. Perhaps to be expected, but I am very pleased by this!

For 850hPa T over part of North America (70-110W, 40-65N), however, the range in the early part of the series is larger and while the comparison with the ncep/ncar shows reasonable agreement in terms of the year-to-year variations, there are instances where the ncep/ncar data are outside the range of the 20cr ensemble, even in the most recent years.

Presumably ncep/ncar is a better estimate in recent times as it includes radiosonde/satellite observations, but would the range of the 20cr be expected to include the other reanalysis? Or is that not expected based on the biases in the underlying model?

Any guidance would be very gratefully received.

Ideally you would expect the 20CR range to cover the other reanalyses in most instances if both are unbiased and NCEP/NCAR is error free. Temperatures in 20CR have problems over the Arctic, so there might be biases in the northernmost part of your domain (and they have a seasonal cycle), and generally that region is not very well represented prior to around 1940. But other than that, I am not aware of problems so if you could name the instances that might help further. Stefan Brönnimann

I'm new to this site and have just prepared a page in my own 'area'. After some fiddling around, I seemed to get something going, though will need some fine-tuning as I re-learn Wiki. It deals with an issue regarding some trends in the US Reanalysis data sets and I've obtained some EC data for comparison. Hence it will have some focus on comparison, so I suppose the heading of this thread is apt.

My question is: If I have done the right thing by doing it in my area, how do I get it linked into the main site for others to have a read?

Sorry for troubling you but I sincerely think is an important issue for the folks here to consider.

I warmly appreciate what the site is about and look forward to helping the cause.


Dave, Happy to help. Welcome! Please be very specific when you refer to something. What is "EC data"? This could refer to Environment Canada or sometimes to the European Centre for Medium Range Weather Forecasts. When you are authenticated, look at the Help area. Directions for how to link and how to include images should be there, e.g., Please put Help related comments under the Help area if at all possible. Then the Help pages can themselves be updated as your questions are answered. best wishes, gil compo

To answer the question about the link, it should probably be linked from here On there top bar, there is a help link. If you click it, it tells you how to a) create the page and b) link to it. If you have problems, comment and we will try to help/improve directions. Cathy

Ashwini Jain (not verified)

Tue, 11/05/2013 - 12:07

I am using wind velocity and I want to calculate it at a height of 80mtrs from the ground . So could you tell me that at what height from the ground these data have been taken????


Tue, 11/05/2013 - 12:42

In reply to by Ashwini Jain (not verified)

Typically, there is meteorology produced at 2m and 10m above the surface. MERRA produces winds at 50m, and also the lowest model level. the lowest model level does not have a fixed height above the surface, but the geopotential height of the lowest model level is also provided. pressure level wind profile data is also provided, but again, not at a fixed height, geopotential height is provided. 80m is an atypical height for most reanalysis output. In the case of MERRA, you would need the 2 lowest model levels to interpolate to that height. These are available in 3D data files, but, there would be a substantial amount of work to get the interpolated winds.

Another general answer to your question is to read the documentation associated with the reanalysis system that you are using to determine which height the data are at, in case it is not obvious (such as u2m and v2m).

I hope this helps,

Add new comment

This question is for testing whether or not you are a human visitor and to prevent automated spam submissions.