In this paper homogenisation methods are blindly tested on a dataset with unprecedented realism. All the most used and best algorithms have participated.
The two main conclusions are.
1. Homogenisation improves climate data. Because the test was blind and because of the realism of the data, this can now be stated with confidence.
2. Modern algorithms, which are designed to also work with an inhomogeneous reference, are clearly better than traditional ones. It needed a realistic benchmark dataset with surrogate climate networks to see this difference clearly.
A longer introduction to this paper can be found on the blog of Victor Venema:
A direct link to the (open access) article is:
The benchmark dataset is kept available for people who would like to test their algorithms (in a non-blind fashion):