The article on benchmarking homogenization algorithms for monthly temperature and precipitation data has been published by Climate of the Past. For more information, see also:
- Our press release
- My blog post, an informal introduction to the article
- The open access article itself from Climate of the Past
Inhomogeneities in climate records
Climate records (e.g. temperature and precipitation measurements) are affected by changes in measurement conditions, e.g., modernisation of the instrumentation, relocation of the weather stations, changes in observation rules, automation, etc. These inhomogeneities are of the same order of magnitude as anthropogenic climate change or slow climatic cycles. The aim of a homogenisation procedure is to detect and correct these changes. Inhomogeneities of this kind can be detected as jumps in the difference between observations from two nearby stations. The measurements of neighbouring stations are usually strongly correlated and jumps in the difference between these measurements indicate a change in the conditions of one station, whereas climatic changes are expected to affect both stations. By analysing a larger network of stations, these jumps can in general be attributed to a specific station.
COST Action on homogenisation
The COST Action ES0601: Advances in homogenisation methods of climate series: an integrated approach (HOME) aims at improving homogenisation algorithms. The main activity of HOME is a blind validation of monthly homogenization algorithms based on a benchmark dataset. Additionally, we are working on validating the correction methods of daily homogenization methods.
Within the Daily Stew project, we are working on improving the detection of inhomogeneities by using daily data; currently detection is still typically performed on monthly or annual means. Without performing any corrections, the project aims to use the data of the homogeneous subperiods to estimate trends in extreme weather directly.
Long range dependence
The main aim of homogenization is to be able to study trends and decadal variability with more accuracy. However, the community working on long range dependence (LRD) almost never reported whether they used homogenized data or not. In Rust et al. (2008) we showed that homogenization has a strong influence on the power law exponent used to quantify LRD. Without homogenization the exponent is artificially high. People should thus always use homogenized data to study LRD.
Homogenisation mailing list
We have started an email distribution list on homogenisation on climate data. The list is most useful if all people interested in homogenization are one it. Thus please consider subscribing.
ReferencesFewer jumps, less memory: homogenized temperature records and long memory
JGR-Atmospheres, 113, D19110, doi:10.1029/2008JD009919, 2008.
Henning Rust, Olivier Mestre, and Victor Venema