Automatic Homogenization of Time Series: How to Use Metadata?

Round 1

Reviewer 1 Report

The topic of the article deals with a very specific and narrow topic, even within the framework of scientific studies dealing with homogenization. Metadata is very important for a correct assessment of climate development and finding out whether changes in the time series are due to natural variability or just a change of instrument or relocation of the station. In most cases, high-quality metadata is missing, and with the help of perhaps data rescue, there is an effort to supplement the missing data.Therefore, most homogenization software must work with little or no metadata. This article evaluates well what the homogenization results are in the case of an automatic system, for example the ACMANT software, metadata can improve the results and also under what conditions.The author of the article is certainly an expert in homogenization and his results can be rated as high quality.

Comment: 

row 118: time series cover the same period and no gaps = it is not "real world" - the results may be different from the real data that is normally worked with.

row 205: To describe that the ANOVA method gives the best results is misleading. It always depends on specific data, even within the framework of comparing the quality of correction methods.

Figure - the graphic editing is bad, the description of the composite images is missing - add there (a), (b), etc.

row 344: use the passive voice. The author writes in the first person plural and is the only author.

row 389: Metadata is very important even with a large number of time series, for the reason that, for example, automation could have occurred here at the same time (+/- 1-5 years). How do you solve this problem, since the change will be the same at all stations and the detection methods will not be able to detect these breaks?

Figure 5 - use full description of the figure. It is not possible to refer to the previous image.

row 443: The author also tested options where the metadata is complete. This is usually not realistic at all, at most 50% of breaks are described by metadata.

row: 461-461: Metadata is of primary importance in confirming that the break detected by the methods is indeed there and can be adjust without the risk of artificially breaking the series.

 

 

 

 

Author Response

See attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

The author has provided a good summary of the state-of-the-art on automatic homogenization of climate time series. However, the principal contribution is likely to be in presenting a new method of how incorporating metadata into the automatic homogenization process (by using ACMANT) based on the ANOVA correction model. The document ends with good advice on the use of metadata in this matter.

 

Author Response

Thank you for the review and the acknowledgement of the paper.