Break detection of annual Swiss temperature series

Kuglitsch, F. G., R. Auchmann, R. Bleisch, S. Brönnimann, O. Martius, and M. Stewart, 2012: Break detection of annual Swiss temperature series. J. Geophys. Res., 117, D13105, doi: 10.1029/2012JD017729.

Abstract. Instrumental temperature series are often affected by artificial breaks (“break points”) due to (e.g.,) changes in station location, land-use, or instrumentation. The Swiss climate observation network offers a high number and density of stations, many long and relatively complete daily to sub-daily temperature series, and well-documented station histories (i.e., metadata). However, for many climate observation networks outside of Switzerland, detailed station histories are missing, incomplete, or inaccessible. To correct these records, the use of reliable statistical break detection methods is necessary. Here, we apply three statistical break detection methods to high-quality Swiss temperature series and use the available metadata to assess the methods. Due to the complex terrain in Switzerland, we are able to assess these methods under specific local conditions such as the Foehn or crest situations. We find that the temperature series of all stations are affected by artificial breaks (average = 1 break point / 48 years) with discrepancies in the abilities of the methods to detect breaks. However, by combining the three statistical methods, almost all of the detected break points are confirmed by metadata. In most cases, these break points are ascribed to a combination of factors in the station history.

Synthesis

  1. By T. Otally Independent and Andy Editor

    This is clearly an important study, especially at the time, showing on real data that break detection works for the Swiss network. The high quality of the Swiss metadata make this paper extra valuable.

    Our apologies for assessment 3 by Mock Adriansen. It does not show much engagement with the study and the arguments are not pertinent. As such it will be ignored for this synthesis. If this had not been a mock example it would even be removed.

    Having read the assessments by Victor Venema and Adrian Mockton the arguments seem to be valid and we see no reason to deviate from the proposed grades.

    Impact on the larger scientific community. [60]
    Contribution to the scientific field of the journal. [75]
    The technical quality of the paper. [70]
    Importance at the time of publishing. [80]
    Importance of the research program. [-]

Reviews

  1. The article uses the high-quality Swiss climate network with very good metadata to validate three break detection methods and proposes a combined break detection method. The break detection methods stem from the homogenisation methods RHTestsV3, PRODIGE and Toreti. Due to the good metadata the paper can also report in detail on the causes of the detected breaks.

    The study is important for the validation of homogenisation methods. By using real data it is complementary to studies using stochastic data (such as the HOME benchmarking study). The advantage of using real data is that the properties of the inhomogeneities are by definition realistic. The disadvantage is that the presence of inhomogeneities is more uncertain. The study clearly shows that homogenisation methods successfully find real breaks. For the combined method 94% of the breaks found are supported by metadata. One cannot expect to find 100% because even in Switzerland metadata is not perfect and some real breaks are likely not found.

    The study also compares the three detection methods. I am not convinced that the study shows that the break test of PRODIGE overestimates the number of breaks. It could also be that PRODIGE is more sensitive and was able to find smaller breaks and thus more breaks. The main example in the paper, the homogenisation of the morning temperature reading of station Basel, suggests that several of the breaks found by PRODIGE are supported by metadata or by breaks in other parameters. If the study had found that all methods detect about the same number of breaks, but one of them has a higher false alarm rate, that would have shown differences in quality. In the Swiss case I am not sure we can decide without simulated data.

    For this comparison of the three methods it is also important to note the special characteristics of the Swiss network. Not only the metadata, but also the homogeneity of the Swiss data is exceptional with only one break per 48 years. Typical for Europe is one break every 20 years and for Northern America one break per 15 years. Many Swiss series will thus have only one or no breakpoint, which likely favours single-breakpoint methods. The mountainous landscape in Switzerland means that reference series will be more different from the candidate than in a comparable flat region. For other networks the results of a comparison could be different.

    More detailed comments can be found in the web annotations. (These annotations were made in a separate stream for the Homogenisation Journal; they should have been readable for all, but this is apparently not the case. Please follow this link to join the group.)

    Impact on the larger scientific community. [60]
    I presume the larger climatological community expected homogenisation methods to work.

    Contribution to the scientific field of the journal. [75]
    The study is an important independent way to show that homogenisation detection methods work. Its importance is limited by the limited importance of break detection alone. In the HOME benchmarking study only moderate correlations were found between break detection scores and the climatological quality of the homogenise data (Venema et al., 2012).

    The technical quality of the paper. [70]
    The paper is mostly of high quality, but some claims seem not to be sufficiently supported by the evidence.

    Importance at the time of publishing. [80]
    Recently we have several papers using real data to validate homogenisation methods. At the time, in 2012, this was the main paper doing so systematically and thus an important trail blazer.

    Importance of the research program. [-]
    Not relevant. Single paper.

    References
    Venema, V., O. Mestre, E. Aguilar, I. Auer, J.A. Guijarro, P. Domonkos, G. Vertacnik, T. Szentimrey, P. Stepanek, P. Zahradnicek, J. Viarre, G. Müller-Westermeier, M. Lakatos, C.N. Williams, M.J. Menne, R. Lindau, D. Rasol, E. Rustemeier, K. Kolokythas, T. Marinova, L. Andresen, F. Acquaotta, S. Fratianni, S. Cheval, M. Klancar, M. Brunetti, Ch. Gruber, M. Prohom Duran, T. Likso, P. Esteban, Th. Brandsma. Benchmarking homogenization algorithms for monthly data. Climate of the Past, 8, pp. 89-115, doi: 10.5194/cp-8-89-2012, 2012.

  2. Example review by Adrian Mockton

    Having read the paper carefully, as well as the assessment by the ever brilliant Victor Venema, I concur completely and have nothing to add.

    Impact on the larger scientific community. [60]

    Contribution to the scientific field of the journal. [75]

    The technical quality of the paper. [70]

    Importance at the time of publishing. [80]

    Importance of the research program. [-]
    Not relevant. Single paper.

  3. Example review by Mock Adriansen

    Nothing good ever comes out of Switzerland.

    Impact on the larger scientific community. [10]

    Contribution to the scientific field of the journal. [10]

    The technical quality of the paper. [10]

    Importance at the time of publishing. [10]

    Importance of the research program. [10]

Leave a Reply

Everyone is welcome to make comments on this paper below. The comments are pre-moderated (will only appear after approval by the editors) to ensure they are on topic.

Your email address will not be published. Required fields are marked *