CoChat
Sign Up
Cited by 349Climate variability and modelsUrban Heat Island MitigationPrecipitation Measurement and Analysis

Homogenization of Temperature Series via Pairwise Comparisons

Open access
Menne, Matthew J., Williams, Claude N.·2009·Journal of Climate
DOI 10.1175/2008jcli2263.1
Open this paper in CoChat — free accountOpen PDF

Abstract

Abstract An automated homogenization algorithm based on the pairwise comparison of monthly temperature series is described. The algorithm works by forming pairwise difference series between serial monthly temperature values from a network of observing stations. Each difference series is then evaluated for undocumented shifts, and the station series responsible for such breaks is identified automatically. The algorithm also makes use of station history information, when available, to improve the identification of artificial shifts in temperature data. In addition, an evaluation is carried out to distinguish trend inhomogeneities from abrupt shifts. When the magnitude of an apparent shift attributed to a particular station can be reliably estimated, an adjustment is made for the target series. The pairwise algorithm is shown to be robust and efficient at detecting undocumented step changes under a variety of simulated scenarios with step- and trend-type inhomogeneities. Moreover, the approach is shown to yield a lower false-alarm rate for undocumented changepoint detection relative to the more common use of a reference series. Results from the algorithm are used to assess evidence for trend inhomogeneities in U.S. monthly temperature data.

Integrity signals

Registry

Crossref · 10.1175/2008jcli2263.1

Cites

44 of 44 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Claims the paper makes

0 of 42 cited sources checked · 0 supported · 0 flagged

  • Findingown data

    The pairwise algorithm is robust and efficient at detecting undocumented step changes under a variety of simulated scenarios with step- and trend-type inhomogeneities.

    “The pairwise algorithm is shown to be robust and efficient at detecting undocumented step changes under a variety of simulated scenarios with step-and trend-type inhomogeneities.”
    Quote in the paperSubstantial claim
    • The authors state this as an assumption: A statistically significant and persistent violation of relative homogeneity is artificial or has origins other than background variations in weather and climate.
  • Findingown data

    The pairwise approach yields a lower false-alarm rate for undocumented changepoint detection relative to the more common use of a reference series.

    “Moreover, the approach is shown to yield a lower false-alarm rate for undocumented changepoint detection relative to the more common use of a reference series.”
    Quote in the paperSubstantial claim
    • The authors state this as an assumption: A statistically significant and persistent violation of relative homogeneity is artificial or has origins other than background variations in weather and climate.
  • Findingown data

    The adjusted series all have trends much closer to the true climate trend of zero.

    “In general, the adjusted series all have trends much closer to the true ''climate'' trend of zero.”
    Quote in the paperSubstantial claim
  • Findingown data

    The pairwise approach limits the number of false alarms whenever neighboring series are impacted by undocumented changepoints, as evidenced by low FAR for cases 4 and 6 relative to the reference series approach.

    “the pairwise approach limits the number of false alarms whenever the neighboring series are impacted by undocumented changepoints as evidenced by the low FAR for cases 4 and 6 relative to the reference series approach.”
    Quote in the paperSubstantial claim
  • Findingown data

    The pairwise algorithm has a much higher success rate in identifying homogeneous target series than the reference series approach, as indicated by higher CRC percentages for cases 1, 3, and 5.

    “In general, the pairwise algorithm has a much higher success rate in identifying homogeneous target series than the reference series approach as indicated by the higher CRC percentages for cases 1, 3, and 5.”
    Quote in the paperSubstantial claim
  • May combine two claims. 29
  • Classified as probably not a substantial claim. 5

  • May combine two claims.

    Findingown data

    Under M5, the CRCs are comparable to the M4 simulations, but the pairwise algorithm tended to misclassify the D t f g series as M3 or M4 in roughly equal proportions.

    “In this scenario, the CRCs are comparable to the M4 simulations, but in this case, the pairwise algorithm tended to misclassify the D t f g series as M3 or M4 in roughly equal proportions.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    The RMSE for changepoint estimates in series with trends is about as good as in the series with no trend inhomogeneities, suggesting that the model identification is reasonably successful at identifying step changes that occur within local trends.

    “Table 2 also indicates that the RMSE for changepoint estimates in series with trends is about as good as in the series with no trend inhomogeneities, which suggests that the model identification is reasonably successful at identifying step changes that occur within local trends.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingreasoning

    Under M2 scenarios, the pairwise algorithm correctly classified more than 85% of the D t f g series when b was greater than or equal to 0.01, but less than 50% when b = 0.005.

    “Under the M2 scenarios, the pairwise algorithm correctly classified more than 85% of the D t f g series when b was greater than or equal to 0.01 (a slope yielding a change of 1s in 100 time steps), but less than 50% when b 5 0.005 (a change of 0.5s in 100 time steps).”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    The BIC correctly reclassifies most of these breaks as M2, but also cannot always distinguish a trend (M2) from a step change (M3, M4, or M5).

    “The BIC correctly reclassifies most of these breaks as M2, but also cannot always distinguish a trend (M2) from a step change (M3, M4, or M5).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    In the pairwise algorithm, most (>88%) of the D t f g series with a step change of 1s or greater were correctly identified as M3, and the CRC exceeds 90% in such cases.

    “In the pairwise algorithm, most (.88%) of the D t f g series with a step change of 1s or greater were correctly identified as M3, and the CRC exceeds 90% in such cases.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    Changes in excess of 0.5σ are readily detected, and most misses are generally less than 0.5σ.

    “In both cases, changes in excess of 0.5s are readily detected, and most misses are generally less than 0.5s.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingnone given

    More than 99.5% of the monthly temperature series from the 100 nearest neighbors in the U.S. Cooperative Observer Network are correlated at r >= 1/2, and r >= 0.8 in 90% of cases.

    “This restriction effectively occurs in practice for monthly temperature values from the U.S. Cooperative Observer Network where more than 99.5% of the monthly temperature series from the 100 nearest neighbors are correlated at this level (and r $ 0.8 in 90% of cases).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingreasoning

    The higher hit rate in the monthly simulations is 67% (Table 2) compared to a little less than 50% shown in Fig. 8 when 10 neighbors are available.

    “This accounts for the higher hit rate in the ''monthly'' simulations, that is, 67% (Table 2) compared to the rate of a little less than 50% shown in Fig. 8 when 10 neighbors are available.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    The largest undetected changepoint was preceded 10 time steps earlier by another undetected changepoint, and the two changepoints essentially masked one another.

    “Furthermore, the largest undetected changepoint (10.696s) was preceded 10 time steps earlier by another undetected changepoint of 20.451s; that is, the two changepoints essentially masked one another.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingreasoning

    The pairwise algorithm classifies only 91% of M2 target series as homogeneous (no step) when b = 0.01 and 86.9% when b = 0.02.

    “Consequently, the pairwise algorithm classifies only 91% of M2 target series as homogeneous (no step) when b 5 0.01 and 86.9% when b 5 0.02.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingreasoning

    The pairwise algorithm's higher success rate in identifying homogeneous target series is true when neighbor series are themselves homogeneous (cases 1 and 5), and especially when all neighbors have changepoints (case 3).

    “This is true when the neighbor series are themselves homogeneous as in cases 1 and 5, but especially when all the neighbors have changepoints as in case 3, which cause numerous inhomogeneities in the reference series.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    Using M3 for all step-change adjustments removes the impact of most trend inhomogeneities because the error for the adjusted trends is roughly the same for the step-only and steps and trends simulations.

    “From a comparison of the RMSE for the adjusted trends in Tables 2 and 6, it is evident that using M3 for all step-change adjustments removes the impact of most trend inhomogeneities because the error for the adjusted trends is roughly the same for the step-only and steps and trends simulations.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    In HCN temperature series, about 50% of the step changes follow M3 (step change with no trend), approximately 40% follow M5 (step change accompanied by a trend change), and about 10% follow M4 (step change within a general trend).

    “An analysis of the more than 100 000 D t f g series segments used to calculate the shift magnitudes for HCN temperature series indicates that about 50% of the step changes follow M3 (step change with no trend), while approximately 40% follow M5 (step change accompanied by a trend change) and about 10% follow M4 (step change within a general trend).”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    With M3-only adjustments, the trend at Cheesman is clearly in sync with the average of trends in surrounding series, although the sawtooth signature remains in the adjusted data.

    “Although the sawtooth signature remains in the adjusted data, the trend at Cheesman using the M3 adjustments is clearly in sync with the average of trends in surrounding series.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    M5 rarely minimizes the BIC when there are no relative trends in the simulated data, but M4 is identified as the 'best' model in over 16% of cases when the shifts are treated as undocumented.

    “Based on these results, it appears that M5 rarely minimizes the BIC when there are no relative trends in the simulated data, but M4 is identified as the ''best'' model in over 16% of cases when the shifts are treated as undocumented.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    The percentage of correctly identified M4 difference series is about 70% at b = 0.02 when n = 240 and c = 120 versus 50% for n = 100 and c = 50.

    “For example, the percentage of correctly identified M4 difference series is about 70% at b 5 0.02 when n 5 240 and c 5 120 versus 50% for n 5 100 and c 5 50.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    When there is perfect knowledge of all shifts in the monthly steps and trends simulations, the frequency of M3 solutions is approximately correct, although M4 is selected too often at the expense of M5.

    “However, when there is perfect knowledge of all shifts in the monthly steps and trends simulations, the models minimize the BIC in way that suggests that the frequency of M3 solutions is approximately correct (although M4 is selected too often at the expense of M5).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    Relative to the more traditional use of a climate reference series, a pairwise approach to undocumented changepoint detection reduces the number of false alarms in general and is particularly successful at identifying homogeneous segments.

    “Relative to the more traditional use of a climate reference series, a pairwise approach to undocumented changepoint detection reduces the number of false alarms in general and is particularly successful at identifying homogeneous segments.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Premisecitation

    The procedure explicitly looks for both abrupt jumps and local, unrepresentative trends in temperature series.

    “Last, the procedure explicitly looks for both abrupt ''jumps'' as well as local, unrepresentative trends in the temperature series (DeGaetano 2006).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites DeGaetano 2006 source

      Not checked in the free run.

  • May combine two claims.

    Interpretationreasoning

    Conducting a separate assessment of reference series homogeneity fails to exploit the enhanced sensitivity of relative homogeneity testing, and many small-amplitude changepoints may go undetected in the reference series only to be later attributed to the target series.

    “Unfortunately, conducting a separate assessment of reference series homogeneity fails to exploit the enhanced sensitivity of relative homogeneity testing, and many smallamplitude changepoints may go undetected in the reference series only to be later attributed to the target series.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationnone given

    If the trends in two correlated temperature series (bX and bY) are not equal, then a local, unrepresentative trend (creeping inhomogeneity) is present in one or both series.

    “This assumption is evaluated further in subsequent steps because if b X 6 ¼ b Y , then a local, unrepresentative trend (i.e., creeping inhomogeneity) is present in X t f g and/or Y t f g.”
    Quote in the paperNot a substantial claim
    • The authors state this as an assumption: The trends in two correlated temperature series (bX and bY) are approximately equal in general.

  • May combine two claims.

    Interpretationreasoning

    The increase in false alarms when trend inhomogeneities are present occurs because the beginning or end of a trend inhomogeneity is often identified as a step change and short interval trends of about 24 months or less tend to be virtually indistinguishable from step changes and are adjusted as an abrupt change.

    “The increase in false alarms when trend inhomogeneities are present occurs for two main reasons.First, the beginning or end of a trend inhomogeneity is often identified as a step change by the pairwise algorithm.Second, short interval trends of about 24 months or less tend to be virtually indistinguishable from step changes and are therefore adjusted as an abrupt change.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Premisecitation

    The recursive testing approach is associated with a low probability of false changepoint detection without requiring an analyst to interpret the results.

    “First, the approach is associated with a low probability of false changepoint detection without requiring an analyst to interpret the results (cf.Caussinus and Mestre 2004).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Caussinus and Mestre 2004

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    The use of a reference series is the most widely employed approach to relative changepoint detection.

    “The use of a reference series is the most widely employed approach to relative changepoint detection, and MW05 evaluated the implications of such an approach for undocumented changepoint detection.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites MW05

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    Use of such tests in the presence of trends can lead to falsely detected step changes as well as inaccurate estimates of the magnitude of a shift when it occurs within a general trend.

    “Use of such tests in the presence of trends can lead to falsely detected step changes as well as to inaccurate estimates of the magnitude of a shift when it occurs within a general trend (DeGaetano 2006;Pielke et al. 2007).”
    Quote in the paperSubstantial claim0 of 2 sources checked
    • cites DeGaetano 2006 source

      Not checked in the free run.

    • cites Pielke et al. 2007 source

      Not checked in the free run.

  • May combine two claims.

    Interpretationreasoning

    Misclassification of M2 as flat does not impact the CRC because there is no step change assigned to the target.

    “This kind of misclassification, however, does not impact the CRC because there is no step change assigned to the target.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Interpretationreasoning

    The power of pairwise detection can be further improved by increasing the sample size between changepoints, which can be achieved by testing serial monthly values rather than annual or seasonal averages.

    “Moreover, the power of pairwise detection can be further improved by increasing the sample size between changepoints, which can be achieved by testing serial monthly values rather than annual or seasonal averages.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Interpretationreasoning

    To maximize the power of changepoint detection, station histories should be exploited whenever possible.

    “It follows that to maximize the power of changepoint detection, station histories should be exploited whenever possible.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Interpretationown data

    In the U.S. surface temperature record, there is evidence of relative trends between series, but some apparent trends may be an artifact of unidentified, small shifts.

    “Results from applying the pairwise algorithm to observed temperature series suggest that while there is evidence of relative trends between series in the U.S. surface temperature record, some apparent trends may be an artifact of unidentified, small shifts.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Interpretationreasoning

    The skill of changepoint detection and identification increases with increasing correlation between series, which reduces the variance of the D t f g series.

    “In addition, the skill of changepoint detection and identification increases with increasing correlation between series, which reduces the variance of the D t f g series.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationreasoning

    The temperature series that result using M3-only adjustments arguably approximate the best theoretical climate series for each location, but the local trend signal is nevertheless aliased out of the original series, thus limiting the use of the adjusted series in some attribution studies of observed temperature change.

    “Still, while the temperature series that result using the M3-only adjustments arguably approximate the best theoretical climate series for each location, the local trend signal is nevertheless aliased out of the original series, thus limiting the use of the adjusted series in some attribution studies of observed temperature change.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisenone given

    The pairwise method is similar to the graph theory approach used by Christy et al. (2006) except that the pairwise algorithm makes no attempt to compare climate series that do not overlap in time.

    “In this regard, the pairwise method is similar to the graph theory approach used by Christy et al. (2006) except that the pairwise algorithm makes no attempt to compare climate series that do not overlap in time.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Christy et al. (2006) source

      Not checked in the free run.

  • May combine two claims.

    Interpretationreasoning

    The step-only model necessarily leads to the aliasing of any associated trend inhomogeneity onto the estimate of the step change, but a reliable estimate of the background climate signal is obtained.

    “While this necessarily leads to the aliasing of any associated trend inhomogeneity onto the estimate of the step change, a reliable estimate of the background climate signal is obtained.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationnone given

    If a change is implemented on exactly the same date at all stations, relative homogeneity testing will not be effective.

    “Of course, if a change is implemented on exactly the same date at all stations, relative homogeneity testing will not be effective.”
    Quote in the paperNot a substantial claim
34 of 34

  • The authors state this as an assumption.

    Similar variations in climate occur at nearby locations because of the spatial correlation inherent to meteorological fields.

    “The assumption is that similar variations in climate occur at nearby locations because of the spatial correlation inherent to meteorological fields (Livezey and Chen 1983).”

  • The authors state this as an assumption.

    A statistically significant and persistent violation of relative homogeneity is artificial or has origins other than background variations in weather and climate.

    “A statistically significant and persistent violation of relative homogeneity is presumed to be artificial or, at least, to have origins other than the background variations in weather and climate.”

  • The authors state this as an assumption.

    The trends in two correlated temperature series (bX and bY) are approximately equal in general.

    “In reality, it is unrealistic to assume that b X and b Y are stationary in t given the nature of multidecadal variations in climate series; however, it may be that b X ' b Y in general.”

  • The authors state this as an assumption.

    The periodicity in the differenced series equation (4) is negligible, especially because the series are first deseasonalized.

    “At present, the periodicity in ( 4) is considered to be negligible, especially since X t f g and Y t f g are first deseasonalized.”

  • The paper relies on this without support.

    Changes to the circumstances behind a series of climate observations are practically inevitable at some point during the period of record.

    “Unfortunately, changes to the circumstances behind a series of climate observations are practically inevitable at some point during the period of record.”

  • The paper relies on this without support.

    Discontinuities in a climate series can be induced by virtually any change in instrumentation or observation practice.

    “Discontinuities in a climate series can be induced by virtually any change in instrumentation or observation practice.”

  • The paper relies on this without support.

    The distribution of step-change estimates is not necessarily symmetric.

    “because the distribution of step-change estimates is not necessarily symmetric, the median estimate is used to adjust the target series.”

  • The paper relies on this without support.

    The magnitude of a step change will not be accurately estimated if the type of changepoint has been misidentified.

    “The magnitude of a step change will not be accurately estimated if the type of changepoint has been misidentified.”

  • The paper relies on this without support.

    No general assessment of the frequency of the various types of changepoints in observed temperature series has been conducted.

    “A number of recent studies have focused on the impact of land use change on the temperature record (e.g., Peterson and Owen 2005;Kalnay et al. 2006;Parker 2006;Pielke et al. 2007), yet no general assessment of the frequency of the various types of changepoints in observed temperature series has been conducted.”
9 of 9

Cites

Flagged first, then by citations

Cited by

Sorted by citations

Related

Authors
Menne, M. J., & Williams, C. N.
Affiliations
National Oceanic and Atmospheric Administration (US)
Corresponding author
Menne, Matthew J.
ORCID
Menne, Matthew J.
Venue
Journal of Climate
Publisher
American Meteorological Society
Published
2009-04-01
Volume
22
Issue
7
Pages
1700-1717
ISSN
  • 0894-8755 (ISSN-L)
  • 1520-0442
DOI
10.1175/2008jcli2263.1
OpenAlex
W2153133730
Cited by
349
References
44
Citations by year
2026: 18 · 2025: 15 · 2024: 11 · 2023: 17 · 2022: 33 · 2021: 20 · 2020: 19 · 2019: 19 · 2018: 24 · 2017: 14 · 2016: 20 · 2015: 25 · 2014: 21 · 2013: 27 · 2012: 36
FWCI
8.96
Type
journal article
Language
en
Open access
Locations
Journal of Climate (publishedVersion)
Funders
  • U.S. Department of Energy
  • Biological and Environmental Research
Topics
  • Physical Sciences › Environmental Science › Global and Planetary Change › Climate variability and models
  • Physical Sciences › Environmental Science › Environmental Engineering › Urban Heat Island Mitigation
  • Physical Sciences › Earth and Planetary Sciences › Atmospheric Science › Precipitation Measurement and Analysis
Keywords
pairwise comparison, monthly temperature data, change point detection, station history, false alarm rate

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-11.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-11.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review