A framework for benchmarking of homogenisation algorithm performance on the global scale

Willett, K. M.; Williams, Claude N.; Jolliffe, Ian T.; Lund, Robert; Alexander, Lisa V.; Brönnimann, Stefan; Vincent, Lucie A.; Easterbrook, Steve; Venema, Victor; Berry, David I.; Warren, Rachel; Lopardo, G.; Auchmann, Renate; Aguilar, Enric; Menne, Matthew J.; Gallagher, Colin; Hausfather, Zeke; Thorarinsdottir, Thordis L.; Thorne, Peter

doi:10.5194/gi-3-187-2014

Cited by 35 publications

(27 citation statements)

References 42 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The International Surface Temperature Initiative aims at computing the uncertainties remaining in homogenized data and will perform a benchmarking mimicking the global observational network, which therefore includes a realistic range of SNRs (Willett et al, 2014;Thorne et al, 2011).…”

Section: Discussionmentioning

confidence: 99%

The joint influence of break and noise variance on the break detection capability in time series homogenization

Lindau

Venema

2018

Adv. Stat. Clim. Meteorol. Oceanogr.

Self Cite

View full text Add to dashboard Cite

Abstract. Instrumental climate records of the last centuries suffer from multiple breaks due to relocations and changes in measurement techniques. These breaks are detected by relative homogenization algorithms using the difference time series between a candidate and a reference. Modern multiple changepoint methods use a decomposition approach where the segmentation explaining most variance defines the breakpoints, while a stop criterion restricts the number of breaks. In this study a pairwise multiple breakpoint algorithm consisting of these two components is tested with simulated data for a range of signal-to-noise ratios (SNRs) found in monthly temperature station datasets. The results for low SNRs obtained by this algorithm do not differ much from random segmentations; simply increasing the stop criterion to reduce the number of breaks is shown to not be helpful. This can be understood by considering that, in case of multiple breakpoints, even a random segmentation can explain about half of the break variance. We derive analytical equations for the explained noise and break variance for random and optimal segmentations. From these we conclude that reliable break detection at low but realistic SNRs needs a new approach. The problem is relevant because the uncertainty of the trends of individual stations is shown to be climatologically significant also for these small SNRs. An important side result is a new method to determine the break variance and the number of breaks in a difference time series by studying the explained variance for random break positions. We further discuss the changes from monthly to annual scale which increase the SNR by more than a factor of 3.

show abstract

Section: Discussionmentioning

confidence: 99%

The joint influence of break and noise variance on the break detection capability in time series homogenization

Lindau

Venema

2018

Adv. Stat. Clim. Meteorol. Oceanogr.

Self Cite

View full text Add to dashboard Cite

show abstract

“…To assess its performance, we applied our breakpoint detection algorithm to a benchmark of synthetic daily temperature series developed within the International Surface Temperature Initiative (ISTI) framework (Willett et al ., ; Killick, ). The synthetic series are contaminated by inhomogeneities having the same statistical properties of those found in real temperature series, with the difference that the position of each breakpoint is known a priori.…”

Section: Data Production Methodsmentioning

confidence: 99%

The EUSTACE global land station daily air temperature dataset

Brugnara

Good

Squintu

et al. 2019

Geoscience Data Journal

View full text Add to dashboard Cite

We describe a global dataset of quality‐controlled in situ daily air temperature observations covering the period 1850–2015, developed in the framework of the EUSTACE (EU Surface Temperature for All Corners of Earth) project (http://www.eustaceproject.org). The dataset includes a total of 35,364 daily series of maximum and minimum temperature obtained from seven different collections. About 97% of the series are publicly available in a common format, while the remaining 3% can be obtained from the original data providers. Unlike other similar products, duplicates have been removed without blending of series, which simplifies data traceability and improves the temporal homogeneity of the individual series at the cost of a smaller average length. Residual artificial signals (breakpoints) in the series caused by station relocations, changes in instrumentation, etc., have been detected by means of the combination of four breakpoint detection tests, four variables and three temporal aggregations. The combined results give not only the most probable position of the breakpoints, but also a measure of their likelihood. The reliability of the detection was estimated for each year of each target series, based on the number of reference series and on their correlation with the target series. Moreover, its general performance was evaluated through a benchmark of synthetic series. This product will be combined with datasets of marine and ice in situ air temperature observations and with measurements from satellite to produce the first complete global statistical reconstruction of daily near‐surface air temperature.

show abstract

“…could be used to assess the false alarm rate of a changepoint detection procedure (Easterling and Peterson, ). Indeed, such a “benchmarking” exercise is currently being pursued by the International Surface Temperature Initiative (Willett et al ., ). Third, a simulated set of clean (changepoint free) stations may be useful for validating climate model downscaling methods. Finally, the statistical models fitted here could be extrapolated to future climate scenarios—say, used to generate a record for the future that has the same station fluctuation characteristics, but is one degree hotter than today's climate on average.…”

Section: Introductionmentioning

confidence: 97%

“…could be used to assess the false alarm rate of a changepoint detection procedure (Easterling and Peterson, 1995). Indeed, such a "benchmarking" exercise is currently being pursued by the International Surface Temperature Initiative (Willett et al, 2014). • Third, a simulated set of clean (changepoint free) stations may be useful for validating climate model downscaling methods.…”

Section: Introductionmentioning

confidence: 99%

Simulation of temperature series and small networks from data

Washington

Seymour

Lund

et al. 2019

Intl Journal of Climatology

View full text Add to dashboard Cite

It is often desirable to simulate a single temperature series or a collection (network) of temperature series. Accurate simulations can enhance our understanding of temperature trends and variabilities. Simulation can also be used to generate data with known specifications, which are useful in assessing climate data processing routines such as quality control and homogenization algorithms. Possessing multiple realistic temperature series is often beneficial as only one natural record of our climate is available. The current popularity of general circulation models (GCMs) demonstrates how important simulation techniques are. However, even with sophisticated downscaling, it is difficult to replicate station temperature data that have realistic seasonal cycles as well as realistic temporal and spatial correlations. This paper reviews and studies statistical time series methods that can replicate a single series or a network of series from data. The methods are purely statistical—no atmospheric dynamics are involved—and attempt to produce replicates of the records under study. The work here develops (a) methods that simulate temperatures from a fixed season (say a particular day or month of the year) that match the distributional characteristics of the observed data; (b) methods for simulating an entire series that matches the series' temporal autocorrelations and seasonal cycle; and (c) methods for simulating a network of series that reproduce the data's observed seasonal cycles and spatial and temporal autocorrelations. Applications are given throughout, including one where a GCM series and local station data are used in tandem to describe long‐term trends and inject realistic station‐level short‐term fluctuations. This paper can be used as a tutorial for the simulation of a single climate observation, an entire climate series, or a network of multiple climate series simultaneously. Extensions of the ideas that involve GCMs are also examined.

show abstract

A framework for benchmarking of homogenisation algorithm performance on the global scale

Cited by 35 publications

References 42 publications

The joint influence of break and noise variance on the break detection capability in time series homogenization

The joint influence of break and noise variance on the break detection capability in time series homogenization

The EUSTACE global land station daily air temperature dataset

Simulation of temperature series and small networks from data

Contact Info

Product

Resources

About