Pith. sign in

REVIEW 4 major objections 4 minor 41 references

Change point detection in ERA5 ground temperature time series

T0 review · 4 major / 4 minor · reviewed 2026-08-06 · deepseek-v4-flash

Pith's one-line read A two-slope model with one change point describes local land temperature evolution better than a single trend on about half of Earth's land, with the change clustering near 1980.

desk verdict Honest and useful map of regional trend changes, but the headline claims need multiple-testing control and a direct check of the reanalysis discontinuity. read the letter →

arxiv 2507.15045 v1 pith:CAYKT2BR submitted 2025-07-20 stat.AP physics.data-an

classification stat.APphysics.data-an MSC 62P12
keywords changepointdetectionERA5warmingtrendtwo-slopefitmodelselectionBIClocalclimatetippingpoints
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper asks whether local land temperature records from 1950 to 2021 are better described by one straight trend or by two straight trends that meet at a change year. Using ERA5 2-meter annual temperatures on a 1-degree grid, it fits a continuous two-slope model at every land point and compares it with a single linear fit using Akaike's and Bayesian information criteria plus a significance test. It finds that for roughly half of global land area, the two-slope model wins, and detected change points cluster around 1980 plus or minus 10 years, with many locations showing stronger warming after the change and some showing slowdown. If correct, this means single-trend summaries of local warming hide a widespread acceleration, and it raises the question of whether some local climates have already passed a tipping point.

What carries the argument

The central object is the continuous two-slope model: two linear segments constrained to meet at a single change point, so the model captures a change in warming trend rather than a jump in temperature. At each candidate change year the slopes and intercepts are obtained analytically by a constrained least-squares fit, and the selected change year is the one minimizing the root-mean-square error. Model selection between the dual-linear and single-linear descriptions is carried out with the Bayesian information criterion, with Akaike's criterion checked as an alternative, and a null-hypothesis test based on the absolute difference of the two slopes provides a 95% confidence decision.

What would settle it

Re-run the same grid-cell fit on ERA5 data restricted to 1979 onward; if the cluster of change points around 1980 disappears or shifts, the reported timing is an artifact of combining pre-satellite and satellite-era data.

Watch

Extended reading notes

Core claim

The paper argues that a continuous dual-linear fit, two straight lines that merge at a single change year, is the parsimonious description of local 2-meter temperature evolution on ERA5 land grid points from 1950 to 2021. Comparing this model with a single-trend fit by BIC and a 95% confidence hypothesis test, it finds that for roughly half of the global land area, area-weighted, the two-slope model is preferred. For the majority of those grid points the change year falls between about 1970 and 1990, clustering near 1980, and the fitted second slopes are mostly positive and often larger than the first, indicating accelerated warming in many regions while some regions slow down or cool. The same two-slope fit applied to global mean land temperature finds a change around 1976 to 1980, from slight cooling to strong warming.

Load-bearing premise

The load-bearing premise is that each 72-year local temperature series is adequately described by exactly two straight-line segments that meet at one change year.

Editorial extensions

If this is right

  • About half of the global land area is better described by a two-slope fit than by a single linear trend, so one-trend descriptions of local warming are incomplete for those regions.
  • Detected change years lie mostly between 1970 and 1990, with a concentration near 1980, making those years candidate past tipping events in the paper's interpretation.
  • In many regions, especially northern land masses, the second slope is larger than the first, meaning local warming accelerated; in some regions the second slope is smaller or negative, meaning warming slowed or reversed.
  • The global mean land temperature series itself shows a change around 1976 to 1980, from a slightly negative to a strongly positive trend, consistent with the local picture.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The paper's own Siberia example implies that a three-segment model may be the better description there; extending the BIC comparison to three or more continuous segments would show how many apparent 1980 change points are actually two separate changes, such as one in the 1970s and one in the 1990s.
  • If the acceleration around 1980 is real, projections that anchor a single historical trend will tend to underestimate near-future local warming in the affected regions; that extrapolation is not made in the paper.
  • A checkable prediction following from the paper is that grid points with a larger second slope should mostly continue on that steeper slope in years after 2021 unless another change point appears.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 4 minor

Summary. The paper applies a continuous two-slope (piecewise-linear with a single change point) least-squares model to annual ERA5 2-meter temperature time series from 1950 to 2021 on roughly 20,000 land grid points. The two linear segments are constrained to merge at the change point, and the change year is selected by minimizing the RMS error over a grid of candidate years (1960-2010). The model is compared against a single linear trend using BIC, AIC, and a null-hypothesis test based on the absolute slope difference. The authors report that about half of the global land area shows a statistically significant trend change, produce global maps of the detected change years and segment slopes, and find that many change points cluster around 1980±10, which they connect to the question of whether climate tipping points have been passed.

Significance. If the central empirical claim holds, the paper provides a useful global inventory of local warming-trend accelerations and decelerations, with maps that could inform regional climate adaptation. The method has some genuine strengths: the constrained least-squares solution is derived in closed form in Appendix A, and the synthetic experiments in Section III give an honest assessment of the detection uncertainty (for realistic noise, 80% of change points are within ±8 years). The comparison of BIC and null-hypothesis testing on synthetic data is also a good practice. However, the headline result (the 1980±10 concentration of change points and the '50% of land area' claim) is currently not robustly established, because the analysis starts in 1950 despite the well-known 1979 satellite-assimilation discontinuity in ERA5, because no multiple-testing correction is applied to the ~20,000 per-grid tests, because the change-year and slope maps carry no uncertainty information, and because the two-segment model is acknowledged to be misspecified in at least some regions.

major comments (4)
  1. [Section VI and Section IV] The decision to analyze the full 1950-2021 period is justified in Section VI by the statement that 'the global mean surface temperature shows a change point in the 7th decade,' but this justification is circular if that global-mean change point is itself produced by the well-known 1979 satellite-data discontinuity in ERA5. The abstract's headline '1980 ±10' and the histogram in Fig. 8 peak in exactly the same period. Because no independent station-based validation is provided, the central empirical claim cannot be separated from a reanalysis artifact. Please repeat the analysis for the 1979-2021 subset, and validate the detected change years and slope changes against station-based gridded products such as Berkeley Earth or CRUTEM, or otherwise demonstrate that the spatial pattern of detected change points is not aligned with the spatial distribution of observation density changes.
  2. [Section V] The claim that 'the temperature in approximately 50% of the global land area has experienced a statistically significant change in trend over these years (in terms of 95% confidence)' is based on per-grid-point tests applied to roughly 20,000 series. Under the null hypothesis of no change, about 5% of tests would be expected to reject by chance even if no grid point had a real change; spatial correlation makes the effective number of independent tests smaller but still large. No multiple-testing correction (e.g., false discovery rate) or field-significance test is reported. The 50% figure and the 'many grid points' language in the abstract therefore overstate the evidence. Please provide a multiple-testing-adjusted assessment of the fraction of land area with significant trend changes.
  3. [Section VI and Section III] The paper itself acknowledges model misspecification: Section VI and Fig. 10 show a Siberia example in which two comparable RMS minima exist and a three-segment fit is 'more appropriate,' and Section III shows that, under a no-change-point null, the procedure produces spurious change points preferentially near the series ends. This means that the detected 'change year' and the 'acceleration around 1980' summary can be artifacts of fitting a single breakpoint to a smoothly accelerating or nonlinear temperature evolution, rather than evidence of a discrete regime shift. The paper does not compare the two-segment model against a smooth alternative (e.g., a quadratic trend or LOESS) or against multi-change-point models for the full grid. Please add such a diagnostic, at least for representative regions, to show that the detected change points are not simply the piecewise-linear approximation to a continuous acceleration.
  4. [Section V and Section III] The maps in Figs. 7 and 9 and the histogram in Fig. 8 present point estimates without any uncertainty quantification, even though Section III shows that the change year is estimated with an 80% interval of ±8 years for realistic noise levels and Section IV explicitly provides formulas (Eqs. 3-4) for trend uncertainties in the presence of short- and long-range correlations. The year-to-year fluctuations in Fig. 8 are described as 'much larger than statistical estimation errors,' but no confidence bands are shown. Without uncertainty measures, the reader cannot distinguish real spatial-temporal structure from estimation noise. Please provide per-grid confidence intervals or masks showing where the change year is well constrained, and add error bands to the cumulative distribution in Fig. 8.
minor comments (4)
  1. [Abstract and Section I] There are several typos: 'indentify' in the abstract, 'annomalies' and 'temperture' in Section I, and 'veryfied' and 'fector' in Section II. Please correct these.
  2. [Section IV, Eqs. (3)-(4)] The notation in Eq. (3) is not fully defined in the text: the symbol d appears in Eq. (3) and is related to H later, but the text says 'd = H - 1/2'; Eq. (4) uses d in the hypergeometric function without explaining the parameter range or the role of ϕ. A brief definition of all symbols directly after Eq. (3) would improve readability.
  3. [Section V] In the paragraph after Fig. 6, the sentence 'more than 44% of them are located in Antarctica' is ambiguous because 'them' could refer to grid points where the single-linear model is preferred or to the total land area; please rephrase.
  4. [General] The manuscript does not state whether the code and the processed change-point maps will be made available, nor does it specify the exact ERA5 data version and download date beyond the CDS reference. Given that the paper is a data-driven statistical analysis, a data and code availability statement would strengthen reproducibility.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: the grid-point change points are fitted outputs, and the sole self-referential start-year caveat is a limitation rather than a construction.

full rationale

The central result is obtained by direct least-squares fits of annual ERA5 grid-point series to a continuous two-segment model, with model selection by BIC and a Monte Carlo null test; the slopes and change years are fitted outputs, not inputs. The synthetic calibration uses noise parameters from independent station data (Potsdam) and from long-range-correlated surrogates, which is standard practice rather than circularity. The SSP comparison in Sec. IV is a consistency check on independent projections and is not used to derive the grid-point change-point map. The self-citations [17,30] supply correlation parameters and a variance formula used for error bars and context; the main detection does not reduce to these parameters. The one self-referential note is in Sec. VI, where the paper acknowledges the 1979 satellite-data discontinuity and justifies the 1950 start by the global-mean ERA5 change point in the 1970s; that global change point is itself a fit to the same data and therefore cannot independently exclude a data artifact. This is a genuine limitation and a correctness risk, but it does not by construction force the values of the fitted grid-point change points, so I do not treat it as a circular derivation of the paper's central empirical claim.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The model uses four fitted parameters per grid point (two slopes, one offset, one change time) plus calibration choices for synthetic nulls: noise amplitude, Hurst exponent, and minimum segment length. The main assumptions are the two-segment model, the validity of ERA5, and the representativeness of the synthetic noise. No new physical entities are postulated.

free parameters (6)
  • Change point year T = varies per grid point; searched over 1960 to 2010
    Optimized for each grid point by minimizing RMS error; central to the change-year maps in Figures 7 and 8.
  • Segment slopes and intercepts (a1, b1, a2, b2) = fit per grid point
    Least-squares coefficients of the two linear segments, with values shown in Figure 9.
  • Synthetic noise standard deviation sigma = 0.45 K
    Estimated from Potsdam annual temperatures and used to generate the null-distribution ensembles in Section III.
  • Hurst exponent H = 0.65 and 0.80 in synthetic nulls; d=0.29 for the ERA5 global series
    Used to generate long-range correlated null time series; controls the false-positive rate of BIC and of the hypothesis test.
  • Minimum segment length = 10 years
    Ad hoc robustness choice restricting change point candidates to 1960 to 2010, which changes the set of possible change years.
  • LOESS bandwidth = 42 years
    Used only in the Potsdam motivation figure, not in the main gridpoint analysis.
assumptions (5)
  • domain assumption Each grid point's annual temperature series is a piecewise linear signal with at most one change point plus stationary noise.
    Invoked in Sections II, V, and VI; the Siberia example with two RMS minima shows this premise can fail.
  • domain assumption ERA5 annual mean 2m temperatures are valid local temperature proxies from 1950 onward, including the pre-satellite era.
    Used in Sections II and V; the authors acknowledge in Section VI that pre-1979 reanalysis is less reliable.
  • domain assumption The null distribution of the slope-difference statistic is produced by Gaussian or ARFIMA noise with parameters representative of all grid points.
    Section III builds 95% confidence intervals from sigma=0.45K and H=0.65 or 0.80; significance statements inherit this assumption.
  • domain assumption The two fitted segments merge continuously at the change point because natural processes are continuous.
    Section II; this excludes jump models and can bias slopes if a genuine discontinuity exists.
  • standard math Standard BIC and AIC formulas with the Gaussian likelihood approximation apply at N=72.
    Equations (1) and (2) in Section III; finite-sample behavior is tested in synthetic data.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Change point detection in ERA5 ground temperature time series." pith.science (2026). https://pith.science/paper/CAYKT2BR

@misc{pith2026250715045,
  author       = {Pith},
  title        = {Pith review of: Change point detection in ERA5 ground temperature time series},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/CAYKT2BR}},
  note         = {Machine review of arXiv:2507.15045}
}
read the original abstract

We analyze the ERA5 reanalysis 2-meter temperature time series on all land grid points using change point analysis. We fit two linear slopes to the data with the constraint that they merge at the point in time where the slope changes. We compare such fits to a standard linear regression in two ways: We use Akaike's and the Bayesian information criteria for model selection, and we test against the null hypothesis of no change of the trend value. For those grid points where the dual linear fit is superior, we construct maps of the time when the trend changes, and of the warming trends in both time intervals. In doing so, we indentify areas where warming speeds up, but find as well areas where warming slows down. We thereby contribute to the characterization of local effects of climate change. We find that many grid points exhibit a change to a much stronger warming trend around the 1980s. This raises the question of whether the climate system has already passed some tipping point.

Figures

Figures reproduced from arXiv: 2507.15045 by the authors.

Figure 1
Figure 1. FIG. 1. The anomalies of annual mean temperatures of Pots [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. FIG. 2. Deterministic trend (orange) and standard deviation (light blue (daily) and light orange (annual) shadows) of artificial [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. FIG. 3. The distribution of BIC [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: FIG. 4. Applying the dual-linear fit to the global temperature time series. (a) The blue curve is the average temperature of the [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: FIG. 5. The dual-linear fit applied to three different climate projections. The fit was done to data in the years 1850-2040, [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: FIG. 6. The difference between the Bayesian scores (BIC) [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]
Figure 7
Figure 7. Figure 7: FIG. 7. The years of change detected by the optimal dual [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]
Figure 8
Figure 8. Figure 8: FIG. 8. Left: A histogram of the number of grid points for which the change point lies in the respective year. The large [PITH_FULL_IMAGE:figures/full_fig_p009_8.png]
Figure 9
Figure 9. Figure 9: FIG. 9. (a) The slope of the first fitted line in [PITH_FULL_IMAGE:figures/full_fig_p010_9.png]
Figure 10
Figure 10. Figure 10: FIG. 10. Raw data and fits for one grid point each in the west and the east of Siberia, where we detected two very different [PITH_FULL_IMAGE:figures/full_fig_p011_10.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

41 extracted references · 41 canonical work pages

  1. [1]

    H. Akaike. A new look at the statistical model identifica- tion. IEEE Transactions on Automatic Control , 19:716– 723, 1974

  2. [2]

    Aminikhanghahi and D

    S. Aminikhanghahi and D. J. Cook. A survey of methods for time series change point detection. Knowledge and Information Systems , 51:339–367, 2017

  3. [3]

    Anderson and K

    D. Anderson and K. Burnham. Model Selection and Multi-model Inference. Springer-Verlag, second edition, 2004

  4. [4]

    Arrhenius

    S. Arrhenius. Xxxi. on the influence of carbonic acid in the air upon the temperature of the ground.Philosophical Magazine and Journal of Science , 41:237–276, 1896

  5. [5]

    Bromwich, J

    D. Bromwich, J. Nicolas, A. Monaghan, M. Lazzara, L. Keller, G. Weidner, and A. Wilson. Erratum: Corri- gendum: Central west antarctica among the most rapidly warming regions on earth. Nature Geoscience, 7:76–76, 2014

  6. [6]

    Cowtan, P

    K. Cowtan, P. Jacobs, P. Thorne, and R. Wilkinson. Sta- tistical analysis of coverage error in simple global temper- ature estimators. Dynamics and Statistics of the Climate System, 3, 2018

  7. [7]

    Climate data center of the german weather service

    Deutscher Wetterdienst. Climate data center of the german weather service. https://www.dwd.de. Usage under CC BY 4.0. Potsdam annual mean temperatures downloaded from https://opendata.dwd.de/climate_ environment/CDC/observations\_germany/climate/ annual/kl/historical/, station No. 03987, data set jahreswerte KL 03987 18930101 20241231 hist.zip containing ...

  8. [8]

    Klimastatusbericht

    Deutscher Wetterdienst. Klimastatusbericht

Show all 41 references
  1. [9]

    J. Ding, V. Tarokh, and Y. Yang. Model selection tech- niques: An overview. IEEE Signal Processing Magazine , 35:16–34, 2018

  2. [10]

    (available only in German)

  3. [11]

    Estrada, D

    F. Estrada, D. Kim, and P. Perron. Spatial variations in the warming trend and the transition to more severe weather in midlatitudes. Scientific Reports, 11:145, 2021

  4. [12]

    J. F. Eichner, E. Koscielny-Bunde, A. Bunde, S. Havlin, and H.-J. Schellnhuber. Power-law persistence and trends in the atmosphere: a detailed study of long temperature records. Physical Review E , 68:046133, 2003

  5. [13]

    Falvey and R

    M. Falvey and R. Garreaud. Regional cooling in a warm- ing world: Recent temperature trends in the southeast pacific and along the west coast of subtropical south 13 america (1979–2006). Journal of Geophysical Research , 114, 2009

  6. [14]

    The family of era5 datasets

    European Centre for Medium-Range Weather Fore- casts (ECMWF). The family of era5 datasets. https://confluence.ecmwf.int/display/CKB/The+ family+of+ERA5+datasets. as viewed in January 2025

  7. [15]

    Hersbach, B

    H. Hersbach, B. Bell, P. Berrisford, S. Hirahara, A. Hor´ anyi, J. Mu˜ noz-Sabater, J. Nicolas, C. Peubey, R. Radu, D. Schepers, et al. The era5 global reanalysis. Quarterly Journal of the Royal Meteorological Society , 146:1999–2049, 2020

  8. [16]

    Hersbach, B

    H. Hersbach, B. Bell, P. Berrisford, G. Biavati, A. Hor´ anyi, J. Mu˜ noz-Sabater, J. Nicolas, C. Peubey, R. Radu, I. Rozum, D. Schepers, A. Simmons, C. Soci, D. Dee, and J. Th´ epaut. Era5 hourly data on single levels from 1940 to present, 2023. Copernicus Climate Change Serv...

  9. [17]

    Kassel and H

    J. Kassel and H. Kantz. Statistical inference of one- dimensional persistent nonlinear time series and applica- tion to predictions. Physical Review Research, 4:013206, 2022

  10. [18]

    Intergovernmental panel on climate change

    IPCC. Intergovernmental panel on climate change. https://www.ipcc.ch. Contains links to all reports

  11. [19]

    Koscielny-Bunde, H

    E. Koscielny-Bunde, H. E. Roman, A. Bunde, and S. Havlin. Long-range power-law correlations in local daily temperature fluctuations. Philosophical Magazine B, 77:1331–1340, 1998

  12. [20]

    Koscielny-Bunde, A

    E. Koscielny-Bunde, A. Bunde, and S. Havlin. Analysis of daily temperature fluctuations. Physica A: Statistical Mechanics and its Applications , 231:393–396, 1996

  13. [21]

    Lenton, H

    T. Lenton, H. Held, E. Kriegler, J. Hall, W. Lucht, S. Rahmstorf, and H. Schellnhuber. Tipping elements in the earth’s climate system. Proceedings of the National Academy of Sciences, 105:1786–1793, 2008

  14. [22]

    T. Lenton. Early warning of climate tipping points. Na- ture Climate Change , 1:201–209, 2011

  15. [23]

    Massah and H

    M. Massah and H. Kantz. Confidence intervals for time averages in the presence of long-range correlations, a case study on earth surface temperature anomalies. Geophys- ical Research Letters, 43:9243–9249, 2016

  16. [24]

    Mahlstein, R

    I. Mahlstein, R. Knutti, S. Solomon, and R. Portmann. Early onset of significant local warming in low latitude countries. Environmental Research Letters , 6:034009, 2011

  17. [25]

    Global mean surface tem- perature records

    Multiple Sources. Global mean surface tem- perature records. https://en.wikipedia.org/ wiki/Global_surface_temperature, https: //gml.noaa.gov/data/data.php, https:// crudata.uea.ac.uk/cru/data/temperature/, https://data.giss.nasa.gov/gistemp/, https: //www.nsstc.uah.edu/data/m...

  18. [26]

    Mudelsee

    M. Mudelsee. Trend analysis of climate time series: A review of methods. Earth-Science Reviews, 190:310–322, 2019

  19. [27]

    Global temperature vital signs

    NASA. Global temperature vital signs. https://climate.nasa.gov/vital-signs/ global-temperature/, 2024

  20. [28]

    Era5-land: A state-of-the-art global reanalysis dataset for land applications

    Joaqu ´ ın Mu˜ noz-Sabater, Emanuel Dutra, Anna Agust ´ ı- Panareda, Cl´ ement Albergel, Giacomo Arduini, and Gi- anpaolo Balsamo. Era5-land: A state-of-the-art global reanalysis dataset for land applications. Earth System Science Data, 13:4349–4383, 2021

  21. [29]

    C. Peng, S. Buldyrev, S. Havlin, M. Simons, H. Stan- ley, and A. Goldberger. Mosaic organization of dna nu- cleotides. Physical Review E , 49:1685–1689, 1994

  22. [30]

    J. Patz, D. Campbell-Lendrum, T. Holloway, and J. Fo- ley. Impact of regional climate change on human health. Nature, 438:310–317, 2005

  23. [31]

    Rahmstorf, G

    S. Rahmstorf, G. Foster, and N. Cahill. Global tempera- ture evolution: recent trends and some pitfalls. Environ- mental Research Letters, 12:054001, 2017

  24. [32]

    Phillips, M

    E. Phillips, M. H¨ oll, H. Kantz, and Y. Zhou. Trend anal- ysis in the presence of short-and long-range correlations with application to regional warming. Physical Review E, 108, 2023

  25. [33]

    G. Schwarz. Estimating the dimension of a model.Annals of Statistics , 6:461–464, 1978

  26. [34]

    Retamales-Mu˜ noz, C

    G. Retamales-Mu˜ noz, C. Dur´ an-Alarc´ on, and C. Mat- tar. Recent land surface temperature patterns in antarc- tica using satellite and reanalysis data. Journal of South American Earth Sciences, 95:102304, 2019

  27. [35]

    Stoica and Y

    P. Stoica and Y. Selen. Model-order selection: a review of information criterion rules. IEEE Signal Processing Magazine, 21:36–47, 2004

  28. [36]

    Steig, D

    E. Steig, D. Schneider, S. Rutherford, M. Mann, J. Comiso, and D. Shindell. Warming of the antarctic ice-sheet surface since the 1957 international geophysical year. Nature, 457:459–462, 2009

  29. [37]

    You-Ren, H

    D. You-Ren, H. Bjørn, and F. Samset. Evaluating global and regional land warming trends in the past decades with both modis and era5-land land surface temperature data. Remote Sensing of Environment , 280, 2022

  30. [38]

    M. Stone. An asymptotic equivalence of choice of model by cross-validation and akaike’s criterion. Journal of the Royal Statistical Society. Series B (Methodological) , 39:44–47, 1977

  31. [39]

    Zhang and Y

    T. Zhang and Y. Huang. Estimating the impacts of warming trends on wheat and maize in china from 1980 to 2008 based on county level data.International Journal of Climatology, 33:699–708, 2013

  32. [40]

    Zhang, B

    H. Zhang, B. Huang, J. Lawrimore, M. Menne, and T. Smith. Noaa global surface temperature dataset (noaaglobaltemp), version 4.0. https://www.ncei. noaa.gov/metadata/geoportal/rest/metadata/ item/gov.noaa.ncdc%253AC01704/html, 2019. https://www.ncei.noaa.gov/access/monitoring/ ...

  33. [2024]

    https://www.dwd.de/DE/leistungen/ klimastatusbericht/publikationen/ksb_2024.pdf,

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.