<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE article PUBLIC "-//NLM//DTD Journal Publishing with OASIS Tables v3.0 20080202//EN" "https://jats.nlm.nih.gov/nlm-dtd/publishing/3.0/journalpub-oasis3.dtd">
<article xmlns:xlink="http://www.w3.org/1999/xlink" xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:oasis="http://docs.oasis-open.org/ns/oasis-exchange/table" xml:lang="en" dtd-version="3.0" article-type="research-article">
  <front>
    <journal-meta><journal-id journal-id-type="publisher">GMD</journal-id><journal-title-group>
    <journal-title>Geoscientific Model Development</journal-title>
    <abbrev-journal-title abbrev-type="publisher">GMD</abbrev-journal-title><abbrev-journal-title abbrev-type="nlm-ta">Geosci. Model Dev.</abbrev-journal-title>
  </journal-title-group><issn pub-type="epub">1991-9603</issn><publisher>
    <publisher-name>Copernicus Publications</publisher-name>
    <publisher-loc>Göttingen, Germany</publisher-loc>
  </publisher></journal-meta>
    <article-meta>
      <article-id pub-id-type="doi">10.5194/gmd-19-10917-2026</article-id><title-group><article-title>The stippled grid points are statistically significant: (mis)uses  of false discovery rate correction for geospatial data</article-title><alt-title>(Mis)uses of false discovery rate correction</alt-title>
      </title-group>
      <contrib-group>
        <contrib contrib-type="author" corresp="yes" rid="aff1 aff2">
          <name><surname>Schutte</surname><given-names>Michael Konrad</given-names></name>
          <email>michael.schutte@geo.uu.se</email>
        <ext-link>https://orcid.org/0009-0003-8761-1639</ext-link></contrib>
        <contrib contrib-type="author" corresp="no" rid="aff1 aff2">
          <name><surname>Olivetti</surname><given-names>Leonardo</given-names></name>
          
        <ext-link>https://orcid.org/0009-0003-4904-4362</ext-link></contrib>
        <contrib contrib-type="author" corresp="no" rid="aff3 aff4">
          <name><surname>Pons</surname><given-names>Flavio Maria Emanuele</given-names></name>
          
        </contrib>
        <contrib contrib-type="author" corresp="no" rid="aff1 aff2 aff5">
          <name><surname>Messori</surname><given-names>Gabriele</given-names></name>
          
        <ext-link>https://orcid.org/0000-0002-2032-5211</ext-link></contrib>
        <aff id="aff1"><label>1</label><institution>Department of Earth Sciences, Uppsala University, Uppsala, Sweden</institution>
        </aff>
        <aff id="aff2"><label>2</label><institution>Swedish Centre for Impacts of Climate Extremes (climes), Uppsala University, Uppsala, Sweden</institution>
        </aff>
        <aff id="aff3"><label>3</label><institution>Laboratoire des Sciences du Climat et de l'Environnement, UMR 8212 CEA-CNRS-UVSQ,  Université Paris-Saclay, IPSL, 91191 Gif-sur-Yvette, France</institution>
        </aff>
        <aff id="aff4"><label>4</label><institution>Science Partners, Paris, France</institution>
        </aff>
        <aff id="aff5"><label>5</label><institution>Department of Meteorology, Stockholm University, Stockholm, Sweden</institution>
        </aff>
      </contrib-group>
      <author-notes><corresp id="corr1">Michael Konrad Schutte (michael.schutte@geo.uu.se)</corresp></author-notes><pub-date><day>17</day><month>November</month><year>2026</year></pub-date>
      
      <volume>19</volume>
      <issue>22</issue>
      <fpage>10917</fpage><lpage>10932</lpage>
      <history>
        <date date-type="received"><day>17</day><month>April</month><year>2026</year></date>
           <date date-type="rev-request"><day>21</day><month>May</month><year>2026</year></date>
           <date date-type="rev-recd"><day>11</day><month>November</month><year>2026</year></date>
           <date date-type="accepted"><day>12</day><month>November</month><year>2026</year></date>
      </history>
      <permissions>
        <copyright-statement>Copyright: © 2026 Michael Konrad Schutte et al.</copyright-statement>
        <copyright-year>2026</copyright-year>
      <license license-type="open-access"><license-p>This work is licensed under the Creative Commons Attribution 4.0 International License. To view a copy of this licence, visit <ext-link ext-link-type="uri" xlink:href="https://creativecommons.org/licenses/by/4.0/">https://creativecommons.org/licenses/by/4.0/</ext-link></license-p></license></permissions><self-uri xlink:href="https://gmd.copernicus.org/articles/19/10917/2026/gmd-19-10917-2026.html">This article is available from https://gmd.copernicus.org/articles/19/10917/2026/gmd-19-10917-2026.html</self-uri><self-uri xlink:href="https://gmd.copernicus.org/articles/19/10917/2026/gmd-19-10917-2026.pdf">The full text article is available as a PDF file from https://gmd.copernicus.org/articles/19/10917/2026/gmd-19-10917-2026.pdf</self-uri>
      <abstract><title>Abstract</title>

      <p id="d2e140">Peer-reviewed articles in the geosciences routinely assess statistical significance in spatially distributed data. Statistical significance is often assessed independently at each grid point, while formal adjustment for multiple testing is applied less consistently. Although several approaches to account for multiple testing exist, their application to geosciences data is not always straightforward, as these data often exhibit spatially coherent signals.</p>

      <p id="d2e143">In this work we revisit the Benjamini-Hochberg (BH) false discovery rate (FDR) correction, which is a commonly used method to account for multiple testing, in the context of spatially structured datasets. We first highlight how neglecting multiple-testing correction can substantially inflate the number of false positives. We further show that the BH correction approach proposed in literature for application in geosciences can yield counterintuitive results when applied to spatially coherent signals that cover large fractions of a spatial domain. To illustrate the latter point, we provide an example based on near-surface air temperature composites following sudden stratospheric warmings. We show that when anomalies are spatially coherent, restricting the spatial domain can increase the BH-corrected significance threshold. Consequently, the same underlying field can display a larger number of statistically significant grid points solely due to domain selection, despite unchanged data. We explain this behavior from the point of view of the rank-based structure of the BH correction and discuss its implications for spatial inference and uncertainty quantification in the geosciences. Furthermore, a numerical experiment confirms that the larger the fraction of true positives the higher the BH-corrected significance threshold becomes for a set FDR level.</p>

      <p id="d2e146">Building on these insights, we outline practical recommendations for transparent and robust significance assessment in geoscientific applications. These include clearly documenting multiple-testing corrections when adjusted pointwise significance is shown, cautious interpretation of adjusted thresholds, careful avoidance of a posteriori domain selection and considering spatially aware alternatives such as regional or cluster-based inference when appropriate. Overall, our results highlight both the need to account for multiple-testing, and potential issues with a naïve application and interpretation of the BH correction. We hope that our work may contribute to more robust statistical testing in the geosciences.</p>
  </abstract>
    
<funding-group>
<award-group id="gs1">
<funding-source>Horizon 2020</funding-source>
<award-id>948309</award-id>
<award-id>101231771</award-id>
</award-group>
</funding-group>
</article-meta>
  </front>
<body>
      

<sec id="Ch1.S1" sec-type="intro">
  <label>1</label><title>Introduction</title>
      <p id="d2e158">Highlighting regions with statistically significant signals on maps is very common in the analysis of geospatial data, including model-evaluation, observational diagnostics and event composites <xref ref-type="bibr" rid="bib1.bibx14 bib1.bibx16 bib1.bibx10" id="paren.1"><named-content content-type="pre">e.g.,</named-content></xref>. This approach supports hypothesis testing and interpretation of results, and can also translate into real-world decision-making, e.g., in the case of forecast evaluation <xref ref-type="bibr" rid="bib1.bibx22 bib1.bibx36 bib1.bibx42" id="paren.2"><named-content content-type="pre">e.g.,</named-content></xref>. However, the common practice of applying a pointwise significance test at each data point (i.e. repeated local tests) introduces a fundamental statistical issue: when hypothesis tests are conducted multiple times, the probability of finding apparently significant results purely by chance increases sharply. For example, if 100 independent grid points are tested at the 5 % significance level, one would expect on average five points to appear “significant” even in the absence of any true signal (i.e. false rejections of the null hypothesis). However, for any single sample, this number may be higher. The expected proportion of incorrectly rejected null hypotheses among all rejected hypotheses is known as false discovery rate (FDR). Moreover, the chance of at least one spurious rejection is extremely high: 99.41 % when performing 100 local tests. This point was eloquently made by <xref ref-type="bibr" rid="bib1.bibx48" id="text.3"/>. The geosciences community has been aware of this issue for well over a century <xref ref-type="bibr" rid="bib1.bibx44" id="paren.4"/> and the issue has since been revisited regularly <xref ref-type="bibr" rid="bib1.bibx31 bib1.bibx24 bib1.bibx23 bib1.bibx46 bib1.bibx48" id="paren.5"><named-content content-type="pre">e.g.,</named-content></xref>.</p>
      <p id="d2e182">To avoid inflating the number of false rejections, adjustment for multiple testing is required. A variety of approaches have been proposed to address this issue, including Family-Wise Error Rate (FWER) control <xref ref-type="bibr" rid="bib1.bibx44" id="paren.6"><named-content content-type="pre">e.g.,</named-content></xref>, field significance methods <xref ref-type="bibr" rid="bib1.bibx31" id="paren.7"><named-content content-type="pre">e.g.,</named-content></xref> and, more recently, the Benjamini–Hochberg FDR correction <xref ref-type="bibr" rid="bib1.bibx5" id="paren.8"/>, hereinafter BH correction. The BH correction aims to limit the expected proportion of falsely rejected hypotheses among all tests, and is widely applied in many fields such as genomics or neuroimaging <xref ref-type="bibr" rid="bib1.bibx34 bib1.bibx1" id="paren.9"><named-content content-type="pre">e.g.,</named-content></xref>. The original paper by <xref ref-type="bibr" rid="bib1.bibx5" id="text.10"/> has accumulated more than 125 000 citations according to Google Scholar. Following the review and recommendations of <xref ref-type="bibr" rid="bib1.bibx48" id="text.11"/>, the BH correction has also become the de facto gold standard for multiple-testing correction in geoscientific applications. Nonetheless, it is still implemented sporadically in the field. To provide a heuristic illustration we randomly sampled fifty articles from meteorology and climate science cited by the IPCC Sixth Assessment Report, Chapter 11 <xref ref-type="bibr" rid="bib1.bibx39" id="paren.12"/>, which discusses how anthropogenic climate change affects weather and climate extreme events. We found that 18 studies assessed statistical significance on a map, of which 17 without multiple-testing correction, while only one article implemented FDR correction. These numbers are consistent with previous empirical observations of the same issue <xref ref-type="bibr" rid="bib1.bibx48" id="paren.13"><named-content content-type="pre">e.g.,</named-content></xref> and further motivate recent efforts to implement FDR correction methods in climate model testing <xref ref-type="bibr" rid="bib1.bibx25" id="paren.14"/>.</p>
      <p id="d2e221">However, even the application and interpretation of the BH correction can be problematic. The BH correction is often misunderstood as adjusting the significance threshold for each individual grid point, making it “harder” for a point to be considered significant. The last author of this study has himself implicitly adopted this fallacious interpretation in some of his past work. While BH correction does control the global error rate, it does not provide any guarantee for individual grid points. In some applications, the resulting BH-corrected threshold may even be higher than the uncorrected threshold, if the pointwise significance level is lower than the chosen FDR level. That is, some grid points become significant only after the BH correction. This effect is highly sensitive to the choice of domain. Interpreting BH-corrected significance on maps as if each highlighted grid point were locally significant can therefore be misleading.</p>
      <p id="d2e224">There exist adaptive refinements of the BH correction that adapt to properties of the data, either by estimating the proportion of true null hypotheses to increase power <xref ref-type="bibr" rid="bib1.bibx40 bib1.bibx7" id="paren.15"/>, or by modifying the procedure to remain valid under dependence between tests <xref ref-type="bibr" rid="bib1.bibx6 bib1.bibx33" id="paren.16"/>. However, we leave a deeper analysis of their behavior to future work. Indeed, <xref ref-type="bibr" rid="bib1.bibx13" id="text.17"/> found that the BH correction still outperforms the method developed by <xref ref-type="bibr" rid="bib1.bibx6" id="text.18"/> that accounts for test dependence. Moreover, the comparatively limited application of FDR corrections in the geosciences speaks for focusing on the BH correction, which offers the best trade-off among effectiveness, simplicity, and established use in the field.</p>
      <p id="d2e240">In the following section, we provide a more detailed explanation of the BH correction, and subsequently highlight that: (i) any depiction of pointwise statistical significance on maps must be accompanied by an explicit and appropriate multiple-testing correction; and (ii) the BH correction can yield misleading thresholds if misunderstood or deliberately exploited. We illustrate these points with an example based on meteorological reanalysis data and a numerical experiment. Finally, we conclude with some practical recommendations for transparent and robust significance testing in the geosciences and its interpretation.</p>
</sec>
<sec id="Ch1.S2">
  <label>2</label><title>The false discovery rate</title>
      <p id="d2e251">Under the assumption of independent tests, the number of falsely rejected null hypotheses is a random variable following a binomial distribution with probability <inline-formula><mml:math id="M1" display="inline"><mml:mi mathvariant="italic">α</mml:mi></mml:math></inline-formula>, the chosen significance level <xref ref-type="bibr" rid="bib1.bibx48" id="paren.19"/>. Consequently, the probability of obtaining at least one false positive approaches unity as the number of tests increases. For the commonly used 5 % significance level, the probability of no false rejections is <inline-formula><mml:math id="M2" display="inline"><mml:mrow><mml:mo>(</mml:mo><mml:mn mathvariant="normal">1</mml:mn><mml:mo>-</mml:mo><mml:mn mathvariant="normal">0.05</mml:mn><mml:msup><mml:mo>)</mml:mo><mml:mi>n</mml:mi></mml:msup></mml:mrow></mml:math></inline-formula>, where <inline-formula><mml:math id="M3" display="inline"><mml:mi>n</mml:mi></mml:math></inline-formula> is the number of tests. In high-resolution gridded maps containing thousands of grid points, at least one false rejection is therefore almost certain. Moreover, the expected number of false rejections scales linearly with <inline-formula><mml:math id="M4" display="inline"><mml:mi>n</mml:mi></mml:math></inline-formula> (the expected value of the binomial distribution is <inline-formula><mml:math id="M5" display="inline"><mml:mrow><mml:mi>n</mml:mi><mml:mo>⋅</mml:mo><mml:mi mathvariant="italic">α</mml:mi></mml:mrow></mml:math></inline-formula>), implying that hundreds of spuriously significant grid points may appear in a single high-resolution map (e.g., on average about 500 for <inline-formula><mml:math id="M6" display="inline"><mml:mrow><mml:mi>n</mml:mi><mml:mo>=</mml:mo><mml:mn mathvariant="normal">10</mml:mn><mml:mspace width="0.125em" linebreak="nobreak"/><mml:mn mathvariant="normal">000</mml:mn></mml:mrow></mml:math></inline-formula> grid points).</p>
      <p id="d2e325">A commonly-used approach in the geosciences to address this issue is the BH correction <xref ref-type="bibr" rid="bib1.bibx5" id="paren.20"/> for FDR, which seeks to limit the expected proportion of falsely rejected hypotheses. For example, an FDR level of <inline-formula><mml:math id="M7" display="inline"><mml:mrow><mml:msub><mml:mi mathvariant="italic">α</mml:mi><mml:mi mathvariant="normal">FDR</mml:mi></mml:msub><mml:mo>=</mml:mo><mml:mn mathvariant="normal">0.1</mml:mn></mml:mrow></mml:math></inline-formula> implies that, on average, no more than 10 % of rejected null hypotheses are false positives. In practice, the <inline-formula><mml:math id="M8" display="inline"><mml:mi>p</mml:mi></mml:math></inline-formula> values from the local tests are first sorted in ascending order and the BH-corrected threshold <inline-formula><mml:math id="M9" display="inline"><mml:mrow><mml:msubsup><mml:mi>p</mml:mi><mml:mi mathvariant="normal">FDR</mml:mi><mml:mo>*</mml:mo></mml:msubsup></mml:mrow></mml:math></inline-formula> is defined as

          <disp-formula id="Ch1.E1" content-type="numbered"><label>1</label><mml:math id="M10" display="block"><mml:mrow><mml:msubsup><mml:mi>p</mml:mi><mml:mi mathvariant="normal">FDR</mml:mi><mml:mo>*</mml:mo></mml:msubsup><mml:mo>=</mml:mo><mml:msub><mml:mi mathvariant="normal">max</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mo>=</mml:mo><mml:mn mathvariant="normal">1</mml:mn><mml:mo>,</mml:mo><mml:mspace linebreak="nobreak" width="0.125em"/><mml:mi mathvariant="normal">…</mml:mi><mml:mo>,</mml:mo><mml:mspace width="0.125em" linebreak="nobreak"/><mml:msub><mml:mi>N</mml:mi><mml:mi mathvariant="normal">tests</mml:mi></mml:msub></mml:mrow></mml:msub><mml:mfenced close="]" open="["><mml:mrow><mml:msub><mml:mi>p</mml:mi><mml:mi>i</mml:mi></mml:msub><mml:mo>:</mml:mo><mml:msub><mml:mi>p</mml:mi><mml:mi>i</mml:mi></mml:msub><mml:mo>≤</mml:mo><mml:mfenced close=")" open="("><mml:mstyle displaystyle="true"><mml:mfrac style="display"><mml:mi>i</mml:mi><mml:mrow><mml:msub><mml:mi>N</mml:mi><mml:mi mathvariant="normal">tests</mml:mi></mml:msub></mml:mrow></mml:mfrac></mml:mstyle></mml:mfenced><mml:msub><mml:mi mathvariant="italic">α</mml:mi><mml:mi mathvariant="normal">FDR</mml:mi></mml:msub></mml:mrow></mml:mfenced><mml:mo>,</mml:mo></mml:mrow></mml:math></disp-formula>

        so that all <inline-formula><mml:math id="M11" display="inline"><mml:mi>p</mml:mi></mml:math></inline-formula> values smaller than or equal to the largest <inline-formula><mml:math id="M12" display="inline"><mml:mi>p</mml:mi></mml:math></inline-formula> value satisfying this condition are considered statistically significant. If no local <inline-formula><mml:math id="M13" display="inline"><mml:mi>p</mml:mi></mml:math></inline-formula> value satisfies this condition, the field as a whole does not reach statistical significance, implying that all apparent local discoveries may be spurious – an interpretation analogous to that of field significance tests. Following Eq. (<xref ref-type="disp-formula" rid="Ch1.E1"/>), controlling FDR means that the threshold adjusts dynamically depending on the proportion of low <inline-formula><mml:math id="M14" display="inline"><mml:mi>p</mml:mi></mml:math></inline-formula> values in the sample. While this property is intrinsic to the BH correction, it is seldom acknowledged when applied in practice.</p>
      <p id="d2e470">The appeal of the BH correction for geoscientific applications lies primarily in its ease of implementation and apparent ease of interpretation, as well as its robustness to violations of the assumption that local tests are independent <xref ref-type="bibr" rid="bib1.bibx48" id="paren.21"/>. Indeed, while the BH correction assumes local test independence, numerical experiments by <xref ref-type="bibr" rid="bib1.bibx48" id="text.22"/> show that the correction performs well even in the presence of strong spatial autocorrelation – as is commonly the case in many geophysical fields. This contrasts earlier approaches such as field significance testing, which are highly sensitive to spatial dependence <xref ref-type="bibr" rid="bib1.bibx31" id="paren.23"/>.</p>
      <p id="d2e482">However, while the BH correction provides a valid control of the global error rate, it does not necessarily render individual tests more conservative in a local sense. Based on simulations with spatially autocorrelated data, <xref ref-type="bibr" rid="bib1.bibx48" id="text.24"/> recommends selecting an <inline-formula><mml:math id="M15" display="inline"><mml:mrow><mml:msub><mml:mi mathvariant="italic">α</mml:mi><mml:mi mathvariant="normal">FDR</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> approximately twice as large as the target field significance level in order to avoid overly restrictive thresholds. This recommendation is well-motivated for large-scale maps in which statistically significant signals are sparse and spatially restricted, yet it exposes an important limitation of the BH correction. Because the BH-corrected threshold <inline-formula><mml:math id="M16" display="inline"><mml:mrow><mml:msubsup><mml:mi>p</mml:mi><mml:mi mathvariant="normal">FDR</mml:mi><mml:mo>*</mml:mo></mml:msubsup></mml:mrow></mml:math></inline-formula> depends on the rank ordering of local <inline-formula><mml:math id="M17" display="inline"><mml:mi>p</mml:mi></mml:math></inline-formula> values, it can become counterintuitively permissive when applied to a spatial subset exhibiting a prevalent signal. In practice, selecting an <inline-formula><mml:math id="M18" display="inline"><mml:mrow><mml:msub><mml:mi mathvariant="italic">α</mml:mi><mml:mi mathvariant="normal">FDR</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> larger than the nominal pointwise significance level (e.g., <inline-formula><mml:math id="M19" display="inline"><mml:mrow><mml:msub><mml:mi mathvariant="italic">α</mml:mi><mml:mi mathvariant="normal">FDR</mml:mi></mml:msub><mml:mo>=</mml:mo><mml:mn mathvariant="normal">0.1</mml:mn></mml:mrow></mml:math></inline-formula> for a target field significance level of 5 % as recommended by <xref ref-type="bibr" rid="bib1.bibx48" id="altparen.25"/>) can lead to corrected thresholds that exceed said pointwise significance level. Applications in the geosciences literature often implicitly or explicitly assume that the BH correction makes the test more stringent, without acknowledging that commonly-used parameter choices may produce the counterintuitive outcome that more grid points appear statistically significant on a map following BH correction than when using uncorrected local tests. This becomes particularly problematic if the BH correction is applied a posteriori, after inspecting the results of uncorrected grid-point-level tests, in which case it may be regarded as p-hacking.</p>
      <p id="d2e550">The BH correction operates by controlling a set-level property of the rejected hypotheses, rather than adjusting the error probability at each grid point. When BH-corrected results are visualized in the same manner as pointwise significance maps, this distinction is easily obscured, encouraging local interpretations that the method does not support. In contrast to pointwise testing, which bounds the probability of false positives at each location, the BH correction constrains only the expected fraction of false positives among all declared discoveries. As a result, FDR-based significance should not be interpreted as a local measure of statistical significance. This point is illustrated with a real-world case study in the next section.</p>
</sec>
<sec id="Ch1.S3">
  <label>3</label><title>Example of assessing statistical significance with meteorological data</title>
      <p id="d2e561">We now illustrate how the BH correction can yield corrected thresholds that exceed the nominal local significance level, and how markedly different apparent patterns of statistical significance can result from the same underlying data, solely as a consequence of changing the analysis domain. This is what would occur if the domain were selected a posteriori after inspecting the results. Employing an example from meteorology, we show composites of 2 m air temperature (t2m) anomalies, averaged over the 60 d following 26 Sudden Stratospheric Warming events <xref ref-type="bibr" rid="bib1.bibx4 bib1.bibx29" id="paren.26"><named-content content-type="pre">SSWs; e.g.,</named-content></xref>. It is well documented that SSWs often result in near-surface responses, for instance showing below-average temperatures over northern Eurasia <xref ref-type="bibr" rid="bib1.bibx27 bib1.bibx10" id="paren.27"><named-content content-type="pre">e.g.,</named-content></xref>. However, considerable uncertainty remains about the exact location and magnitude of these anomalies due to variability in stratosphere-troposphere coupling <xref ref-type="bibr" rid="bib1.bibx35 bib1.bibx28" id="paren.28"><named-content content-type="pre">e.g.,</named-content></xref>. In the following, we illustrate how domain choice affects the visual interpretation of statistical significance in this example.</p>
<sec id="Ch1.S3.SS1">
  <label>3.1</label><title>Temperature data and <inline-formula><mml:math id="M20" display="inline"><mml:mi>p</mml:mi></mml:math></inline-formula> value computation</title>
      <p id="d2e594">The temperature data is obtained from ERA5 reanalysis <xref ref-type="bibr" rid="bib1.bibx20" id="paren.29"/>, a widely used data set in meteorology. We use t2m data over the Northern Hemisphere north of 20° N with a horizontal resolution of 0.5° between 1980 and 2022. After linearly detrending the t2m data, temperature anomalies are computed by subtracting a smoothed seasonal cycle (15 d centered running mean). The 26 SSW dates between 1980 and 2022 are obtained from the Sudden Stratospheric Warming Compendium <xref ref-type="bibr" rid="bib1.bibx10 bib1.bibx9" id="paren.30"/>, based on the conventional definition of reversal of 10 hPa zonal winds at 60° N <xref ref-type="bibr" rid="bib1.bibx11" id="paren.31"/>.</p>
      <p id="d2e606">To test the statistical significance of the t2m anomalies, we assess at each grid point whether the mean of the 26 individual 60 d mean t2m anomalies following the SSW onset dates is significantly colder than climatology using a one-sided, one-sample Student's <inline-formula><mml:math id="M21" display="inline"><mml:mi>t</mml:mi></mml:math></inline-formula> test:

            <disp-formula id="Ch1.E2" content-type="numbered"><label>2</label><mml:math id="M22" display="block"><mml:mrow><mml:mi>t</mml:mi><mml:mo>=</mml:mo><mml:mstyle displaystyle="true"><mml:mfrac style="display"><mml:mrow><mml:mover accent="true"><mml:mi>x</mml:mi><mml:mo mathvariant="normal">‾</mml:mo></mml:mover><mml:mo>-</mml:mo><mml:msub><mml:mi mathvariant="italic">μ</mml:mi><mml:mn mathvariant="normal">0</mml:mn></mml:msub></mml:mrow><mml:mrow><mml:mi>s</mml:mi><mml:mo>/</mml:mo><mml:msqrt><mml:mi>n</mml:mi></mml:msqrt></mml:mrow></mml:mfrac></mml:mstyle><mml:mo>,</mml:mo></mml:mrow></mml:math></disp-formula>

          where <inline-formula><mml:math id="M23" display="inline"><mml:mover accent="true"><mml:mi>x</mml:mi><mml:mo mathvariant="normal">‾</mml:mo></mml:mover></mml:math></inline-formula> and <inline-formula><mml:math id="M24" display="inline"><mml:mi>s</mml:mi></mml:math></inline-formula> are the sample mean and sample standard deviation of the <inline-formula><mml:math id="M25" display="inline"><mml:mrow><mml:mi>n</mml:mi><mml:mo>=</mml:mo><mml:mn mathvariant="normal">26</mml:mn></mml:mrow></mml:math></inline-formula> individual event-wise 60 d mean anomalies at a given grid point, and <inline-formula><mml:math id="M26" display="inline"><mml:mrow><mml:msub><mml:mi mathvariant="italic">μ</mml:mi><mml:mn mathvariant="normal">0</mml:mn></mml:msub><mml:mo>=</mml:mo><mml:mn mathvariant="normal">0</mml:mn></mml:mrow></mml:math></inline-formula> is the population mean under the null hypothesis of no SSW-related response, consistent with the detrended and deseasonalized nature of the input data. The resulting <inline-formula><mml:math id="M27" display="inline"><mml:mi>p</mml:mi></mml:math></inline-formula> value is obtained from the Student's <inline-formula><mml:math id="M28" display="inline"><mml:mi>t</mml:mi></mml:math></inline-formula>-distribution with <inline-formula><mml:math id="M29" display="inline"><mml:mrow><mml:mi>n</mml:mi><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn><mml:mo>=</mml:mo><mml:mn mathvariant="normal">25</mml:mn></mml:mrow></mml:math></inline-formula> degrees of freedom, testing the alternative hypothesis that the mean anomaly is significantly colder than climatology (<inline-formula><mml:math id="M30" display="inline"><mml:mrow><mml:mover accent="true"><mml:mi>x</mml:mi><mml:mo mathvariant="normal">‾</mml:mo></mml:mover><mml:mo>&lt;</mml:mo><mml:msub><mml:mi mathvariant="italic">μ</mml:mi><mml:mn mathvariant="normal">0</mml:mn></mml:msub></mml:mrow></mml:math></inline-formula>). The 60 d time window follows observational and modeling evidence that SSWs influence tropospheric circulation for up to 60 d after their onset <xref ref-type="bibr" rid="bib1.bibx3 bib1.bibx35" id="paren.32"><named-content content-type="pre">e.g.,</named-content></xref>. We note that temperature variability is not stationary across the November–March window, such that individual SSW events falling in different months are drawn from different background-variability regimes. While this introduces mild heterogeneity in the pooled sample variance estimate, it does not affect the domain-dependence of the BH-corrected threshold illustrated below, which is a structural property of the BH correction.</p>
</sec>
<sec id="Ch1.S3.SS2">
  <label>3.2</label><title>Statistical significance assessment</title>
      <p id="d2e757">Assessing statistical significance with independent tests for a 5 % threshold at each grid point highlights broad regions of significant cold temperature anomalies (Fig. <xref ref-type="fig" rid="F1"/>a). Even relatively weak anomalies, for example over the North Atlantic or over North America, are marked as statistically significant. Regardless of the lower temperature variability over oceans than over land, one may suspect that the significance of some signals is spurious.</p>

      <fig id="F1" specific-use="star"><label>Figure 1</label><caption><p id="d2e764">Composites of 2 m air temperature anomalies averaged over 60 d following SSWs. Statistical significance of cold anomalies is indicated by horizontal hatching for all grid points north of 20° N with respect to: <bold>(a)</bold> A local 5 % significance level using a one-sided Student's <inline-formula><mml:math id="M31" display="inline"><mml:mi>t</mml:mi></mml:math></inline-formula> test and <bold>(b)</bold> adjusted with the BH correction. Panel <bold>(c)</bold> shows the same hatching as in the red box in panel <bold>(b)</bold>, while panel <bold>(d)</bold> shows horizontal hatching for the BH-corrected significance with respect to only northern European grid points. The subtitles in panels <bold>(b)</bold> to <bold>(d)</bold> further indicate the BH-corrected <inline-formula><mml:math id="M32" display="inline"><mml:mi>p</mml:mi></mml:math></inline-formula> value as <inline-formula><mml:math id="M33" display="inline"><mml:mrow><mml:msup><mml:mi>p</mml:mi><mml:mo>*</mml:mo></mml:msup></mml:mrow></mml:math></inline-formula>. A zoomed-in map of northern Europe for the local 5 % significance level is found in Fig. <xref ref-type="fig" rid="FA2"/>.</p></caption>
          <graphic xlink:href="https://gmd.copernicus.org/articles/19/10917/2026/gmd-19-10917-2026-f01.png"/>

        </fig>

      <p id="d2e822">Applying the BH correction as recommended by <xref ref-type="bibr" rid="bib1.bibx48" id="text.33"/> with <inline-formula><mml:math id="M34" display="inline"><mml:mrow><mml:mi mathvariant="italic">α</mml:mi><mml:mo>=</mml:mo><mml:mn mathvariant="normal">0.1</mml:mn></mml:mrow></mml:math></inline-formula> to all grid points between 20 and 90° N reduces the number of points marked as statistically significant (Fig. <xref ref-type="fig" rid="F1"/>b). In agreement with earlier studies, northern Eurasia stands out as the region with the clearest signal of significant below-average temperatures following SSWs <xref ref-type="bibr" rid="bib1.bibx27 bib1.bibx10" id="paren.34"><named-content content-type="pre">e.g.,</named-content></xref>. In this case, the BH correction suppresses points with relatively weak anomalies, as these points could be linked to false rejections, while retaining the signal with the highest amplitude over northern Eurasia.</p>
      <p id="d2e848">We now restrict our attention to t2m anomalies over northern Europe (55–70° N, 4–32° E; Fig. <xref ref-type="fig" rid="F1"/>c), which has been highlighted as a region experiencing colder than usual temperatures following SSWs in a number of studies <xref ref-type="bibr" rid="bib1.bibx26 bib1.bibx27 bib1.bibx32" id="paren.35"><named-content content-type="pre">e.g.,</named-content></xref>. We note that this domain was identified partly through inspection of the hemispheric composite in Fig. <xref ref-type="fig" rid="F1"/>b, which is an a posteriori choice of the kind our study cautions against, and one that here serves as an example of the issue we aim to highlight. The behavior described below is moreover qualitatively consistent for other regions (e.g., Fig. <xref ref-type="fig" rid="FA1"/>). If the BH correction is applied only to grid points within northern Europe, the cold anomalies over nearly the entire region appear statistically significant (Fig. <xref ref-type="fig" rid="F1"/>d). This is in sharp contrast to the case where the BH correction is applied on the full hemispheric analysis domain (cf. Fig. <xref ref-type="fig" rid="F1"/>c and d). The number of hatched grid points in the former case is even larger than the uncorrected significance testing case (cf. Figs. <xref ref-type="fig" rid="F1"/>d to <xref ref-type="fig" rid="FA2"/>), which is linked to the pointwise significance level of 5 % being lower than the FDR level <inline-formula><mml:math id="M35" display="inline"><mml:mrow><mml:mi mathvariant="italic">α</mml:mi><mml:mo>=</mml:mo><mml:mn mathvariant="normal">0.1</mml:mn></mml:mrow></mml:math></inline-formula>. The discrepancy occurs even though the underlying temperature anomaly field is identical in all cases. It arises from the dependence of the BH-corrected threshold <inline-formula><mml:math id="M36" display="inline"><mml:mrow><mml:msubsup><mml:mi>p</mml:mi><mml:mi mathvariant="normal">FDR</mml:mi><mml:mo>*</mml:mo></mml:msubsup></mml:mrow></mml:math></inline-formula> on the rank ordering of <inline-formula><mml:math id="M37" display="inline"><mml:mi>p</mml:mi></mml:math></inline-formula> values. Restricting the analysis to a region dominated by low <inline-formula><mml:math id="M38" display="inline"><mml:mi>p</mml:mi></mml:math></inline-formula> values increases the adjusted threshold, allowing a larger fraction of points to satisfy the corrected value.</p>
      <p id="d2e911">To better understand this behavior, we examine the ranked <inline-formula><mml:math id="M39" display="inline"><mml:mi>p</mml:mi></mml:math></inline-formula> values and corresponding thresholds of negative t2m anomalies (Fig. <xref ref-type="fig" rid="F2"/>). When the BH correction is applied to all grid points between 20 and 90° N, the intersection defining <inline-formula><mml:math id="M40" display="inline"><mml:mrow><mml:msup><mml:mi>p</mml:mi><mml:mo>*</mml:mo></mml:msup></mml:mrow></mml:math></inline-formula> (Eq. <xref ref-type="disp-formula" rid="Ch1.E1"/>) is found at <inline-formula><mml:math id="M41" display="inline"><mml:mrow><mml:msup><mml:mi>p</mml:mi><mml:mo>*</mml:mo></mml:msup><mml:mo>≈</mml:mo><mml:mn mathvariant="normal">0.65</mml:mn></mml:mrow></mml:math></inline-formula> %, well below 5 % (Fig. <xref ref-type="fig" rid="F2"/>a). In contrast, when the BH correction is applied only to northern Europe, <inline-formula><mml:math id="M42" display="inline"><mml:mrow><mml:msup><mml:mi>p</mml:mi><mml:mo>*</mml:mo></mml:msup><mml:mo>≈</mml:mo><mml:mn mathvariant="normal">7.97</mml:mn></mml:mrow></mml:math></inline-formula> % is an order of magnitude higher (Fig. <xref ref-type="fig" rid="F2"/>b). This difference arises because the <inline-formula><mml:math id="M43" display="inline"><mml:mi>p</mml:mi></mml:math></inline-formula> values over northern Europe are generally small, increasing the fraction of points below the <inline-formula><mml:math id="M44" display="inline"><mml:mrow><mml:mstyle displaystyle="false"><mml:mfrac style="text"><mml:mi>i</mml:mi><mml:mi>N</mml:mi></mml:mfrac></mml:mstyle><mml:mi mathvariant="italic">α</mml:mi></mml:mrow></mml:math></inline-formula> line when fewer grid points are considered. While the absolute distribution of <inline-formula><mml:math id="M45" display="inline"><mml:mi>p</mml:mi></mml:math></inline-formula> values over northern Europe remains unchanged, their relative ranks within the smaller test family differ, resulting in a substantially more permissive threshold.</p>

      <fig id="F2" specific-use="star"><label>Figure 2</label><caption><p id="d2e1001">Ranked <inline-formula><mml:math id="M46" display="inline"><mml:mi>p</mml:mi></mml:math></inline-formula> values (colored line) and the BH-corrected threshold (black line) for: <bold>(a)</bold> all grid points north of 20° N (Fig. <xref ref-type="fig" rid="F1"/>b) and <bold>(b)</bold> northern Europe only (Fig. <xref ref-type="fig" rid="F1"/>d). In both panels, the red dotted line indicates the significance threshold obtained from the respective FDR correction, and the black dashed line indicates the 5 % significance level. Histograms to the right of each panel show the distribution of <inline-formula><mml:math id="M47" display="inline"><mml:mi>p</mml:mi></mml:math></inline-formula> values. Note that both panels only show <inline-formula><mml:math id="M48" display="inline"><mml:mi>p</mml:mi></mml:math></inline-formula> values below 0.14 for improved readability. Figure <xref ref-type="fig" rid="FA3"/> shows all <inline-formula><mml:math id="M49" display="inline"><mml:mi>p</mml:mi></mml:math></inline-formula> values, including those above 0.14.</p></caption>
          <graphic xlink:href="https://gmd.copernicus.org/articles/19/10917/2026/gmd-19-10917-2026-f02.png"/>

        </fig>

</sec>
</sec>
<sec id="Ch1.S4">
  <label>4</label><title>The domain dependence of BH correction</title>
      <p id="d2e1060">The example above illustrates that the BH correction as commonly implemented in geoscience applications can yield less restrictive thresholds than the uncorrected statistical test when applied to spatial domains dominated by coherent signals. This behavior follows directly from the choice of <inline-formula><mml:math id="M50" display="inline"><mml:mrow><mml:mi mathvariant="italic">α</mml:mi><mml:mo>=</mml:mo><mml:mn mathvariant="normal">0.1</mml:mn></mml:mrow></mml:math></inline-formula> being larger than the pointwise significance level of 5 % and from the way the BH correction is defined <xref ref-type="bibr" rid="bib1.bibx5" id="paren.36"/>, which identifies the critical <inline-formula><mml:math id="M51" display="inline"><mml:mi>p</mml:mi></mml:math></inline-formula> value threshold <inline-formula><mml:math id="M52" display="inline"><mml:mrow><mml:msup><mml:mi>p</mml:mi><mml:mo>*</mml:mo></mml:msup></mml:mrow></mml:math></inline-formula> as the largest <inline-formula><mml:math id="M53" display="inline"><mml:mi>p</mml:mi></mml:math></inline-formula> value satisfying Eq. (<xref ref-type="disp-formula" rid="Ch1.E1"/>).</p>
      <p id="d2e1106">When only a small fraction of <inline-formula><mml:math id="M54" display="inline"><mml:mi>p</mml:mi></mml:math></inline-formula> values in the full set are sufficiently low, the BH criterion is satisfied only at relatively early ranks in the ordered <inline-formula><mml:math id="M55" display="inline"><mml:mi>p</mml:mi></mml:math></inline-formula> value distribution, resulting in a comparatively strict (low) threshold. Conversely, when the analysis is restricted to a domain dominated by sufficiently small <inline-formula><mml:math id="M56" display="inline"><mml:mi>p</mml:mi></mml:math></inline-formula> values, the ratio <inline-formula><mml:math id="M57" display="inline"><mml:mstyle displaystyle="false"><mml:mfrac style="text"><mml:mi>i</mml:mi><mml:mi>N</mml:mi></mml:mfrac></mml:mstyle></mml:math></inline-formula> changes so that the intersection occurs at a higher <inline-formula><mml:math id="M58" display="inline"><mml:mi>p</mml:mi></mml:math></inline-formula> value, effectively relaxing the threshold for declaring significance. In other words, the same underlying field can yield an apparently higher fraction of grid points identified as statistically significant on a reduced spatial domain than both the BH correction applied to a larger domain and the uncorrected case.</p>
      <p id="d2e1148">Crucially, this behavior reflects a fundamental tension between a set-level error criterion and the local interpretation encouraged by significance maps. Coherent anomalies, such as large-scale temperature responses, generate clusters of similar <inline-formula><mml:math id="M59" display="inline"><mml:mi>p</mml:mi></mml:math></inline-formula> values that alter the rank structure on which the BH correction operates. This behavior only applies to domains with a relatively large proportion of small <inline-formula><mml:math id="M60" display="inline"><mml:mi>p</mml:mi></mml:math></inline-formula> values and when <inline-formula><mml:math id="M61" display="inline"><mml:mrow><mml:msub><mml:mi mathvariant="italic">α</mml:mi><mml:mi mathvariant="normal">FDR</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> is larger than the nominal pointwise significance level. If either of these conditions does not hold, the BH correction has the more intuitive effect of reducing or leaving unchanged the significance threshold <inline-formula><mml:math id="M62" display="inline"><mml:mrow><mml:msup><mml:mi>p</mml:mi><mml:mo>*</mml:mo></mml:msup></mml:mrow></mml:math></inline-formula>. While this FDR control method remains valid under many forms of spatial dependence, its sensitivity to rank ordering implies that spatial context and domain choice can strongly influence the resulting significance threshold, and hence the apparent robustness and spatial extent of detected signals.</p>
</sec>
<sec id="Ch1.S5">
  <label>5</label><title>Numerical experiment</title>
      <p id="d2e1196">To complement the meteorological example and investigate the sensitivity of the BH-corrected threshold to key properties of the tested field in a controlled setting, we conduct a numerical experiment with synthetic data. We mainly consider the role of the fraction of the true signal <inline-formula><mml:math id="M63" display="inline"><mml:mrow><mml:msub><mml:mi>f</mml:mi><mml:mi mathvariant="normal">s</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula>, but also include sensitivity analysis to other parameters such as the relationship between the domain size and the correlation length in the data, and the signal-to-noise ratio (SNR).</p>
<sec id="Ch1.S5.SS1">
  <label>5.1</label><title>Experimental design</title>
      <p id="d2e1217">We generate synthetic fields on a square grid with <inline-formula><mml:math id="M64" display="inline"><mml:mrow><mml:msub><mml:mi>n</mml:mi><mml:mi mathvariant="normal">g</mml:mi></mml:msub><mml:mo>×</mml:mo><mml:msub><mml:mi>n</mml:mi><mml:mi mathvariant="normal">g</mml:mi></mml:msub><mml:mo>=</mml:mo><mml:msub><mml:mi>N</mml:mi><mml:mi mathvariant="normal">g</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> grid points. The background field has value 0. Similar to <xref ref-type="bibr" rid="bib1.bibx48" id="text.37"/>, the signal consists of a circular patch of grid points with value 1, occupying a fraction <inline-formula><mml:math id="M65" display="inline"><mml:mrow><mml:msub><mml:mi>f</mml:mi><mml:mi mathvariant="normal">s</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> of the total domain, defined by the radius <inline-formula><mml:math id="M66" display="inline"><mml:mrow><mml:mi>r</mml:mi><mml:mo>=</mml:mo><mml:msqrt><mml:mrow><mml:msub><mml:mi>f</mml:mi><mml:mi mathvariant="normal">s</mml:mi></mml:msub><mml:msub><mml:mi>N</mml:mi><mml:mi mathvariant="normal">g</mml:mi></mml:msub><mml:mo>/</mml:mo><mml:mi mathvariant="italic">π</mml:mi></mml:mrow></mml:msqrt></mml:mrow></mml:math></inline-formula>. A circular patch provides a geometrically simple and parameter-free signal region that can be scaled through <inline-formula><mml:math id="M67" display="inline"><mml:mrow><mml:msub><mml:mi>f</mml:mi><mml:mi mathvariant="normal">s</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula>, making it straightforward to control the fraction of the true signal. Values of <inline-formula><mml:math id="M68" display="inline"><mml:mrow><mml:msub><mml:mi>f</mml:mi><mml:mi mathvariant="normal">s</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> ranging from 0.01 to 0.9 (Table <xref ref-type="table" rid="T1"/>) span the full range from a sparse, localized signal broadly analogous to the hemispheric case in Sect. <xref ref-type="sec" rid="Ch1.S3.SS2"/> where only a small fraction of grid points exhibit relatively strong anomalies, to a spatially dominant response analogous to the northern Europe subdomain where most grid points carry relatively small <inline-formula><mml:math id="M69" display="inline"><mml:mi>p</mml:mi></mml:math></inline-formula> values.</p>

<table-wrap id="T1" specific-use="star"><label>Table 1</label><caption><p id="d2e1322">Summary of parameter values used in the numerical experiment.</p></caption><oasis:table frame="topbot"><oasis:tgroup cols="3">
     <oasis:colspec colnum="1" colname="col1" align="left"/>
     <oasis:colspec colnum="2" colname="col2" align="left"/>
     <oasis:colspec colnum="3" colname="col3" align="left"/>
     <oasis:thead>
       <oasis:row rowsep="1">
         <oasis:entry colname="col1">Parameter</oasis:entry>
         <oasis:entry colname="col2">Values</oasis:entry>
         <oasis:entry colname="col3">Physical interpretation</oasis:entry>
       </oasis:row>
     </oasis:thead>
     <oasis:tbody>
       <oasis:row>
         <oasis:entry colname="col1"><inline-formula><mml:math id="M70" display="inline"><mml:mrow><mml:msub><mml:mi>n</mml:mi><mml:mi mathvariant="normal">g</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula></oasis:entry>
         <oasis:entry colname="col2">30, 50, 100</oasis:entry>
         <oasis:entry colname="col3">Grid size; <inline-formula><mml:math id="M71" display="inline"><mml:mrow><mml:msub><mml:mi>N</mml:mi><mml:mi mathvariant="normal">g</mml:mi></mml:msub><mml:mo>=</mml:mo><mml:msubsup><mml:mi>n</mml:mi><mml:mi mathvariant="normal">g</mml:mi><mml:mn mathvariant="normal">2</mml:mn></mml:msubsup></mml:mrow></mml:math></inline-formula> total grid points</oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1"><inline-formula><mml:math id="M72" display="inline"><mml:mrow><mml:msub><mml:mi>f</mml:mi><mml:mi mathvariant="normal">s</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula></oasis:entry>
         <oasis:entry colname="col2">0.01, 0.1, 0.25, 0.5, 0.9</oasis:entry>
         <oasis:entry colname="col3">Fraction of grid points with true signal</oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1"><inline-formula><mml:math id="M73" display="inline"><mml:mi mathvariant="italic">ν</mml:mi></mml:math></inline-formula></oasis:entry>
         <oasis:entry colname="col2">0.1, 0.5, 1, 1.5, 2, 5, 10, 50</oasis:entry>
         <oasis:entry colname="col3">Noise wavenumber; <inline-formula><mml:math id="M74" display="inline"><mml:mrow><mml:mi mathvariant="italic">λ</mml:mi><mml:mo>=</mml:mo><mml:msub><mml:mi>n</mml:mi><mml:mi mathvariant="normal">g</mml:mi></mml:msub><mml:mo>/</mml:mo><mml:mo>(</mml:mo><mml:mn mathvariant="normal">2</mml:mn><mml:mi mathvariant="italic">π</mml:mi><mml:mi mathvariant="italic">ν</mml:mi><mml:mo>)</mml:mo></mml:mrow></mml:math></inline-formula></oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1">SNR</oasis:entry>
         <oasis:entry colname="col2">0.5, 0.75, 1, 2</oasis:entry>
         <oasis:entry colname="col3">Signal-to-noise ratio; <inline-formula><mml:math id="M75" display="inline"><mml:mrow><mml:mi mathvariant="italic">σ</mml:mi><mml:mo>=</mml:mo><mml:mn mathvariant="normal">1</mml:mn><mml:mo>/</mml:mo><mml:mi mathvariant="normal">SNR</mml:mi></mml:mrow></mml:math></inline-formula></oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1"><inline-formula><mml:math id="M76" display="inline"><mml:mi>M</mml:mi></mml:math></inline-formula></oasis:entry>
         <oasis:entry colname="col2">20, 1000</oasis:entry>
         <oasis:entry colname="col3">Sample size: number of simulated fields</oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1"><inline-formula><mml:math id="M77" display="inline"><mml:mi mathvariant="italic">α</mml:mi></mml:math></inline-formula></oasis:entry>
         <oasis:entry colname="col2">0.02, 0.05, 0.10, 0.20</oasis:entry>
         <oasis:entry colname="col3">FDR level</oasis:entry>
       </oasis:row>
     </oasis:tbody>
   </oasis:tgroup></oasis:table></table-wrap>

      <p id="d2e1521">To this signal we add a spatially correlated Gaussian noise field with standard deviation <inline-formula><mml:math id="M78" display="inline"><mml:mrow><mml:mi mathvariant="italic">σ</mml:mi><mml:mo>=</mml:mo><mml:mn mathvariant="normal">1</mml:mn><mml:mo>/</mml:mo><mml:mi mathvariant="normal">SNR</mml:mi></mml:mrow></mml:math></inline-formula>, so that larger SNR values correspond to stronger signals relative to the background variability. Gaussian noise is a natural and widely used choice for synthetic geophysical fields, as many atmospheric variables are approximately normally distributed after removing the seasonal cycle <xref ref-type="bibr" rid="bib1.bibx43 bib1.bibx47" id="paren.38"/>. It is furthermore consistent with the distributional assumptions underlying the Student's <inline-formula><mml:math id="M79" display="inline"><mml:mi>t</mml:mi></mml:math></inline-formula> test used to compute local <inline-formula><mml:math id="M80" display="inline"><mml:mi>p</mml:mi></mml:math></inline-formula> values, and provides a tractable, analytically well-understood null field. Spatial correlation is introduced by generating white noise in spectral space and multiplying each Fourier mode by a spectral filter before transforming back to physical space, yielding a field with power spectral density <inline-formula><mml:math id="M81" display="inline"><mml:mrow><mml:mi>S</mml:mi><mml:mo>(</mml:mo><mml:mi mathvariant="italic">ω</mml:mi><mml:mo>)</mml:mo><mml:mo>=</mml:mo><mml:mi>exp⁡</mml:mi><mml:mo>(</mml:mo><mml:mo>-</mml:mo><mml:mstyle displaystyle="false"><mml:mfrac style="text"><mml:mrow><mml:mo>|</mml:mo><mml:mi mathvariant="italic">ω</mml:mi><mml:mo>|</mml:mo><mml:msub><mml:mi>n</mml:mi><mml:mi mathvariant="normal">g</mml:mi></mml:msub></mml:mrow><mml:mi mathvariant="italic">ν</mml:mi></mml:mfrac></mml:mstyle><mml:mo>)</mml:mo></mml:mrow></mml:math></inline-formula>, where <inline-formula><mml:math id="M82" display="inline"><mml:mrow><mml:mo>|</mml:mo><mml:mi mathvariant="italic">ω</mml:mi><mml:mo>|</mml:mo><mml:mo>=</mml:mo><mml:msqrt><mml:mrow><mml:msubsup><mml:mi mathvariant="italic">ω</mml:mi><mml:mi>x</mml:mi><mml:mn mathvariant="normal">2</mml:mn></mml:msubsup><mml:mo>+</mml:mo><mml:msubsup><mml:mi mathvariant="italic">ω</mml:mi><mml:mi>y</mml:mi><mml:mn mathvariant="normal">2</mml:mn></mml:msubsup></mml:mrow></mml:msqrt></mml:mrow></mml:math></inline-formula> is the radial wavenumber, and <inline-formula><mml:math id="M83" display="inline"><mml:mi mathvariant="italic">ν</mml:mi></mml:math></inline-formula> is the noise wavenumber, corresponding to a spatial autocorrelation length of approximately <inline-formula><mml:math id="M84" display="inline"><mml:mrow><mml:mi mathvariant="italic">λ</mml:mi><mml:mo>≈</mml:mo><mml:mstyle displaystyle="false"><mml:mfrac style="text"><mml:mrow><mml:msub><mml:mi>n</mml:mi><mml:mi mathvariant="normal">g</mml:mi></mml:msub></mml:mrow><mml:mrow><mml:mn mathvariant="normal">2</mml:mn><mml:mi mathvariant="italic">π</mml:mi><mml:mi mathvariant="italic">ν</mml:mi></mml:mrow></mml:mfrac></mml:mstyle></mml:mrow></mml:math></inline-formula>. Implementing the correlation structure in the spectral space provides exact and independent control over the power at each wavenumber, which is not achievable with smoothing kernels in the physical space, where the resulting spectral properties are determined implicitly by the kernel shape and width. The noise wavenumber <inline-formula><mml:math id="M85" display="inline"><mml:mi mathvariant="italic">ν</mml:mi></mml:math></inline-formula> used in the numerical experiment spans from large-scale structures (<inline-formula><mml:math id="M86" display="inline"><mml:mrow><mml:mi mathvariant="italic">ν</mml:mi><mml:mo>=</mml:mo><mml:mn mathvariant="normal">0.1</mml:mn></mml:mrow></mml:math></inline-formula>, <inline-formula><mml:math id="M87" display="inline"><mml:mrow><mml:mi mathvariant="italic">λ</mml:mi><mml:mo>≈</mml:mo><mml:mn mathvariant="normal">1.6</mml:mn><mml:msub><mml:mi>n</mml:mi><mml:mi mathvariant="normal">g</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula>) to near-white noise (<inline-formula><mml:math id="M88" display="inline"><mml:mrow><mml:mi mathvariant="italic">ν</mml:mi><mml:mo>=</mml:mo><mml:mn mathvariant="normal">50</mml:mn></mml:mrow></mml:math></inline-formula>, <inline-formula><mml:math id="M89" display="inline"><mml:mrow><mml:mi mathvariant="italic">λ</mml:mi><mml:mo>≈</mml:mo><mml:mn mathvariant="normal">0.003</mml:mn><mml:msub><mml:mi>n</mml:mi><mml:mi mathvariant="normal">g</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula>), covering the range from planetary-scale to grid-scale spatial dependence typical of geophysical fields.</p>
      <p id="d2e1729">The parameter values explored in the experiment are summarized in Table <xref ref-type="table" rid="T1"/>. For each of the 480 parameter combinations of <inline-formula><mml:math id="M90" display="inline"><mml:mrow><mml:msub><mml:mi>N</mml:mi><mml:mi mathvariant="normal">g</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula>, <inline-formula><mml:math id="M91" display="inline"><mml:mrow><mml:msub><mml:mi>f</mml:mi><mml:mi mathvariant="normal">s</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula>, <inline-formula><mml:math id="M92" display="inline"><mml:mi mathvariant="italic">ν</mml:mi></mml:math></inline-formula> and SNR, we generate <inline-formula><mml:math id="M93" display="inline"><mml:mi>M</mml:mi></mml:math></inline-formula> independent realizations of the field, each consisting of the same deterministic signal and a newly drawn noise field. These realizations take the role of the individual events or ensemble members in a real application, such as the 26 SSW events in Sect. <xref ref-type="sec" rid="Ch1.S3"/>. At each grid point, <inline-formula><mml:math id="M94" display="inline"><mml:mi>M</mml:mi></mml:math></inline-formula> simulated values form the sample entering the significance test described below.</p>
      <p id="d2e1780">We consider two sample sizes for <inline-formula><mml:math id="M95" display="inline"><mml:mi>M</mml:mi></mml:math></inline-formula>. For <inline-formula><mml:math id="M96" display="inline"><mml:mrow><mml:mi>M</mml:mi><mml:mo>=</mml:mo><mml:mn mathvariant="normal">1000</mml:mn></mml:mrow></mml:math></inline-formula> (all realizations), the signal is separated from the noise by a wide margin at the signal grid points (for <inline-formula><mml:math id="M97" display="inline"><mml:mrow><mml:mi mathvariant="normal">SNR</mml:mi><mml:mo>=</mml:mo><mml:mn mathvariant="normal">0.5</mml:mn></mml:mrow></mml:math></inline-formula>, the expected <inline-formula><mml:math id="M98" display="inline"><mml:mi>t</mml:mi></mml:math></inline-formula>-statistic is about 16), so that the dependence of the BH-corrected threshold on the signal fraction can be studied independently of limited detection power. A single test is performed per parameter combination, so the spread across combinations reported below reflects parameter variability. For <inline-formula><mml:math id="M99" display="inline"><mml:mrow><mml:mi>M</mml:mi><mml:mo>=</mml:mo><mml:mn mathvariant="normal">20</mml:mn></mml:mrow></mml:math></inline-formula>, closer to typical sample sizes in geoscientific composite and ensemble analyses, detection power is limited. To quantify the sampling variability in this low-power regime (<inline-formula><mml:math id="M100" display="inline"><mml:mrow><mml:mi mathvariant="normal">SNR</mml:mi><mml:mo>=</mml:mo><mml:mn mathvariant="normal">0.5</mml:mn></mml:mrow></mml:math></inline-formula>, <inline-formula><mml:math id="M101" display="inline"><mml:mrow><mml:mi>M</mml:mi><mml:mo>=</mml:mo><mml:mn mathvariant="normal">20</mml:mn></mml:mrow></mml:math></inline-formula>), the test is repeated for 100 random subsets of 20 realizations, drawn without replacement from the 1000 realizations, for every combination of the remaining parameters (<inline-formula><mml:math id="M102" display="inline"><mml:mrow><mml:msub><mml:mi>N</mml:mi><mml:mi mathvariant="normal">g</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula>, <inline-formula><mml:math id="M103" display="inline"><mml:mrow><mml:msub><mml:mi>f</mml:mi><mml:mi mathvariant="normal">s</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula>, <inline-formula><mml:math id="M104" display="inline"><mml:mi mathvariant="italic">ν</mml:mi></mml:math></inline-formula>).</p>
      <p id="d2e1887">At each grid point, we test the null hypothesis that the mean over the <inline-formula><mml:math id="M105" display="inline"><mml:mi>M</mml:mi></mml:math></inline-formula> simulated values is zero against the alternative that it is larger than zero, using the one-sided, one-sample Student's <inline-formula><mml:math id="M106" display="inline"><mml:mi>t</mml:mi></mml:math></inline-formula>-test of Eq. (<xref ref-type="disp-formula" rid="Ch1.E2"/>) with <inline-formula><mml:math id="M107" display="inline"><mml:mrow><mml:mi>n</mml:mi><mml:mo>=</mml:mo><mml:mi>M</mml:mi></mml:mrow></mml:math></inline-formula>, as in the SSW example (Sect. <xref ref-type="sec" rid="Ch1.S3"/>). As the simulated fields are independent and the noise is Gaussian, the assumptions of the <inline-formula><mml:math id="M108" display="inline"><mml:mi>t</mml:mi></mml:math></inline-formula>-test hold at each grid point by construction. The spatial correlation of the noise affects only the dependence between the tests. The resulting <inline-formula><mml:math id="M109" display="inline"><mml:mrow><mml:msubsup><mml:mi>N</mml:mi><mml:mi mathvariant="normal">g</mml:mi><mml:mn mathvariant="normal">2</mml:mn></mml:msubsup></mml:mrow></mml:math></inline-formula> <inline-formula><mml:math id="M110" display="inline"><mml:mi>p</mml:mi></mml:math></inline-formula> values are corrected with the BH procedure (Eq. <xref ref-type="disp-formula" rid="Ch1.E1"/>) for four different FDR levels, spanning values from below to above the standard recommendation of <xref ref-type="bibr" rid="bib1.bibx48" id="text.39"/>.</p>
      <p id="d2e1953">To evaluate the behavior of the BH correction across this parameter space, we record for each application of the test and BH correction (i.e., each parameter combination for <inline-formula><mml:math id="M111" display="inline"><mml:mrow><mml:mi>M</mml:mi><mml:mo>=</mml:mo><mml:mn mathvariant="normal">1000</mml:mn></mml:mrow></mml:math></inline-formula> and each of the 100 subsets for <inline-formula><mml:math id="M112" display="inline"><mml:mrow><mml:mi>M</mml:mi><mml:mo>=</mml:mo><mml:mn mathvariant="normal">20</mml:mn></mml:mrow></mml:math></inline-formula>) the identified threshold <inline-formula><mml:math id="M113" display="inline"><mml:mrow><mml:msup><mml:mi>p</mml:mi><mml:mo>*</mml:mo></mml:msup></mml:mrow></mml:math></inline-formula> (obtained as in Sect. <xref ref-type="sec" rid="Ch1.S3"/>), and the number of true signal grid points correctly (TP) and background grid points incorrectly (FP) flagged as significant. From these we compute the realized false discovery proportion,

            <disp-formula id="Ch1.E3" content-type="numbered"><label>3</label><mml:math id="M114" display="block"><mml:mrow><mml:mi mathvariant="normal">FDP</mml:mi><mml:mo>=</mml:mo><mml:mstyle displaystyle="true"><mml:mfrac style="display"><mml:mi mathvariant="normal">FP</mml:mi><mml:mrow><mml:mi mathvariant="normal">FP</mml:mi><mml:mo>+</mml:mo><mml:mi mathvariant="normal">TP</mml:mi></mml:mrow></mml:mfrac></mml:mstyle><mml:mo>,</mml:mo></mml:mrow></mml:math></disp-formula>

          and power,

            <disp-formula id="Ch1.E4" content-type="numbered"><label>4</label><mml:math id="M115" display="block"><mml:mrow><mml:mi mathvariant="normal">Power</mml:mi><mml:mo>=</mml:mo><mml:mstyle displaystyle="true"><mml:mfrac style="display"><mml:mi mathvariant="normal">TP</mml:mi><mml:mrow><mml:msub><mml:mi>N</mml:mi><mml:mi mathvariant="normal">sig</mml:mi></mml:msub></mml:mrow></mml:mfrac></mml:mstyle><mml:mo>,</mml:mo></mml:mrow></mml:math></disp-formula>

          where <inline-formula><mml:math id="M116" display="inline"><mml:mrow><mml:msub><mml:mi>N</mml:mi><mml:mi mathvariant="normal">sig</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> is the total number of true signal grid points in the domain.</p>
</sec>
<sec id="Ch1.S5.SS2">
  <label>5.2</label><title>Numerical results</title>
      <p id="d2e2058">We now summarize the sensitivity of the three diagnostics as a function of signal fraction <inline-formula><mml:math id="M117" display="inline"><mml:mrow><mml:msub><mml:mi>f</mml:mi><mml:mi mathvariant="normal">s</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula>.</p>
      <p id="d2e2072">The BH-corrected significance threshold <inline-formula><mml:math id="M118" display="inline"><mml:mrow><mml:msup><mml:mi>p</mml:mi><mml:mo>*</mml:mo></mml:msup></mml:mrow></mml:math></inline-formula> increases monotonically with <inline-formula><mml:math id="M119" display="inline"><mml:mrow><mml:msub><mml:mi>f</mml:mi><mml:mi mathvariant="normal">s</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> and is higher for larger <inline-formula><mml:math id="M120" display="inline"><mml:mi mathvariant="italic">α</mml:mi></mml:math></inline-formula> (Fig. <xref ref-type="fig" rid="F3"/>a). This directly reproduces in a controlled setting the mechanism illustrated by the meteorological example: as the fraction of truly significant grid points grows, the rank-based threshold becomes more permissive. For <inline-formula><mml:math id="M121" display="inline"><mml:mrow><mml:msub><mml:mi>f</mml:mi><mml:mi mathvariant="normal">s</mml:mi></mml:msub><mml:mo>≲</mml:mo><mml:mn mathvariant="normal">0.1</mml:mn></mml:mrow></mml:math></inline-formula>, <inline-formula><mml:math id="M122" display="inline"><mml:mrow><mml:msup><mml:mi>p</mml:mi><mml:mo>*</mml:mo></mml:msup></mml:mrow></mml:math></inline-formula> remains well below the 5 % local significance level across all <inline-formula><mml:math id="M123" display="inline"><mml:mi mathvariant="italic">α</mml:mi></mml:math></inline-formula> choices, while for <inline-formula><mml:math id="M124" display="inline"><mml:mrow><mml:msub><mml:mi>f</mml:mi><mml:mi mathvariant="normal">s</mml:mi></mml:msub><mml:mo>&gt;</mml:mo><mml:mn mathvariant="normal">0.1</mml:mn></mml:mrow></mml:math></inline-formula> the BH-corrected threshold <inline-formula><mml:math id="M125" display="inline"><mml:mrow><mml:msup><mml:mi>p</mml:mi><mml:mo>*</mml:mo></mml:msup></mml:mrow></mml:math></inline-formula> can substantially exceed the 5% level, particularly for larger <inline-formula><mml:math id="M126" display="inline"><mml:mi mathvariant="italic">α</mml:mi></mml:math></inline-formula>. For the low-power regime, BH-corrected significance thresholds are consistently lower than in the full parameter space but follow the same dependence on <inline-formula><mml:math id="M127" display="inline"><mml:mrow><mml:msub><mml:mi>f</mml:mi><mml:mi mathvariant="normal">s</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula>.</p>

      <fig id="F3" specific-use="star"><label>Figure 3</label><caption><p id="d2e2186">Results of the numerical experiment as a function of signal fraction <inline-formula><mml:math id="M128" display="inline"><mml:mrow><mml:msub><mml:mi>f</mml:mi><mml:mi mathvariant="normal">s</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula>. Panels <bold>(a)</bold>–<bold>(c)</bold> show the median (markers) and interquartile range (error bars) of the BH-corrected threshold <inline-formula><mml:math id="M129" display="inline"><mml:mrow><mml:msup><mml:mi>p</mml:mi><mml:mo>*</mml:mo></mml:msup></mml:mrow></mml:math></inline-formula>, the realized false discovery proportion (FDP), and power, respectively, aggregated across all 480 parameter combinations with <inline-formula><mml:math id="M130" display="inline"><mml:mrow><mml:mi>M</mml:mi><mml:mo>=</mml:mo><mml:mn mathvariant="normal">1000</mml:mn></mml:mrow></mml:math></inline-formula> (saturated lines) and across the low-power regime (<inline-formula><mml:math id="M131" display="inline"><mml:mrow><mml:mi mathvariant="normal">SNR</mml:mi><mml:mo>=</mml:mo><mml:mn mathvariant="normal">0.5</mml:mn></mml:mrow></mml:math></inline-formula>, <inline-formula><mml:math id="M132" display="inline"><mml:mrow><mml:mi>M</mml:mi><mml:mo>=</mml:mo><mml:mn mathvariant="normal">20</mml:mn></mml:mrow></mml:math></inline-formula>; pale). Panel <bold>(d)</bold> shows the FDP–Power trade-off as a scatter plot, with markers indicating the median and error bars the interquartile range across all combinations (solid) and the low-power regime (pale). Colors indicate the choice of FDR level <inline-formula><mml:math id="M133" display="inline"><mml:mi mathvariant="italic">α</mml:mi></mml:math></inline-formula> as provided in the legend in panel <bold>(a)</bold>. In panel <bold>(c)</bold>, the median power across all parameter combinations is shown as a single gray line, as all choices of <inline-formula><mml:math id="M134" display="inline"><mml:mi mathvariant="italic">α</mml:mi></mml:math></inline-formula> yield near-identical power close to unity.</p></caption>
          <graphic xlink:href="https://gmd.copernicus.org/articles/19/10917/2026/gmd-19-10917-2026-f03.png"/>

        </fig>

      <p id="d2e2285">FDP is approximately constant across low-to-moderate signal fractions and decreases for larger signal fractions (Fig. <xref ref-type="fig" rid="F3"/>b), reflecting the increasing dominance of true positives among all rejections as the signal covers a larger portion of the domain. This pattern is consistent across both the full parameter space and the low-power regime (<inline-formula><mml:math id="M135" display="inline"><mml:mrow><mml:mi mathvariant="normal">SNR</mml:mi><mml:mo>=</mml:mo><mml:mn mathvariant="normal">0.5</mml:mn></mml:mrow></mml:math></inline-formula>, <inline-formula><mml:math id="M136" display="inline"><mml:mrow><mml:mi>M</mml:mi><mml:mo>=</mml:mo><mml:mn mathvariant="normal">20</mml:mn></mml:mrow></mml:math></inline-formula>), confirming that it is a property of the BH correction method rather than an artifact of specific parameter choices.</p>
      <p id="d2e2314">Power is close to unity in most parameter combinations (Fig. <xref ref-type="fig" rid="F3"/>c), indicating that for typical settings all choices of <inline-formula><mml:math id="M137" display="inline"><mml:mi mathvariant="italic">α</mml:mi></mml:math></inline-formula> achieve near-perfect detection. This is expected for <inline-formula><mml:math id="M138" display="inline"><mml:mrow><mml:mi>M</mml:mi><mml:mo>=</mml:mo><mml:mn mathvariant="normal">1000</mml:mn></mml:mrow></mml:math></inline-formula>, where signal grid points yield <inline-formula><mml:math id="M139" display="inline"><mml:mi>p</mml:mi></mml:math></inline-formula> values far below the BH threshold for all <inline-formula><mml:math id="M140" display="inline"><mml:mi mathvariant="italic">α</mml:mi></mml:math></inline-formula> (Sect. <xref ref-type="sec" rid="Ch1.S5.SS1"/>). Furthermore, the high power may be related to the sharp signal boundary in the experimental setup, which yields in several cases well-separated <inline-formula><mml:math id="M141" display="inline"><mml:mi>p</mml:mi></mml:math></inline-formula> value distributions for signal and background grid points. In the low-power regime, however, power rises from near zero at <inline-formula><mml:math id="M142" display="inline"><mml:mrow><mml:msub><mml:mi>f</mml:mi><mml:mi mathvariant="normal">s</mml:mi></mml:msub><mml:mo>=</mml:mo><mml:mn mathvariant="normal">0.01</mml:mn></mml:mrow></mml:math></inline-formula> to between about 0.3 (<inline-formula><mml:math id="M143" display="inline"><mml:mrow><mml:mi mathvariant="italic">α</mml:mi><mml:mo>=</mml:mo><mml:mn mathvariant="normal">0.02</mml:mn></mml:mrow></mml:math></inline-formula>) and 0.9 (<inline-formula><mml:math id="M144" display="inline"><mml:mrow><mml:mi mathvariant="italic">α</mml:mi><mml:mo>=</mml:mo><mml:mn mathvariant="normal">0.2</mml:mn></mml:mrow></mml:math></inline-formula>) at <inline-formula><mml:math id="M145" display="inline"><mml:mrow><mml:msub><mml:mi>f</mml:mi><mml:mi mathvariant="normal">s</mml:mi></mml:msub><mml:mo>=</mml:mo><mml:mn mathvariant="normal">0.9</mml:mn></mml:mrow></mml:math></inline-formula>, and is higher for larger <inline-formula><mml:math id="M146" display="inline"><mml:mi mathvariant="italic">α</mml:mi></mml:math></inline-formula> (Fig. <xref ref-type="fig" rid="F3"/>c). A more permissive FDR level raises <inline-formula><mml:math id="M147" display="inline"><mml:mrow><mml:msup><mml:mi>p</mml:mi><mml:mo>*</mml:mo></mml:msup></mml:mrow></mml:math></inline-formula> and thus admits more rejections, which increases power but also the realized FDP. When the signal is weak compared to the noise and the sample is small, the <inline-formula><mml:math id="M148" display="inline"><mml:mi>p</mml:mi></mml:math></inline-formula> values of signal and background grid points overlap, so that detecting more signal points inevitably comes at the cost of more false positives.</p>
      <p id="d2e2444">The FDP–power trade-off (Fig. <xref ref-type="fig" rid="F3"/>d) is consistent with the results for FDP and power discussed before: in the case of <inline-formula><mml:math id="M149" display="inline"><mml:mrow><mml:mi>M</mml:mi><mml:mo>=</mml:mo><mml:mn mathvariant="normal">1000</mml:mn></mml:mrow></mml:math></inline-formula>, all <inline-formula><mml:math id="M150" display="inline"><mml:mi mathvariant="italic">α</mml:mi></mml:math></inline-formula> choices achieve near-perfect power and the meaningful distinction between them lies in FDP control; in the low-power regime, a genuine trade-off emerges.</p>
      <p id="d2e2468">Overall, <inline-formula><mml:math id="M151" display="inline"><mml:mrow><mml:msub><mml:mi>f</mml:mi><mml:mi mathvariant="normal">s</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> emerges as the parameter most strongly controlling the behavior of the BH correction. Sensitivity analysis of <inline-formula><mml:math id="M152" display="inline"><mml:mrow><mml:msup><mml:mi>p</mml:mi><mml:mo>*</mml:mo></mml:msup></mml:mrow></mml:math></inline-formula>, FDP and power to the remaining parameters (<inline-formula><mml:math id="M153" display="inline"><mml:mrow><mml:msub><mml:mi>N</mml:mi><mml:mi mathvariant="normal">g</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula>, <inline-formula><mml:math id="M154" display="inline"><mml:mi mathvariant="italic">ν</mml:mi></mml:math></inline-formula>, SNR, <inline-formula><mml:math id="M155" display="inline"><mml:mi>M</mml:mi></mml:math></inline-formula>, Fig. <xref ref-type="fig" rid="FA4"/>) shows that none exhibit statistically robust trends across the full parameter space, with large interquartile ranges suggesting that <inline-formula><mml:math id="M156" display="inline"><mml:mrow><mml:msub><mml:mi>f</mml:mi><mml:mi mathvariant="normal">s</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> is the primary driver of BH-corrected <inline-formula><mml:math id="M157" display="inline"><mml:mrow><mml:msup><mml:mi>p</mml:mi><mml:mo>*</mml:mo></mml:msup></mml:mrow></mml:math></inline-formula>, FDP and power behavior for a given <inline-formula><mml:math id="M158" display="inline"><mml:mi mathvariant="italic">α</mml:mi></mml:math></inline-formula>.</p>

<table-wrap id="T2" specific-use="star"><label>Table 2</label><caption><p id="d2e2553">Practical recommendations for assessing statistical significance in geospatial data.</p></caption><oasis:table frame="topbot"><oasis:tgroup cols="3">
     <oasis:colspec colnum="1" colname="col1" align="justify" colwidth="4cm"/>
     <oasis:colspec colnum="2" colname="col2" align="justify" colwidth="8cm"/>
     <oasis:colspec colnum="3" colname="col3" align="justify" colwidth="4cm"/>
     <oasis:thead>
       <oasis:row rowsep="1">
         <oasis:entry colname="col1" align="left">Recommendation</oasis:entry>
         <oasis:entry colname="col2" align="left">Explanation</oasis:entry>
         <oasis:entry colname="col3" align="left">Examples</oasis:entry>
       </oasis:row>
     </oasis:thead>
     <oasis:tbody>
       <oasis:row rowsep="1">
         <oasis:entry colname="col1" align="left">Specify statistical assumptions transparently</oasis:entry>
         <oasis:entry colname="col2" align="left">State null hypothesis, test sidedness, test statistic, and significance level. If pointwise significance testing is appropriate, report the number of performed tests. If a multiple-testing correction is applied, also report all relevant information for correct interpretation and reproducibility (methodology, adjusted threshold <inline-formula><mml:math id="M159" display="inline"><mml:mrow><mml:msup><mml:mi>p</mml:mi><mml:mo>*</mml:mo></mml:msup></mml:mrow></mml:math></inline-formula> and significance level).</oasis:entry>
         <oasis:entry colname="col3" align="left">(Scientific best practice)</oasis:entry>
       </oasis:row>
       <oasis:row rowsep="1">
         <oasis:entry colname="col1" align="left">Visualize FDR behavior</oasis:entry>
         <oasis:entry colname="col2" align="left">If FDR correction is applied, show sorted <inline-formula><mml:math id="M160" display="inline"><mml:mi>p</mml:mi></mml:math></inline-formula> values together with the <inline-formula><mml:math id="M161" display="inline"><mml:mrow><mml:mi mathvariant="italic">α</mml:mi><mml:mstyle displaystyle="false"><mml:mfrac style="text"><mml:mi>i</mml:mi><mml:mi>N</mml:mi></mml:mfrac></mml:mstyle></mml:mrow></mml:math></inline-formula> decision line to illustrate how the rejection threshold is obtained.</oasis:entry>
         <oasis:entry colname="col3" align="left"><xref ref-type="bibr" rid="bib1.bibx48" id="text.40"/>; Fig. <xref ref-type="fig" rid="F2"/></oasis:entry>
       </oasis:row>
       <oasis:row rowsep="1">
         <oasis:entry colname="col1" align="left">Present effect sizes</oasis:entry>
         <oasis:entry colname="col2" align="left">Statistical significance alone does not indicate practical relevance. Report physical anomaly magnitude or a standardized effect size (e.g., Cohen's <inline-formula><mml:math id="M162" display="inline"><mml:mi>d</mml:mi></mml:math></inline-formula>) alongside <inline-formula><mml:math id="M163" display="inline"><mml:mi>p</mml:mi></mml:math></inline-formula> value maps.</oasis:entry>
         <oasis:entry colname="col3" align="left"><xref ref-type="bibr" rid="bib1.bibx2" id="text.41"/>, <xref ref-type="bibr" rid="bib1.bibx45" id="text.42"/>; Fig. <xref ref-type="fig" rid="F1"/></oasis:entry>
       </oasis:row>
       <oasis:row rowsep="1">
         <oasis:entry colname="col1" align="left">Avoid a posteriori domain selection</oasis:entry>
         <oasis:entry colname="col2" align="left">Define spatial domain and masks before significance testing; a posteriori domain selection alters the hypothesis family, inferred threshold and inflates the de facto number of tests being performed <xref ref-type="bibr" rid="bib1.bibx18" id="paren.43"/></oasis:entry>
         <oasis:entry colname="col3" align="left">Sect. <xref ref-type="sec" rid="Ch1.S3.SS2"/>.</oasis:entry>
       </oasis:row>
       <oasis:row rowsep="1">
         <oasis:entry colname="col1" align="left">Interpret large adjusted thresholds cautiously</oasis:entry>
         <oasis:entry colname="col2" align="left">When anomalies with relatively high amplitudes cover large fractions of the domain, BH-corrected thresholds can become counterintuitively permissive; optionally cap thresholds (e.g. at 5 %) and report this explicitly.</oasis:entry>
         <oasis:entry colname="col3" align="left">
                      <xref ref-type="bibr" rid="bib1.bibx37" id="text.44"/>
                    </oasis:entry>
       </oasis:row>
       <oasis:row rowsep="1">
         <oasis:entry colname="col1" align="left">Check applicability of pointwise tests</oasis:entry>
         <oasis:entry colname="col2" align="left">Avoid pointwise testing when the sample size is limited or spatial coherence dominates the signal.</oasis:entry>
         <oasis:entry colname="col3" align="left">
                      <xref ref-type="bibr" rid="bib1.bibx13 bib1.bibx28" id="text.45"/>
                    </oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1" align="left">Use spatially aware methods when appropriate</oasis:entry>
         <oasis:entry colname="col2" align="left">When spatial coherence is expected, consider regional averages, cluster-based tests, or object-based approaches reducing either the FDR level <inline-formula><mml:math id="M164" display="inline"><mml:mi mathvariant="italic">α</mml:mi></mml:math></inline-formula> or <inline-formula><mml:math id="M165" display="inline"><mml:mi>N</mml:mi></mml:math></inline-formula>.</oasis:entry>
         <oasis:entry colname="col3" align="left"><xref ref-type="bibr" rid="bib1.bibx46" id="text.46"/>, <xref ref-type="bibr" rid="bib1.bibx41" id="text.47"/>, <xref ref-type="bibr" rid="bib1.bibx30" id="text.48"/>, <xref ref-type="bibr" rid="bib1.bibx17" id="text.49"/></oasis:entry>
       </oasis:row>
     </oasis:tbody>
   </oasis:tgroup></oasis:table></table-wrap>

</sec>
</sec>
<sec id="Ch1.S6">
  <label>6</label><title>Some practical recommendations</title>
      <p id="d2e2768">The previous sections demonstrate that statistical significance assessment of geospatial data when conducting large number of tests is complex, and that controlling for the global error rate comes with its own pitfalls. In particular, the widely-used BH correction can invite inappropriate local interpretations when results are visualized on maps, leading to strongly domain-dependent conclusions. Table <xref ref-type="table" rid="T2"/> summarizes a set of practical recommendations aimed at improving robustness, interpretability, and transparency in geoscientific significance assessment. We briefly elaborate on these points below.</p>
      <p id="d2e2773">A first requirement is transparent reporting of all statistical assumptions. Studies should clearly specify the null hypothesis, test sidedness, test statistic, and significance level. Furthermore, if pointwise testing is appropriate, the pointwise significance level and the number of tests performed should be reported. Likewise, one should explicitly state whether multiple-testing correction has been applied and, if so, the resulting adjusted threshold. Visualizing the behavior of the FDR correction procedure, for example through ranked <inline-formula><mml:math id="M166" display="inline"><mml:mi>p</mml:mi></mml:math></inline-formula>-value plots as shown in Fig. <xref ref-type="fig" rid="F2"/>, can substantially improve interpretability and facilitate open discussion of statistical decisions.</p>
      <p id="d2e2785">A related point concerns what significance actually tells us. A low <inline-formula><mml:math id="M167" display="inline"><mml:mi>p</mml:mi></mml:math></inline-formula> value indicates that the observed value of the chosen test statistics supports rejection of the null hypothesis in favor of the specified alternative with low probability of error, but it says nothing about whether the underlying signal is large enough to matter physically. This gap widens with sample size: even physically negligible anomalies may become statistically significant given a large enough sample size <xref ref-type="bibr" rid="bib1.bibx45" id="paren.50"/>. We therefore recommend reporting effect sizes, such as the physical anomaly magnitude or a standardized measure like Cohen's <inline-formula><mml:math id="M168" display="inline"><mml:mi>d</mml:mi></mml:math></inline-formula> <xref ref-type="bibr" rid="bib1.bibx12" id="paren.51"/>, alongside significance maps. In most geoscientific research, showing physical anomalies is already common practice. For example, Fig. <xref ref-type="fig" rid="F1"/> already does this by design, with color shading conveying the magnitude of t2m anomalies and horizontal hatching conveying statistical significance.</p>
      <p id="d2e2810">Equally important is defining the testing domain a priori. Selecting or modifying spatial domains after inspecting results changes the family of hypotheses being tested and can substantially alter the adjusted significance threshold. As demonstrated in Sect. <xref ref-type="sec" rid="Ch1.S3.SS2"/>, such a posteriori choices can result in misleading interpretations and must be avoided.</p>
      <p id="d2e2816">An additional question is whether pointwise testing is appropriate at all. Geophysical signals may exhibit strong spatial coherence, while sample sizes remain limited. Sometimes, the sample size may even be too low to perform statistical testing with sufficient statistical power, which should be stated clearly when applicable. Furthermore, grid-point testing may overemphasize unnecessary spatial detail to the detriment of statistical power. In many applications, the scientific question concerns the existence of a regional-scale anomaly rather than the significance of individual grid points. When spatial coherence is expected, spatially aware approaches can provide a more consistent alternative to pointwise hypothesis testing. Instead of evaluating each grid point independently, the analysis can be performed on spatially aggregated quantities that better reflect the underlying physical structure of the system. Examples include regional averages, spatial clusters, or object-based approaches, in which coherent features such as circulation structures or contiguous regions exceeding a predefined threshold are identified and treated as statistical units in the analysis. Again, such choices should be made a priori based on the research question prompting the statistical analysis and should not follow preliminary significance testing or data inspection. By reducing the dimensionality of the problem, these approaches decrease the number of independent tests and align statistical inference more closely with physically meaningful structures.</p>
      <p id="d2e2820">The regional mean temperature analysis presented above provides a simple illustration. Considering only land points, the regional mean 60 d t2m anomaly over northern Europe is negative and can be assessed using a single hypothesis test (<inline-formula><mml:math id="M169" display="inline"><mml:mrow><mml:mi>p</mml:mi><mml:mo>=</mml:mo><mml:mn mathvariant="normal">0.045</mml:mn></mml:mrow></mml:math></inline-formula> obtained from a one-sided, one-sample Student's <inline-formula><mml:math id="M170" display="inline"><mml:mi>t</mml:mi></mml:math></inline-formula> test for an area-weighted mean anomaly of <inline-formula><mml:math id="M171" display="inline"><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">0.99</mml:mn></mml:mrow></mml:math></inline-formula> K). In this case, inference focuses directly on the question of whether a coherent regional response exists, rather than on hundreds of individual grid-point tests whose outcomes are strongly correlated.</p>
      <p id="d2e2852">Another practical aspect concerns the choice of the FDR level itself. In most applications, the FDR level <inline-formula><mml:math id="M172" display="inline"><mml:mi mathvariant="italic">α</mml:mi></mml:math></inline-formula> is selected a priori (e.g. <inline-formula><mml:math id="M173" display="inline"><mml:mrow><mml:mi mathvariant="italic">α</mml:mi><mml:mo>=</mml:mo><mml:mn mathvariant="normal">0.1</mml:mn></mml:mrow></mml:math></inline-formula> following <xref ref-type="bibr" rid="bib1.bibx48" id="altparen.52"/>) and treated as independent of the data structure. However, the behavior of the BH correction depends not only on <inline-formula><mml:math id="M174" display="inline"><mml:mi mathvariant="italic">α</mml:mi></mml:math></inline-formula>, but also on the <inline-formula><mml:math id="M175" display="inline"><mml:mi>p</mml:mi></mml:math></inline-formula> values. One straightforward option to reduce the BH-corrected threshold is simply to choose a smaller fixed <inline-formula><mml:math id="M176" display="inline"><mml:mi mathvariant="italic">α</mml:mi></mml:math></inline-formula> (e.g., <inline-formula><mml:math id="M177" display="inline"><mml:mrow><mml:mi mathvariant="italic">α</mml:mi><mml:mo>=</mml:mo><mml:mn mathvariant="normal">0.05</mml:mn></mml:mrow></mml:math></inline-formula> instead of 0.1). Authors may also  consider capping the adjusted threshold at a level defined a priori (e.g. 5 %), stating this choice explicitly. Since applying the BH correction with the suggested <inline-formula><mml:math id="M178" display="inline"><mml:mrow><mml:mi mathvariant="italic">α</mml:mi><mml:mo>=</mml:mo><mml:mn mathvariant="normal">0.1</mml:mn></mml:mrow></mml:math></inline-formula> can yield adjusted <inline-formula><mml:math id="M179" display="inline"><mml:mi>p</mml:mi></mml:math></inline-formula> value thresholds larger than the conventional local 5 % level (see example in Sect. <xref ref-type="sec" rid="Ch1.S3.SS2"/>), setting such an upper limit may reduce the risk of overconfident interpretations of statistical significance. Alternatively, <inline-formula><mml:math id="M180" display="inline"><mml:mi mathvariant="italic">α</mml:mi></mml:math></inline-formula> could be adjusted dynamically based on properties of the tested field, such as the fraction of nominally significant tests or the ratio between domain size and spatial autocorrelation length scale. Adaptive refinements of this kind have been discussed in the wider FDR literature <xref ref-type="bibr" rid="bib1.bibx6 bib1.bibx40 bib1.bibx7 bib1.bibx33" id="paren.53"/>.</p>
      <p id="d2e2943">One may also consider opting for other approaches than the BH correction, for instance by addressing the Family-Wise Error Rate (FWER) instead of the FDR. However, FWER-controlling methods derived under an independence assumption tend to become overly conservative under positive spatial correlation <xref ref-type="bibr" rid="bib1.bibx31" id="paren.54"/>. One can account for spatially correlated signals, for example, by modifying the Walker criterion by replacing the total number of tests <inline-formula><mml:math id="M181" display="inline"><mml:mi>N</mml:mi></mml:math></inline-formula> by an effective number of independent tests <xref ref-type="bibr" rid="bib1.bibx46 bib1.bibx44" id="paren.55"/>:

          <disp-formula id="Ch1.E5" content-type="numbered"><label>5</label><mml:math id="M182" display="block"><mml:mrow><mml:msubsup><mml:mi>p</mml:mi><mml:mi mathvariant="normal">Walker</mml:mi><mml:mo>*</mml:mo></mml:msubsup><mml:mo>=</mml:mo><mml:mn mathvariant="normal">1</mml:mn><mml:mo>-</mml:mo><mml:msup><mml:mfenced open="(" close=")"><mml:mrow><mml:mn mathvariant="normal">1</mml:mn><mml:mo>-</mml:mo><mml:msub><mml:mi mathvariant="italic">α</mml:mi><mml:mn mathvariant="normal">0</mml:mn></mml:msub></mml:mrow></mml:mfenced><mml:mrow><mml:mn mathvariant="normal">1</mml:mn><mml:mo>/</mml:mo><mml:msub><mml:mi>N</mml:mi><mml:mi mathvariant="normal">eff</mml:mi></mml:msub></mml:mrow></mml:msup><mml:mo>.</mml:mo></mml:mrow></mml:math></disp-formula>

        Here, <inline-formula><mml:math id="M183" display="inline"><mml:mrow><mml:msub><mml:mi>N</mml:mi><mml:mi mathvariant="normal">eff</mml:mi></mml:msub><mml:mo>&lt;</mml:mo><mml:mi>N</mml:mi></mml:mrow></mml:math></inline-formula> represents the reduced, effective number of degrees of freedom implied by spatial dependence among grid points. There exist different ways to estimate <inline-formula><mml:math id="M184" display="inline"><mml:mrow><mml:msub><mml:mi>N</mml:mi><mml:mi mathvariant="normal">eff</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula>, e.g., from the spatial autocorrelation structure, or from other definitions of spatial objects or clusters <xref ref-type="bibr" rid="bib1.bibx8 bib1.bibx19 bib1.bibx15 bib1.bibx17" id="paren.56"/>. Framed in this way, Eq. (<xref ref-type="disp-formula" rid="Ch1.E5"/>) can be interpreted as a pragmatic compromise between fully pointwise inference and purely regional testing.</p>
      <p id="d2e3034">However, estimating <inline-formula><mml:math id="M185" display="inline"><mml:mrow><mml:msub><mml:mi>N</mml:mi><mml:mi mathvariant="normal">eff</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> reliably remains challenging. Spatial correlations are typically non-stationary and anisotropic, and their structure varies across variables, regions, and temporal scales. As noted by <xref ref-type="bibr" rid="bib1.bibx48" id="text.57"/>, these difficulties limit the robustness of methods that rely on a single global estimate of independence. In this sense, both adapting the FDR level and estimating <inline-formula><mml:math id="M186" display="inline"><mml:mrow><mml:msub><mml:mi>N</mml:mi><mml:mi mathvariant="normal">eff</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> can be viewed as attempts to reconcile multiple-testing control with the effective dimensionality imposed by spatial dependence.</p>
</sec>
<sec id="Ch1.S7" sec-type="conclusions">
  <label>7</label><title>Conclusions</title>
      <p id="d2e3070">The peer-reviewed geosciences literature routinely presents statistical significance based on pointwise testing on spatial maps. Multiple-testing correction is essential for any inference drawn from such analyses. Yet, such correction should be applied with care. In this work, we have highlighted a specific caveat of the widely adopted Benjamini-Hochberg (BH) false discovery rate (FDR) correction <xref ref-type="bibr" rid="bib1.bibx5 bib1.bibx48" id="paren.58"/>. When a large fraction of grid points exhibit strong anomalies, the BH correction can yield a relatively permissive corrected significance threshold. This behavior is expected from the formulation of the BH correction and from the common choice of using an FDR level <inline-formula><mml:math id="M187" display="inline"><mml:mi mathvariant="italic">α</mml:mi></mml:math></inline-formula> higher than the pointwise significance level. It can however lead to overconfident conclusions if statistical significance is interpreted pointwise. This becomes an issue specifically when many studies implicitly or explicitly interpret FDR correction as automatically leading to stricter thresholds.</p>
      <p id="d2e3083">This permissiveness does not imply that FDR correction procedures are inherently flawed. Rather, it reflects a mismatch between the aim of the BH correction method  and the interpretation of statistical significance that is typically given when testing spatially distributed data. The FDR-adjusted significance threshold is designed to control the global error rate, not false discovery at single grid points. Because the BH-adjusted threshold depends on the global rank distribution of <inline-formula><mml:math id="M188" display="inline"><mml:mi>p</mml:mi></mml:math></inline-formula> values, it is sensitive to spatial coherence, domain definition, and the proportion of affected grid points. As a result, statistical conclusions derived from maps can change in ways that appear counterintuitive when the tested spatial domain is modified, even if the underlying physical signal remains unchanged. This becomes an issue when analysis domains are chosen a posteriori following exploratory analyses, or when FDR-corrected significance is interpreted as showing locally significant grid points.</p>
      <p id="d2e3094">We have proposed a set of practical recommendations to improve the robustness and transparency of statistical assessments in geospatial analyses. These include clearly reporting the applied correction method along with all relevant information, cautious interpretation of adjusted thresholds, and considering spatially aware alternatives when appropriate. Such considerations are crucial to ensure reproducibility, comparability and interpretability across studies.</p>
      <p id="d2e3097">Ultimately, significance assessment in geospatial data involves a trade-off between interpretability and statistical rigor. Pointwise inference provides intuitive spatial detail but risks overstating confidence, whereas spatially aggregated approaches offer stronger error control at the expense of spatial specificity. Developing methods that balance these competing goals while remaining computationally efficient and interpretable remains an important direction for future research.</p>
</sec>

      
      </body>
    <back><app-group>

<app id="App1.Ch1.S1">
  <label>Appendix A</label><title>Additional figures</title>

      <fig id="FA1"><label>Figure A1</label><caption><p id="d2e3114">As Fig. <xref ref-type="fig" rid="F1"/> but for a domain across eastern Russia ranging between 110–150° E, and 58–73° N.</p></caption>
        
        <graphic xlink:href="https://gmd.copernicus.org/articles/19/10917/2026/gmd-19-10917-2026-f04.png"/>

      </fig>

<fig id="FA2"><label>Figure A2</label><caption><p id="d2e3130">Composites of 2 m air temperature anomalies averaged over 60 d following SSWs. Horizontal hatching indicates statistical significance of cold anomalies with respect to a local 5 % significance level, obtained with a one-sided Student's <inline-formula><mml:math id="M189" display="inline"><mml:mi>t</mml:mi></mml:math></inline-formula> test.</p></caption>
        
        <graphic xlink:href="https://gmd.copernicus.org/articles/19/10917/2026/gmd-19-10917-2026-f05.png"/>

      </fig>

      <fig id="FA3"><label>Figure A3</label><caption><p id="d2e3151">As Fig. <xref ref-type="fig" rid="F2"/>, but showing all <inline-formula><mml:math id="M190" display="inline"><mml:mi>p</mml:mi></mml:math></inline-formula> values, including those above 0.14.</p></caption>
        
        <graphic xlink:href="https://gmd.copernicus.org/articles/19/10917/2026/gmd-19-10917-2026-f06.png"/>

      </fig>

<fig id="FA4"><label>Figure A4</label><caption><p id="d2e3174">Sensitivity of the BH-corrected threshold <inline-formula><mml:math id="M191" display="inline"><mml:mrow><mml:msup><mml:mi>p</mml:mi><mml:mo>*</mml:mo></mml:msup></mml:mrow></mml:math></inline-formula> <bold>(a–d)</bold>, realized false discovery proportion FDP <bold>(e–h)</bold>, and power <bold>(i–l)</bold> to domain size <inline-formula><mml:math id="M192" display="inline"><mml:mrow><mml:msub><mml:mi>N</mml:mi><mml:mi mathvariant="normal">g</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> <bold>(a, e, i)</bold>, signal-to-noise ratio SNR <bold>(b, f, j)</bold>, sample size <inline-formula><mml:math id="M193" display="inline"><mml:mrow><mml:mi>M</mml:mi><mml:mo>∈</mml:mo><mml:mo mathvariant="italic">{</mml:mo><mml:mn mathvariant="normal">20</mml:mn><mml:mo>,</mml:mo><mml:mn mathvariant="normal">50</mml:mn><mml:mo>,</mml:mo><mml:mn mathvariant="normal">250</mml:mn><mml:mo>,</mml:mo><mml:mn mathvariant="normal">1000</mml:mn><mml:mo mathvariant="italic">}</mml:mo></mml:mrow></mml:math></inline-formula> (number of simulated fields per grid point; <bold>c, g, k</bold>), and noise wavenumber <inline-formula><mml:math id="M194" display="inline"><mml:mi mathvariant="italic">ν</mml:mi></mml:math></inline-formula> <bold>(d, h, l)</bold>. In each panel, markers show the median and error bars the interquartile range across all remaining parameter combinations. Colors indicate the choice of FDR level <inline-formula><mml:math id="M195" display="inline"><mml:mi mathvariant="italic">α</mml:mi></mml:math></inline-formula> as in Fig. <xref ref-type="fig" rid="F3"/>, given in the legend in panel <bold>(a)</bold>. In the bottom row, the median power across all parameter combinations is shown as a single purple line, as all choices of <inline-formula><mml:math id="M196" display="inline"><mml:mi mathvariant="italic">α</mml:mi></mml:math></inline-formula> yield near-identical power close to unity.</p></caption>
        
        <graphic xlink:href="https://gmd.copernicus.org/articles/19/10917/2026/gmd-19-10917-2026-f07.png"/>

      </fig>

</app>
  </app-group><notes notes-type="codedataavailability"><title>Code and data availability</title>

      <p id="d2e3288">ERA5 data is freely available from <ext-link xlink:href="https://doi.org/10.24381/cds.bd0915c6" ext-link-type="DOI">10.24381/cds.bd0915c6</ext-link> <xref ref-type="bibr" rid="bib1.bibx21" id="paren.59"/>. The code to create the figures and for the numerical experiment is available from <ext-link xlink:href="https://doi.org/10.5281/zenodo.20280930" ext-link-type="DOI">10.5281/zenodo.20280930</ext-link> <xref ref-type="bibr" rid="bib1.bibx38" id="paren.60"/>.</p>
  </notes><notes notes-type="authorcontribution"><title>Author contributions</title>

      <p id="d2e3306">MKS: Conceptualization, Methodology, Formal analysis, Investigation, Data Curation, Writing – Original Draft, Writing – Review &amp; Editing, Software, Visualization. LO: Conceptualization, Methodology, Investigation, Writing – Original Draft, Writing – Review &amp; Editing, Visualization. FMEP: Conceptualization, Methodology, Investigation, Writing – Review &amp; Editing, Software, Visualization. GM: Conceptualization, Methodology, Writing – Review &amp; Editing, Visualization, Supervision, Funding acquisition.</p>
  </notes><notes notes-type="competinginterests"><title>Competing interests</title>

      <p id="d2e3315">The contact author has declared that none of the authors has any competing interests.</p>
  </notes><notes notes-type="disclaimer"><title>Disclaimer</title>

      <p id="d2e3321">Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. The authors bear the ultimate responsibility for providing appropriate place names. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.</p>
  </notes><ack><title>Acknowledgements</title><p id="d2e3327">We would like to thank Daniel Stephen Wilks, Davide Faranda, and Antonio Segalini for their helpful feedback on the manuscript. During the preparation of this work the authors used ChatGPT and Claude in order to improve the readability and language of the manuscript. After using this tool/service, the authors reviewed and edited the content as needed and take full responsibility for the content of the published article. The data handling was enabled by resources provided by the National Academic Infrastructure for Supercomputing in Sweden (NAISS), partially funded by the Swedish Research Council Vetenskapsrådet through grant agreement no. 2022-06725.</p></ack><notes notes-type="financialsupport"><title>Financial support</title>

      <p id="d2e3332">MS, LO, and GM acknowledge funding from the European Research Council (ERC) under the European Union's Horizon 2020 research and innovation program (grant agreement no. 948309, CENÆ project) and Horizon Europe research and innovation program (grant agreement no. 101231771, DYCLEX project). The publication of this article was funded by the  Swedish Research Council, Forte, Formas, and Vinnova.</p>
  </notes><notes notes-type="reviewstatement"><title>Review statement</title>

      <p id="d2e3343">This paper was edited by Paul Ullrich and reviewed by three anonymous referees.</p>
  </notes><ref-list>
    <title>References</title>

      <ref id="bib1.bibx1"><label>Alberton et al.(2020)Alberton, Nichols, Gamba, and Winkler</label><mixed-citation>Alberton, B. A., Nichols, T. E., Gamba, H. R., and Winkler, A. M.: Multiple testing correction over contrasts for brain imaging, NeuroImage, 216, 116 760, <ext-link xlink:href="https://doi.org/10.1016/j.neuroimage.2020.116760" ext-link-type="DOI">10.1016/j.neuroimage.2020.116760</ext-link>, 2020.</mixed-citation></ref>
      <ref id="bib1.bibx2"><label>Amrhein et al.(2019)Amrhein, Greenland, and McShane</label><mixed-citation>Amrhein, V., Greenland, S., and McShane, B.: Scientists rise up against statistical significance, Nature, 567, 305–307, <ext-link xlink:href="https://doi.org/10.1038/d41586-019-00857-9" ext-link-type="DOI">10.1038/d41586-019-00857-9</ext-link>, 2019.</mixed-citation></ref>
      <ref id="bib1.bibx3"><label>Baldwin and Dunkerton(2001)</label><mixed-citation>Baldwin, M. P. and Dunkerton, T. J.: Stratospheric Harbingers of Anomalous Weather Regimes, Science, 294, 581–584, <ext-link xlink:href="https://doi.org/10.1126/science.1063315" ext-link-type="DOI">10.1126/science.1063315</ext-link>, 2001.</mixed-citation></ref>
      <ref id="bib1.bibx4"><label>Baldwin et al.(2021)Baldwin, Ayarzagüena, Birner, Butchart, Butler, Charlton-Perez, Domeisen, Garfinkel, Garny, Gerber, Hegglin, Langematz, and Pedatella</label><mixed-citation>Baldwin, M. P., Ayarzagüena, B., Birner, T., Butchart, N., Butler, A. H., Charlton-Perez, A. J., Domeisen, D. I. V., Garfinkel, C. I., Garny, H., Gerber, E. P., Hegglin, M. I., Langematz, U., and Pedatella, N. M.: Sudden Stratospheric Warmings, Rev. Geophys., 59, e2020RG000708, <ext-link xlink:href="https://doi.org/10.1029/2020RG000708" ext-link-type="DOI">10.1029/2020RG000708</ext-link>, 2021.</mixed-citation></ref>
      <ref id="bib1.bibx5"><label>Benjamini and Hochberg(1995)</label><mixed-citation>Benjamini, Y. and Hochberg, Y.: Controlling the false discovery rate: a practical and powerful approach to multiple testing, J. Roy. Stat. Soc. B, 57, 289–300, <ext-link xlink:href="https://doi.org/10.1111/j.2517-6161.1995.tb02031.x" ext-link-type="DOI">10.1111/j.2517-6161.1995.tb02031.x</ext-link>, 1995.</mixed-citation></ref>
      <ref id="bib1.bibx6"><label>Benjamini and Yekutieli(2001)</label><mixed-citation> Benjamini, Y. and Yekutieli, D.: The Control of the False Discovery Rate in Multiple Testing under Dependency, Ann. Stat., 29, 1165–1188, 2001.</mixed-citation></ref>
      <ref id="bib1.bibx7"><label>Benjamini et al.(2006)Benjamini, Krieger, and Yekutieli</label><mixed-citation> Benjamini, Y., Krieger, A. M., and Yekutieli, D.: Adaptive Linear Step-up Procedures That Control the False Discovery Rate, Biometrika, 93, 491–507, 2006.</mixed-citation></ref>
      <ref id="bib1.bibx8"><label>Bretherton et al.(1999)Bretherton, Widmann, Dymnikov, Wallace, and Bladé</label><mixed-citation>Bretherton, C. S., Widmann, M., Dymnikov, V. P., Wallace, J. M., and Bladé, I.: The Effective Number of Spatial Degrees of Freedom of a Time-Varying Field, J. Clim., 12, 1990 – 2009, <ext-link xlink:href="https://doi.org/10.1175/1520-0442(1999)012&lt;1990:TENOSD&gt;2.0.CO;2" ext-link-type="DOI">10.1175/1520-0442(1999)012&lt;1990:TENOSD&gt;2.0.CO;2</ext-link>, 1999. </mixed-citation></ref>
      <ref id="bib1.bibx9"><label>Butler(2025)</label><mixed-citation>Butler, A.: Table of major mid-winter SSWs in reanalyses products, NOAA, <uri>https://csl.noaa.gov/groups/csl8/sswcompendium/majorevents.html</uri> (last access: 6 October 2026), 2025.</mixed-citation></ref>
      <ref id="bib1.bibx10"><label>Butler et al.(2017)Butler, Sjoberg, Seidel, and Rosenlof</label><mixed-citation>Butler, A. H., Sjoberg, J. P., Seidel, D. J., and Rosenlof, K. H.: A sudden stratospheric warming compendium, Earth Syst. Sci. Data, 9, 63–76, <ext-link xlink:href="https://doi.org/10.5194/essd-9-63-2017" ext-link-type="DOI">10.5194/essd-9-63-2017</ext-link>, 2017.</mixed-citation></ref>
      <ref id="bib1.bibx11"><label>Charlton and Polvani(2007)</label><mixed-citation>Charlton, A. J. and Polvani, L. M.: A New Look at Stratospheric Sudden Warmings. Part I: Climatology and Modeling Benchmarks, J. Climate, 20, 449–469, <ext-link xlink:href="https://doi.org/10.1175/JCLI3996.1" ext-link-type="DOI">10.1175/JCLI3996.1</ext-link>, 2007.</mixed-citation></ref>
      <ref id="bib1.bibx12"><label>Cohen(1988)</label><mixed-citation>Cohen, J.: Statistical Power Analysis for the Behavioral Sciences, 2nd Edn., Routledge, New York, <ext-link xlink:href="https://doi.org/10.4324/9780203771587" ext-link-type="DOI">10.4324/9780203771587</ext-link>, 1988.</mixed-citation></ref>
      <ref id="bib1.bibx13"><label>Cortés et al.(2020)Cortés, Mahecha, Reichstein, and Brenning</label><mixed-citation>Cortés, J., Mahecha, M., Reichstein, M., and Brenning, A.: Accounting for multiple testing in the analysis of spatio-temporal environmental data, Environ. Ecol. Stat., 27, 293–318, <ext-link xlink:href="https://doi.org/10.1007/s10651-020-00446-4" ext-link-type="DOI">10.1007/s10651-020-00446-4</ext-link>, 2020.</mixed-citation></ref>
      <ref id="bib1.bibx14"><label>Deser et al.(2012)Deser, Phillips, Bourdette, and Teng</label><mixed-citation>Deser, C., Phillips, A., Bourdette, V., and Teng, H.: Uncertainty in climate change projections: the role of internal variability, Clim. Dynam., 38, 527–546, <ext-link xlink:href="https://doi.org/10.1007/s00382-010-0977-x" ext-link-type="DOI">10.1007/s00382-010-0977-x</ext-link>, 2012.</mixed-citation></ref>
      <ref id="bib1.bibx15"><label>Dong et al.(2025)Dong, Ullrich, Lee, Gleckler, Chang, and O'Brien</label><mixed-citation>Dong, B., Ullrich, P., Lee, J., Gleckler, P., Chang, K., and O'Brien, T. A.: A new metrics framework for quantifying and intercomparing atmospheric rivers in observations, reanalyses, and climate models, Geosci. Model Dev., 18, 961–976, <ext-link xlink:href="https://doi.org/10.5194/gmd-18-961-2025" ext-link-type="DOI">10.5194/gmd-18-961-2025</ext-link>, 2025.</mixed-citation></ref>
      <ref id="bib1.bibx16"><label>Eade et al.(2014)Eade, Smith, Scaife, Wallace, Dunstone, Hermanson, and Robinson</label><mixed-citation>Eade, R., Smith, D., Scaife, A., Wallace, E., Dunstone, N., Hermanson, L., and Robinson, N.: Do seasonal-to-decadal climate predictions underestimate the predictability of the real world?, Geophys. Res. Lett., 41, 5620–5628, <ext-link xlink:href="https://doi.org/10.1002/2014GL061146" ext-link-type="DOI">10.1002/2014GL061146</ext-link>, 2014.</mixed-citation></ref>
      <ref id="bib1.bibx17"><label>Faranda and Alberti(2026)</label><mixed-citation>Faranda, D. and Alberti, T.: Investigating the Role of Climate Change in the 3 May 2025 Western Europe Hailstorm Using Atmospheric Analogs, Atmos. Sci. Lett., 27, e70016, <ext-link xlink:href="https://doi.org/10.1002/asl2.70016" ext-link-type="DOI">10.1002/asl2.70016</ext-link>, 2026.</mixed-citation></ref>
      <ref id="bib1.bibx18"><label>Gelman and Loken(2013)</label><mixed-citation>Gelman, A. and Loken, E.: The garden of forking paths: Why multiple comparisons can be a problem, even when there is no “fishing expedition” or <uri>https://sites.stat.columbia.edu/gelman/research/unpublished/p_hacking.pdf</uri> (last access: 8 November 2026), Department of Statistics, Columbia University, unpublished manuscript, 2013.</mixed-citation></ref>
      <ref id="bib1.bibx19"><label>Griffith(2013)</label><mixed-citation>Griffith, D. A.: Establishing Qualitative Geographic Sample Size in the Presence of Spatial Autocorrelation, Ann. Assoc. Am. Geogr., 103, 1107–1122, <ext-link xlink:href="https://doi.org/10.1080/00045608.2013.776884" ext-link-type="DOI">10.1080/00045608.2013.776884</ext-link>, 2013.</mixed-citation></ref>
      <ref id="bib1.bibx20"><label>Hersbach et al.(2020)Hersbach, Bell, Berrisford, Hirahara, Horányi, Muñoz-Sabater, Nicolas, Peubey, Radu, Schepers, Simmons, Soci, Abdalla, Abellan, Balsamo, Bechtold, Biavati, Bidlot, Bonavita, De Chiara, Dahlgren, Dee, Diamantakis, Dragani, Flemming, Forbes, Fuentes, Geer, Haimberger, Healy, Hogan, Hólm, Janisková, Keeley, Laloyaux, Lopez, Lupu, Radnoti, de Rosnay, Rozum, Vamborg, Villaume, and Thépaut</label><mixed-citation>Hersbach, H., Bell, B., Berrisford, P., Hirahara, S., Horányi, A., Muñoz-Sabater, J., Nicolas, J., Peubey, C., Radu, R., Schepers, D., Simmons, A., Soci, C., Abdalla, S., Abellan, X., Balsamo, G., Bechtold, P., Biavati, G., Bidlot, J., Bonavita, M., De Chiara, G., Dahlgren, P., Dee, D., Diamantakis, M., Dragani, R., Flemming, J., Forbes, R., Fuentes, M., Geer, A., Haimberger, L., Healy, S., Hogan, R. J., Hólm, E., Janisková, M., Keeley, S., Laloyaux, P., Lopez, P., Lupu, C., Radnoti, G., de Rosnay, P., Rozum, I., Vamborg, F., Villaume, S., and Thépaut, J.-N.: The ERA5 global reanalysis, Q. J. Roy. Meteorol. Soc., 146, 1999–2049, <ext-link xlink:href="https://doi.org/10.1002/qj.3803" ext-link-type="DOI">10.1002/qj.3803</ext-link>, 2020.</mixed-citation></ref>
      <ref id="bib1.bibx21"><label>Hersbach et al.(2025)Hersbach, Bell, Berrisford, Biavati, G., Muñoz Sabater, Nicolas, Peubey, Radu, Rozum, Schepers, Simmons, Soci, Dee, and Thépaut</label><mixed-citation>Hersbach, H., Bell, B., Berrisford, P., Biavati, G., Horányi, A., Muñoz Sabater, J., Nicolas, J., Peubey, C., Radu, R., Rozum, I., Schepers, D., Simmons, A., Soci, C., Dee, D., and Thépaut, J.-N.: ERA5 hourly data on pressure levels from 1940 to present, Copernicus Climate Change Service (C3S) Climate Data Store (CDS) [data set], <ext-link xlink:href="https://doi.org/10.24381/cds.bd0915c6" ext-link-type="DOI">10.24381/cds.bd0915c6</ext-link>, 2025.</mixed-citation></ref>
      <ref id="bib1.bibx22"><label>Jeuring et al.(2024)Jeuring, Samuelsen, Lamers, Müller, Åge Hjøllo, Bertino, and Hagen</label><mixed-citation>Jeuring, J., Samuelsen, E. M., Lamers, M., Müller, M., Åge Hjøllo, B., Bertino, L., and Hagen, B.: Map-Based Ensemble Forecasts for Maritime Operations: An Interactive Usability Assessment with Decision Scenarios, Weather Clim. Soc., 16, 235–256, <ext-link xlink:href="https://doi.org/10.1175/WCAS-D-23-0076.1" ext-link-type="DOI">10.1175/WCAS-D-23-0076.1</ext-link>, 2024.</mixed-citation></ref>
      <ref id="bib1.bibx23"><label>Katz(2002)</label><mixed-citation>Katz, R. W.: Sir Gilbert Walker and a Connection between El Niño and Statistics, Stat. Sci., 17, <ext-link xlink:href="https://doi.org/10.1214/ss/1023799000" ext-link-type="DOI">10.1214/ss/1023799000</ext-link>, 2002.</mixed-citation></ref>
      <ref id="bib1.bibx24"><label>Katz and Brown(1991)</label><mixed-citation>Katz, R. W. and Brown, B. G.: The problem of multiplicity in research on teleconnections, Int. J. Climatol., 11, 505–513, <ext-link xlink:href="https://doi.org/10.1002/joc.3370110504" ext-link-type="DOI">10.1002/joc.3370110504</ext-link>, 1991.</mixed-citation></ref>
      <ref id="bib1.bibx25"><label>Kelleher and Mahajan(2026)</label><mixed-citation>Kelleher, M. E. and Mahajan, S.: Enhanced climate reproducibility testing with false discovery rate correction, Earth Syst. Dynam., 17, 23–39, <ext-link xlink:href="https://doi.org/10.5194/esd-17-23-2026" ext-link-type="DOI">10.5194/esd-17-23-2026</ext-link>, 2026.</mixed-citation></ref>
      <ref id="bib1.bibx26"><label>King et al.(2019)King, Butler, Jucker, Earl, and Rudeva</label><mixed-citation>King, A. D., Butler, A. H., Jucker, M., Earl, N. O., and Rudeva, I.: Observed relationships between sudden stratospheric warmings and European climate extremes, J. Geophys. Res.-Atmos., 124, 13943–13961, <ext-link xlink:href="https://doi.org/10.1029/2019JD030480" ext-link-type="DOI">10.1029/2019JD030480</ext-link>, 2019.</mixed-citation></ref>
      <ref id="bib1.bibx27"><label>Kolstad et al.(2010)Kolstad, Breiteig, and Scaife</label><mixed-citation>Kolstad, E. W., Breiteig, T., and Scaife, A. A.: The association between stratospheric weak polar vortex events and cold air outbreaks in the Northern Hemisphere, Q. J. Roy. Meteorol. Soc., 136, 886–893, <ext-link xlink:href="https://doi.org/10.1002/qj.620" ext-link-type="DOI">10.1002/qj.620</ext-link>, 2010.</mixed-citation></ref>
      <ref id="bib1.bibx28"><label>Kolstad et al.(2022)Kolstad, Lee, Butler, Domeisen, and Wulff</label><mixed-citation>Kolstad, E. W., Lee, S. H., Butler, A. H., Domeisen, D. I., and Wulff, C. O.: Diverse surface signatures of stratospheric polar vortex anomalies, J. Geophys. Res.-Atmos., 127, e2022JD037422, <ext-link xlink:href="https://doi.org/10.1029/2022JD037422" ext-link-type="DOI">10.1029/2022JD037422</ext-link>, 2022.</mixed-citation></ref>
      <ref id="bib1.bibx29"><label>Lee et al.(2025)Lee, Butler, and Manney</label><mixed-citation>Lee, S. H., Butler, A. H., and Manney, G. L.: Two major sudden stratospheric warmings during winter 2023/2024, Weather, 80, 45–53, <ext-link xlink:href="https://doi.org/10.1002/wea.7656" ext-link-type="DOI">10.1002/wea.7656</ext-link>, 2025.</mixed-citation></ref>
      <ref id="bib1.bibx30"><label>Lembo et al.(2026)Lembo, Messori, Faranda, Galfi, Graversen, and Pons</label><mixed-citation>Lembo, V., Messori, G., Faranda, D., Galfi, V. M., Graversen, R. G., and Pons, F. E.: Concurrent heat waves and their linkage to large-scale meridional heat transports through planetary-scale waves, Weather Clim. Dynam., 7, 453–473, <ext-link xlink:href="https://doi.org/10.5194/wcd-7-453-2026" ext-link-type="DOI">10.5194/wcd-7-453-2026</ext-link>, 2026.</mixed-citation></ref>
      <ref id="bib1.bibx31"><label>Livezey and Chen(1983)</label><mixed-citation>Livezey, R. E. and Chen, W. Y.: Statistical Field Significance and its Determination by Monte Carlo Techniques, Mon. Weather Rev., 111, 46–59, <ext-link xlink:href="https://doi.org/10.1175/1520-0493(1983)111&lt;0046:SFSAID&gt;2.0.CO;2" ext-link-type="DOI">10.1175/1520-0493(1983)111&lt;0046:SFSAID&gt;2.0.CO;2</ext-link>, 1983.</mixed-citation></ref>
      <ref id="bib1.bibx32"><label>Monnin et al.(2022)Monnin, Kretschmer, and Polichtchouk</label><mixed-citation>Monnin, E., Kretschmer, M., and Polichtchouk, I.: The role of the timing of sudden stratospheric warmings for precipitation and temperature anomalies in Europe, Int. J. Climatol., 42, 3448–3462, <ext-link xlink:href="https://doi.org/10.1002/joc.7426" ext-link-type="DOI">10.1002/joc.7426</ext-link>, 2022.</mixed-citation></ref>
      <ref id="bib1.bibx33"><label>Mrkvička and Myllymäki(2023)</label><mixed-citation>Mrkvička, T. and Myllymäki, M.: False discovery rate envelopes, Stat. Comput., 33, 109, <ext-link xlink:href="https://doi.org/10.1007/s11222-023-10275-7" ext-link-type="DOI">10.1007/s11222-023-10275-7</ext-link>, 2023.</mixed-citation></ref>
      <ref id="bib1.bibx34"><label>Noble(2009)</label><mixed-citation>Noble, W. S.: How does multiple testing correction work?, Nat. Biotechnol., 27, 1135–1137, <ext-link xlink:href="https://doi.org/10.1038/nbt1209-1135" ext-link-type="DOI">10.1038/nbt1209-1135</ext-link>, 2009.</mixed-citation></ref>
      <ref id="bib1.bibx35"><label>Oehrlein et al.(2021)Oehrlein, Polvani, Sun, and Deser</label><mixed-citation>Oehrlein, J., Polvani, L. M., Sun, L., and Deser, C.: How well do we know the surface impact of sudden stratospheric warmings?, Geophys. Res. Lett., 48, e2021GL095 493, <ext-link xlink:href="https://doi.org/10.1029/2021GL095493" ext-link-type="DOI">10.1029/2021GL095493</ext-link>, 2021.</mixed-citation></ref>
      <ref id="bib1.bibx36"><label>Pappenberger et al.(2019)Pappenberger, Cloke, and Baugh</label><mixed-citation>Pappenberger, F., Cloke, H. L., and Baugh, C. A.: Cartograms for use in forecasting weather-driven natural hazards, Cartogr. J., 56, 134–145, <ext-link xlink:href="https://doi.org/10.1080/00087041.2018.1534358" ext-link-type="DOI">10.1080/00087041.2018.1534358</ext-link>, 2019.</mixed-citation></ref>
      <ref id="bib1.bibx37"><label>Schutte et al.(2026a)Schutte, Olivetti, Krouma, and Messori</label><mixed-citation>Schutte, M. K., Olivetti, L., Krouma, M., and Messori, G.: Stratospheric and tropospheric contributions to North American cold temperatures and subsequent North Atlantic jet stream anomalies, Weather Clim. Ext., 52, 100901, <ext-link xlink:href="https://doi.org/10.1016/j.wace.2026.100901" ext-link-type="DOI">10.1016/j.wace.2026.100901</ext-link>, 2026a.</mixed-citation></ref>
      <ref id="bib1.bibx38"><label>Schutte et al.(2026b)Schutte, Olivetti, Pons, and Messori</label><mixed-citation>Schutte, M. K., Olivetti, L., Pons, F. M. E., and Messori, G.: Code and data for “The Stippled Gridpoints are Statistically Significant: (Mis)uses of False Discovery Rate Correction for Geospatial Data”, Zenodo [code and data set], <ext-link xlink:href="https://doi.org/10.5281/zenodo.20280930" ext-link-type="DOI">10.5281/zenodo.20280930</ext-link>, 2026b.</mixed-citation></ref>
      <ref id="bib1.bibx39"><label>Seneviratne et al.(2021)Seneviratne, Zhang, Adnan, Badi, Dereczynski, A. Di Luca, Iskandar, Kossin, Lewis, Otto, Pinto, Satoh, Vicente-Serrano, Wehner, and Zhou</label><mixed-citation>Seneviratne, S., Zhang, X., Adnan, M., Badi, W., Dereczynski, C., A. Di Luca, S. G., Iskandar, I., Kossin, J., Lewis, S., Otto, F., Pinto, I., Satoh, M., Vicente-Serrano, S., Wehner, M., and Zhou, B.: Weather and Climate Extreme Events in a Changing Climate, in: Climate Change 2021: The Physical Science Basis. Contribution of Working Group I to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change, edited by: Masson-Delmotte, V., Zhai, P., Pirani, A., Connors, S. L., Péan, C., Berger, S., Caud, N., Chen, Y., Goldfarb, L., Gomis, M. I., Huang, M., Leitzell, K., Lonnoy, E.,  Matthews, J. B. R., Maycock, T. K., Waterfield, T., Yelekçi, O., Yu, R., and Zhou, B., Cambridge University Press, 1513–1766, <ext-link xlink:href="https://doi.org/10.1017/9781009157896.013" ext-link-type="DOI">10.1017/9781009157896.013</ext-link>, 2021.</mixed-citation></ref>
      <ref id="bib1.bibx40"><label>Storey(2002)</label><mixed-citation>Storey, J. D.: A Direct Approach to False Discovery Rates, J. Roy. Stat. Soc. B, 64, 479–498, <ext-link xlink:href="https://doi.org/10.1111/1467-9868.00346" ext-link-type="DOI">10.1111/1467-9868.00346</ext-link>, 2002.</mixed-citation></ref>
      <ref id="bib1.bibx41"><label>Sun et al.(2015)Sun, Reich, Tony Cai, Guindani, and Schwartzman</label><mixed-citation>Sun, W., Reich, B. J., Tony Cai, T., Guindani, M., and Schwartzman, A.: False discovery control in large-scale spatial multiple testing, J. Roy. Stat. Soc. B, 77, 59–83, <ext-link xlink:href="https://doi.org/10.1111/rssb.12064" ext-link-type="DOI">10.1111/rssb.12064</ext-link>, 2015.</mixed-citation></ref>
      <ref id="bib1.bibx42"><label>Taggart and Wilke(2025)</label><mixed-citation>Taggart, R. J. and Wilke, D. J.: Warnings based on risk matrices: a coherent framework with consistent evaluation, Nat. Hazards Earth Syst. Sci., 25, 2657–2677, <ext-link xlink:href="https://doi.org/10.5194/nhess-25-2657-2025" ext-link-type="DOI">10.5194/nhess-25-2657-2025</ext-link>, 2025.</mixed-citation></ref>
      <ref id="bib1.bibx43"><label>von Storch and Zwiers(1999)</label><mixed-citation>von Storch, H. and Zwiers, F. W.: Statistical Analysis in Climate Research, Cambridge University Press, Cambridge, <ext-link xlink:href="https://doi.org/10.1017/CBO9780511612336" ext-link-type="DOI">10.1017/CBO9780511612336</ext-link>, 1999.</mixed-citation></ref>
      <ref id="bib1.bibx44"><label>Walker(1914)</label><mixed-citation> Walker, G. T.: Correlation in seasonal variations of weather, III: On the criterion for the reality of relationships or periodicities, Memoirs of the India Meteorological Department, 21, 13–15, 1914.</mixed-citation></ref>
      <ref id="bib1.bibx45"><label>Wasserstein et al.(2019)Wasserstein, Schirm, and Lazar</label><mixed-citation>Wasserstein, R. L., Schirm, A. L., and Lazar, N. A.: Moving to a world beyond “<inline-formula><mml:math id="M197" display="inline"><mml:mrow><mml:mi>p</mml:mi><mml:mo>&lt;</mml:mo><mml:mn mathvariant="normal">0.05</mml:mn></mml:mrow></mml:math></inline-formula>”, Am. Stat., 73, 1–19, <ext-link xlink:href="https://doi.org/10.1080/00031305.2019.1583913" ext-link-type="DOI">10.1080/00031305.2019.1583913</ext-link>, 2019.</mixed-citation></ref>
      <ref id="bib1.bibx46"><label>Wilks(2006)</label><mixed-citation>Wilks, D. S.: On “Field Significance” and the False Discovery Rate, J. Appl. Meteorol. Clim., 45, 1181–1189, <ext-link xlink:href="https://doi.org/10.1175/JAM2404.1" ext-link-type="DOI">10.1175/JAM2404.1</ext-link>, 2006.</mixed-citation></ref>
      <ref id="bib1.bibx47"><label>Wilks(2011)</label><mixed-citation> Wilks, D. S.: Statistical Methods in the Atmospheric Sciences, in: vol. 100 of International Geophysics, 3rd Edn.,Academic Press, ISBN 978-0-12-385022-5, 2011.</mixed-citation></ref>
      <ref id="bib1.bibx48"><label>Wilks(2016)</label><mixed-citation>Wilks, D. S.: “The Stippling Shows Statistically Significant Grid Points”: How Research Results are Routinely Overstated and Overinterpreted, and What to Do about It, B. Am. Meteorol. Soc., 97, 2263–2273, <ext-link xlink:href="https://doi.org/10.1175/BAMS-D-15-00267.1" ext-link-type="DOI">10.1175/BAMS-D-15-00267.1</ext-link>, 2016.</mixed-citation></ref>

  </ref-list></back>
    <!--<article-title-html>The stippled grid points are statistically significant: (mis)uses  of false discovery rate correction for geospatial data</article-title-html>
<abstract-html/>
<ref-html id="bib1.bib1"><label>Alberton et al.(2020)Alberton, Nichols, Gamba, and
Winkler</label><mixed-citation>
      
Alberton, B. A., Nichols, T. E., Gamba, H. R., and Winkler, A. M.: Multiple
testing correction over contrasts for brain imaging, NeuroImage, 216,
116&thinsp;760, <a href="https://doi.org/10.1016/j.neuroimage.2020.116760" target="_blank">https://doi.org/10.1016/j.neuroimage.2020.116760</a>, 2020.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib2"><label>Amrhein et al.(2019)Amrhein, Greenland, and
McShane</label><mixed-citation>
      
Amrhein, V., Greenland, S., and McShane, B.: Scientists rise up against
statistical significance, Nature, 567, 305–307,
<a href="https://doi.org/10.1038/d41586-019-00857-9" target="_blank">https://doi.org/10.1038/d41586-019-00857-9</a>, 2019.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib3"><label>Baldwin and Dunkerton(2001)</label><mixed-citation>
      
Baldwin, M. P. and Dunkerton, T. J.: Stratospheric Harbingers of Anomalous
Weather Regimes, Science, 294, 581–584, <a href="https://doi.org/10.1126/science.1063315" target="_blank">https://doi.org/10.1126/science.1063315</a>, 2001.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib4"><label>Baldwin et al.(2021)Baldwin, Ayarzagüena, Birner, Butchart, Butler,
Charlton-Perez, Domeisen, Garfinkel, Garny, Gerber, Hegglin, Langematz, and
Pedatella</label><mixed-citation>
      
Baldwin, M. P., Ayarzagüena, B., Birner, T., Butchart, N., Butler, A. H.,
Charlton-Perez, A. J., Domeisen, D. I. V., Garfinkel, C. I., Garny, H.,
Gerber, E. P., Hegglin, M. I., Langematz, U., and Pedatella, N. M.: Sudden
Stratospheric Warmings, Rev. Geophys., 59, e2020RG000708,
<a href="https://doi.org/10.1029/2020RG000708" target="_blank">https://doi.org/10.1029/2020RG000708</a>, 2021.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib5"><label>Benjamini and Hochberg(1995)</label><mixed-citation>
      
Benjamini, Y. and Hochberg, Y.: Controlling the false discovery rate: a
practical and powerful approach to multiple testing, J. Roy. Stat. Soc. B, 57,
289–300, <a href="https://doi.org/10.1111/j.2517-6161.1995.tb02031.x" target="_blank">https://doi.org/10.1111/j.2517-6161.1995.tb02031.x</a>, 1995.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib6"><label>Benjamini and Yekutieli(2001)</label><mixed-citation>
      
Benjamini, Y. and Yekutieli, D.: The Control of the False Discovery Rate in
Multiple Testing under Dependency, Ann. Stat., 29, 1165–1188, 2001.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib7"><label>Benjamini et al.(2006)Benjamini, Krieger, and
Yekutieli</label><mixed-citation>
      
Benjamini, Y., Krieger, A. M., and Yekutieli, D.: Adaptive Linear Step-up
Procedures That Control the False Discovery Rate, Biometrika, 93, 491–507,
2006.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib8"><label>Bretherton et al.(1999)Bretherton, Widmann, Dymnikov, Wallace, and
Bladé</label><mixed-citation>
      
Bretherton, C. S., Widmann, M., Dymnikov, V. P., Wallace, J. M., and Bladé,
I.: The Effective Number of Spatial Degrees of Freedom of a Time-Varying
Field, J. Clim., 12, 1990 – 2009,
<a href="https://doi.org/10.1175/1520-0442(1999)012&lt;1990:TENOSD&gt;2.0.CO;2" target="_blank">https://doi.org/10.1175/1520-0442(1999)012&lt;1990:TENOSD&gt;2.0.CO;2</a>, 1999.


    </mixed-citation></ref-html>
<ref-html id="bib1.bib9"><label>Butler(2025)</label><mixed-citation>
      
Butler, A.: Table of major mid-winter SSWs in reanalyses products, NOAA,
<a href="https://csl.noaa.gov/groups/csl8/sswcompendium/majorevents.html" target="_blank"/> (last access: 6 October 2026),
2025.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib10"><label>Butler et al.(2017)Butler, Sjoberg, Seidel, and
Rosenlof</label><mixed-citation>
      
Butler, A. H., Sjoberg, J. P., Seidel, D. J., and Rosenlof, K. H.: A sudden
stratospheric warming compendium, Earth Syst. Sci. Data, 9, 63–76,
<a href="https://doi.org/10.5194/essd-9-63-2017" target="_blank">https://doi.org/10.5194/essd-9-63-2017</a>, 2017.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib11"><label>Charlton and Polvani(2007)</label><mixed-citation>
      
Charlton, A. J. and Polvani, L. M.: A New Look at Stratospheric Sudden
Warmings. Part I: Climatology and Modeling Benchmarks, J. Climate, 20,
449–469, <a href="https://doi.org/10.1175/JCLI3996.1" target="_blank">https://doi.org/10.1175/JCLI3996.1</a>, 2007.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib12"><label>Cohen(1988)</label><mixed-citation>
      
Cohen, J.: Statistical Power Analysis for the Behavioral Sciences, 2nd Edn., Routledge, New York, <a href="https://doi.org/10.4324/9780203771587" target="_blank">https://doi.org/10.4324/9780203771587</a>, 1988.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib13"><label>Cortés et al.(2020)Cortés, Mahecha, Reichstein, and
Brenning</label><mixed-citation>
      
Cortés, J., Mahecha, M., Reichstein, M., and Brenning, A.: Accounting for
multiple testing in the analysis of spatio-temporal environmental data,
Environ. Ecol. Stat., 27, 293–318, <a href="https://doi.org/10.1007/s10651-020-00446-4" target="_blank">https://doi.org/10.1007/s10651-020-00446-4</a>,
2020.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib14"><label>Deser et al.(2012)Deser, Phillips, Bourdette, and
Teng</label><mixed-citation>
      
Deser, C., Phillips, A., Bourdette, V., and Teng, H.: Uncertainty in climate
change projections: the role of internal variability, Clim. Dynam., 38,
527–546, <a href="https://doi.org/10.1007/s00382-010-0977-x" target="_blank">https://doi.org/10.1007/s00382-010-0977-x</a>, 2012.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib15"><label>Dong et al.(2025)Dong, Ullrich, Lee, Gleckler, Chang, and
O'Brien</label><mixed-citation>
      
Dong, B., Ullrich, P., Lee, J., Gleckler, P., Chang, K., and O'Brien, T. A.: A new metrics framework for quantifying and intercomparing atmospheric rivers
in observations, reanalyses, and climate models, Geosci. Model Dev., 18,
961–976, <a href="https://doi.org/10.5194/gmd-18-961-2025" target="_blank">https://doi.org/10.5194/gmd-18-961-2025</a>, 2025.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib16"><label>Eade et al.(2014)Eade, Smith, Scaife, Wallace, Dunstone, Hermanson,
and Robinson</label><mixed-citation>
      
Eade, R., Smith, D., Scaife, A., Wallace, E., Dunstone, N., Hermanson, L., and
Robinson, N.: Do seasonal-to-decadal climate predictions underestimate the
predictability of the real world?, Geophys. Res. Lett., 41, 5620–5628,
<a href="https://doi.org/10.1002/2014GL061146" target="_blank">https://doi.org/10.1002/2014GL061146</a>, 2014.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib17"><label>Faranda and Alberti(2026)</label><mixed-citation>
      
Faranda, D. and Alberti, T.: Investigating the Role of Climate Change in the 3
May 2025 Western Europe Hailstorm Using Atmospheric Analogs, Atmos. Sci.
Lett., 27, e70016, <a href="https://doi.org/10.1002/asl2.70016" target="_blank">https://doi.org/10.1002/asl2.70016</a>, 2026.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib18"><label>Gelman and Loken(2013)</label><mixed-citation>
      
Gelman, A. and Loken, E.: The garden of forking paths: Why multiple comparisons can be a problem, even when there is no “fishing expedition” or
<a href="https://sites.stat.columbia.edu/gelman/research/unpublished/p_hacking.pdf" target="_blank"/> (last access: 8 November 2026), Department of Statistics, Columbia University, unpublished manuscript, 2013.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib19"><label>Griffith(2013)</label><mixed-citation>
      
Griffith, D. A.: Establishing Qualitative Geographic Sample Size in the
Presence of Spatial Autocorrelation, Ann. Assoc. Am. Geogr., 103, 1107–1122,
<a href="https://doi.org/10.1080/00045608.2013.776884" target="_blank">https://doi.org/10.1080/00045608.2013.776884</a>, 2013.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib20"><label>Hersbach et al.(2020)Hersbach, Bell, Berrisford, Hirahara, Horányi,
Muñoz-Sabater, Nicolas, Peubey, Radu, Schepers, Simmons, Soci, Abdalla,
Abellan, Balsamo, Bechtold, Biavati, Bidlot, Bonavita, De Chiara, Dahlgren,
Dee, Diamantakis, Dragani, Flemming, Forbes, Fuentes, Geer, Haimberger,
Healy, Hogan, Hólm, Janisková, Keeley, Laloyaux, Lopez, Lupu, Radnoti,
de Rosnay, Rozum, Vamborg, Villaume, and Thépaut</label><mixed-citation>
      
Hersbach, H., Bell, B., Berrisford, P., Hirahara, S., Horányi, A.,
Muñoz-Sabater, J., Nicolas, J., Peubey, C., Radu, R., Schepers, D., Simmons,
A., Soci, C., Abdalla, S., Abellan, X., Balsamo, G., Bechtold, P., Biavati,
G., Bidlot, J., Bonavita, M., De Chiara, G., Dahlgren, P., Dee, D.,
Diamantakis, M., Dragani, R., Flemming, J., Forbes, R., Fuentes, M., Geer,
A., Haimberger, L., Healy, S., Hogan, R. J., Hólm, E., Janisková, M.,
Keeley, S., Laloyaux, P., Lopez, P., Lupu, C., Radnoti, G., de Rosnay, P.,
Rozum, I., Vamborg, F., Villaume, S., and Thépaut, J.-N.: The ERA5 global
reanalysis, Q. J. Roy. Meteorol. Soc., 146, 1999–2049, <a href="https://doi.org/10.1002/qj.3803" target="_blank">https://doi.org/10.1002/qj.3803</a>, 2020.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib21"><label>Hersbach et al.(2025)Hersbach, Bell, Berrisford, Biavati, G.,
Muñoz Sabater, Nicolas, Peubey, Radu, Rozum, Schepers, Simmons, Soci, Dee,
and Thépaut</label><mixed-citation>
      
Hersbach, H., Bell, B., Berrisford, P., Biavati, G., Horányi, A.,
Muñoz Sabater, J., Nicolas, J., Peubey, C., Radu, R., Rozum, I., Schepers,
D., Simmons, A., Soci, C., Dee, D., and Thépaut, J.-N.: ERA5 hourly data on
pressure levels from 1940 to present, Copernicus Climate Change Service (C3S) Climate Data Store (CDS) [data set], <a href="https://doi.org/10.24381/cds.bd0915c6" target="_blank">https://doi.org/10.24381/cds.bd0915c6</a>, 2025.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib22"><label>Jeuring et al.(2024)Jeuring, Samuelsen, Lamers, Müller, Åge
Hjøllo, Bertino, and Hagen</label><mixed-citation>
      
Jeuring, J., Samuelsen, E. M., Lamers, M., Müller, M., Åge Hjøllo, B.,
Bertino, L., and Hagen, B.: Map-Based Ensemble Forecasts for Maritime
Operations: An Interactive Usability Assessment with Decision Scenarios,
Weather Clim. Soc., 16, 235–256, <a href="https://doi.org/10.1175/WCAS-D-23-0076.1" target="_blank">https://doi.org/10.1175/WCAS-D-23-0076.1</a>, 2024.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib23"><label>Katz(2002)</label><mixed-citation>
      
Katz, R. W.: Sir Gilbert Walker and a Connection between El Niño and
Statistics, Stat. Sci., 17, <a href="https://doi.org/10.1214/ss/1023799000" target="_blank">https://doi.org/10.1214/ss/1023799000</a>, 2002.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib24"><label>Katz and Brown(1991)</label><mixed-citation>
      
Katz, R. W. and Brown, B. G.: The problem of multiplicity in research on
teleconnections, Int. J. Climatol., 11, 505–513,
<a href="https://doi.org/10.1002/joc.3370110504" target="_blank">https://doi.org/10.1002/joc.3370110504</a>, 1991.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib25"><label>Kelleher and Mahajan(2026)</label><mixed-citation>
      
Kelleher, M. E. and Mahajan, S.: Enhanced climate reproducibility testing with false discovery rate correction, Earth Syst. Dynam., 17, 23–39,
<a href="https://doi.org/10.5194/esd-17-23-2026" target="_blank">https://doi.org/10.5194/esd-17-23-2026</a>, 2026.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib26"><label>King et al.(2019)King, Butler, Jucker, Earl, and
Rudeva</label><mixed-citation>
      
King, A. D., Butler, A. H., Jucker, M., Earl, N. O., and Rudeva, I.: Observed
relationships between sudden stratospheric warmings and European climate
extremes, J. Geophys. Res.-Atmos., 124, 13943–13961,
<a href="https://doi.org/10.1029/2019JD030480" target="_blank">https://doi.org/10.1029/2019JD030480</a>, 2019.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib27"><label>Kolstad et al.(2010)Kolstad, Breiteig, and Scaife</label><mixed-citation>
      
Kolstad, E. W., Breiteig, T., and Scaife, A. A.: The association between
stratospheric weak polar vortex events and cold air outbreaks in the Northern
Hemisphere, Q. J. Roy. Meteorol. Soc., 136, 886–893, <a href="https://doi.org/10.1002/qj.620" target="_blank">https://doi.org/10.1002/qj.620</a>,
2010.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib28"><label>Kolstad et al.(2022)Kolstad, Lee, Butler, Domeisen, and
Wulff</label><mixed-citation>
      
Kolstad, E. W., Lee, S. H., Butler, A. H., Domeisen, D. I., and Wulff, C. O.:
Diverse surface signatures of stratospheric polar vortex anomalies, J.
Geophys. Res.-Atmos., 127, e2022JD037422, <a href="https://doi.org/10.1029/2022JD037422" target="_blank">https://doi.org/10.1029/2022JD037422</a>, 2022.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib29"><label>Lee et al.(2025)Lee, Butler, and Manney</label><mixed-citation>
      
Lee, S. H., Butler, A. H., and Manney, G. L.: Two major sudden stratospheric
warmings during winter 2023/2024, Weather, 80, 45–53,
<a href="https://doi.org/10.1002/wea.7656" target="_blank">https://doi.org/10.1002/wea.7656</a>, 2025.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib30"><label>Lembo et al.(2026)Lembo, Messori, Faranda, Galfi, Graversen, and
Pons</label><mixed-citation>
      
Lembo, V., Messori, G., Faranda, D., Galfi, V. M., Graversen, R. G., and Pons, F. E.: Concurrent heat waves and their linkage to large-scale meridional heat transports through planetary-scale waves, Weather Clim. Dynam., 7, 453–473, <a href="https://doi.org/10.5194/wcd-7-453-2026" target="_blank">https://doi.org/10.5194/wcd-7-453-2026</a>, 2026.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib31"><label>Livezey and Chen(1983)</label><mixed-citation>
      
Livezey, R. E. and Chen, W. Y.: Statistical Field Significance and its
Determination by Monte Carlo Techniques, Mon. Weather Rev., 111,
46–59, <a href="https://doi.org/10.1175/1520-0493(1983)111&lt;0046:SFSAID&gt;2.0.CO;2" target="_blank">https://doi.org/10.1175/1520-0493(1983)111&lt;0046:SFSAID&gt;2.0.CO;2</a>, 1983.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib32"><label>Monnin et al.(2022)Monnin, Kretschmer, and
Polichtchouk</label><mixed-citation>
      
Monnin, E., Kretschmer, M., and Polichtchouk, I.: The role of the timing of
sudden stratospheric warmings for precipitation and temperature anomalies in
Europe, Int. J. Climatol., 42, 3448–3462, <a href="https://doi.org/10.1002/joc.7426" target="_blank">https://doi.org/10.1002/joc.7426</a>, 2022.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib33"><label>Mrkvička and Myllymäki(2023)</label><mixed-citation>
      
Mrkvička, T. and Myllymäki, M.: False discovery rate envelopes,
Stat. Comput., 33, 109, <a href="https://doi.org/10.1007/s11222-023-10275-7" target="_blank">https://doi.org/10.1007/s11222-023-10275-7</a>, 2023.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib34"><label>Noble(2009)</label><mixed-citation>
      
Noble, W. S.: How does multiple testing correction work?, Nat. Biotechnol., 27,
1135–1137, <a href="https://doi.org/10.1038/nbt1209-1135" target="_blank">https://doi.org/10.1038/nbt1209-1135</a>, 2009.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib35"><label>Oehrlein et al.(2021)Oehrlein, Polvani, Sun, and
Deser</label><mixed-citation>
      
Oehrlein, J., Polvani, L. M., Sun, L., and Deser, C.: How well do we know the
surface impact of sudden stratospheric warmings?, Geophys. Res. Lett., 48,
e2021GL095&thinsp;493, <a href="https://doi.org/10.1029/2021GL095493" target="_blank">https://doi.org/10.1029/2021GL095493</a>, 2021.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib36"><label>Pappenberger et al.(2019)Pappenberger, Cloke, and
Baugh</label><mixed-citation>
      
Pappenberger, F., Cloke, H. L., and Baugh, C. A.: Cartograms for use in
forecasting weather-driven natural hazards, Cartogr. J., 56,
134–145, <a href="https://doi.org/10.1080/00087041.2018.1534358" target="_blank">https://doi.org/10.1080/00087041.2018.1534358</a>, 2019.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib37"><label>Schutte et al.(2026a)Schutte, Olivetti, Krouma, and
Messori</label><mixed-citation>
      
Schutte, M. K., Olivetti, L., Krouma, M., and Messori, G.: Stratospheric and
tropospheric contributions to North American cold temperatures and subsequent
North Atlantic jet stream anomalies, Weather Clim. Ext., 52, 100901,
<a href="https://doi.org/10.1016/j.wace.2026.100901" target="_blank">https://doi.org/10.1016/j.wace.2026.100901</a>, 2026a.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib38"><label>Schutte et al.(2026b)Schutte, Olivetti, Pons, and
Messori</label><mixed-citation>
      
Schutte, M. K., Olivetti, L., Pons, F. M. E., and Messori, G.: Code and data
for “The Stippled Gridpoints are Statistically Significant: (Mis)uses of
False Discovery Rate Correction for Geospatial Data”, Zenodo [code and data set], <a href="https://doi.org/10.5281/zenodo.20280930" target="_blank">https://doi.org/10.5281/zenodo.20280930</a>, 2026b.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib39"><label>Seneviratne et al.(2021)Seneviratne, Zhang, Adnan, Badi, Dereczynski, A. Di Luca, Iskandar, Kossin, Lewis, Otto, Pinto, Satoh, Vicente-Serrano, Wehner, and Zhou</label><mixed-citation>
      
Seneviratne, S., Zhang, X., Adnan, M., Badi, W., Dereczynski, C., A. Di Luca,
S. G., Iskandar, I., Kossin, J., Lewis, S., Otto, F., Pinto, I., Satoh, M.,
Vicente-Serrano, S., Wehner, M., and Zhou, B.: Weather and Climate Extreme
Events in a Changing Climate, in: Climate Change 2021: The Physical Science
Basis. Contribution of Working Group I to the Sixth Assessment Report of the
Intergovernmental Panel on Climate Change, edited by: Masson-Delmotte, V., Zhai, P., Pirani, A., Connors, S. L., Péan, C., Berger, S., Caud, N., Chen, Y., Goldfarb, L., Gomis, M. I., Huang, M., Leitzell, K., Lonnoy, E.,  Matthews, J. B. R., Maycock, T. K., Waterfield, T., Yelekçi, O., Yu, R., and Zhou, B., Cambridge University Press, 1513–1766, <a href="https://doi.org/10.1017/9781009157896.013" target="_blank">https://doi.org/10.1017/9781009157896.013</a>, 2021.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib40"><label>Storey(2002)</label><mixed-citation>
      
Storey, J. D.: A Direct Approach to False Discovery Rates, J. Roy. Stat. Soc. B, 64, 479–498,
<a href="https://doi.org/10.1111/1467-9868.00346" target="_blank">https://doi.org/10.1111/1467-9868.00346</a>, 2002.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib41"><label>Sun et al.(2015)Sun, Reich, Tony Cai, Guindani, and
Schwartzman</label><mixed-citation>
      
Sun, W., Reich, B. J., Tony Cai, T., Guindani, M., and Schwartzman, A.: False
discovery control in large-scale spatial multiple testing, J. Roy. Stat. Soc. B, 77, 59–83, <a href="https://doi.org/10.1111/rssb.12064" target="_blank">https://doi.org/10.1111/rssb.12064</a>, 2015.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib42"><label>Taggart and Wilke(2025)</label><mixed-citation>
      
Taggart, R. J. and Wilke, D. J.: Warnings based on risk matrices: a coherent
framework with consistent evaluation, Nat. Hazards Earth Syst. Sci., 25,
2657–2677, <a href="https://doi.org/10.5194/nhess-25-2657-2025" target="_blank">https://doi.org/10.5194/nhess-25-2657-2025</a>, 2025.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib43"><label>von Storch and Zwiers(1999)</label><mixed-citation>
      
von Storch, H. and Zwiers, F. W.: Statistical Analysis in Climate Research,
Cambridge University Press, Cambridge, <a href="https://doi.org/10.1017/CBO9780511612336" target="_blank">https://doi.org/10.1017/CBO9780511612336</a>, 1999.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib44"><label>Walker(1914)</label><mixed-citation>
      
Walker, G. T.: Correlation in seasonal variations of weather, III: On the
criterion for the reality of relationships or periodicities, Memoirs of the
India Meteorological Department, 21, 13–15, 1914.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib45"><label>Wasserstein et al.(2019)Wasserstein, Schirm, and
Lazar</label><mixed-citation>
      
Wasserstein, R. L., Schirm, A. L., and Lazar, N. A.: Moving to a world beyond
“<i>p</i> &lt; 0.05”, Am. Stat., 73, 1–19,
<a href="https://doi.org/10.1080/00031305.2019.1583913" target="_blank">https://doi.org/10.1080/00031305.2019.1583913</a>, 2019.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib46"><label>Wilks(2006)</label><mixed-citation>
      
Wilks, D. S.: On “Field Significance” and the False Discovery
Rate, J. Appl. Meteorol. Clim., 45, 1181–1189,
<a href="https://doi.org/10.1175/JAM2404.1" target="_blank">https://doi.org/10.1175/JAM2404.1</a>, 2006.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib47"><label>Wilks(2011)</label><mixed-citation>
      
Wilks, D. S.: Statistical Methods in the Atmospheric Sciences, in: vol. 100 of International Geophysics, 3rd Edn.,Academic Press, ISBN 978-0-12-385022-5,
2011.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib48"><label>Wilks(2016)</label><mixed-citation>
      
Wilks, D. S.: “The Stippling Shows Statistically Significant Grid Points”:
How Research Results are Routinely Overstated and Overinterpreted, and What
to Do about It, B. Am. Meteorol. Soc., 97, 2263–2273,
<a href="https://doi.org/10.1175/BAMS-D-15-00267.1" target="_blank">https://doi.org/10.1175/BAMS-D-15-00267.1</a>, 2016.

    </mixed-citation></ref-html>--></article>
