{"id":379735,"date":"2025-05-24T19:38:51","date_gmt":"2025-05-24T17:38:51","guid":{"rendered":"https:\/\/climatescience.press\/?p=379735"},"modified":"2025-05-24T19:38:52","modified_gmt":"2025-05-24T17:38:52","slug":"ipcc-climate-models-proven-to-lack-predictive-ability","status":"publish","type":"post","link":"https:\/\/climatescience.press\/?p=379735","title":{"rendered":"IPCC Climate Models Proven to Lack Predictive\u00a0Ability"},"content":{"rendered":"\n
\"\"<\/figure>\n\n\n\n

<\/p>\n\n\n\n

From Science Matters<\/a><\/p>\n\n\n\n

By\u00a0Ron Clutz<\/a><\/p>\n\n\n

\n
\"\"<\/figure>\n<\/div>\n\n\n

The recently published paper is Are Climate Model Forecasts Useful for Policy Making?<\/strong><\/a> by Kesten C. Green and Willie Soon. Excerpts in italics with my bolds and added images.<\/p>\n\n\n\n

Effect of Variable Choice on Reliability and Predictive Validity<\/strong><\/em><\/p>\n\n\n\n

Abstract<\/strong><\/em><\/h4>\n\n\n\n

For a model to be useful for policy decisions,<\/strong> statistical fit is insufficient.<\/strong> Evidence that the model provides out-of-estimation-sample forecasts that are more accurate and reliable<\/strong> than those from plausible alternative models, including a simple benchmark, is necessary.<\/strong><\/em><\/p>\n\n\n

\n
\"\"<\/figure>\n<\/div>\n\n\n

The UN\u2019s IPCC advises governments with forecasts<\/strong> of global average temperature drawn from models based on hypotheses of causality.<\/strong> Specifically, manmade warming principally from carbon dioxide<\/strong> emissions  (Anthro) tempered by<\/strong> the effects of volcanic eruptions (Volcanic)<\/strong> and<\/strong> by variations in the  Sun\u2019s energy (Solar).<\/strong> Out-of-sample forecasts from that model<\/strong>, with and without the IPCC\u2019s favoured measure of Solar, were compared with forecasts from models that excluded human influence<\/strong> and included Volcanic and one of two independent measures of Solar. The models were used to forecast Northern Hemisphere land temperatures<\/strong> and\u2014to avoid urban heat island effects\u2014rural only temperatures. Benchmark forecasts were obtained by extrapolating estimation sample median temperatures.<\/em><\/p>\n\n\n\n

The independent solar models reduced forecast errors<\/strong> relative to those of the benchmark model for all eight combinations<\/strong> of four estimation periods and the two temperature variables tested. The models that included the IPCC\u2019s Anthro variable reduced errors for only three<\/strong> of the eight combinations and produced extreme forecast errors<\/strong> from most model estimation periods. The correlation between estimation sample statistical fit and forecast accuracy was -0.26. Further tests might identify better models: Only one extrapolation model and only two of many possible independent solar models were tested, and combinations of forecasts from different methods were not examined.<\/em><\/p>\n\n\n

\n
\"\"<\/figure>\n<\/div>\n\n\n

The anthropogenic models\u2019 unreliability would appear to void policy relevance<\/strong>. In practice, even <\/em>the models validated in this study may fail to improve accuracy relative to na\u00efve forecasts<\/strong> due to <\/em>uncertainty over the future causal variable values. Our findings emphasize that out-of-sample <\/em>forecast errors, not statistical fit, should be used to choose between models (hypotheses).<\/em><\/p>\n\n\n\n

Background<\/strong><\/em><\/h4>\n\n\n\n

In their attempts to achieve the IPCC objective of identifying a human cause<\/strong> for temperature changes\u2014specifically \u201cglobal warming\u201d\u2014the IPCC researchers have framed the problem<\/strong> as one of \u201cattributing\u201d<\/strong> changes in the Earth\u2019s temperature to the respective contributions of putative anthropogenic (\u201cAnthro\u201d)<\/strong> principally carbon dioxide emissions<\/strong> altering the composition of the atmosphere\u2014and natural<\/strong> influences\u2014principally aerosols<\/strong> from volcanic eruptions altering the composition of the atmosphere (\u201cVolcanic\u201d), and total solar irradiance<\/strong>, or TSI, variations (\u201cSolar\u201d).<\/em><\/p>\n\n\n\n

Given the task they were set, the IPCC researchers have devoted
much of their efforts into developing estimates of the Anthro variable.<\/strong><\/em><\/p>\n\n\n\n

The IPCC\u2019s most recent, AR6<\/strong>, report (IPCC, 2021) only considered one estimate of Solar<\/strong> for the purpose of attribution (Matthes et al., 2017) and made no allowance for<\/strong> the effect of urban heat islands<\/strong> on the temperature measures they used (Connolly et al., 2021, 2023; Soon et al., 2023). Moreover, a study of the statistical attribution or \u201cfingerprinting\u201d approach<\/strong> used by IPCC researchers (e.g., Allen and Tett, 1999; Hasselmann, et al., 1995; Hegerl et al., 1997; Santer et al.,1995) concluded that the approach was invalid<\/strong> (McKitrick, 2022). The IPCC authors\u2019 analyses failed to meet the assumptions of the method they used, and they failed to correctly implement the method.<\/em><\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n

In sum, the objective given to the IPCC researchers and the approach that they have taken suggests that plausible alternative hypotheses on the causes of terrestrial temperature changes<\/strong> may not<\/strong> have been adequately tested<\/strong>, as is required by the scientific method (Armstrong and Green, 2022). That concern is consistent with Armstrong and Green\u2019s (2022) observation that government sponsorship of research can create incentives that may influence researchers\u2019 choices of hypotheses to test and how they test them.<\/em><\/p>\n\n\n\n

1.1 Alternative hypotheses on Solar<\/strong><\/em><\/h5>\n\n\n\n

To address the first of the foregoing limitations in the IPCC attribution studies\u2014failure to fairly <\/em>test alternative TSI estimates\u2014Connolly et al. (2021, 2023) comprehensively reviewed alternative estimates of TSI<\/strong> covering the 169 years from 1850 to 2018. In addition to the Matthes, et al. <\/em>(2017) TSI estimates series used by the IPCC (2021)\u2014henceforth \u201cIPCC Solar\u201d\u2014Connolly et al. <\/em>(2023) identified 27 alternative Solar time series.<\/strong><\/em><\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n

The alternative estimates of Solar correlate quite well with the TSI used in the AR6 report\u2014Pearson\u2019s r values range between 0.39 and 0.97 with a median of 0.82\u2014but the degree of TSI variation in Watts per square metre (Wm-2) differs considerably between the estimates.<\/strong> The ranges of the individual alternative TSI estimate series vary between 0.49 and 4.64 Wm-2, with a median range of 1.77 Wm-2, while IPCC Solar has a range of only 0.19 Wm-2.<\/em><\/p>\n\n\n\n

In this study, we consider two plausible TSI reconstructions<\/strong> from Connolly et al. (2023). Those from Hoyt and Schatten (1993) and from Bard et al. (2000), which Connolly et al. (2023) updated to the year 20182. The former TSI record (\u201cH1993 Solar\u201d) was based on the so-called multiproxy\u2014i.e., equatorial solar rotation rate, sunspot structure, the decay rate of individual sunspots, the number of sunspots without umbrae, and the length and decay rate of the 11-yr sunspot activity cycle\u2014reconstruction of the solar irradiance history.<\/em><\/p>\n\n\n\n

1.2 Alternative hypotheses on temperature estimation<\/strong><\/em><\/h5>\n\n\n\n

The IPCC\u2019s attribution studies do not account for the direct effects of human activities on local temperatures (heat islands)<\/strong>\u2014the second weakness addressed in this study. For example, heating and cooling of building interiors, electricity generation, manufacturing, freight and transport, asphalt and concrete, and where vegetation and open water have been removed or added. Where temperature readings are taken close to such human sources of heat or absence of natural cooling, they cannot properly reflect the individual effects of human emissions of carbon dioxide, etc., that the IPCC are concerned about (their Anthro variable), the Volcanic variable, and TSI.<\/em><\/p>\n\n\n\n

To address this second limitation<\/strong> in the IPCC attribution studies, Connolly et al. (2021, 2023) developed four alternative estimates of surface temperatures that were intended to avoid heat island effects<\/strong>. They were based on rural only<\/strong> weather station readings, sea surface<\/strong> temperature readings, tree-ring<\/strong> width measurements, and glacier length<\/strong> measurements. For comparison with the approach used by the IPCC, they also<\/strong> developed an all-land temperature estimates<\/strong> series for the Northern Hemisphere.<\/em><\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n
1.5 Hypotheses tested<\/strong><\/em><\/h5>\n\n\n\n

The foregoing discussion suggests the following hypotheses, which are tested in this study.<\/em><\/p>\n\n\n\n