{"id":304416,"date":"2024-02-26T11:24:00","date_gmt":"2024-02-26T10:24:00","guid":{"rendered":"https:\/\/climatescience.press\/?p=304416"},"modified":"2024-02-26T11:24:03","modified_gmt":"2024-02-26T10:24:03","slug":"top-climate-model-improved-to-show-enso-skill","status":"publish","type":"post","link":"https:\/\/climatescience.press\/?p=304416","title":{"rendered":"Top Climate Model Improved to Show ENSO\u00a0Skill"},"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\n\n\n

Previous posts (linked at end) discuss how the climate model from RAS (Russian Academy of Science)<\/strong> has evolved through several versions. The interest arose because of its greater ability to replicate the past temperature history.<\/strong> The model is part of the CMIP program which is now going the next step to CMIP7, and is one of the first to test with a new climate simulation. Improvements to the latest model, INMCM60, show an enhanced ability to replicate ENSO oscillations in the Pacific ocean, which have significant climate impacts world wide.<\/p>\n\n\n\n

This news comes by way of a new paper published in the Russian Journal of Numerical Analysis and Mathematical Modelling February 2024.  The title is ENSO phase locking, asymmetry and predictability in the INMCM Earth system model<\/strong>, <\/a> Seleznev et al. (2024) Excerpts in italics with my bolds and images from the article.<\/p>\n\n\n\n

Abstract:<\/strong><\/em><\/h5>\n\n\n\n

Advanced numerical climate models<\/strong> are known to exhibit biases<\/strong> in simulating some features of El Ni\u00f1o\u2013Southern Oscillation (ENSO)<\/strong> which is a key mode of inter-annual climate variability. In this study we analyze<\/strong> how two fundamental features of observed ENSO \u2013 asymmetry<\/strong> between hot and cold states and phase-locking<\/strong> to the annual cycle \u2013 are re\ufb02ected in two di\ufb00erent versions of the INMCM Earth system model (state-of-the-art Earth system model participating in the Coupled Model Intercomparison Project).<\/em><\/p>\n\n\n\n

We identify the above ENSO features<\/strong> using the conventional empirical orthogonal functions (EOF) analysis<\/strong> which is applied to both observed and simulated upper ocean heat content (OHC) data in the tropical Paci\ufb01c. We obtain that the observed tropical Paci\ufb01c OHC variability<\/strong> is described well by two leading EOF-modes which roughly re\ufb02ect the fundamental recharge-discharge mechanism of ENSO.<\/strong> These modes exhibit strong seasonal cycles associated with ENSO phase locking while the revealed nonlinear dependencies between amplitudes of these cycles re\ufb02ect ENSO asymmetry.<\/em><\/p>\n\n\n\n

We also assess and compare predictability of observed and simulated ENSO<\/strong> based on linear inverse modeling. We \ufb01nd that the improved INMCM6 model has signi\ufb01cant bene\ufb01ts in simulating described features of observed ENSO as compared with the previous INMCM5 model. The improvements of the INMCM6 model providing such bene\ufb01ts arediscussed. We argue that proper cloud parametrization scheme is crucial for accurate simulation of ENSO dynamics with numerical climate models<\/strong><\/em><\/p>\n\n\n\n

Introduction<\/strong><\/em><\/h5>\n\n\n\n

El Ni\u00f1o\u2013Southern Oscillation (ENSO) is the most prominent mode of inter-annual climate variability<\/strong> which originates in the tropical Paci\ufb01c, but has a global impact<\/strong> [41]. Accurately simulating ENSO is still a challenging task for global climate modelers [3,5,15,25]. In the comprehensive study [35] large-ensemble climate model simulations provided by the Coupled Model Intercomparison Project phases 5 (CMIP5)and 6 (CMIP6) were analyzed. It was found that the CMIP6 models signi\ufb01cantly outperform those fromCMIP5 for 8 out of 24 ENSO-relevant metrics<\/strong>, especially regarding the simulation of ENSO spatial patterns, diversity and teleconnections. Nevertheless, some important aspects<\/strong> of the observed ENSO are still not satisfactorily simulated<\/strong> by the most of state-of-the-art models [7,38,49]. In this study we are aimed at examination of how two such aspects \u2013 ENSO asymmetry and ENSO phase-locking<\/strong> to the annual cycle \u2013are re\ufb02ected in the INMCM Earth system model [44, 45].<\/em><\/p>\n\n\n\n

The asymmetry between hot (El Nino) and cold (La Nina) states<\/strong> is a fundamental feature in the observed ENSO occurrences [39]. El Ni\u00f1o events are often stronger than La Ni\u00f1a events, while the latter ones tend to be more persistent<\/strong> [10]. Such an asymmetry is generally attributed to nonlinear feedbacks between sea surface temperatures (SSTs), thermocline and winds in the tropical Paci\ufb01c [2,19,28]. The alternative conceptions highlight the role of tropical instability waves [1] and fast atmospheric processes associated with irregular zonal wind anomalies [24]. ENSO phase-locking is identi\ufb01ed as the tendency of ENSO-events to peak in boreal winter.<\/strong><\/em><\/p>\n\n\n\n

Several studies [11,17,34] argue that the phase-locking is associated with seasonal changes in thermocline depth, ocean upwelling velocity, and cloud feedback processes.<\/strong> These processes collectively contribute to the coupling strength modulation between ocean and atmosphere, which, in the context of conceptual ENSO models [4,18], provides seasonal modulation of stability (in the sense of decay rate) of the \u201cENSO oscillator\u201d. Another theory [20,42] supposes the phase-locking results from nonlinear interactions between the seasonal forcing and the inherent ENSO cycle. Both the asymmetry and phaselocking e\ufb00ects are typically captured by low-dimensional data-driven ENSO models<\/strong> [14, 21, 26, 29, 37].<\/em><\/p>\n\n\n\n

In this work we identify the ENSO features discussed above via the analysis of upper ocean heat content (OHC) variability in the the tropical Paci\ufb01c<\/strong>. The recent study [37] analyzed high-resolution reanalysis dataset of the tropical Paci\ufb01c (10N \u2013 10S, 120E \u2013 80W) OHC anomalies in the 0\u2013300 m depth<\/strong> layer using the standard empirical orthogonal function (EOF) decomposition [16]. It was found that observed OHC variability is e\ufb00ectively captured by two leading EOFs<\/strong>, which roughly describe the fundamental recharge-discharge mechanism of ENSO [18]. The time series of the corresponding principal components (PCs) demonstrate strong seasonal cycles, re\ufb02ecting ENSO phase-locking, while the revealed inter-annual nonlinear dependencies between these cycles can be associated with ENSO asymmetry [37].<\/em><\/p>\n\n\n\n

Here we apply similar analysis to the OHC data simulated by two di\ufb00erent versions of INMCM<\/strong> Earth system model. The \ufb01rst is the INMCM5<\/strong> model [45] from CMIP6, and the second is the perspective INMCM6<\/strong> [44] model with improved parameterization of clouds, large-scale condensation and aerosols.<\/strong> Along with the traditional EOF decomposition we invoke the linear inverse modeling to assess and compare predictability of ENSO from observed and simulated data.<\/em><\/p>\n\n\n\n

The paper is organized as follows.<\/strong>\u00a0Sect. 2<\/strong>\u00a0describes the\u00a0datasets<\/strong>\u00a0we analyze: OHC reanalysis dataset and OHC data obtained from the ensemble simulations of global climate with two versions of INMCM model. Data preparation, including separation of the forced and internal variability, is also discussed. The ensemble EOF analysis is represented, which is used for identifying the meaningful processes contributing to observed and simulated ENSO dynamics.\u00a0Sect. 3<\/strong>\u00a0presents the\u00a0results<\/strong>\u00a0we obtain in analyzing both\u00a0observed and simulated OHC data<\/strong>. In\u00a0Sect. 4<\/strong>\u00a0we\u00a0summarize and discuss<\/strong>\u00a0the obtained results, particularly regarding the signi\ufb01cant bene\ufb01ts of new version of INMCM model (INMCM6) in simulating key features of observed ENSO.<\/em><\/p>\n\n\n

\n
\"\"
Fig. 1: Two leading EOFs of the observed tropical Paci\ufb01c upper ocean heat content (OHC) variability<\/em><\/figcaption><\/figure>\n<\/div>\n\n
\n
\"\"
Fig. 2: Two leading EOFs of the INMCM5 ensemble of tropical Paci\ufb01c upper ocean heat content simulations<\/em><\/figcaption><\/figure>\n<\/div>\n\n
\n
\"\"
Fig. 3: The same as in Fig. 2 but for INMCM6 model simulations<\/em><\/figcaption><\/figure>\n<\/div>\n\n\n

The corresponding spatial patterns<\/strong> in Fig. 1<\/strong> have clear interpretation. The \ufb01rst contributes to the central and eastern tropical Paci\ufb01c,<\/strong> where most signi\ufb01cant variations of sea surface temperature (SST) during El Ni\u00f1o\/La Nina events occur [9]. The second<\/strong> predominates mainly in the western tropical Paci\ufb01c<\/strong> and can be associated with the OHC accumulation and discharge before and during the El Ni\u00f1o events [48].<\/em><\/p>\n\n\n\n

What we can see from Fig. 2<\/strong> is that the two leading EOFs of OHC variability simulated by the INMCM5 model do not correspond to the observed ones.<\/strong> The corresponding time series and spatial patterns exhibit smaller-scale features, as compared to those we obtain from the reanalysys data, indicating their noisier spatio-temporal nature.<\/em><\/p>\n\n\n\n

The two leading EOFs of the improved INMCM6 model (Fig. 3)<\/strong>, by contrast, capture well both the spatial and temporal features of observed EOFs.<\/strong> In the next section we focus on furtheranalysis of these EOFs assuming that they contain the most meaningful information about ENSO dynamics.<\/em><\/p>\n\n\n\n

Discussion<\/strong><\/em><\/h5>\n\n\n\n

In this study we have analyzed how two di\ufb00erent versions of the INMCM model<\/strong> [44,45] (state-of-the-art Earth system model participating in the Coupled Model Intercomparison Project, CMIP) simulate some features of El Ni\u00f1o\u2013Southern Oscillation (ENSO) which is a key mode of the global climate<\/strong>. We identi\ufb01ed the ENSO features via the EOF analysis applied to both observed and simulated upper ocean heat content(OHC) variability in the the tropical Paci\ufb01c.<\/strong> It was found that the observed tropical Paci\ufb01c OHC variability is captured well by two leading modes (EOFs) which re\ufb02ect the fundamental recharge-discharge mechanism of ENSO<\/strong> involving a recharge and discharge of OHC along the equator caused by a disequilibrium between zonal winds and zonal mean thermocline depth. These modes are phase-shifted and exhibit the strong seasonal cycles associated with ENSO phase locking.<\/strong> The inter-annual dependencies between amplitudes of the revealed ESNO seasonal cycles are strongly nonlinear which re\ufb02ects the asymmetry between hot (ElNino) and cold (La Nina) states<\/strong> of observed ENSO. We found that the INMCM5 model<\/strong> (the previous version of the INMCM model from CMIP6) poorly reproduces the leading modes<\/strong> of observed ENSO and re\ufb02ect neither the observed ENSO phase locking nor asymmetry. At the same time, the perspective INMCM6 model demonstrates signi\ufb01cant improvement in simulating<\/strong> these key features of observed ENSO. The analysis of ENSO predictability based on linear inverse modeling indicates that the improved INMCM6 model re\ufb02ects well the ENSO spring predictability barrier<\/strong> and therefore could potentially have an advantage in long range weather prediction as compared with the INMCM5.<\/em><\/p>\n\n\n\n

Such bene\ufb01ts of<\/strong> the new version of the INMCM model (INMCM6)<\/strong> in simulating observed ENSO dynamics can be provided by using more relevant parametrization of sub-grid scale processes.<\/strong> Particularly, the di\ufb00erence in the amplitude of OHC anomaly associated with ENSO between INMCM5 and INMCM6 shown in Fig.2-3 can be explained mainly by the di\ufb00erence in cloud parameterization<\/strong> in these models. In short, in INMCM5 El-Nino event leads to increase of middle and low clouds<\/strong> over central and eastern Paci\ufb01c that leads to cooling because of decrease in surface incoming shortwave radiation.<\/strong><\/em><\/p>\n\n\n\n

While\u00a0decrease in low clouds and increase in high clouds in INMCM6<\/strong>\u00a0over El-Nino region during positive phase of ENSO\u00a0lead to further upper ocean warming<\/strong>\u00a0[43]. This is consistent with the recent study [36] which argued that erroneous cloud feedback arising from a dominant contribution of low-level clouds may lead to heat \ufb02ux feedback bias in the tropical Paci\ufb01c, which play a key role in ENSO dynamics. Fast decrease in OHC in central Paci\ufb01c after El-Nino maximum in INMCM6 can probably occur because of too shallow mixed layer in equatorial Paci\ufb01c in the model, that leads to fast surface cooling after renewal of upwelling and further increase of tradewinds.\u00a0Summarizing the above we can conclude that proper cloud parameterization scheme is crucial for accurate simulation of observed ENSO with numerical climate models.<\/strong><\/em><\/p>\n\n\n

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

Background on INMCM6<\/strong><\/p>\n\n\n\n

\n
New 2023 INMCM RAS Climate Model First Results<\/a><\/blockquote>