ENSO dynamics in current climate models: an investigation using nonlinear dimensionality reduction

[1] The temporal and spatial characteristics of decadal-scale variability in the Northern Hemisphere (NH) cool-season (October–March) Arctic precipitation are identified from both the Climate Prediction Center (CPC) Merged Analysis of Precipitation (CMAP) and the Global Precipitation Climatology Project (GPCP) precipitation data sets. This decadal variability is shown to be partly connected to the decadal-scale variations in tropical central Pacific sea surface temperatures (SSTs) that are primarily associated with a decadal modulation of the El Niño–Southern Oscillation (ENSO), i.e., transitions between periods favoring typical eastern Pacific warming (EPW) events and periods favoring central Pacific warming (CPW) events. Regression and composite analyses reveal that increases of central Pacific SSTs drive a stationary Rossby wave train that destructively interferes with the wave number-1 component of the extratropical planetary wave. This destructive interference is opposite to the mean effect of typical EPW on the extratropical planetary wave. It leads to suppressed upward propagation of wave energy into the polar stratosphere, a stronger stratospheric polar vortex, and a tendency toward a positive phase of the Arctic Oscillation (AO). The positive AO tendency is synchronized on the decadal scale with a poleward shift of the NH storm tracks, particularly in the North Atlantic. Storm track variations further induce changes in the amount of moisture transported into the Arctic by synoptic eddies. The fluctuations in the eddy moisture transport ultimately contribute to the observed decadal-scale variations in the total Arctic precipitation in the NH cool season.

Sea surface temperature (SST) variability in the equatorial Pacific plays a crucial role in shaping global climate patterns. As El Niño–Southern Oscillation (ENSO)‐related extremes intensify, understanding SST fluctuations and their connection to ENSO becomes increasingly important. However, limited instrumental records hinder the ability to simulate SST variability and distinguish natural variability from anthropogenic trends. Tree rings provide annual data essential for reconstructing past SST and ENSO dynamics. While previous research has successfully reconstructed SST variability in the central equatorial Pacific, capturing signals from the eastern Pacific remains more challenging. This study analyzes P . rodolfovasquezii trees from the central Peruvian Andes (11.72°S, 75.14°W) using ring width and stable oxygen and carbon isotopes to assess their sensitivity to Pacific SST variability. We focus on the two leading modes of SST variability derived from Empirical Orthogonal Functions: EOF1 representing SST variability in the Niño 3.4 region, and EOF2 depicting a west‐east dipole between the Niño 4 and 1 + 2 regions, explaining 66.22% and 10.79% of the variance in tropical Pacific SSTs, respectively. We developed a 129‐year multiproxy tree‐ring record. Trees' sensitivity to local and regional hydroclimate is reflected in the temporal variability of their isotopic content. Tree‐ring oxygen isotopes showed significant correlations with both SST modes, although their ability to capture EOF2 variability was limited—likely due to the trees' distance from the influence region of the Niño 1 + 2. Overall, oxygen isotopes outperformed ring width and carbon isotopes, underscoring their potential for future SST reconstructions in the tropical Andes.

This study investigates the relationship between tropical Pacific sea surface temperature (SST) variability and cyclogenesis over the Gulf Stream region of the North Atlantic. A cyclone identification scheme and Lagrangian trajectories are used to compare preferred cyclogenesis locations and precyclogenesis flow paths associated with three patterns of tropical Pacific SST variability: eastern Pacific (EP) El Niño, central Pacific (CP) El Niño, and La Niña. During EP El Niño and La Niña winters, the upper-level precyclogenesis flow takes a subtropical path over North America and Gulf Stream cyclogenesis predominantly occurs under the North Atlantic jet entrance, which is the climatologically preferred location. In contrast, during CP El Niño winters, when the warmest SST anomalies occur in the central tropical Pacific, the precyclogenesis flow takes a northern path across North America and Gulf Stream cyclogenesis tends to occur farther north under the jet exit. The shift in preferred cyclogenesis is consistent with changes in transient upstream flow perturbations, detected using potential vorticity (PV) streamer frequencies, which are associated with the stationary wave response. Compared to EP El Niño winters, CP El Niño winters exhibit fewer southward-extending streamers and cyclonic (LC2) flow behavior, resulting in precyclogenesis air bypassing the right entrance of the North Atlantic jet. Downstream, Gulf Stream cyclones penetrate deeper into high Arctic latitudes during CP El Niño winters than in other cases. The results highlight distinct signatures of tropical SST anomalies on synoptic-scale atmospheric features and could help constrain future changes in the North Atlantic storm track and the associated poleward heat transport.

In recent years, the data tends to be very large and complex and it becomes very difficult and tedious to work with large datasets containing huge number of features. That's where Dimensionality Reduction comes into play. Dimensionality Reduction is a pre-processing step in various fields such as machine learning, data mining, statistics etc. and is effective in removing irrelevant and highly redundant data. In this paper, the author's performed a vast literature survey and aims to provide an adequate application based understanding of various dimensionality reduction techniques and to work as a guide to choose right approach of Dimensionality Reduction for better performance in different applications. Here, the authors have also performed detailed experiments on two different datasets for comparative analysis between various linear and non-linear dimensionality reduction techniques to figure out the effectiveness of the techniques used. PCA, a linear dimensionality reduction technique, outperformed all other techniques used in the experiments. In fact, almost all the linear dimensionality reduction techniques outperformed the non-linear techniques on both datasets by a huge error percentage margin.

Complex-valued neural networks for nonlinear complex principal component analysis

Linear dimensionality reduction by maximizing the Chernoff distance in the transformed space

This study investigates the relationship between the central tropical Pacific (CTP) sea surface temperature (SST) and the surface air temperature (SAT) variability un-related to canonical El Niño-Southern Oscillation (ENSO) over mid-to-high latitude Eurasia during boreal summers over the past half-century. The results show that their relationship experienced a decadal shift around the early 1980s. Before the early 1980s, the Eurasian SAT-CTP SST connection was weak; after that time, the relationship became stronger, and the SAT anomalies exhibited a significant wave-like pattern over Eurasia. Such a decadal change in the Eurasian SAT-CTP SST relationship could be attributed to decadal changes in the mean state and variability of CTP SST. The warmer mean state and enhanced SST variability after the early 1980s reinforced the convective activities over the tropical Pacific, leading to significantly anomalous divergence/convergence and Rossby wave sources over the North Pacific. This outcome further excited the wave train propagating along the Northern Hemisphere zonal jet stream to northern Eurasia and then affected the surface heat fluxes and atmospheric circulations over the region, resulting in wave-like SATs over Eurasia. However, during the period before the early 1980s, the CTP SST had a weak impact on the North Pacific atmospheric circulation and was consequently not able to excite the wave train pattern to impact the Eurasian atmospheric circulation and SATs. The physical processes linking the CTP SST and Eurasian SAT are further confirmed by numerical simulations. The results of this study are valuable to understanding the variability of summer Eurasian SATs.

A theoretical analysis for comparing two linear dimensionality reduction (LDR) techniques, namely Fisher's discriminant (FD) and Loog-Duin (LD) dimensionality reduciton, is presented. The necessary and sufficient conditions for which FD and LD provide the same linear transformation are discussed and proved. To derive these conditions, it is first shown that the two criteria preserve the same maximum value after a diagonalization process is applied, and then the necessary and sufficient conditions for various cases, including coincident covariance matrices, coincident prior probabilities, and for when one of the covariances is the identity matrix. A measure for comparing the two criteria is derived from the necessary and sufficient conditions, and used to empirically show that the conditions are statistically related to the classification error for a post-processing quadratic classifier and the Chernoff distance in the transformed space.

Persistent, multi-year shifts in atmospheric circulations and their associated influence on regional climates have profound impacts on physical, biological, and socioeconomic systems. The Pacific Decadal Precession (PDP), an atmospheric mode of variability consisting of a lower tropospheric height dipole which rotates counterclockwise over several years in the North Pacific, describes a series of such shifts in atmospheric circulations. One phase of the PDP, the north-south (N-S) phase, is hypothesized to be partially driven by central tropical Pacific (CP) sea surface temperature (SST) variability, but robust assessment of this dynamical connection in climate models remains to be done. In this study, we investigate this hypothesis with analyses in both reanalysis and selected models from the Coupled Model Intercomparison Project Phase 6 (CMIP6) archive. We show that the emergence of the N-S phase is both related to and influenced by tropical Pacific decadal SST variability, specifically variability associated with CP El Niño-Southern Oscillation (ENSO) events. When examining the pre-industrial runs of the CMIP6 models, we find that most models cannot recover the characteristic cyclonic precession of the dipoles of the PDP, instead featuring only amplitude and sign changes of the N-S phase, Moreover, the models do not replicate the dynamical connections between the tropical Pacific and this North Pacific mode of climate variability. Our results suggest that primary reasons for this inconsistency are that models inaccurately simulate both the SST pattern associated with the PDP, shared low-frequency power associated with CP ENSO events, and incorrect Rossby wavetrains emanating from the tropical Pacific into the North Pacific on quasi-decadal timescales. Taken together, our analyses offer another benchmark by which to test the fidelity of the climate model simulations in capturing Pacific decadal climate variability in order to improve decadal-to-centennial climate projections.

Pronounced model-observation discrepancies in the changes of tropical Pacific zonal sea surface temperature gradient during the satellite era imply systematic model deficiencies. However, the relatively short high-quality instrumental record hampers robustly determining the response of tropical Pacific sea surface temperature to greenhouse gas increases. By adopting paleoclimate proxy records along with a series of climate model simulations, here we show that the zonal gradient is likely to decrease under sustained strong forcing. Paleoclimate proxy records indicate an overall increase of the zonal gradient over time, which has been accompanied by global-mean cooling associated with decreasing carbon dioxide concentrations. Model simulations are found to broadly reproduce the distinct contrast between warmer high carbon dioxide climates and the opposite climates, albeit with large inter-model discrepancy. The qualitative agreement among paleoclimate proxy records and modeled representations therefore lends some important credence to the sign of model-projected future tropical Pacific mean state change.

Many features of the El Nino-Southern Oscillation (ENSO) display significant interdecadal changes. These include general characteristics such as amplitude, period, and developing features, and also nonlinearities, especially the El Nino-La Nina asymmetry. A review of previous studies on the interdecadal changes in the ENSO nonlinearities is provided. In particular, the methods for measuring ENSO nonlinearities, their possible driving mechanisms, and their interdecadal changes are discussed. Two methods for measuring ENSO nonlinearities are introduced; the maximum potential intensity, which refers to the upper and lower bounds of the cold tongue temperature, and the skewness, which represents the asymmetry of a probability density function. For example, positive skewness (a strong El Nino vs. a weak La Nina) of the tropical Pacific sea surface temperature (SST) anomalies is dominant over the eastern tropical Pacific, with an increase seen during recent decades (e.g., 1980–2000). This positive skewness can be understood as a result of several nonlinear processes. These include the warming effect on both El Nino and La Nina by nonlinear dynamic heating (NDH), which intensifies El Nino and suppresses La Nina; the asymmetric negative feedback due to tropical oceanic instability waves, which has a relatively stronger influence on the La Nina event; the nonlinear physics of the ocean mixed layer; the Madden-Julian-Oscillation/Westerly-Wind-Burst and ENSO interaction; the biological-physical feedback process; and the nonlinear responses of the tropical atmospheric convection to El Nino and La Nina conditions. The skewness of the tropical eastern Pacific SST anomalies and the intensities of the above-mentioned mechanisms have both experienced clear decadal changes in a dynamically associated manner. In particular, there is a dynamic linkage between the decadal changes in the El Nino-La Nina asymmetry and those in NDH. This linkage is based on the recent decadal changes in mean climate states, which provided a favorable condition for thermocline feedback rather than for zonal advection feedback, and thus promoted the eastward propagation of the ENSO-related atmospheric and oceanic fields. The eastward propagating ENSO mode easily produces a positive NDH, resulting in asymmetric ENSO events in which El Nino conditions are stronger than La Nina conditions.

Previous studies argued that the Pacific Meridional Mode (PMM) impacts tropical cyclone (TC) genesis variability over the southeastern part of the western North Pacific (SE‐WNP). Here, we find that the statistical relationship between PMM and SE‐WNP TC genesis frequency is dominated by their co‐variability on decadal timescales. The decadal component of the PMM exhibits very similar temporal and spatial features to quasi‐decadal tropical Pacific sea surface temperature (SST) variability. The latter can affect SE‐WNP TC activity via changes in both zonal vertical wind shear and low‐level vorticity. In contrast, the interannual component of the PMM exhibits no statistically significant correlation with SE‐WNP TC genesis. Furthermore, observations show that both interannual and decadal variability of SE‐WNP TC activity are well correlated with the commonly used Niño3.4 El Niño‐Southern Oscillation index. Thus, equatorial Pacific SST variability is the dominant source of SE‐WNP TC activity predictability on different timescales.

<p>The dry half of the year from May to October over the Maritime continent is a crucial period. Because a severely dry year could lead to unprecedented damages due to forest fires. The amount of rainfall during this period is closely tied to the sea surface temperature (SST) evolution over the tropical Pacific Ocean. The future change in the tropical Pacific SST has uncertainty with majority of the climate models showing a higher warming over eastern tropical Pacific. However multiple climate models show a higher warming rate over the western tropical Pacific, known as the warm pool region. This feature can also be seen in observations over the last 120 years. The effect of this differing tropical Pacific SST evolution can have contrasting influence over the maritime continent precipitation. Therefore, this study provides a risk-based estimate of the future dry season maritime continent precipitation, which is based on the possible evolution of tropical Pacific SST through physical climate storyline approach. We provide storylines of future dry period precipitation using the outputs of 38 climate models from Coupled Model Intercomparison Project version 6 (CMIP6). Our results suggest that a higher basin wide tropical Pacific warming with higher warming in the eastern tropical Pacific than the west would lead to a drier condition over the Islands of Sumatra and Java. Whereas lower increase in basin wide tropical Pacific SST with a warmer western tropical Pacific, which we saw in observations, would lead to wetter condition over the Islands of Borneo and Neu Guinea.</p><p> </p>

Most investigations of relationships between tropical Pacific sea surface temperature anomaly (SSTA) events and regional climate patterns have assumed the teleconnections to be linear, whereby the climate patterns associated with cold SSTA events are considered to be similar in structure and morphology but opposite in sign to those linked to warm SSTA events. In contrast, and motivated by early evidence of nonlinearity in the above regard, this study identifies characteristic (i.e., composite) calendar monthly central and eastern North American precipitation patterns separately for warm and cold SSTA events in different regions of the tropical Pacific (central, eastern, west-central“horseshoe,” far western) identified through principal component analysis. The precipitation anomaly patterns are computed from an approximately 1° lat–long set of monthly station data for 1950–92. Their robustness and nonlinearity are established using local, regional, and field statistical significance tests and a va...

Analysis of population genetic variation and structure is a common practice for genome-wide studies, including association mapping, ecology, and evolution studies in several crop species. In this study, machine learning (ML) clustering methods, K-means (KM), and hierarchical clustering (HC), in combination with non-linear and linear dimensionality reduction techniques, deep autoencoder (DeepAE) and principal component analysis (PCA), were used to infer population structure and individual assignment of maize inbred lines, i.e., dent field corn (n = 97) and popcorn (n = 86). The results revealed that the HC method in combination with DeepAE-based data preprocessing (DeepAE-HC) was the most effective method to assign individuals to clusters (with 96% of correct individual assignments), whereas DeepAE-KM, PCA-HC, and PCA-KM were assigned correctly 92, 89, and 81% of the lines, respectively. These findings were consistent with both Silhouette Coefficient (SC) and Davies–Bouldin validation indexes. Notably, DeepAE-HC also had better accuracy than the Bayesian clustering method implemented in InStruct. The results of this study showed that deep learning (DL)-based dimensional reduction combined with ML clustering methods is a useful tool to determine genetically differentiated groups and to assign individuals into subpopulations in genome-wide studies without having to consider previous genetic assumptions.