Displays

The concentration of trace elements such as Ni, Co, and Cu in a stalagmite is determined by (i) the amount of these elements present in so-called organic-metal complexes (OMCs) that trap the ionic forms of such elements in the dripwater, and (ii) the amount that is able to decay from the OMCs into the aqueous phase, from where the elements can adsorb to the growing stalagmite surface (and remain captured within the stalagmite crystal structure). A statistical treatment of the decay of a population of trace element ions from OMCs allow us to model the rates at which the dripwater dropped from the roof of the cave on to the stalagmite’s surface. The problem is however made challenging due to: (i) the lack of reliable monitoring data that quantifies the relationship between OMC trace metal ion concentration and stalagmite trace metal ion concentration, and (ii) the presence of chronological uncertainties in our estimates of trace element concentrations at past time points from the depth-based measurements along the stalagmite. We present here a semi-heuristic, semi-theoretical approach that estimates dripwater rates using a theoretical model based on the population-level chemical kinetics of trace element decay from OMCs, and a heuristic choice of calibration data sets based on precipitation and temperature from nearby weather station data. Our approach is applied to trace metal data from the Heshang Cave in southeastern China, and we are able to reconstruct a driprate proxy time series — a first quantitative hydrological proxy record presented along with well-defined estimates of uncertainty.

How to cite: Goswami, B., Hartland, A., Hu, C., Hoepker, S., Fox, B. R. S., Marwan, N., and Breitenbach, S. F. M.: Paleo-drip rates from trace metal concentrations in stalagmites: An inverse modeling problem with data uncertainties, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-19959, https://doi.org/10.5194/egusphere-egu2020-19959, 2020.

The tropical Pacific has experienced a new type of El Niño, which has occurred particularly frequently during the last decade and is referred to as the central Pacific (CP) El Niño. Various coupled models with different degrees of complexities have been used to make real-time El Niño predictions, but large uncertainties still exist in the forecasts. It is still not yet known how much of the uncertainty is specifically related to the new CP type of El Niño and how much is common to both this type and the conventional Eastern Pacific (EP) type of El Niño. In this study, the deterministic performance of an El Niño-Southern Oscillation (ENSO) ensemble prediction system (EPS) is examined for these two types of El Niño. Ensemble hindcasts are performed for the nine EP El Niño events and twelve CP El Niño events that have occurred since 1950. The results show that (1) the skill scores for the EP events are significantly better than those for the CP events at all lead times; (2) the systematic forecast biases come mostly from the prediction of the CP events; and (3) the systematic error is characterized by an overly warm eastern Pacific during the spring season, indicating a stronger spring prediction barrier for the CP El Niño. Further improvements of coupled atmosphere-ocean models in CP El Niño prediction should be recognized as a major challenge and high-priority task for the climate prediction community.

How to cite: Zheng, F., Yu, J.-Y., and Zhu, J.: Contrasting the skills and biases of deterministic predictions for the two types of El Niño, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-4966, https://doi.org/10.5194/egusphere-egu2020-4966, 2020.

We propose Lacunarity as a novel recurrence quantification measure and apply it in the context of dynamical regime transitions. Many complex real-world systems exhibit abrupt regime shifts. We carry out a recurrence plot based analysis for different paradigmatic systems and thermoacoustic combustion time series in order to demonstrate the ability of our method to detect dynamical transitions on variable temporal scales. Lacunarity is usually interpreted as a measure of ‘gappiness’ of an arbitrary spatial pattern. In application to recurrence plots, it quantifies the degree of heterogenity in the temporal recurrent patterns. Our method succeeds to distinguish states of varying dynamical complexity in presence of noise and short time series length. In contrast to traditional recurrence quantifiers, no specification of minimal line lengths is required and features beyond the scope of line structures can be accounted for. Applied to acoustic pressure fluctuation time series, it captures both the rich variability in dynamical complexity and detects shifts of characteristic time scales.

How to cite: Braun, T., Marwan, N., Unni, V. R., Sujith, R. I., and Kurths, J.: Detection of Dynamical Regime Transitions with Lacunarity as a Multiscale Recurrence Quantification Measure, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-3475, https://doi.org/10.5194/egusphere-egu2020-3475, 2020.

We suggest a method for nonlinear analysis of atmospheric circulation regimes in the middle latitudes. The method is based on the kernel principal component analysis allowing to separate principal modes of dynamics entangled in data. We propose a new kernel function accounting specifics of large-scale wave patterns in the mid-latitude atmosphere. First, capabilities of the method are shown by the analysis of the 3-layer quasi-geostrophic model of the Northern hemisphere atmosphere: a statistically significant set of modes can be detected by the method from relatively short (several thousand days) time series. Next, we consider reanalysis data of wintertime geopotential height anomalies over the Northern hemisphere from 1950 to the present. The principal components obtained uncover several recurrent and persistent wave structures which are associated with different weather regimes. We find that there is a pronounced inter-annual and decadal variability in the dominance of different modes in different years. Possible climatic and external forcings which impact such variability as well as long-term predictability of anomalous weather seasons based on the obtained components are discussed.

How to cite: Mukhin, D. and Hannachi, A.: Detecting regimes of the mid-latitude atmospheric circulation by nonlinear data decomposition, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-10984, https://doi.org/10.5194/egusphere-egu2020-10984, 2020.

Whether it is turbulence fluid flows or climate variability, there is a big gap between our ability to develop understanding of underlying phenomena/processes and our ability to produce skillful predictions. We focus on near-term prediction of climate as an example. In this context, the state-of-the-art is such that we are able to predict how 30-year global averages of surface temperature will change, but we are unable to predict shorter time scale regional changes.  We investigate a range of deep learning approaches to the problem ranging from reservoir computing to deep convolutional Long Short-Term Memory network architectures. The best performing architectures are seen to be capable of predicting an Earth System Model’s leading modes of global temperature variability with prediction lead times of up to a year. This approach is proposed as a useful practical tool for climate prediction. Further insight into the difficulty of the prediction problem is provided by considering the Lorenz '63 model: Long prediction horizons seen when the system is fully observed is seen to be progressively degraded as the system is less thoroughly observed, while noting the difficulty of fully observing the earth system

How to cite: Nadiga, B.: Learnt variability as a tool for climate prediction and predictability, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-19079, https://doi.org/10.5194/egusphere-egu2020-19079, 2020.

Observations indicate that there exist two types of El Niño events: one is the EP-El Niño with a warming center in the eastern tropical Pacific, and the other is the CP-El Niño with large positive SST anomalies in the central tropical Pacific. Most current numerical models show low skills in identifying the El Niño diversity. The present study examines the dynamical properties of the ENSO forecast system NFSV-ICM which combines an intermediate complexity ENSO model (ICM) with a nonlinear forcing singular vector (NFSV)-tendency perturbation forecast model. This system is able to distinguish different types of El Niño in simulations and predictions. It is shown that the NFSV-ICM system is able to capture the horizontal distribution of the SST anomalies and their amplitudes in the mature phase of not only EP-El Niño but also CP-El Niño. At the same time, the NFSV-ICM is able to describe the evolution of SST anomalies associated with the two types of El Niño up to at least two-season lead time, while the corresponding forecasts with the ICM is only limited to at most one-season lead time. These improvements are associated with the modifications of atmospheric and ocean processes described by the ICM through the NFSV-tendency perturbations. In particular, the thermocline and zonal advection feedback are strongly modified and improve the conditions of emergence of both the EP- and CP-El Niño events. The NFSV-ICM therefore provides a useful platform for studying ENSO dynamics and predictability associated with El Niño diversities.

How to cite: Tao, L., Duan, W., and Vannitsem, S.: Improving the forecast skill of El Nino diversity: A nonlinear forcing singular vector approach, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-2113, https://doi.org/10.5194/egusphere-egu2020-2113, 2020.

Every point on the Earth’s surface is a dynamical system which behaves in a complex way while interacting with other dynamical systems. Network theory captures this feature of climate to study the collective behaviour of these interacting systems giving new insights into the problem. Recently, climate networks have been a promising approach to the study of climate phenomena such as El Niño, Indian monsoon, etc. These phenomena, however, occur over a long period of time. Weather phenomena such as tropical cyclones (TCs) that are relatively short-lived, destructive events are a major concern to life and property especially for densely populated coastlines such as in the North Indian Ocean (NIO) basin. Here, we study TCs in the NIO basin by constructing climate networks using the ERA5 Sea Surface Temperature and Air temperature at 1000 hPa. We analyze these networks using the percolation framework for the post-monsoon (October-November-December) season which experiences a high frequency of TCs every year. We find significant signatures of TCs in the network structure which appear as abrupt discontinuities in the percolation-based parameters during the period of a TC. This shows the potential of climate networks towards forecasting of tropical cyclones.

This project has received funding from the European Union’s Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie grant agreement No 813844.

How to cite: Gupta, S., Kurths, J., and Pappenberger, F.: Study of Tropical Cyclones in the North Indian Ocean basin using Percolation in Climate Networks, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-5916, https://doi.org/10.5194/egusphere-egu2020-5916, 2020.

The sensitive area of targeted observation for the short-term prediction of the vertical thermal structure in the summer Yellow Sea is investigated by utilizing the Conditional Nonlinear Optimal Perturbation (CNOP) method and a adjoint-free algorithm with the Regional Ocean Modeling System. We use a vertical integration scheme of temperature to locate the sensitive area, in which reducing the initial errors are expected to yield great improvements in vertical thermal structure prediction of the verification area. We perform a series of sensitivity experiments to evaluate the effectiveness of the identified sensitive area. Our results show that, initially adding random perturbations in the sensitive area have the greatest negative effects on the prediction than in other areas (eg. the verification area, regions east and northeast of the verification area). Moreover, Observing System Simulation Experiments (OSSEs) indicate that, eliminating the initial errors in the sensitive area can lead to a more refined prediction than in other selected areas (including the verification area itself). Our study suggests that implementing targeted observation is a feasible way to improve the short-term prediction of the vertical thermal structure in the summer Yellow Sea.

How to cite: Liu, K., Liu, J., Hu, H., Guo, W., and Cui, B.: Identifying the sensitive area in targeted observation for improving the vertical thermal structure prediction in the summer Yellow Sea, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-2408, https://doi.org/10.5194/egusphere-egu2020-2408, 2020.

The use of coupled Backward Lyapunov vectors (BLv) for ensemble forecast is demonstrated in a coupled ocean-atmosphere system of reduced order, the Modular Arbitrary Order Ocean-Atmosphere sytem (MAOOAM). It is found that the best set of BLvs to build a coupled ocean-atmosphere forecasting system are the ones associated with near-neutral or slightly negative Lyapunov exponents. This counter intuitive result is related to the fact that these sets display larger projections on the ocean variables than the others, leading to an appropriate spread for the ocean, and at the same time a rapid transfer of these errors toward the most unstable BLvs affecting predominantly the atmosphere is experienced. The latter dynamics is a natural property of any generic perturbation in nonlinear chaotic dynamical systems, allowing for a reliable spread with the atmosphere too. The implications of these results for operational ensemble forecasts in coupled ocean-atmosphere systems are discussed.  

How to cite: Vannitsem, S. and Duan, W.: On the use of near-neutral backward Lyapunov vectors to get reliable ensemble forecasts in coupled ocean-atmosphere systems, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-2901, https://doi.org/10.5194/egusphere-egu2020-2901, 2020.

The present study uses the nonlinear singular vector (NFSV) approach to identify the optimally-growing tendency perturbations of the Weather Research and Forecasting (WRF) model for tropical cyclone (TC) intensity forecasts. For nine selected TC cases, the NFSV-tendency perturbations of the WRF model, including components of potential temperature and/or moisture, are calculated when TC intensities are forecasted with a 24-hour lead time, and their respective potential temperature components are demonstrated to have more impact on the TC intensity forecasts. The perturbations coherently show barotropic structure around the central location of the TCs at the 24-hour lead time, and their dominant energies concentrate in the middle layers of the atmosphere. Moreover, such structures do not depend on TC intensities and subsequent development of the TC. The NFSV-tendency perturbations may indicate that the model uncertainty that is represented by tendency perturbations but associated with the inner-core of TCs, makes larger contributions to the TC intensity forecast uncertainty. Further analysis shows that the TC intensity forecast skill could be greatly improved as preferentially superimposing an appropriate tendency perturbation associated with the sensitivity of NFSVs to correct the model, even if using a WRF with coarse resolution.

How to cite: Qin, X., Duan, W., and Xu, H.: Sensitivity on tendency perturbations of tropical cyclone short-range intensity forecasts generated by WRF, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-2942, https://doi.org/10.5194/egusphere-egu2020-2942, 2020.

Different types of El Niño-Southern Oscillation (ENSO) predictions are sensitive to the initial errors in different key areas in the Pacific Ocean. And it is known that the prediction can be improved by removing the initial errors by using assimilation methods. However yet, few studies have quantified to what extent can different types of ENSO predictions be improved by assimilating variable in different key areas. In Hou et.al (2019), 4 types of ocean temperature initial error patterns were classified for two types of El Niño prediction. It was indicated that initial errors in the north Pacific, covering the Victoria Mode region, along with south Pacific, covering the South Pacific Meridional Mode region, and subsurface layer of western equatorial Pacific have strong influence on the ENSO prediction. Following the data analysis method and the initial error patterns they proposed, we assimilate ocean temperature in these three key areas of Pacific Ocean by using CMIP5 pi-control dataset and particle filter method. Most EP- and CP-El Niño predictions in December are improved after assimilating the ocean temperature in southeast Pacific, north Pacific and western equatorial Pacific from January to March. Specially, for the prediction ensemble which contains EP(CP)-type-1 initial errors, the EP(CP)-El Niño prediction skill raises the most after assimilating the Tropical Pacific temperature, comparing with the result of assimilating the south Pacific and north Pacific. As for the prediction ensemble which contains EP-type-2 initial errors, which present similar pattern to EP-type-1 but with opposite sign, the EP-El Niño prediction skill increases the most by assimilating the north Pacific temperature. The results verify that the initial errors in the north Pacific exert contrary influences on the ENSO prediction with that in the southeast Pacific and western tropical Pacific. In addition, the initial errors in the north Pacific is more of concern for the SST prediction in the central tropical Pacific in December, while those in the southeast Pacific and tropical western Pacific is more related to the SST prediction in the central-eastern tropical Pacific. In conclusion, to better predict the types of El Niño, attentions should be paid to the initial ocean temperature accuracy not only in the tropical Pacific but also in the north and south Pacific. 

How to cite: Hou, M. and Zhi, X.: Evaluating the effect of tropical and extratropical Pacific initial errors on two types of El Niño prediction using particle filter approach, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-3293, https://doi.org/10.5194/egusphere-egu2020-3293, 2020.

Extreme rainfall events are often coupled across long spatial distances due to atmospheric teleconnections. Revealing such linkages is important for our understanding of extreme events and related atmospheric circulation patterns, but also for enhancing the forecast skill of such events [1]. Here, we present recent results [2] on how complex networks can be employed to discover extreme rainfall teleconnections from high-resolution satellite data. Our method allows to quantitatively distinguish regional weather systems from global-scale teleconnections coupling the individual weather systems. Several lines of evidence suggest that the most relevant mechanisms for global-scale teleconnections of extreme rainfall events are related to atmospheric Rossby waves. We exemplify our approach with a focus on extreme rainfall events in the mountain regions of South-Central Asia (including Northern Pakistan and India), and show that they are statistically significantly coupled to preceding events in Europe as well as succeeding events in eastern China. An analysis of the corresponding atmospheric circulation patterns shows that a previously revealed, quasi-stationary Rossby wave termed the ‘silk road pattern’ [3] is responsible for this instance of long-range coupling between extreme rainfall events. Overall, our findings give new insights into the connections between atmospheric Rossby waves and extreme rainfall events, and thus into the potential predictability of related natural hazards. Moreover, they give promising clues in how to constrain state-of-the-art climate models with respect to their simulation of extreme rainfall.

[1] B. Hoskins: The potential for skill across the range of the seamless weather-climate prediction problem: A stimulus for our science, QJRMS 2013

[2] N. Boers, B. Goswami, A. Rheinwalt, B. Bookhagen, B. Hoskins, J. Kurths: Complex networks reveal global pattern of extreme-rainfall teleconnections, Nature 2019

[3] T. Enomoto, B. Hoskins, Y. Matsuda: The formation mechanism of the Bonin high in August, QJRMS 2003

How to cite: Boers, N., Goswami, B., Rheinwalt, A., Bookhagen, B., Hoskins, B., and Kurths, J.: Global-scale teleconnections of extreme rainfall revealed by complex networks, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-7885, https://doi.org/10.5194/egusphere-egu2020-7885, 2020.

The skills of most ENSO prediction models have declined significantly since 2000. This decline may be due to a weakening of the correlation between tropical predictors and ENSO. Moreover, the effects of extratropical ocean variability on ENSO have increased during this period. To improve ENSO predictability, we investigate the influence of the tropical-extratropical Atlantic and Pacific sea surface temperature(SST) on ENSO during the periods of pre-2000 and post-2000. We find that the influence of the northern tropical Atlantic(NTA) SST on ENSO has significantly increase since 2000. Meanwhile, there is a much earlier and stronger SST responses between NTA SST and ENSO over the central-eastern Pacific during June–July–August in the post-2000 period compared with the pre-2000 period. Furthermore, the extratropical Pacific SST predictors for ENSO still retain a ~10-month lead time after 2000. We use SST signals in the extratropical Atlantic and Pacific to predict ENSO using a statistical prediction model. These results reveal a significant improvement in ENSO prediction skills. These results indicate that the Atlantic and Pacific SSTAs can make substantial contributions to ENSO prediction, and can be further used to enhance ENSO predictability after 2000.

How to cite: Shi, L., Ding, R., and Tseng, Y.: Contributions of tropical-extratropical oceans to the prediction skill of ENSO after 2000, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-3319, https://doi.org/10.5194/egusphere-egu2020-3319, 2020.

The optimal perturbation method is a beneficial way to generate ensemble members to be used in ensemble forecasting. With orthogonal optimal perturbation, orthogonal conditional nonlinear optimal perturbations (O-CNOPs) generating initial perturbations and orthogonal nonlinear forcing singular vectors (O-NFSVs) generating model perturbations are two kinds of skillful ensemble forecasting methods. There is main disadvantage that O-CNOPs and O-NFSVs generate optimal perturbation members may need a lot of time, but in practical weather prediction, the ensemble members usually need to be generated quickly. In order to benefit from O-CNOPs and O-NFSVs, as well as considering the cost of calculation, therefore, we present a way with the big data and machine learning thinking to simplify the process of the optimal perturbation ensemble methods. Using the historical samples and their optimal perturbations to establish a database, we look for the historical sample which is analogous to what need to be forecasted currently from the database by using the convolutional neural network (CNN). In comparison with using optimization algorithm to get O-CNOPs and O-NFSVs directly, this way gets O-CNOPs and O-NFSVs faster which still obtain acceptable prediction performance. In addition, once the CNN model is trained completely, the cost of time for prediction will be saved. We illustrate the advantage by numerical simulations of a Lorenz 96 model.

Further more, based on above study, some comparison of the ensemble forecasting skill of O-CNOPs and O-NFSVs has been done, and there are three results for the reference: (1) in the early stage (1-6 days), the O-CNOPs method perform more skillfully, and in the later stage (6-12 days), the O-NFSVs method perform more skillfully; (2) within 1-5 days, if the development of analysis error is bigger than or close to the average value of the analysis error development of historical samples, the O-CNOPs method is preferred, else the O-NFSVs method is preferred; (3) within 0-3 days, if the development of energy is bigger than or close to the average value of the energy development of the historical samples, the O-CNOPs method is preferred, else the O-NFVS method is preferred. Next, further work is required to examine and explore more and deeper research using machine learning in ensemble forecasting studies of atmosphere and other systems.

How to cite: Ma, J. and Duan, W.: Preliminary application of machine learning in ensemble forecasting, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-4017, https://doi.org/10.5194/egusphere-egu2020-4017, 2020.

According to recent study [1], the current state-of-the-art climate models lack the substantial part of internal multidecadal climate signal which is observed in the 20th century surface air temperature reanalysis data as a global stadium wave (GSW). In the presented work we further investigate this phenomenon using the recently developed method [2] of empirical spatio-temporal data decomposition into linear dynamical modes (LDMs). The important property of LDMs is their ability to take into account the time scales of the system evolution (they are extracted from observed dataset by the Bayesian optimization technique) better than some other linear techniques, e.g. traditional empirical orthogonal function decomposition. Like any linear decomposition, it provides the time series of principal components and corresponding spatial patterns.
We modify the initially developed LDM decomposition to make it possible to take into account a prescribed external forcing (like CO2 emissions, sun activity etc.) and then find part of variability which may be considered as an internal climate dynamics decomposed into set of modes with different time scales, and hence may be helpful in GSW interpretation. The results of applying the method to the 20th century surface air temperature with different ways of forcing inclusion will be presented and discussed.

1. Kravtsov, S., Grimm, C., & Gu, S. (2018). Global-scale multidecadal variability missing in state-of-the-art climate models. Npj Climate and Atmospheric Science, 1(1), 34. https://doi.org/10.1038/s41612-018-0044-6
2. Gavrilov, A., Seleznev, A., Mukhin, D., Loskutov, E., Feigin, A., & Kurths, J. (2018). Linear dynamical modes as new variables for data-driven ENSO forecast. Climate Dynamics. https://doi.org/10.1007/s00382-018-4255-7

How to cite: Gavrilov, A., Kravtsov, S., Mukhin, D., Loskutov, E., and Feigin, A.: Application of linear dynamical mode decomposition to surface air temperature in 20th century, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-11463, https://doi.org/10.5194/egusphere-egu2020-11463, 2020.

      Considering the effects of initial uncertainty on the ENSO forecast, ensemble forecasts method is applied in the latest version of ENSO forecast system in National Marine Environmental Forecasting Center (NMEFC, China). The currently operational ENSO forecasts system of NMEFC is established based on the CESM model, with initialization and data assimilation.

      First, leading five Singular Vectors (SV) are obtained using the climatological SST empirical singular vector method, and a SV based ensemble forecasts system is . However, the SVs can only present the initial errors that have the fasted error growth rates in a linear assumption, while ENSO and its forecasting system both are nonlinear. So, Conditional Nonlinear Optimal Perturbations (CNOP), which is has the largest error growth at the prediction time in a nonlinear scenario, is used to replace the leading SV, while other 4 SVs are kept to construct a CNOP-SV based ensemble forecast system. The hindcasts of ENSO from 1982 to 2017 shows that, the ENSO prediction skills of both SV based and CNOP-SV based ENSO ensemble forecasts are improved when compared with the old forecasting system, moreover, the CNOP-SV based ensemble forecast system has a much larger spread, showing higher prediction skills.

How to cite: Zhou, Q., Zhang, Y., Hu, J., and Duan, W.: Applying Conditional Nonlinear Optimal Perturbation (CNOP) in the ensemble ENSO forecast system, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-6339, https://doi.org/10.5194/egusphere-egu2020-6339, 2020.

The “spring predictability barrier” (SPB) is a well-known characteristic of ENSO prediction, which has been widely studied for El Niño events. However, due to the nonlinearity of the coupled ocean–atmosphere system and the asymmetries between El Niño and La Niña, it is worthy to investigate the SPB for La Niña events and reveal their differences with El Niño. This study investigates the season-dependent predictability of sea surface temperature (SST) for La Niña events by exploring initial error growth in a perfect model scenario within the Community Earth System Model. The results show that for the prediction through the spring season, the prediction errors caused by initial errors have a season-dependent evolution and induce an SPB for La Niña events. Two types of initial errors that often yield the SPB phenomenon are identified: the first are type-1 initial errors showing positive SST errors in the central-eastern equatorial Pacific accompanied by a large positive error in the upper layers of the eastern equatorial Pacific. The second are type-2 errors presenting an SST pattern with positive errors in the southeastern equatorial Pacific and a west–east dipole pattern in the subsurface ocean. The type-1 errors exhibit an evolving mode similar to the growth phase of an El Niño-like event, while the type-2 initially experience a La Niña-like decay and then a transition to the growth phase of an El Niño-like event. Both types of initial errors cause positive prediction errors for Niño3 SST and under-predict the corresponding La Niña events. The resultant prediction errors of type-1 errors are owing to the growth of the initial errors in the upper layers of the eastern equatorial Pacific. For the type-2 errors, the prediction errors originate from the initial errors in the subsurface layers of the western equatorial Pacific. These two regions may represent the sensitive areas of targeted observation for La Niña prediction. In addition, the type-2 errors in the equatorial regions are enlarged by the recharge process from 10°N in the central Pacific during the eastward propagation. Therefore, the off-equatorial regions around 10°N in the central Pacific may represent another sensitive area of La Niña prediction. Additional observations may be prioritized in these identified sensitive areas to better predict La Niña events.

How to cite: Hu, J., Duan, W., and Zhou, Q.: Season-dependent predictability and error growth dynamics for La Nina predictions, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-6584, https://doi.org/10.5194/egusphere-egu2020-6584, 2020.

We investigate the Middle Pleistocene Transition (MPT) - a rapid change in the periodicity of the Pleistocene glacial cycles from 41 kyr to about 100 kyr, which occurred about a million years ago - using the data-driven model [1]. Here we estimate stability of the model using a novel concept of interval stability [2-4], referring to the behavior of the perturbed model during a finite time interval. In a few words we define the class of 'safe' perturbations after which the system (our data-driven model) returns back to the initial dynamical regime and 'unsafe' perturbation of minimal amplitude needed to disrupt the system.

We demonstrate that the MPT is likely associated with decreasing of the climate system's interval stability to rapid disturbances (millennial and shorter). This confirms the statement made in the paper [1] that the main factor in the onset of the long-period glacial cycles is strongly nonlinear oscillations induced by the short-scale variability.

1. D. Mukhin, A. Gavrilov, E. Loskutov, J. Kurths, A. Feigin. Bayesian Data Analysis for Revealing Causes of the Middle Pleistocene Transition. Scientific Reports, 9 7328 (2019).
2. P. Menck, J. Heitzig, N. Marwan, J. Kurths. How basin stability complements the linear-stability paradigm. Nature Phys, 9 89–92 (2013).
3. V. Klinshov, V. Nekorkin, J. Kurths. Stability threshold approach for complex dynamical systems. New Journal Physics, 18 013004 (2016).
4. V. Klinshov, S. Kirillov, J. Kurths, V. Nekorkin. Interval stability for complex systems. New Journal Physics, 18 013004 (2018).

How to cite: Loskutov, E., Vdovin, V., Gavrilov, A., Mukhin, D., and Feigin, A.: The Middle Pleistocene Transition: estimation of the interval stability by data-driven model, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-11277, https://doi.org/10.5194/egusphere-egu2020-11277, 2020.

Previous studies show that the kinetic energy of mesoscale eddies (MEs) accounts for more than 80% of the global ocean energy. The theoretical study and numerical simulation of MEs will enable us to better understand the dynamics of ocean circulation. Weiss and Grooms (2017) found that assimilating uniform observations taken over MEs is much better than assimilating a subset of observations on a regular grid for improving prediction skill of SSH associated with ocean state. In the present study, we use a conditional nonlinear optimal perturbation (CNOP) approach to investigate the sensitivity of the ocean state sea surface height (SSH) predictions on MEs with a two-layer quasi-geostrophic model and show the optimal assimilating scheme. In the study, the CNOPs of SSH predictions are first computed. It is found that, if one regards the regions covered by the grid points with large values of CNOPs as sensitive area of SSH predictions, the sensitive areas are mainly located on MEs. Furthermore, the stronger the MEs, the more the MEs grid points covered by the sensitive area. Especially, these grid points associated with sensitive areas are not uniformly distributed over the MEs. It is obvious that the predictions of SSH are quite sensitive to the initialization of MEs (especially that of the particular region of large values of CNOPs for strong MEs, rather than of the uniformly distributed grid points over MEs). Therefore, an appropriate initialization of MEs is much helpful for improving the prediction accuracy of SSH. And the CNOPs of SSH prediction here may provide useful information on how to improve initialization of MEs.

How to cite: Jiang, L. and Duan, W.: Target observation of mesoscale eddies in the ocean, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-6633, https://doi.org/10.5194/egusphere-egu2020-6633, 2020.

By analyzing the seismicity in the new time domain termed natural time [1],  the entropy changes of seismicity before major earthquakes have been studied. It was found [2-5] that the key quantity is the entropy change ΔS under time reversal, which is minimized a few months before major earthquakes such as the M9.0 Tohoku earthquake [2] on 11 March 2011 and the M8.2 Chiapas earthquake [3] in Mexico on 7 September 2017; accompanied by an abrupt increase of its fluctuations [4,5]. Here we discuss how these fluctuations may lead to a procedure through which the occurrence time of an impending mainshock can be estimated [6].

References

1. Varotsos P.A., Sarlis N.V. and Skordas E.S., Natural Time Analysis: The new view of time. Precursory Seismic Electric Signals, Earthquakes and other Complex Time-Series (Springer-Verlag, Berlin Heidelberg) 2011.

2. N. V. Sarlis, E. S. Skordas, and P. A. Varotsos, "A remarkable change of the entropy of seismicity in natural time under time reversal before the super-giant M9 Tohoku earthquake on 11 March 2011", EPL (Europhysics Letters), 124 (2018), 29001.

3. N. V. Sarlis, E. S. Skordas P. A. Varotsos, A. Ramírez-Rojas, E. L. Flores-Márquez, "Natural time analysis: On the deadly Mexico M8.2 earthquake on 7 September 2017", Physica A 506 (2018), 625-634.

4. P. A. Varotsos, N. V. Sarlis and E. S. Skordas, "Tsallis Entropy Index q and the Complexity Measure of Seismicity in Natural Time under Time Reversal before the M9 Tohoku Earthquake in 2011", Entropy 20 (2018), 757.

5. A. Ramírez-Rojas, E. L. Flores-Márquez, N. V. Sarlis and P. A. Varotsos, "The Complexity Measures Associated with the Fluctuations of the Entropy in Natural Time before the Deadly México M8.2 Earthquake on 7 September 2017", Entropy 20 (2018), 477.

6. E. S. Skordas, N. V. Sarlis and P. A. Varotsos “Identifying the occurrence time of an impending major earthquake by means of the fluctuations of the entropy change under time reversal”, EPL (Europhysics Letters), in press.

How to cite: Skordas, E. S., Sarlis, N. V., Lazaridou-Varotsos, M. S., and Varotsos, P. A.: Natural time analysis: Estimation of the occurrence time of a major earthquake from the entropy changes of the preceding seismicity., EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-7649, https://doi.org/10.5194/egusphere-egu2020-7649, 2020.

We suggest a new method for construction of data-driven dynamical models from observed multidimensional time series. The method is based on a recurrent neural network (RNN) with specific structure, which allows for the joint reconstruction of both a low-dimensional embedding for dynamical components in the data and an operator describing the low-dimensional evolution of the system. The key link of the method is a Bayesian optimization of both model structure and the hypothesis about the data generating law, which is needed for constructing the cost function for model learning.  The form of the model we propose allows us to construct a stochastic dynamical system of moderate dimension that copies dynamical properties of the original high-dimensional system. An advantage of the proposed method is the data-adaptive properties of the RNN model: it is based on the adjustable nonlinear elements and has easily scalable structure. The combination of the RNN with the Bayesian optimization procedure efficiently provides the model with statistically significant nonlinearity and dimension.
The method developed for the model optimization aims to detect the long-term connections between system’s states – the memory of the system: the cost-function used for model learning is constructed taking into account this factor. In particular, in the case of absence of interaction between the dynamical component and noise, the method provides unbiased reconstruction of the hidden deterministic system. In the opposite case when the noise has strong impact on the dynamics, the method yield a model in the form of a nonlinear stochastic map determining the Markovian process with memory. Bayesian approach used for selecting both the optimal model’s structure and the appropriate cost function allows to obtain the statistically significant inferences about the dynamical signal in data as well as its interaction with the noise components.
Data driven model derived from the relatively short time series of the QG3 model – the high dimensional nonlinear system producing chaotic behavior – is shown be able to serve as a good simulator for the QG3 LFV components. The statistically significant recurrent states of the QG3 model, i.e. the well-known teleconnections in NH, are all reproduced by the model obtained. Moreover, statistics of the residence times of the model near these states is very close to the corresponding statistics of the original QG3 model. These results demonstrate that the method can be useful in modeling the variability of the real atmosphere.

The work was supported by the Russian Science Foundation (Grant No. 19-42-04121).

How to cite: Feigin, A., Seleznev, A., Mukhin, D., Gavrilov, A., and Loskutov, E.: Bayesian recurrent neural network as a tool for reconstructing dynamical systems from multidimensional data, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-8787, https://doi.org/10.5194/egusphere-egu2020-8787, 2020.

   A typhoon is a high-impact weather phenomenon that causes serious damage to people and property when landing on the Korean peninsula. Therefore, accurate prediction of typhoon intensity and track is very important in establishing damage prevention measures. 
   KMA has applied its own typhoon initialization process (KMA bogusing) for Global Data Assimilation and Prediction System (GDPS) to produce realistic initial fields since 2010, when the Unified Model (UM) of UK Met Office (UKMO) was introduced as an operational model. If the typhoon intensity of the background is weaker than the observation, the KMA bogussing process generates horizontally spread mean sea level pressures by using TC warning center’s advisory and use it for data assimilation. 
   In June 2018, GDPS has been upgraded based on PS40 N1280 of UKMO with significantly increased its horizontal resolution from 17 km to 10 km. The new version of GDPS showed improved performance in TC intensity predictions. Since the model simulates tropical cyclone intensity strong enough, we investigated the impact of typhoon initialization on the predictability of the new version GDPS.

How to cite: Lim, J. O., Boo, K.-O., Kong, H., Jun, S., Park, J.-H., and Kang, H.-S.: Evaluation of the impact of typhoon initialization on numerical weather forecasts., EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-11974, https://doi.org/10.5194/egusphere-egu2020-11974, 2020.

Targeted observation is an appealing procedure to improve oceanic model predictions by taking additional assimilation of collected measurements. However, studies on targeted observation in the oceanic field have been largely based on modeling efforts, and there is a need for field validating observations. Here, we report the preparatory work of a field campaign, which is designed based on the identified sensitive area by the Conditional Nonlinear Optimal Perturbation (CNOP) approach, to improve the short-range summer thermal structures prediction in the Yellow Sea (YS). We firstly simulated the hindcasting (2016-2018) temperature structures in the summertime, and found that the locations of the sensitive areas are generally consistent in space for each hindcast year. Then, we introduced the technique of multiple-assimilation and the definition of time-varying sensitive area, and designed observing strategies for the YS summer campaign. Observing System Simulation Experiments (OSSEs) were conducted prior to address the plan on field campaign in the Yellow Sea in August 2019. Results show that, reducing the initial errors in the sensitive area can lead to more improvement on thermal structures prediction than that in other area.

How to cite: Liu, J., Guo, W., Cui, B., Liu, K., and Hu, H.: Targeted observations based on identified sensitive areas by CNOP to improve the thermal structures prediction in the summer Yellow Sea: preparatory work for the campaign in the field, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-12376, https://doi.org/10.5194/egusphere-egu2020-12376, 2020.

Aerosols and their transformation process in atmosphere have significant effects on climate. Transformation process is a complex combination of physical and chemical reactions. Multiple oxidizing agents and other factors, such as radiation, affect the transformation process. Characterization of these factors and their strength is a problem, where advanced methods might help to gain more understanding.

In this work, we modeled transformation of wood combustion emission measured in the environmental chamber by using causal modeling (Pearl, 2009). The aim of the study was to use state-of-the-art causal discovery methods to search causal pathways between measured variables: precursors and particle products. The data used in the modelling are introduced in Tiitta et al. (2016).

In addition to wood combustion experiments, we simulated artificial datasets to understand abilities of the model. We wanted to evaluate the accuracy of our model to confirm the correct structure between variables and reproduce the measured transformation. This helps us to understand the model performance in real datasets.

We found that model could reproduce the measured evolution well. The structure between emission parts was not completely matching to prior assumption. Usually incorrect predictors in the modeled structure are highly correlated with correct causes.

References:

Pearl, J.: Causality : Models, Reasoning and Inference., Cambridge University Press., 2009.

Tiitta et al., Atmos. Chem. Phys., 16, 13251-13269, 2016.

How to cite: Leinonen, V., Tiitta, P., Sippula, O., Czech, H., Leskinen, A., Karvanen, J., Isokääntä, S., and Mikkonen, S.: Applying causal discovery algorithm to find predictors for transformation process of wood combustion emission, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-16961, https://doi.org/10.5194/egusphere-egu2020-16961, 2020.

Focusing on the transfer and evolvement of initial perturbation from temperature of ocean to the underwater acoustic propagation,comparing with the remote sensing data,optimizing Ocean-Acoustic Coupled Model, the reliability of this model is verified. On this basis,global and regional error development experiments are carried out by adding perturbation on the initial temperature of a controlled test. The results show that after 5 days evolution of the initial temperature field,the global perturbation of the propagation loss is saturated, and the perturbation structure is basically consistent with the law of the dynamic ocean.For the target area in Kuroshio region, the initial perturbation in the upstream region is the fastest. This conclusion can provide the basis for the adaptive observation of ocean acoustics.

How to cite: Guo, W., Cui, B., and Liu, J.: The Transfer and Evolvement of Initial Temperature Perturbation in Ocean-Acoustic Coupled Process, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-20928, https://doi.org/10.5194/egusphere-egu2020-20928, 2020.

Focusing on the rapid prediction of acoustic field uncertainty in environment with temporal and spatial sound speed perturbation, evolvement of sound speed structure over time is predicted based on the ocean-acoustic coupled model to obtain the uncertainty distribution of the vertical structure of sound speed. Further, a method combining  the arbitrary polynomial chaos expansion with the empirical orthogonal function is proposed to reduce the dimensionality of uncertain parameters and to obtain the uncertainty distribution of the acoustic field. Simulations have shown that the computational complexity can be reduced by 2 orders of magnitude compared to the conventional polynomial chaos expansion while ensures the same precision. Moreover, the computational complexity is not influenced by the complexity of the sound speed profile. The acoustic field and uncertainty predicted in uncertain environment by proposed method also have been tested with the experimental data.

How to cite: Cui, B. and Guo, W.: The Rapid Uncertainty Prediction of the Ocean-Acoustic Coupled Model, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-21445, https://doi.org/10.5194/egusphere-egu2020-21445, 2020.