Browse

Yinxing Liu
,
Zhiwei Zhang
,
Qingguo Yuan
, and
Wei Zhao

Abstract

Meridional heat transport induced by oceanic mesoscale eddies (EHT) plays a significant role in the heat budget of the Southern Ocean (SO) but the decadal trends in EHT and its associated mechanisms are still obscure. Here, this scientific issue is investigated by combining concurrent satellite observations and Estimating the Circulation and Climate of the Ocean, Phase II (ECCO2) reanalysis data over the 24 years between 1993 and 2016. The results reveal that the surface EHTs from both satellite and ECCO2 data consistently show decadal poleward increasing trends in the SO, particularly in the latitude band of the Antarctic Circumpolar Current (ACC). In terms of average in the ACC band, the ECCO2-derived EHT over the upper 1000 m has a linear trend of 1.1 × 10−2 PW decade−1 or 16% per decade compared with its time-mean value of 0.07 PW. Diagnostic analysis based on “mixing length” theory suggests that the decadal strengthening of eddy kinetic energy (EKE) is the dominant mechanism for the increase in EHT in the SO. By performing an energy budget analysis, we further find that the decadal increase in EKE is mainly caused by the strengthened baroclinic instability of large-scale circulation that converts more available potential energy to EKE. For the strengthened baroclinic instability in the SO, it is attributed to the increasing large-scale wind stress work on the large-scale circulation corresponding to the positive phase of the Southern Annular Mode between 1993 and 2016. The decadal trends in EHT identified here may help understand decadal variations of heat storage and sea ice extent in the SO.

Significance Statement

Oceanic mesoscale-eddy-induced meridional heat transport (EHT) is a key process of heat redistribution in the Southern Ocean (SO), but the decadal variations of EHT and the associated mechanisms remain obscure. Here, by analyzing satellite and reanalysis data between 1993 and 2016, we find that the poleward EHT has significant decadal increasing trends in the SO, particularly in the Antarctic Circumpolar Current latitude band. Further analysis suggests that the increasing EHT is mainly caused by enhanced eddy kinetic energy converted by the strengthened baroclinic instability of large-scale circulation, which is attributed to the strengthening winds modulated by the Southern Annular Mode. The above findings may improve our understanding of the decadal variations of heat storage and sea ice extent in the SO.

Restricted access
Zihan Song
,
Shang-Ping Xie
,
Lixiao Xu
,
Xiao-Tong Zheng
,
Xiaopei Lin
, and
Yu-Fan Geng

Abstract

A deep winter mixed layer forms north of the Antarctic Circumpolar Current (ACC) in the Indo-Pacific sectors, while the mixed layer depth (MLD) is shallow in the Atlantic. Using observations and a global atmospheric model, this study investigates the contribution of surface buoyancy flux and background stratification to interbasin MLD variations. The surface heat flux is decomposed into broad-scale and frontal-scale variations. At the broad scale, the meandering ACC path is accompanied by a zonal wavenumber-1 structure of sea surface temperature (SST) with a warmer Pacific than the Atlantic; under the prevailing westerly winds, this temperature contrast results in larger surface heat loss facilitating deeper MLD in the Indo-Pacific sectors than in the Atlantic. In the Indian sector, the intense ACC fronts strengthen surface heat loss compared to the Pacific. The surface freshwater flux pattern largely follows that of evaporation and reinforces the heat flux pattern, especially in the southeast Pacific. A diagnostic relationship is introduced to highlight the role of ACC’s sloping isopycnals in setting a weak submixed layer stratification north of ACC. This weak stratification varies in magnitude across basins. In the Atlantic and western Indian Oceans where the ACC is at a low latitude (∼45°S), solar heating, intrusions of subtropical gyres, and energetic mesoscale eddies together maintain relatively strong stratification. In the southeast Pacific, in comparison, the ACC reaches the southernmost latitude (56°S), far away from the subtropical front. This creates weaker stratification, allowing deep mixed layers to form, aided by surface buoyancy loss.

Restricted access
Hongpei Yang
and
Yu Du

Abstract

During the development of squall lines, low-frequency gravity waves exhibit contrasting behaviors behind and ahead of the system, corresponding to its low-level upshear and downshear sides, respectively. This study employed idealized numerical simulations to investigate how low-level shear and tilted convective heating influence waves during two distinct stages of squall-line evolution. In the initial stage, low-level shear speeds up upshear waves, while it has contrasting effects on the amplitudes of different wave modes, distinguishing it from the Doppler effect. Downshear deep tropospheric downdraft (n = 1 wave) exhibits larger amplitudes, resulting in strengthened low-level inflow and upper-level outflow. However, n = 2 wave with low-level ascent and high-level descent has higher amplitude upshear and exhibits a higher altitude of peak w values downshear, leading to the development of a more extensive upshear low-level cloud deck and a higher altitude of downshear cloud deck. In the mature stage, as the convective updraft greatly tilts rearward (upshear), stronger n = 1 waves occur behind the system, while downshear-propagating n = 2 waves exhibit larger amplitudes. These varying wave behaviors subsequently contribute to the storm-relative circulation pattern. Ahead of the squall line, stronger n = 2 waves and weaker n = 1 waves produce intense outflow concentrated at higher altitudes, along with moderate midlevel inflow and weak low-level inflow. Conversely, behind the system, the remarkable high pressure in the upper troposphere and wake low are attributed to more intense n = 1 waves. Additionally, the cloud anvil features greater width and depth rearward and is situated at higher altitudes ahead of the system due to the joint effects of n = 1 and n = 2 waves.

Significance Statement

Squall lines are a significant source of high-impact weather events, and their development has been partially explained through linear wave dynamics. While the recurrent generation of waves during squall-line evolution has been found, the differentiation of wave behavior behind and ahead of the system, as well as its implications for storm circulation, has remained unclear. This study employs idealized simulations to reveal that during different stages of convection, low-level shear and the tilting of convective heating exert contrasting effects on wave behaviors. Moreover, various wave modes exhibit distinct responses to specific factors, and their combined effect elucidates the structural discrepancies observed both rearward and forward of the convective updraft. These findings could allow a step toward a better understanding of the intricate interaction between waves and convections.

Restricted access
Matthew Patterson
,
Christopher O’Reilly
,
Jon Robson
, and
Tim Woollings

Abstract

The coupled nature of the ocean–atmosphere system frequently makes understanding the direction of causality difficult in ocean–atmosphere interactions. This study presents a method to decompose turbulent surface heat fluxes into a component which is directly forced by atmospheric circulation and a residual which is assumed to be primarily “ocean-forced.” This method is applied to the North Atlantic in a 500-yr preindustrial control run using the Met Office’s HadGEM3-GC3.1-MM model. The method shows that atmospheric circulation dominates interannual to decadal heat flux variability in the Labrador Sea, in contrast to the Gulf Stream where the ocean primarily drives the variability. An empirical orthogonal function analysis identifies several residual heat flux modes associated with variations in ocean circulation. The first of these modes is characterized by the ocean warming the atmosphere along the Gulf Stream and North Atlantic Current and the second by a dipole of cooling in the western subtropical North Atlantic and warming in the subpolar North Atlantic. Lead–lag regression analysis suggests that atmospheric circulation anomalies in prior years partly drive the ocean heat flux modes; however, there is no significant atmospheric circulation response in years following the peaks of the modes. Overall, the heat flux dynamical decomposition method provides a useful way to separate the effects of the ocean and atmosphere on heat flux and could be applied to other ocean basins and to either models or reanalysis datasets.

Significance Statement

Variability of the ocean affects atmospheric circulation and provides a source of long-term predictability for surface weather. However, the atmosphere also affects the ocean. This makes the separation of cause and effect in such atmosphere–ocean interactions difficult. This paper introduces a method to separate “turbulent heat fluxes,” the primary means by which the atmosphere and ocean influence one another, into a component driven by atmospheric variability and a component which is primarily related to ocean variability. The method is tested by applying it to a climate model simulation and is able to identify regions in which the exchange of heat between the ocean and atmosphere is dominated by atmospheric variability and regions which are dominated by the ocean.

Open access
Philine Lou Bommer
,
Marlene Kretschmer
,
Anna Hedström
,
Dilyara Bareeva
, and
Marina M.-C. Höhne

Abstract

Explainable artificial intelligence (XAI) methods shed light on the predictions of machine learning algorithms. Several different approaches exist and have already been applied in climate science. However, usually missing ground truth explanations complicate their evaluation and comparison, subsequently impeding the choice of the XAI method. Therefore, in this work, we introduce XAI evaluation in the climate context and discuss different desired explanation properties, namely, robustness, faithfulness, randomization, complexity, and localization. To this end, we chose previous work as a case study where the decade of annual-mean temperature maps is predicted. After training both a multilayer perceptron (MLP) and a convolutional neural network (CNN), multiple XAI methods are applied and their skill scores in reference to a random uniform explanation are calculated for each property. Independent of the network, we find that XAI methods such as Integrated Gradients, layerwise relevance propagation, and input times gradients exhibit considerable robustness, faithfulness, and complexity while sacrificing randomization performance. Sensitivity methods, gradient, SmoothGrad, NoiseGrad, and FusionGrad, match the robustness skill but sacrifice faithfulness and complexity for the randomization skill. We find architecture-dependent performance differences regarding robustness, complexity, and localization skills of different XAI methods, highlighting the necessity for research task-specific evaluation. Overall, our work offers an overview of different evaluation properties in the climate science context and shows how to compare and benchmark different explanation methods, assessing their suitability based on strengths and weaknesses, for the specific research problem at hand. By that, we aim to support climate researchers in the selection of a suitable XAI method.

Significance Statement

Explainable artificial intelligence (XAI) helps to understand the reasoning behind the prediction of a neural network. XAI methods have been applied in climate science to validate networks and provide new insight into physical processes. However, the increasing number of XAI methods can overwhelm practitioners, making it difficult to choose an explanation method. Since XAI methods’ results can vary, uninformed choices might cause misleading conclusions about the network decision. In this work, we introduce XAI evaluation to compare and assess the performance of explanation methods based on five desirable properties. We demonstrate that XAI evaluation reveals the strengths and weaknesses of different XAI methods. Thus, our work provides climate researchers with the tools to compare, analyze, and subsequently choose explanation methods.

Open access
Reyhaneh Rahimi
,
Praveen Ravirathinam
,
Ardeshir Ebtehaj
,
Ali Behrangi
,
Jackson Tan
, and
Vipin Kumar

Abstract

This paper presents a deep supervised learning architecture for 30-min global precipitation nowcasts with a 4-h lead time. The architecture follows a U-Net structure with convolutional long short-term memory (ConvLSTM) cells empowered by ConvLSTM-based skip connections to reduce information loss due to the pooling operation. The training uses data from the Integrated Multi-satellitE Retrievals for GPM (IMERG) and a few key drivers of precipitation from the Global Forecast System (GFS). The impacts of different training loss functions, including the mean-squared error (regression) and the focal loss (classification), on the quality of precipitation nowcasts are studied. The results indicate that the regression network performs well in capturing light precipitation (<1.6 mm h−1), while the classification network can outperform the regression counterpart for nowcasting of high-intensity precipitation (>8 mm h−1), in terms of the critical success index (CSI). It is uncovered that including the forecast variables can improve precipitation nowcasting, especially at longer lead times in both networks. Taking IMERG as a relative reference, a multiscale analysis, in terms of fractions skill score (FSS), shows that the nowcasting machine remains skillful for precipitation rate above 1 mm h−1 at the resolution of 10 km compared to 50 km for GFS. For precipitation rates greater than 4 mm h−1, only the classification network remains FSS skillful on scales greater than 50 km within a 2-h lead time.

Significance Statement

This study presents a deep neural network architecture for global precipitation nowcasting with a 4-h lead time, using sequences of past satellite precipitation data and simulations from a numerical weather prediction model. The results show that the nowcasting machine can improve short-term predictions of high-intensity global precipitation. The research outcomes will enable us to expand our understanding of how modern artificial intelligence can improve the predictability of extreme weather and benefit flood early warning systems for saving lives and properties.

Restricted access
Matthew C. Wheeler
,
Hanh Nguyen
,
Chris Lucas
,
Zhi-Weng Chua
,
Simon Grainger
,
David A. Jones
,
Michelle L. L’Heureux
,
Ben Noll
,
Tristan Meyers
,
Nicolas C. Fauchereau
,
Alexandre Peltier
,
Thea Turkington
,
Hyung-Jin Kim
, and
Takafumi Umeda
Open access
Xianghui Fang
,
Henk Dijkstra
,
Claudia Wieners
, and
Francesco Guardamagna

Abstract

As the strongest year-to-year fluctuation of the global climate system, El Niño–Southern Oscillation (ENSO) exhibits spatial–temporal diversity, which challenges the classical ENSO theories that mainly focus on the canonical eastern Pacific (EP) type. Besides, the complicated interplay between the interannual anomaly fields and the decadally varying mean state is another difficulty in current ENSO theory. To better account for these issues, the nonlinear two-region recharge paradigm model is extended to a three-region full-field conceptual model to capture the physics in the western Pacific (WP), central Pacific (CP), and EP regions. The results show that the extended conceptual model displays a rich dynamical behavior as parameters setting the efficiencies of upwelling and zonal advection are varied. The model can not only generate El Niño bursting behavior but also simulate the statistical asymmetries between the two types of El Niños and the warm and cold phases of ENSO. Finally, since both the anomaly fields and mean states are simulated by the model, it provides a simple tool to investigate their interactions. The strengthening of the upwelling efficiency, which can be seen as an analogy to a cooling thermocline associated with the oceanic tunnel to the midlatitudes, will increase the zonal gradient of the mean state temperature between the WP and EP, i.e., resembling a negative Pacific decadal oscillation (PDO) pattern along the equatorial Pacific. The influence of the zonal advection efficiency is quite the opposite, i.e., its strengthening will reduce the zonal gradient of the mean state temperature along the equatorial Pacific.

Restricted access
Erica K. Dolinar
and
Jason E. Nachamkin
Open access
Keith L. Seitter
,
Emma Tipton
, and
Paul A. T. Higgins

Abstract

There has been an increase in entrepreneurial activity within the weather, water, and climate (WWC) community over the past decade, with the potential for much more as artificial intelligence/machine learning techniques continue to develop and as new opportunities arise across the weather, climate, and ocean service enterprises. Despite indications of recent growth, this study reports on key challenges that are limiting the community’s ability to achieve the full potential of commercialization of new WWC products and services. Most of these challenges are related to the preparation of those in the WWC community for jobs in the private sector in general and entrepreneurial activities in particular. These results extend and build upon the work of others who have reported on shortcomings in the preparation of students for positions in the private sector, with this study showing that deficits in preparation and awareness of available resources affect potential entrepreneurs well into their career—most researchers are unaware of the resources available to them. Based on a synthesis of input from successful WWC entrepreneurs, many of the challenges could be greatly reduced by relatively minor adjustments to curriculums at universities and through new programs that could be offered by scientific and professional societies to help potential entrepreneurs better take advantage of existing resources as they spin up a new business.

Open access