Deep learning for multi-year ENSO forecasts

A hybrid technique to enhance the rainfall-runoff prediction of physical and data-driven model: a case study of Upper Narmada River Sub-basin, India

Accurate streamflow prediction is crucial for effective water resource management and planning. This study aims to enhance streamflow simulation accuracy in the data-scarce Upper Narmada River Basin (UNB) by proposing a novel hybrid approach, ANNHybrid, which combines a physically-based model (WEAP) with a data-driven model (ANN). The WEAP model was calibrated and validated using observed streamflow data, while the ANN model was trained and tested using meteorological variables and simulated streamflow. The ANNHybrid model integrates simulated flow from both WEAP and ANN to improve prediction accuracy. The results demonstrate that the ANNHybrid model outperforms the standalone WEAP and ANN models, with higher NSE values of 95.5% and 92.3% during training and testing periods, respectively, along with an impressive R2 value of 0.96. The improved streamflow predictions can support better decision-making related to water allocation, reservoir operations, and flood and drought risk assessment. The novelty of this research lies in the development of the ANNHybrid model, which leverages the strengths of both physically-based and data-driven approaches to enhance streamflow simulation accuracy in data-limited regions. The proposed methodology offers a promising tool for sustainable water management strategies in the UNB and other similar catchments.

Machine learning-based extreme event attribution

The observed increase in extreme weather has prompted recent methodological advances in extreme event attribution. We propose a machine learning-based approach that uses convolutional neural networks to create dynamically consistent counterfactual versions of historical extreme events under different levels of global mean temperature (GMT). We apply this technique to one recent extreme heat event (southcentral North America 2023) and several historical events that have been previously analyzed using established attribution methods. We estimate that temperatures during the southcentral North America event were 1.18° to 1.42°C warmer because of global warming and that similar events will occur 0.14 to 0.60 times per year at 2.0°C above preindustrial levels of GMT. Additionally, we find that the learned relationships between daily temperature and GMT are influenced by the seasonality of the forced temperature response and the daily meteorological conditions. Our results broadly agree with other attribution techniques, suggesting that machine learning can be used to perform rapid, low-cost attribution of extreme events.

Preseason maize and wheat yield forecasts for early warning of crop failure

Provided the considerable logistical challenges of anticipatory action and disaster response programs, there is a need for early warning of crop failures at lead times of six to twelve months. But crop yield forecasts at these lead times are virtually nonexistent. By leveraging recent advances in climate forecasting, we demonstrate that global preseason crop yield forecasts are not only possible but are skillful over considerable portions of cropland. Globally, maize and wheat forecasts are skillful at lead times of up to a year ahead of harvest for 15% and 30% of harvested areas, respectively. Forecasts are most skillful in Southeast Africa and Southeast Asia for maize and parts of South and Central Asia, Australia, and Southeast South America for wheat. Wheat forecasts, furthermore, remain skillful at lead times of over 18 months ahead of harvest in some locations. Our results demonstrate that the potential for preseason crop yield forecasts is greater than previously appreciated.

Enhanced nitrous oxide emission factors due to climate change increase the mitigation challenge in the agricultural sector

Effective nitrogen fertilizer management is crucial for reducing nitrous oxide (N2O) emissions while ensuring food security within planetary boundaries. However, climate change might also interact with management practices to alter N2O emission and emission factors (EFs), adding further uncertainties to estimating mitigation potentials. Here, we developed a new hybrid modeling framework that integrates a machine learning model with an ensemble of eight process-based models to project EFs under different climate and nitrogen policy scenarios. Our findings reveal that EFs are dynamically modulated by environmental changes, including climate, soil properties, and nitrogen management practices. Under low-ambition nitrogen regulation policies, EF would increase from 1.18%-1.22% in 2010 to 1.27%-1.34% by 2050, representing a relative increase of 4.4%-11.4% and exceeding the IPCC tier-1 EF of 1%. This trend is particularly pronounced in tropical and subtropical regions with high nitrogen inputs, where EFs could increase by 0.14%-0.35% (relative increase of 11.9%-17%). In contrast, high-ambition policies have the potential to mitigate the increases in EF caused by climate change, possibly leading to slight decreases in EFs. Furthermore, our results demonstrate that global EFs are expected to continue rising due to warming and regional drying-wetting cycles, even in the absence of changes in nitrogen management practices. This asymmetrical influence of nitrogen fertilizers on EFs, driven by climate change, underscores the urgent need for immediate N2O emission reductions and further assessments of mitigation potentials. This hybrid modeling framework offers a computationally efficient approach to projecting future N2O emissions across various climate, soil, and nitrogen management scenarios, facilitating socio-economic assessments and policy-making efforts.

Monthly climate prediction using deep convolutional neural network and long short-term memory

Climate change affects plant growth, food production, ecosystems, sustainable socio-economic development, and human health. The different artificial intelligence models are proposed to simulate climate parameters of Jinan city in China, include artificial neural network (ANN), recurrent NN (RNN), long short-term memory neural network (LSTM), deep convolutional NN (CNN), and CNN-LSTM. These models are used to forecast six climatic factors on a monthly ahead. The climate data for 72 years (1 January 1951-31 December 2022) used in this study include monthly average atmospheric temperature, extreme minimum atmospheric temperature, extreme maximum atmospheric temperature, precipitation, average relative humidity, and sunlight hours. The time series of 12 month delayed data are used as input signals to the models. The efficiency of the proposed models are examined utilizing diverse evaluation criteria namely mean absolute error, root mean square error (RMSE), and correlation coefficient (R). The modeling result inherits that the proposed hybrid CNN-LSTM model achieves a greater accuracy than other compared models. The hybrid CNN-LSTM model significantly reduces the forecasting error compared to the models for the one month time step ahead. For instance, the RMSE values of the ANN, RNN, LSTM, CNN, and CNN-LSTM models for monthly average atmospheric temperature in the forecasting stage are 2.0669, 1.4416, 1.3482, 0.8015 and 0.6292 °C, respectively. The findings of climate simulations shows the potential of CNN-LSTM models to improve climate forecasting. Climate prediction will contribute to meteorological disaster prevention and reduction, as well as flood control and drought resistance.

Enhancing tropical cyclone intensity forecasting with explainable deep learning integrating satellite observations and numerical model outputs

Tropical cyclone (TC) intensity change forecasting remains challenging due to the lack of understanding of the interactions between TC changes and environmental parameters, and the high uncertainties resulting from climate change. This study proposed hybrid convolutional neural networks (hybrid-CNN), which effectively combined satellite-based spatial characteristics and numerical prediction model outputs, to forecast TC intensity with lead times of 24, 48, and 72 h. The models were validated against best track data by TC category and phase and compared with the Korea Meteorological Administrator (KMA)-based TC forecasts. The hybrid-CNN-based forecasts outperformed KMA-based forecasts, exhibiting up to 22%, 110%, and 7% improvement in skill scores for the 24-, 48-, and 72-h forecasts, respectively. For rapid intensification cases, the models exhibited improvements of 62%, 87%, and 50% over KMA-based forecasts for the three lead times. Moreover, explainable deep learning demonstrated hybrid-CNN's potential in predicting TC intensity and contributing to the TC forecasting field.

Explainable El Niño predictability from climate mode interactions

The El Niño-Southern Oscillation (ENSO) provides most of the global seasonal climate forecast skill1-3, yet, quantifying the sources of skilful predictions is a long-standing challenge4-7. Different sources of predictability affect ENSO evolution, leading to distinct global effects. Artificial intelligence forecasts offer promising advancements but linking their skill to specific physical processes is not yet possible8-10, limiting our understanding of the dynamics underpinning the advancements. Here we show that an extended nonlinear recharge oscillator (XRO) model shows skilful ENSO forecasts at lead times up to 16-18 months, better than global climate models and comparable to the most skilful artificial intelligence forecasts. The XRO parsimoniously incorporates the core ENSO dynamics and ENSO's seasonally modulated interactions with other modes of variability in the global oceans. The intrinsic enhancement of ENSO's long-range forecast skill is traceable to the initial conditions of other climate modes by means of their memory and interactions with ENSO and is quantifiable in terms of these modes' contributions to ENSO amplitude. Reforecasts using the XRO trained on climate model output show that reduced biases in both model ENSO dynamics and in climate mode interactions can lead to more skilful ENSO forecasts. The XRO framework's holistic treatment of ENSO's global multi-timescale interactions highlights promising targets for improving ENSO simulations and forecasts.

Bibliometric Analysis of Weather Radar Research from 1945 to 2024: Formations, Developments, and Trends

In the development of meteorological detection technology and services, weather radar undoubtedly plays a pivotal role, especially in the monitoring and early warning of severe convective weather events, where it serves an irreplaceable function. This research delves into the landscape of weather radar research from 1945 to 2024, employing scientometric methods to investigate 13,981 publications from the Web of Science (WoS) core collection database. This study aims to unravel, for the first time, the foundational structures shaping the knowledge domain of weather radar over an 80-year period, exploring general features, collaboration, co-citation, and keyword co-occurrence. Key findings reveal a significant surge in both publications and citations post-1990, peaking in 2022 with 1083 publications and 13832 citations, signaling sustained growth and interest in the field after a period of stagnation. The United States, China, and European countries emerge as key drivers of weather radar research, with robust international collaboration playing a pivotal role in the field's rapid evolution. Analysis uncovers 30 distinct co-citation clusters, showcasing the progression of weather radar knowledge structures. Notably, deep learning emerges as a dynamic cluster, garnering attention and yielding substantial outcomes in contemporary research efforts. Over eight decades, the focus of weather radar investigations has transitioned from hardware and software enhancements to Artificial Intelligence (AI) technology integration and multifunctional applications across diverse scenarios. This study identifies four key areas for future research: leveraging AI technology, advancing all-weather observation techniques, enhancing system refinement, and fostering networked collaborative observation technologies. This research endeavors to support academics by offering an in-depth comprehension of the progression of weather radar research. The findings can be a valuable resource for scholars in efficiently locating pertinent publications and journals. Furthermore, policymakers can rely on the insights gleaned from this study as a well-organized reference point.

Can three-dimensional nitrate structure be reconstructed from surface information with artificial intelligence? - A proof-of-concept study

Nitrate is one of the essential variables in the ocean that is a primary control of the upper ocean pelagic ecosystem. Its three-dimensional (3D) structure is vital for understanding the dynamic and ecosystem. Although several gridded nitrate products exist, the possibility of reconstructing the 3D structure of nitrate from surface data has never been exploited. In this study, we employed two advanced artificial intelligence (AI) networks, U-net and Earthformer, to reconstruct nitrate concentration in the Indian Ocean from surface data. Simulation from an ecosystem model was utilized as the labeling data to train and test the AI networks, with wind vectors, wind stress, sea surface temperature, sea surface chlorophyll-a, solar radiation, and precipitation as the input. We compared the performance of two networks and different pre-processing methods. With the input features decomposed into climatology and anomaly components, the Earthformer achieved optimal reconstruction results with a lower normalized mean square error (NRMSE = 0.1591), spatially and temporally, outperforming U-net (NRMSE = 0.2007) and the climatology prediction (NRMSE = 0.2089). Furthermore, Earthformer was more capable of identifying interannual nitrate anomalies. With a network interpretation technique, we quantified the spatio-temporal importance of every input feature in the best case (Earthformer with decomposed inputs). The influence of different input features on nitrate concentration in the adjacent Java Sea exhibited seasonal variation, stronger than the interannual one. The feature importance highlighted the role of dynamic factors, particularly the wind, matching our understanding of the dynamic controls of the ecosystem. Our reconstruction and network interpretation technique can be extended to other ecosystem variables, providing new possibilities in studies of marine environment and ecology from an AI perspective.

Gauge-Optimal Approximate Learning for Small Data Classification

Small data learning problems are characterized by a significant discrepancy between the limited number of response variable observations and the large feature space dimension. In this setting, the common learning tools struggle to identify the features important for the classification task from those that bear no relevant information and cannot derive an appropriate learning rule that allows discriminating among different classes. As a potential solution to this problem, here we exploit the idea of reducing and rotating the feature space in a lower-dimensional gauge and propose the gauge-optimal approximate learning (GOAL) algorithm, which provides an analytically tractable joint solution to the dimension reduction, feature segmentation, and classification problems for small data learning problems. We prove that the optimal solution of the GOAL algorithm consists in piecewise-linear functions in the Euclidean space and that it can be approximated through a monotonically convergent algorithm that presents-under the assumption of a discrete segmentation of the feature space-a closed-form solution for each optimization substep and an overall linear iteration cost scaling. The GOAL algorithm has been compared to other state-of-the-art machine learning tools on both synthetic data and challenging real-world applications from climate science and bioinformatics (i.e., prediction of the El Niño Southern Oscillation and inference of epigenetically induced gene-activity networks from limited experimental data). The experimental results show that the proposed algorithm outperforms the reported best competitors for these problems in both learning performance and computational cost.

Spatiotemporal characteristics of meteorological drought events in 34 major global river basins during 1901-2021

Meteorological drought is a crucial driver of various types of droughts; thus, identifying the spatiotemporal characteristics of meteorological drought at the basin scale has implications for ecological and water resource security. However, differences in drought characteristics between river basins have not been clearly elucidated. In this study, we identify and compare meteorological drought events in 34 major river basins worldwide using a three-dimensional drought-clustering algorithm based on the standardised precipitation evapotranspiration index on a 12-month scale from 1901 to 2021. Despite synchronous increases in precipitation and potential evapotranspiration (PET), with precipitation increasing by more than three times the PET, 47 % (16/34) of the basins showed a tendency towards drought in over half their basin areas. Drought events occurred frequently, with more than half identified as short-term droughts (lasting less than or equal to three months). Small basins had a larger drought impact area, with major drought events often originating from the basin boundaries and migrating towards the basin centre. Meteorological droughts were driven by changes in sea surface temperature (SST), especially the El Niño Southern Oscillation (ENSO) or other climate indices. Anomalies in SST and atmospheric circulation caused by ENSO events may have led to altered climate patterns in different basins, resulting in drought events. These results provide important insights into the characteristics and mechanisms of meteorological droughts in different river basins worldwide.

Coupling machine learning and physical modelling for predicting runoff at catchment scale

In this paper, we present an approach that combines data-driven and physical modelling for predicting the runoff occurrence and volume at catchment scale. With that aim, we first estimated the runoff volume from recorded storms aided by the Green-Ampt infiltration model. Then, we used machine learning algorithms, namely LightGBM (LGBM) and Deep Neural Network (DNN), to predict the outputs of the physical model fed on a set of atmospheric variables (relative humidity, temperature, atmospheric pressure, and wind velocity) collected before or immediately after the beginning of the storm. Results for a small urban catchment in Madrid show DNN performed better in predicting the runoff occurrence and volume. Moreover, enriching the input primary atmospheric variables with auxiliary variables (e.g., storm intensity data recorded during the first hour, or rain volume and intensity estimates obtained from auxiliary regression methods) largely increased the model performance. We show in this manuscript data-driven algorithms shaped by physical criteria can be successfully generated by allowing the data-driven algorithm learn from the output of physical models. It represents a novel approach for physics-informed data-driven algorithms shifting from common practices in hydrological modelling through machine learning.

Artificial intelligence predicts normal summer monsoon rainfall for India in 2023

Inaccuracy in the All Indian Summer Monsoon Rainfall (AISMR) forecast has major repercussions for India's economy and people's daily lives. Improving the accuracy of AISMR forecasts remains a challenge. An attempt is made here to address this problem by taking advantage of recent advances in machine learning techniques. The data-driven models trained with historical AISMR data, the Niño3.4 index, and categorical Indian Ocean Dipole values outperform the traditional physical models, and the best-performing model predicts that the 2023 AISMR will be roughly 790 mm, which is typical of a normal monsoon year.

Explained predictions of strong eastern Pacific El Niño events using deep learning

Global and regional impacts of El Niño-Southern Oscillation (ENSO) are sensitive to the details of the pattern of anomalous ocean warming and cooling, such as the contrasts between the eastern and central Pacific. However, skillful prediction of such ENSO diversity remains a challenge even a few months in advance. Here, we present an experimental forecast with a deep learning model (IGP-UHM AI model v1.0) for the E (eastern Pacific) and C (central Pacific) ENSO diversity indices, specialized on the onset of strong eastern Pacific El Niño events by including a classification output. We find that higher ENSO nonlinearity is associated with better skill, with potential implications for ENSO predictability in a warming climate. When initialized in May 2023, our model predicts the persistence of El Niño conditions in the eastern Pacific into 2024, but with decreasing strength, similar to 2015-2016 but much weaker than 1997-1998. In contrast to the more typical El Niño development in 1997 and 2015, in addition to the ongoing eastern Pacific warming, an eXplainable Artificial Intelligence analysis for 2023 identifies weak warm surface, increased sea level and westerly wind anomalies in the western Pacific as precursors, countered by warm surface and southerly wind anomalies in the northern Atlantic.

Predicting Gas Adsorption without the Knowledge of Pore Structures: A Machine Learning Method Based on Classical Density Functional Theory

Predicting gas adsorption from the pore structure is an intuitive and widely used methodology in adsorption. However, in real-world systems, the structural information is usually very complicated and can only be approximately obtained from the characterization data. In this work, we developed a machine learning (ML) method to predict gas adsorption form the raw characterization data of N2 adsorption. The ML method is modeled by a convolutional neural network and trained by a large number of data that are generated from a classical density functional theory, and the model gives a very accurate prediction of Ar adsorption. Though the training set is limited to modeling slit pores, the model can be applied to three-dimensional structured pores and real-world materials. The good agreements suggest that there is a universal relationship among adsorption isotherms of different adsorbates, which could be captured by the ML model.

Wildfire precursors show complementary predictability in different timescales

In most of the world, conditions conducive to wildfires are becoming more prevalent. Net carbon emissions from wildfires contribute to a positive climate feedback that needs to be monitored, quantified, and predicted. Here we use a causal inference approach to evaluate the influence of top-down weather and bottom-up fuel precursors on wildfires. The top-down dominance on wildfires is more widespread than bottom-up dominance, accounting for 73.3% and 26.7% of regions, respectively. The top-down precursors dominate in the tropical rainforests, mid-latitudes, and eastern Siberian boreal forests. The bottom-up precursors dominate in North American and European boreal forests, and African and Australian savannahs. Our study identifies areas where wildfires are governed by fuel conditions and hence where fuel management practices may be more effective. Moreover, our study also highlights that top-down and bottom-up precursors show complementary wildfire predictability across timescales. Seasonal or interannual predictions are feasible in regions where bottom-up precursors dominate.

Extreme Events Prediction from Nonlocal Partial Information in a Spatiotemporally Chaotic Microcavity Laser

The forecasting of high-dimensional, spatiotemporal nonlinear systems has made tremendous progress with the advent of model-free machine learning techniques. However, in real systems it is not always possible to have all the information needed; only partial information is available for learning and forecasting. This can be due to insufficient temporal or spatial samplings, to inaccessible variables, or to noisy training data. Here, we show that it is nevertheless possible to forecast extreme event occurrences in incomplete experimental recordings from a spatiotemporally chaotic microcavity laser using reservoir computing. Selecting regions of maximum transfer entropy, we show that it is possible to get higher forecasting accuracy using nonlocal data vs local data, thus allowing greater warning times of at least twice the time horizon predicted from the nonlinear local Lyapunov exponent.

Tropical Cyclone Intensity Probabilistic Forecasting System Based on Deep Learning

Tropical cyclones (TC) are one of the extreme disasters that have the most significant impact on human beings. Unfortunately, intensity forecasting of TC has been a difficult and bottleneck in weather forecasting. Recently, deep learning-based intensity forecasting of TC has shown the potential to surpass traditional methods. However, due to the Earth system’s complexity, nonlinearity, and chaotic effects, there is inherent uncertainty in weather forecasting. Besides, previous studies have not quantified the uncertainty, which is necessary for decision-making and risk assessment. This study proposes an intelligent system based on deep learning, PTCIF, to quantify this uncertainty based on multimodal meteorological data, which, to our knowledge, is the first study to assess the uncertainty of TC based on a deep learning approach. In this study, probabilistic forecasts are made for the intensity of 6–24 hours. Experimental results show that our proposed method is comparable to the forecast performance of weather forecast centers in terms of deterministic forecasts. Moreover, reliable prediction intervals and probabilistic forecasts can be obtained, which is vital for disaster warning and is expected to be a complement to operational models.

Taking advantage of quasi-periodic signals for S2S operational forecast from a perspective of deep learning

The quasi-periodic signals in the earth system could be the predictability source for sub-seasonal to seasonal (S2S) climate prediction because of the connections among the lead-lag time of those signals. The Madden-Julian Oscillation (MJO) is a typical quasi-periodic signal, which is the dominant S2S variability in the tropics. Besides, significantly periodic features in terms of both intensity and location are identified in 10-40 days for the concurrent variation of the subtropical and polar jet streams over Asia in this study. So far, those signals contribute less and are not fully applied to the S2S prediction. The deep learning (DL) approach, especially the long-short term memory (LSTM) networks, has the ability to take advantage of the information at the previous time to improve the prediction after then. This study presents the application of the DL in the postprocessing of S2S prediction using quasi-periodic signals predicted by the operational model to improve the prediction of minimum 2-m air temperature over Asia. With the help of deep learning, it finds the best weights for the ensemble predictions, and the quasi-periodic signals in the atmosphere can further benefit the S2S operational prediction.

A self-attention-based neural network for three-dimensional multivariate modeling and its skillful ENSO predictions

Large biases and uncertainties remain in real-time predictions of El Niño-Southern Oscillation (ENSO) using process-based dynamical models; recent advances in data-driven deep learning algorithms provide a promising mean to achieve superior skill in the tropical Pacific sea surface temperature (SST) modeling. Here, a specific self-attention-based neural network model is developed for ENSO predictions based on the much sought-after Transformer model, named 3D-Geoformer, which is used to predict three-dimensional (3D) upper-ocean temperature anomalies and wind stress anomalies. This purely data-driven and time-space attention-enhanced model achieves surprisingly high correlation skills for Niño 3.4 SST anomaly predictions made 18 months in advance and initiated beginning in boreal spring. Further, sensitivity experiments demonstrate that the 3D-Geoformer model can depict the evolution of upper-ocean temperature and the coupled ocean-atmosphere dynamics following the Bjerknes feedback mechanism during ENSO cycles. Such successful realizations of the self-attention-based model in ENSO predictions indicate its great potential for multidimensional spatiotemporal modeling in geoscience.

Deciphering urban traffic impacts on air quality by deep learning and emission inventory

Air pollution is a major obstacle to future sustainability, and traffic pollution has become a large drag on the sustainable developments of future metropolises. Here, combined with the large volume of real-time monitoring data, we propose a deep learning model, iDeepAir, to predict surface-level PM_2.5 concentration in Shanghai megacity and link with MEIC emission inventory creatively to decipher urban traffic impacts on air quality. Our model exhibits high-fidelity in reproducing pollutant concentrations and reduces the MAE from 25.355 µg/m³ to 12.283 µg/m³ compared with other models. And identifies the ranking of major factors, local meteorological conditions have become a nonnegligible factor. Layer-wise relevance propagation (LRP) is used here to enhance the interpretability of the model and we visualize and analyze the reasons for the different correlation between traffic density and PM_2.5 concentration in various regions of Shanghai. Meanwhile, As the strict and effective industrial emission reduction measurements implementing in China, the contribution of urban traffic to PM_2.5 formation calculated by combining MEIC emission inventory and LRP is gradually increasing from 18.03% in 2011 to 24.37% in 2017 in Shanghai, and the impact of traffic emissions would be ever-prominent in 2030 according to our prediction. We also infer that the promotion of vehicular electrification would achieve further alleviation of PM_2.5 about 8.45% by 2030 gradually. These insights are of great significance to provide the decision-making basis for accurate and high-efficient traffic management and urban pollution control, and eventually benefit people's lives and high-quality sustainable developments of cities.

Explainable deep learning for insights in El Niño and river flows

The El Niño Southern Oscillation (ENSO) is a semi-periodic fluctuation in sea surface temperature (SST) over the tropical central and eastern Pacific Ocean that influences interannual variability in regional hydrology across the world through long-range dependence or teleconnections. Recent research has demonstrated the value of Deep Learning (DL) methods for improving ENSO prediction as well as Complex Networks (CN) for understanding teleconnections. However, gaps in predictive understanding of ENSO-driven river flows include the black box nature of DL, the use of simple ENSO indices to describe a complex phenomenon and translating DL-based ENSO predictions to river flow predictions. Here we show that eXplainable DL (XDL) methods, based on saliency maps, can extract interpretable predictive information contained in global SST and discover SST information regions and dependence structures relevant for river flows which, in tandem with climate network constructions, enable improved predictive understanding. Our results reveal additional information content in global SST beyond ENSO indices, develop understanding of how SSTs influence river flows, and generate improved river flow prediction, including uncertainty estimation. Observations, reanalysis data, and earth system model simulations are used to demonstrate the value of the XDL-CN based methods for future interannual and decadal scale climate projections.

The AI gambit: leveraging artificial intelligence to combat climate change-opportunities, challenges, and recommendations

In this article, we analyse the role that artificial intelligence (AI) could play, and is playing, to combat global climate change. We identify two crucial opportunities that AI offers in this domain: it can help improve and expand current understanding of climate change, and it can contribute to combatting the climate crisis effectively. However, the development of AI also raises two sets of problems when considering climate change: the possible exacerbation of social and ethical challenges already associated with AI, and the contribution to climate change of the greenhouse gases emitted by training data and computation-intensive AI systems. We assess the carbon footprint of AI research, and the factors that influence AI's greenhouse gas (GHG) emissions in this domain. We find that the carbon footprint of AI research may be significant and highlight the need for more evidence concerning the trade-off between the GHG emissions generated by AI research and the energy and resource efficiency gains that AI can offer. In light of our analysis, we argue that leveraging the opportunities offered by AI for global climate change whilst limiting its risks is a gambit which requires responsive, evidence-based, and effective governance to become a winning strategy. We conclude by identifying the European Union as being especially well-placed to play a leading role in this policy response and provide 13 recommendations that are designed to identify and harness the opportunities of AI for combatting climate change, while reducing its impact on the environment.

Multi-task machine learning improves multi-seasonal prediction of the Indian Ocean Dipole

As one of the most predominant interannual variabilities, the Indian Ocean Dipole (IOD) exerts great socio-economic impacts globally, especially on Asia, Africa, and Australia. While enormous efforts have been made since its discovery to improve both climate models and statistical methods for better prediction, current skills in IOD predictions are mostly limited up to three months ahead. Here, we challenge this long-standing problem using a multi-task deep learning model that we name MTL-NET. Hindcasts of the IOD events during the past four decades indicate that the MTL-NET can predict the IOD well up to 7-month ahead, outperforming most of world-class dynamical models used for comparison in this study. Moreover, the MTL-NET can help assess the importance of different predictors and correctly capture the nonlinear relationships between the IOD and predictors. Given its merits, the MTL-NET is demonstrated to be an efficient model for improved IOD prediction.

LSCIDMR: Large-Scale Satellite Cloud Image Database for Meteorological Research

People can infer the weather from clouds. Various weather phenomena are linked inextricably to clouds, which can be observed by meteorological satellites. Thus, cloud images obtained by meteorological satellites can be used to identify different weather phenomena to provide meteorological status and future projections. How to classify and recognize cloud images automatically, especially with deep learning, is an interesting topic. Generally speaking, large-scale training data are essential for deep learning. However, there is no such cloud images database to date. Thus, we propose a large-scale cloud image database for meteorological research (LSCIDMR). To the best of our knowledge, it is the first publicly available satellite cloud image benchmark database for meteorological research, in which weather systems are linked directly with the cloud images. LSCIDMR contains 104 390 high-resolution images, covering 11 classes with two different annotation methods: 1) single-label annotation and 2) multiple-label annotation, called LSCIDMR-S and LSCIDMR-M, respectively. The labels are annotated manually, and we obtain a total of 414 221 multiple labels and 40 625 single labels. Several representative deep learning methods are evaluated on the proposed LSCIDMR, and the results can serve as useful baselines for future research. Furthermore, experimental results demonstrate that it is possible to learn effective deep learning models from a sufficiently large image database for the cloud image classification.

Physics captured by data-based methods in El Niño prediction

On average once every four years, the Tropical Pacific warms considerably during events called El Niño, leading to weather disruptions over many regions on Earth. Recent machine-learning approaches to El Niño prediction, in particular, Convolutional Neural Networks (CNNs), have shown a surprisingly high skill at relatively long lead times. In an attempt to understand this high skill, we here use data from distorted physics simulations with the intermediate-complexity Zebiak-Cane model to determine what aspects of El Niño physics are represented in a specific CNN-based classification method. We find that the CNN can adequately correct for distortions in the ocean adjustment processes, but that the machine-learning method has far more trouble in dealing with distortions in upwelling feedback strength.

Solving transparency in drought forecasting using attention models

Droughts are one of the most devastating and recurring natural disaster due to a multitude of reasons. Among the different drought studies, drought forecasting is one of the key aspects of effective drought management. The occurrence of droughts is related to a multitude of factors which is a combination of hydro-meteorological and climatic factors. These variables are non-linear in nature, and neural networks have been found to effectively forecast drought. However, classical neural nets often succumb to over-fitting due to various lag components among the variables and therefore, the emergence of new deep learning and explainable models can effectively solve this problem. The present study uses an Attention-based model to forecast meteorological droughts (Standard Precipitation Index) at short-term forecast range (1-3 months) for five sites situated in Eastern Australia. The main aim of the work is to interpret the model outcomes and examine how a deep neural network achieves the forecasting results. The plots show the importance of the variables along with its short-term and long-term dependencies at different lead times. The results indicate the importance of large-scale climatic indices at different sequence dependencies specific to the study site, thus providing an example of the necessity to build a spatio-temporal explainable AI model for drought forecasting. The use of such interpretable models would help the decision-makers and planners to use data-driven models as an effective measure to forecast droughts as they provide transparency and trust while using these models.

Subsurface Temperature Reconstruction for the Global Ocean from 1993 to 2020 Using Satellite Observations and Deep Learning

The reconstruction of the ocean’s 3D thermal structure is essential to the study of ocean interior processes and global climate change. Satellite remote sensing technology can collect large-scale, high-resolution ocean observation data, but only at the surface layer. Based on empirical statistical and artificial intelligence models, deep ocean remote sensing techniques allow us to retrieve and reconstruct the 3D ocean temperature structure by combining surface remote sensing observations with in situ float observations. This study proposed a new deep learning method, Convolutional Long Short-Term Memory (ConvLSTM) neural networks, which combines multisource remote sensing observations and Argo gridded data to reconstruct and produce a new long-time-series global ocean subsurface temperature (ST) dataset for the upper 2000 m from 1993 to 2020, which is named the Deep Ocean Remote Sensing (DORS) product. The data-driven ConvLSTM model can learn the spatiotemporal features of ocean observation data, significantly improves the model’s robustness and generalization ability, and outperforms the LighGBM model for the data reconstruction. The validation results show our DORS dataset has high accuracy with an average R2 and RMSE of 0.99/0.34 °C compared to the Argo gridded dataset, and the average R2 and NRMSE validated by the EN4-Profile dataset over the time series are 0.94/0.05 °C. Furthermore, the ST structure between DORS and Argo has good consistency in the 3D spatial morphology and distribution pattern, indicating that the DORS dataset has high quality and strong reliability, and well fills the pre-Argo data gaps. We effectively track the global ocean warming in the upper 2000 m from 1993 to 2020 based on the DORS dataset, and we further examine and understand the spatial patterns, evolution trends, and vertical characteristics of global ST changes. From 1993 to 2020, the average global ocean temperature warming trend is 0.063 °C/decade for the upper 2000 m. The 3D temperature trends revealed significant spatial heterogeneity across different ocean basins. Since 2005, the warming signal has become more significant in the subsurface and deeper ocean. From a remote sensing standpoint, the DORS product can provide new and robust data support for ocean interior process and climate change studies.

Dynamic prediction of global monthly burned area with hybrid deep neural networks

Wildfires not only severely damage the natural environment and global ecological balance but also cause substantial losses to global forest resources and human lives and property. Unprecedented fire events such as Australia's bushfires have alerted us to the fact that wildfire prediction is a critical scientific problem for fire management. Therefore, robust, long-lead models and dynamic predictions of wildfire are valuable for global fire prevention. However, despite decades of effort, the dynamic, effective, and accurate prediction of wildfire remains problematic. There is great uncertainty in predicting the future based on historical and existing spatiotemporal sequence data, but with advances in deep learning algorithms, solutions to prediction problems are being developed. Here, we present a dynamic prediction model of global burned area of wildfire employing a deep neural network (DNN) approach that produces effective wildfire forecasts based on historical time series predictors and satellite-based burned area products. A hybrid DNN that combines long short-term memory and a two-dimensional convolutional neural network (CNN2D-LSTM) was proposed, and CNN2D-LSTM model candidates with four different architectures were designed and compared to construct the optimal architecture for fire prediction. The proposed model was also shown to outperform convolutional neural networks (CNNs) and the fully connected long short-term memory (FcLSTM) approach using the refined index of agreement and evaluation metrics. We produced monthly global burned area spatiotemporal prediction maps and adequately reflected the seasonal peak in fire activity and highly fire-prone areas. Our combined CNN2D-LSTM approach can effectively predict the global burned area of wildfires 1 month in advance and can be generalized to provide seasonal estimates of global fire risk.

Variability of ENSO Forecast Skill in 2-Year Global Reforecasts Over the 20th Century

In order to explore temporal changes of predictability of El Niño Southern Oscillation (ENSO), a novel set of global biennial climate reforecasts for the historical period 1901-2010 has been generated using a modern initialized coupled forecasting system. We find distinct periods of enhanced long-range skill at the beginning and at the end of the twentieth century, and an extended multi-decadal epoch of reduced skill during the 1930s-1950s. Once the forecast skill extends beyond the first spring barrier, the predictability limit is much enhanced and our results provide support for the feasibility of skillful ENSO forecasts up to 18 months. Changes in the mean state, variability (amplitude), persistence, seasonal cycle and predictability suggest that multi-decadal variations in the dynamical characteristics of ENSO rather than the data coverage and quality of the observations have primarily driven the reported non-monotonic skill modulations.

Inversion of Ocean Subsurface Temperature and Salinity Fields Based on Spatio-Temporal Correlation

Ocean observation is essential for studying ocean dynamics, climate change, and carbon cycles. Due to the difficulty and high cost of in situ observations, existing ocean observations are inadequate, and satellite observations are mostly surface observations. Previous work has not adequately considered the spatio-temporal correlation within the ocean itself. This paper proposes a new method—convolutional long short-term memory network (ConvLSTM)—for the inversion of the ocean subsurface temperature and salinity fields with the sea surface satellite observations (sea surface temperature, sea surface salinity, sea surface height, and sea surface wind) and subsurface Argo reanalyze data. Given the time dependence and spatial correlation of the ocean dynamic parameters, the ConvLSTM model can improve inversion models’ robustness and generalizability by considering ocean variability’s significant spatial and temporal correlation characteristics. Taking the 2018 results as an example, our average inversion results in an overall normalized root mean square error (NRMSE) of 0.0568 °C/0.0027 PSS and a correlation coefficient (R) of 0.9819/0.9997 for subsurface temperature (ST)/subsurface salinity (SS). The results show that SSTA, SSSA SSHA, and SSWA together are valuable parameters for obtaining accurate ST/SS estimates, and the use of multiple channels in shallow seas is effective. This study demonstrates that ConvLSTM is superior in modeling the subsurface temperature and salinity fields, fully taking global ocean data’s spatial and temporal correlation into account, and outperforms the classic random forest and LSTM approaches in predicting subsurface temperature and salinity fields.

Forecasting large-scale circulation regimes using deformable convolutional neural networks and global spatiotemporal climate data

Classifying the state of the atmosphere into a finite number of large-scale circulation regimes is a popular way of investigating teleconnections, the predictability of severe weather events, and climate change. Here, we investigate a supervised machine learning approach based on deformable convolutional neural networks (deCNNs) and transfer learning to forecast the North Atlantic-European weather regimes during extended boreal winter for 1–15 days into the future. We apply state-of-the-art interpretation techniques from the machine learning literature to attribute particular regions of interest or potential teleconnections relevant for any given weather cluster prediction or regime transition. We demonstrate superior forecasting performance relative to several classical meteorological benchmarks, as well as logistic regression and random forests. Due to its wider field of view, we also observe deCNN achieving considerably better performance than regular convolutional neural networks at lead times beyond 5–6 days. Finally, we find transfer learning to be of paramount importance, similar to previous data-driven atmospheric forecasting studies.

MIMO: A Unified Spatio-Temporal Model for Multi-Scale Sea Surface Temperature Prediction

Sea surface temperature (SST) is a crucial factor that affects global climate and marine activities. Predicting SST at different temporal scales benefits various applications, from short-term SST prediction for weather forecasting to long-term SST prediction for analyzing El Niño–Southern Oscillation (ENSO). However, existing approaches for SST prediction train separate models for different temporal scales, which is inefficient and cannot take advantage of the correlations among the temperatures of different scales to improve the prediction performance. In this work, we propose a unified spatio-temporal model termed the Multi-In and Multi-Out (MIMO) model to predict SST at different scales. MIMO is an encoder–decoder model, where the encoder learns spatio-temporal features from the SST data of multiple scales, and fuses the learned features with a Cross Scale Fusion (CSF) operation. The decoder utilizes the learned features from the encoder to adaptively predict the SST of different scales. To our best knowledge, this is the first work to predict SST at different temporal scales simultaneously with a single model. According to the experimental evaluation on the Optimum Interpolation SST (OISST) dataset, MIMO achieves the state-of-the-art prediction performance.

Wadenow: A Matlab Toolbox for Early Forecasting of the Velocity Trend of a Rainfall-Triggered Landslide by Means of Continuous Wavelet Transform and Deep Learning

A procedure aimed at forecasting the velocity trend of a landslide for a period of some hours to one or two days is proposed here together with its MATLAB implementation. The method is based on continuous wavelet transform (CWT) and convolutional neural network (CNN) applied to rainfall and velocity time series provided by a real-time monitoring system. It is aimed at recognizing the conditions that induce a strong increase, or even a significant decrease, in the average velocity of the unstable slope. For each evaluation time, the rainfall and velocity scalograms related to the previous days (e.g., two weeks) are computed by means of CWT. A CNN recognizes the velocity trend defined in the training stage corresponds to these scalograms. In this way, forecasts about the start, persistence, and end of a critical event can be provided to the decision makers. An application of the toolbox to a landslide (Perarolo di Cadore landslide, Eastern Alps, Italy) is also briefly described to show how the parameters can be chosen in a real case and the corresponding performance.

Spatiotemporal neural network with attention mechanism for El Niño forecasts

To learn spatiotemporal representations and anomaly predictions from geophysical data, we propose STANet, a spatiotemporal neural network with a trainable attention mechanism, and apply it to El Niño predictions for long-lead forecasts. The STANet makes two critical architectural improvements: it learns spatial features globally by expanding the network's receptive field and encodes long-term sequential features with visual attention using a stateful long-short term memory network. The STANet conducts multitask learning of Nino3.4 index prediction and calendar month classification for predicted indices. In a comparison of the proposed STANet performance with the state-of-the-art model, the accuracy of the 12-month forecast lead correlation coefficient was improved by 5.8% and 13% for Nino3.4 index prediction and corresponding temporal classification, respectively. Furthermore, the spatially attentive regions for the strong El Niño events displayed spatial relationships consistent with the revealed precursor for El Niño occurrence, indicating that the proposed STANet provides good understanding of the spatiotemporal behavior of global sea surface temperature and oceanic heat content for El Niño evolution.

High precision reconstruction of silicon photonics chaos with stacked CNN-LSTM neural networks

Silicon-based optical chaos has many advantages, such as compatibility with complementary metal oxide semiconductor (CMOS) integration processes, ultra-small size, and high bandwidth. Generally, it is challenging to reconstruct chaos accurately because of its initial sensitivity and high complexity. Here, a stacked convolutional neural network (CNN)-long short-term memory (LSTM) neural network model is proposed to reconstruct optical chaos with high accuracy. Our network model combines the advantages of both CNN and LSTM modules. Further, a theoretical model of integrated silicon photonics micro-cavity is introduced to generate chaotic time series for use in chaotic reconstruction experiments. Accordingly, we reconstructed the one-dimensional, two-dimensional, and three-dimensional chaos. The experimental results show that our model outperforms the LSTM, gated recurrent unit (GRU), and CNN models in terms of MSE, MAE, and R-squared metrics. For example, the proposed model has the best value of this metric, with a maximum improvement of 83.29% and 49.66%. Furthermore, 1D, 2D, and 3D chaos were all significantly improved with the reconstruction tasks.

Hotspots of Predictability: Identifying Regions of High Precipitation Predictability at Seasonal Timescales From Limited Time Series Observations

Precipitation prediction at seasonal timescales is important for planning and management of water resources as well as preparedness for hazards such as floods, droughts and wildfires. Quantifying predictability is quite challenging as a consequence of a large number of potential drivers, varying antecedent conditions, and small sample size of high-quality observations available at seasonal timescales, that in turn, increases prediction uncertainty and the risk of model overfitting. Here, we introduce a generalized probabilistic framework to account for these issues and assess predictability under uncertainty. We focus on prediction of winter (Nov-Mar) precipitation across the contiguous United States, using sea surface temperature-derived indices (averaged in Aug-Oct) as predictors. In our analysis we identify "predictability hotspots," which we define as regions where precipitation is inherently more predictable. Our framework estimates the entire predictive distribution of precipitation using copulas and quantifies prediction uncertainties, while employing principal component analysis for dimensionality reduction and a cross validation technique to avoid overfitting. We also evaluate how predictability changes across different quantiles of the precipitation distribution (dry, normal, wet amounts) using a multi-category 3 × 3 contingency table. Our results indicate that well-defined predictability hotspots occur in the Southwest and Southeast. Moreover, extreme dry and wet conditions are shown to be relatively more predictable compared to normal conditions. Our study may help with water resources management in several subregions of the United States and can be used to assess the fidelity of earth system models in successfully representing teleconnections and predictability.

Deep Learning for Predicting Winter Temperature in North China

It is difficult to improve the seasonal prediction skill of winter temperature over North China, owing to the complex dynamics of East Asian winter and the relatively low prediction skill level of current climate models. Deep learning (DL) may be an informative and promising tool to enhance seasonal prediction, particularly in regions where the underlying mechanisms are not clear. Here, using a DL model based on the Convolutional Neural Network (CNN), we have found that the prediction skill for North China winter temperature (NCWT) can be extended up to five months by considering the remote impact of the Northeast Pacific sea-surface temperature (SST) on North China. Based on historical simulations of winter temperatures in North China, we selected six CMIP5 models with relatively small deviations for training the CNN, and the period chosen for training was 1852–1991. The ERA5 data during 1995–2017 were utilized to evaluate the performance of the CNN. Our CNN shows the best performance in a recent 10-year period (2008–2017), showing a significantly improved level of NCWT prediction skill with a correlation skill of 0.65 at a 5-month lead time, which is much better than the forecast skill of the state-of-the-art dynamic seasonal prediction system. Heat map analysis was used to explore the possible physical mechanisms associated with the NCWT anomaly from the perspective of the CNN; the results showed that the SST over the Northeast Pacific is highly relevant to NCWT prediction. The Northeast Pacific warming in the boreal summer is related to the development of the El Niño event in the coming winter, which may induce NCWT anomalies by atmospheric teleconnection. Climate model experiments support the role of Northeast Pacific warming in the boreal summer on NCWT. The improved capability for prediction from using the CNN may help to establish the energy policy for the coming winter and reduce the economic losses from extremely cold in North China.

GAN-rcLSTM: A Deep Learning Model for Radar Echo Extrapolation

The target of radar echo extrapolation is to predict the motion and development of radar echo in the future based on historical radar observation data. For such spatiotemporal prediction problems, a deep learning method based on Long Short-Term Memory (LSTM) networks has been widely used in recent years, although such models generally suffer from weak and blurry prediction. This paper proposes two models called Residual Convolution LSTM (rcLSTM) and Generative Adversarial Networks-rcLSTM (GAN-rcLSTM): The former introduces the residual module, and the latter introduces the discriminator. We use the historical data of 2017 and 2018 in the Jiangsu region as training and test sets. Experiments show that in long sequence forecasts, our model can provide more stable and clear images, while achieving higher CSI scores.

Forecasting the June Ridge Line of the Western Pacific Subtropical High with a Machine Learning Method

The ridge line of the western Pacific subtropical high (WPSHRL) plays an important role in determining the shift in the summer rain belt in eastern China. In this study, we developed a forecast system for the June WPSHRL index based on the latest autumn and winter sea surface temperature (SST). Considering the adverse condition created by the small observed sample size, a very simple neural network (NN) model was selected to extract the non-linear relationship between input predictors (SST) and target predictands (WPSHRL) in the forecast system. In addition, some techniques were used to deal with the small sample size, enhance the stabilization of the forecast skills, and analyze the interpretability of the forecast system. The forecast experiments showed that the linear correlation coefficient between the predictions from the forecast system and their corresponding observations was around 0.6, and about three-fifths of the observed abnormal years (the years with an obviously high or low WPSHRL index) were successfully predicted. Furthermore, sensitivity experiments showed that the forecast system is relatively stable in terms of forecast skill. The above results suggest that the forecast system would be valuable in real-life applications.

The NAO Variability Prediction and Forecasting with Multiple Time Scales Driven by ENSO Using Machine Learning Approaches

Machine learning methods have now become an optional technique in Earth science research, and such data-driven solutions have also made tremendous progress in weather forecasting and climate prediction in recent years. Since climate data are typically time series, the neural network layers, which can identify the intrinsic connections between the points of the sequence and features in two-dimensional data, perform particularly well for climate prediction. The North Atlantic Oscillation (NAO) is a prominent atmospherical mode in the northern hemisphere, with the frequency change characteristic of sea level pressure (SLP) in the North Atlantic sector. One of the reasons why NAO prediction is still challenging is that NAO is also proven to be influenced by other climate circulations, the most significant of which is the interaction between El Niño-Southern Oscillation (ENSO) and NAO. Therefore, sea surface temperature (SST) in the Pacific Ocean used to characterize ENSO is also one of the factors that contribute to the evolution of NAO and can be used as an input factor to predict the NAO. In this paper, the seasonal lag correlation between ENSO and NAO is explored and analyzed. The interaction has been considered in both short-term forecasting and midterm prediction of the NAO variability. The monthly NAO index (NAOI) fluctuation is predicted using the Niño indices based on the RF-Var model, and the accuracy achieves 68% when the lead time is about three months. In addition, integrating multiple physical variables directly related to the NAO and Pacific SST, the short-term NAO forecasting is conducted using a multi-channel neural network named AccNet with trajectory gated recursive unit (TrajGRU) layer. AccNet has the ability to identify the mechanism of the high-frequency variation in several days, and the NAO variability is indicated by SLP. The loss function of AccNet is set to anomaly correlation coefficient (ACC), which is the indicator that verifies spatial correlation in geoscience. Forecasting extreme events of NAO between 2010 and 2021, AccNet presents higher flexibility compared against other structures that can capture spatial-temporal features.

Rapid quantification of intracellular polyhydroxyalkanoates via fluorescence techniques: A critical review

Polyhydroxyalkanoates (PHA) are promising bioplastics with excellent physicochemical properties and biodegradability, whereas PHA products suffer from high manufacturing costs. To reduce costs of PHA production, experiments with mixed microbial cultures and low-cost substrates have been conducted widely, where rapid and robust PHA quantification methods are necessary. Compared with traditional gas chromatography methods, PHA fluorescence quantification (PHA-FQ) methods may be quicker, safer and more suitable for modern experiments with high throughput requirements. However, practical applications of PHA-FQ methods are still limited. Therefore, this review provides a comprehensive understanding of PHA-FQ methods. Performance of PHA-staining fluorochromes, relevant spectral properties, and important staining procedures are summarized. Current developments of PHA-FQ protocols are critically reviewed. Main considerations needed to make PHA-FQ protocol reliable are comprehensively discussed. Finally, potential improvements in various aspects of PHA-FQ methods are highlighted. This review could help researchers develop more effective PHA-FQ methods and facilitate future experiments related to PHA.