Forecast Time Research Articles

Patterns of alcohol use in Belgium, 2013-2020

Abstract Despite numerous efforts worldwide to reduce the harmful effects of alcohol, its impact on health remains significant, contributing significantly to the loss of healthy life years. In Belgium, alcohol use is still a major public health concern, ranking among the highest in alcohol consumption across Europe. This study aims to use available nationally representative data to estimate and evaluate the time trends of alcohol use patterns in Belgium from 2013 to 2018. We used repeated cross-sectional survey data from the Belgian Health Interview Survey (BHIS), which includes self-reported alcohol use and sociodemographic data, to create a time series of annual consumption from 2013 to 2018. This included current and former drinkers, and lifetime abstainers by year, sex, region, and age groups. We applied a Bayesian hierarchical model, using the Integrated Nested Laplace Approximation to create the time series and forecast trends for 2020. In Belgium, the prevalence of current drinking remained at a similar level of 80% from 2013 to 2020, with similar rates observed in the Flemish (FL) and Walloon (WA) regions. However, the Brussels-Capital region (BR) had a slightly lower prevalence of 70%. Trends remained relatively stable across all regions during this period. The national level of former drinkers increased from 5% in 2013 to 8% in 2020, with BR and FL showing similar increases, while WA experienced a decline. Abstainer rates rose from 15% in 2013 to 19% in 2020 nationally, with varying increases in each region. Specifically, abstainer rates in FL increased from 13% to 17% and in WA from 14% to 20%, while BR saw an increase from 19% to 22%. Overall, declining trends were evident, with men exhibiting higher prevalence rates across all categories. Time trends from the BHIS reveal consistent and high use of alcohol in Belgium over 7 years, emphasizing the ongoing public health challenge. Urgent policy interventions are needed to reduce the impact of alcohol on health Key messages • Alcohol use in Belgium is still a pressing public health issue, with high prevalence rates of current drinking in the population. Different patterns of alcohol use can be seen between regions. • Understanding the role of alcohol on population health as an attributable burden can improve health monitoring.

An Improved Neural Network Algorithm for Energy Consumption Forecasting

Accurate and efficient forecasting of energy consumption is a crucial prerequisite for effective energy planning and policymaking. The BP neural network has been widely used in forecasting, machine learning, and various other fields due to its nonlinear fitting ability. In order to improve the prediction accuracy of the BP neural network, this paper introduces the concept of forecast lead time and establishes a mathematical model accordingly. Prior to training the neural network, the input layer data are preprocessed based on the forecast lead time model. The training and forecasting results of the BP neural network when and when not considering forecast lead time are compared and verified. The findings demonstrate that the forecast lead time model can significantly improve the prediction speed and accuracy, proving to be highly applicable for short-term energy consumption forecasting.

Tropical intraseasonal oscillations as key driver and source of predictability for the 2022 Pakistan record-breaking rainfall event

In August 2022, Pakistan experienced unprecedented monsoon rains, leading to devastating floods and landslides affecting millions. While previous research has mainly focused on the contributions of seasonal and synoptic anomalies, this study elucidates the dominant influences of tropical and extratropical intraseasonal oscillations on both the occurrence and subseasonal prediction of this extreme rainfall event. Our scale-decomposed moisture budget analysis revealed that intense rainfall in Pakistan was triggered and sustained by enhanced vertical moisture transport anomalies, primarily driven by interactions between intraseasonal circulation anomalies and the prevailing background moisture field when tropical and mid-latitude systems coincided over Pakistan. Evaluation of subseasonal-to-seasonal prediction models further highlighted the critical role of tropical intraseasonal modes in causing this extreme rainfall event in Pakistan. Models that accurately predicted northward-propagating intraseasonal convection with a forecast lead time of 8–22 days demonstrated good skill in predicting the extreme event over Pakistan.

Predictive Modeling for Enhancing MOOC Completion Rates: A Case Study

In the realm of online asynchronous learning platforms, accurately tracking student performance to predictcourse completion times poses a significant challenge. Completion rates for MOOCs are typically low, with abias towards participants with higher education levels. Understanding factors such as student motivation,engagement, participation, and learning pathway design is crucial for improving student outcomes in onlinecourses. This research developed a predictive framework utilizing advanced deep learning techniques toaccurately forecast course completion times for participants enrolled in an introductory programming course("Python for Beginners" course on the Open Learning Platform of University of Moratuwa Sri Lanka). Byaccurately tracking student performance and leveraging a diverse dataset encompassing demographic andeducational variables, the research seeks to identify factors influencing course completion and predictindividual student outcomes. By utilising deep learning techniques, the prediction performance of the modelwill be improved, ultimately contributing to a more precise forecast of course completion times forparticipants. Evaluation of the model resulted in low Mean Absolute Error (MAE) of 0.0080 and low MeanSquared Error (MSE) of 0.0033 which promises the effectiveness of the developed method in accuratelypredicting course completion times for students. The findings of this study may help increase the successfulcompletion rate of such courses which are delivered in the online asynchronous mode. The study employedadvanced deep learning models optimized through Bayesian methods, highlighting the potential of thesetechniques to enhance MOOC completion rates by offering precise forecasts and actionable insights intostudent engagement. The comprehensive analysis revealed that variables such as 'Current_Lesson', 'SessionTime Category', and 'District_Score' significantly influence completion times. The robust methodologicalframework, including feature engineering, model training, and hyperparameter optimization, sets a precedentfor future research in the field. This research contributes to educational data mining and predictive analytics,offering a scalable approach to improving completion rates and educational outcomes across various onlinelearning platforms. Future research should explore incorporating real-time data and longitudinal studies toenhance model accuracy and generalizability. Additionally, addressing potential biases in the dataset, such asdemographic, prior knowledge, and resource access disparities, is essential to ensure the fair and equitableapplication of the model across diverse student populations. Expanding the research to include a wider rangeof courses and institutions will further validate the model's robustness and applicability in differenteducational contexts.

Advancing Subseasonal Surface Air Temperature and Heat Wave Prediction Skill in China by Incorporating Scale Interaction in a Deep Learning Model

AbstractAccurate subseasonal predictions of high surface air temperature (SAT) and heat wave events 10–30 days in advance are crucial for mitigating the risks of extreme weather; however, they pose a challenge for current operational models. In this study, we trained a convolutional neural network (CNN)‐based deep learning model to exploit the modulations in China's SAT by precursor signals across different timescales to improve predictions of future SAT and heat wave events. This CNN model demonstrated superior capability in capturing the evolution of SAT anomalies and the occurrence of heat wave events with forecast lead times beyond 20 days, compared with that of the operational models of the China Meteorological Administration and European Centre for Medium‐Range Weather Forecasts. Explainability analysis highlighted that subseasonal SAT predictability in China is driven primarily by large‐scale intraseasonal perturbations from both lower‐ and higher‐latitude regions of Eurasia and interannual variability.

Space–Time Forecasting of Heating & Cooling Energy Needs as an Energy Poverty Measure in Romania

Lack of access to basic energy services, known as energy poverty, remains felt in the country, with seasonal changes and an economic divide. The frameworks to measure energy poverty differ spatially and temporally, with climate change and behavioral culture being the essential influencing factors. This paper is focused on heating and cooling energy demands, which can be defined as an energy poverty metric for the propensity to be at risk of energy poverty caused by climate regime. Employing sophisticated statistical space–time forecasting tools, we build a model incorporating spatial and temporal energy consumption volatility across Romanian regions at the NUTS3 level. The model considers climatic conditions and raw data from 45 years (1979–2023) of cooling and heating degree days to determine local trajectories for the next nine years. Identifying high-energy-poverty-risk areas in our research can provide valuable insights for policymakers, enabling them to develop targeted plans for eliminating energy poverty and ensuring equitable access to heating and cooling. The results underline the necessity of differentiated approaches in energy policies and add value to the general understanding of energy poverty issues and conditions, considering the Romanian climatic and socio-economic context.

Forecasting to support EMS tactical planning: what is important and what is not.

Forecasting emergency medical service (EMS) call volumes is critical for resource allocation and planning. The development of many commercial and free software packages has made a variety of forecasting methods accessible. Practitioners, however, are left with little guidance on selecting the most appropriate method for their needs. Using 5 years of data from 3 cities in Alberta, we compute exponential smoothing and benchmark forecasts for 8-hour periods for each ambulance station catchment area and with a forecast horizon of two weeks-a spatio-temporal resolution appropriate for tactical planning. The methods that we consider differ on three spectra: the number and type of time-series components, whether forecasts are computed individually or jointly, and the way in which forecasts at a specific resolution are converted to forecasts at the resolution of interest. We find that it is important to include a weekly seasonal component when forecasting EMS demand. Multiplicative seasonality, however, shows no benefit over additive seasonality. Adding other time-series components (e.g., trend, ARMA errors, Box-Cox transformation) does not improve performance. Spatial resolutions of station catchment area and lower, and temporal resolution of 4-24 hours perform similarly. We adapt an existing hierarchical forecasting framework to a two-dimensional spatio-temporal hierarchy, but find that hierarchical reconciliation of forecasts does not improve performance at the forecast resolution of interest for tactical planning. Neither does jointly forecasting time series. We show that added complexity does not materially improve forecasting performance. The simple methods that we find perform well are easy to implement and interpret, making implementation in practice more likely. In a simulation study we alter the empirical weekly patterns and demonstrate how extreme differences between the weekly seasonality patterns of different regions cause hierarchically-reconciled bottom-up approaches to outperform top-down approaches.

Prediction of Sea Surface Current Around the Korean Peninsula Using Artificial Neural Networks

AbstractThe prediction of sea surface currents is essential for various marine activities such as disaster monitoring, fishing industries, and search and rescue operations. Continuous improvements in numerical models have made it possible to predict more realistic oceans using data assimilation and fine spatial resolution. However, such complicated ocean models require high computational power and time, making them difficult to use for near‐real time forecasting. To compensate for these high computational costs, novel approaches with efficient computational costs combined with numerical model outputs need to be developed. Artificial neural networks could be one of the solutions as they require low computational power for prediction, owing to pre‐trained networks. Here, we present a prediction framework applicable to surface current prediction in the seas around the Korean Peninsula using three‐dimensional (3‐D) convolutional neural networks. The network is based on a 3‐D U‐shaped network structure and is modified to predict sea surface currents using oceanic and atmospheric variables. The forecast procedure is optimized to minimize the error of the next day's sea surface current field with a spatial resolution of 1/24° and its recursively predicting structure allows more days to be predicted. The network performance is evaluated by changing the input days and variables to find the optimal surface‐current‐prediction artificial neural network model. The optimized model shows a root mean squared error of 2.3 cm/s for the first forecast day, demonstrating its strong potential for practical use in the near future.

Development of a city-level surface ozone forecasting system using deep learning techniques and air quality model: Application in eastern China

Utilizing regional air quality models to accurately forecast surface ozone (O3) concentrations, particularly high concentrations, is essential for protecting public health. However, forecasts of air quality model often deviate from site observations due to the limitation of grid resolution and uncertainties from emission sources, meteorological conditions, and chemical reaction mechanisms. Especially, the underestimation is significant under condition of high O3 concentrations. Moreover, such deviations tend to accumulate as forecast lead time increases, compounding the challenges associated with reliable air quality forecast. In this study, we employed AlexNet architecture, a classical convolutional neural network, combined with multiple variables related to meteorology, chemistry, emission and geography to establish a non-linear relationship between grid-scale input variables and site-scale hourly O3 forecast biases in Eastern China, aiming to realize accurate city-level ozone forecast based on a regional air quality prediction model (i.e., Nested Air Quality Prediction Model System, NAQPMS). By assigning weights to high-bias samples and high-concentration samples within the loss function, the proposed Weighted AlexNet model (W_AlexNet) effectively reduced forecast biases and enhanced its capability to predict O3 pollution levels. Compared to NAQPMS, W_AlexNet model demonstrated a 25.71% improvement in RMSE and a 7.17% increase in IOA averagely for hourly O3 (O3-1h) forecasts across four different lead times (24-h, 48-h, 72-h, and 96-h). Notably, W_AlexNet model alleviated the tendency of NAQPMS to underestimate high concentrations and showed a superior performance in improving O3-1h pollution level forecasts, particularly for the 72-h and 96-h lead times. W_AlexNet model can effectively mitigate the bias accumulation effect over increasing lead times, thereby enhancing the reliability of longer-term forecasts. Thus, the W_AlexNet model serves as a post-processing model that can calibrate forecast biases in air quality prediction models, significantly improving the accuracy of O3 high concentration forecasts and providing more precise early warnings of O3 pollution. This underscores its utility in air quality management.

Bayesian inference for link travel time correlation of a bus route

Estimation of link travel time correlation of a bus route is essential to many bus operation applications. Link travel time on a bus route could exhibit complex correlation structures, such as long-range correlations, negative correlations, and time-varying correlations. This paper develops a Bayesian Gaussian model to estimate the link travel time correlation matrix of a bus route using smart-card-like data. Our method overcomes the small-sample-size problem in correlation matrix estimation by borrowing/integrating those incomplete observations from other bus routes. We first conduct a synthetic experiment and results show that the proposed method produces an accurate estimation for correlations with credible intervals. Next, we perform experiments on a real-world bus route with in-out-stop record data; results show that both local and long-range correlations exist on this bus route. Finally, we demonstrate an application of using the estimated covariance matrix to make probabilistic forecasting of link and trip travel time.

A Method Based on Deep Learning for Severe Convective Weather Forecast: CNN-BiLSTM-AM (Version 1.0)

In this study, we propose a model called CNN-BiLSTM-AM that utilizes deep learning techniques to forecast severe convective weather based on ERA5 hourly data and observations. The model integrates a CNN with a Bidirectional Long Short-Term Memory (BiLSTM) system and an Attention Mechanism (AM). The CNN is tasked with extracting features from the input data, while the BiLSTM effectively captures temporal dependencies. The AM enhances the results by considering the impact of past feature states on severe weather phenomena. Additionally, we assess the performance of our model in comparison to traditional network architectures, including ConvLSTM, Predrnn++, CNN, FC-LSTM, and LSTM. Our results indicate that the CNN-BiLSTM-AM model exhibits superior accuracy in precipitation forecasting. Especially with the extension of the forecast time, the model performs well across multiple evaluation metrics. Furthermore, an interpretability analysis of the convective weather mechanisms utilizing machine learning highlights the critical role of total precipitable water (PWAT) in short-term heavy precipitation forecasts. It also emphasizes the significant impact of regional variables on convective weather patterns and the role of convective available potential energy (CAPE) in fostering conditions conducive to convection. These findings not only confirm the effectiveness of deep learning in the automatic identification of severe weather features but also validate the suitability of the sample dataset employed. Given its remarkable performance and robustness, we advocate for the adoption of this model to enhance the forecast of severe convective weather across various business applications.

A parallel recursive framework for modelling time series

Abstract Time series modelling is of significance to several scientific fields. Several approaches based on statistics, machine learning or combinations have been utilized. In order to model and forecast time series a novel parallel framework based on recursive pseudoinverse matrices is proposed. This framework enables the design of arbitrary statistical and machine learning models, adaptively, from a set of potential basis functions. This unification enables compact definition of existing and new models as well as easy implementation for new massively parallel architectures. The choice of appropriate basis functions is analysed and the fitting accuracy, termination criteria and model update operations are presented. A block variant for multivariate time series is also proposed. Parallel GPU implementation and performance optimization of the framework are provided, based on mixed precision arithmetic and matrix operations. The use of different basis functions is showcased with respect to various model univariate and multivariate time series for applications such as regression, frequency estimation and automatic trend detection. Discussions on limitations and future directions of research are also provided.

The Attribution of February Extremes over North America: A Forecast-Based Storyline Study

Abstract The importance of extreme event attribution rises as climate change causes severe damage to populations resulting from unprecedented events. In February 2019, a planetary wave shifted along the U.S.–Canadian border, simultaneously leading to troughing with anomalous cold events and ridging over Alaska and northern Canada with abnormal warm events. Also, a dry-stabilized anticyclonic circulation over low latitudes induced warm extreme events over Mexico and Florida. Most attribution studies compare the climate model simulations under natural or actual forcing conditions and assess probabilistically from a climatological point of view. However, in this study, we use multiple ensembles from an operational forecast model, promising statistical as well as dynamically constrained attribution assessment, often referred to as the storyline approach to extreme event attribution. In the globally averaged results, increasing CO2 concentrations lead to distinct warming signals at the surface, resulting mainly from diabatic heating. Our study finds that CO2-induced warming eventually affects the possibility of extreme events in North America, quantifying the impact of anthropogenic forcing over less than a week’s forecast simulation. Our study assesses the validity of the storyline approach conditional on the forecast lead times, which is hindered by rising noise in CO2 signals and the declining performance of the forecast model. The forecast-based storyline approach is valid for at least half of the land area within a 6-day lead time before the target extreme occurrence. Our attribution results highlight the importance of achieving net-zero emissions ahead of schedule to reduce the occurrence of severe heatwaves.

Impact of WindBorne Observation Assimilation on Prediction of a TPV Merger Case From THINICE

AbstractThe impact of assimilating in situ temperature, moisture and wind observations from a WindBorne Systems balloon, with long duration and adjustable ballast, is evaluated using a case study from the 2022 THINICE field campaign. A case is selected wherein the WindBorne balloon directly sampled a jet streak associated with a tropopause polar vortex (TPV). The observed TPV merged with another TPV at the same time as a downstream Arctic cyclone (AC) redeveloped eastward. The case is used to better understand the role of the observed TPV in the evolution of the downstream AC. The assimilation of the WindBorne observations improves the forecast track and amplitude of the TPV during the ∼1 day forecast period that the TPV can be tracked as a distinct feature. The root mean square error of temperature at 350 hPa is improved by the WindBorne assimilation throughout the 2.5 day forecast period. The surface cyclone track forecast is improved by the WindBorne assimilation during the period of eastward redevelopment of the surface cyclone, and the sea level pressure RMSE in the surrounding region is reduced during the first ∼1.5 days of forecast lead time. Results demonstrate the capability of the long‐duration controllable WindBorne balloons to improve the analysis of the TPV associated with the jet streak, leading to improved forecast of both tropopause‐level and surface pressure features. Additionally, the importance of observing the upstream TPV amplitude for improving forecasts of the process of TPV merging and its impact on the evolution and longevity of the mature AC is confirmed.

Approaches to Modelling Marketing Strategies in E-Commerce

Introduction. Increasing the efficiency of business activities requires developing a marketing strategy using e-commerce to attract a wide range of customers and maximise the company's income. The key marketing metric for most companies is the customer lifetime value indicator, which is why studying approaches to forecasting this indicator in e-commerce requires special attention, considering the changing business operation environment. Aim and tasks. The study aims to develop approaches to modelling the marketing strategy of e-commerce enterprises. The tasks are a comparative characterisation of modern marketing planning tools and an empirical analysis of the possibilities of using the BG/NBD model for marketing planning in e-commerce companies, considering the need to implement scenario marketing planning in unstable conditions. Results. A comparison of the characteristics of the main marketing planning tools was conducted according to features such as content and focus, functional purpose, and planning period, which allows more purposeful use of these tools to solve specific tasks. In order to automatically obtain the required data format for modelling, a Visual Basic program was developed, a forecast of customer behaviour was built, and an indicator of customer transactions was determined for further calculation of customer lifetime value (CLV). During the simulation, four parameters of the BG/NBD model were evaluated, and the maximum value of the logarithmic likelihood function equal to 14460.54 was obtained. The expected number of transactions (forecast time = 26 weeks) was determined for each client from the database. After calculations the information was received, the selected client will make 2.35 transactions during the forecast period of 55-80 weeks. The recommended algorithm for using the BG/NBD model as part of a scenario approach for forecasting the cost of the customer's life cycle allows for more flexible forecasting of digital business revenues and building an optimal marketing strategy. Conclusions. An empirical study of the possibilities of using the BG/NBD model for marketing planning in e-commerce companies showed that, for the simulation of indicators that affect a company's marketing strategy, it is crucial at the initial stage to convert the database of customer purchases into the input format required for modelling. Using a scenario approach allows e-commerce enterprises to carry out variable marketing planning and reduce risks under growing uncertainty. A well-designed marketing strategy that can be adapted to different business development scenarios enables an e-commerce company to increase its sales.

An accurate and efficient forecast framework for fine PM2.5 maps using spatiotemporal recurrent neural networks

Commonly used numerical prediction models for PM2.5 maps suffer from low accuracy and high computation cost, which cannot meet the requirements for fine-scale air pollution control. In this study, we propose a framework based on the spatiotemporal recurrent neural network (PredRNN) to efficiently generate accurate 3-h and 6-km PM2.5 maps with a lead time of 5 days. In this framework, two PredRNN networks are initially utilized to forecast PM2.5 concentration at ground monitoring sites and the spatial distribution of aerosol optical depth (AOD) by assimilating the output of numerical prediction model. Subsequently, the 3-h and 6-km PM2.5 forecasted maps with a lead time of 5 days can be inferred by establishing the regression links between the forecasted results of PM2.5 concentration at ground sites and AOD maps. We evaluate the proposed framework in the Beijing-Tianjin-Hebei urban agglomeration region during 2017–2020. Compared with the numerical prediction products of the Copernicus Atmosphere Monitoring Service, the proposed framework achieves higher accuracy, with R2 of 0.83 at the forecast base time and 0.70 at the fifth day. The spatial information richness is also enhanced by approximately 15.67% according to the information entropy metrics. Notably, the proposed framework only requires 1 min for forecasting 5-days PM2.5 maps. These results demonstrate that our framework can efficiently generate accurate and fine PM2.5 maps with a lead time of 5 days.

A Markov model approach for statistical tropical cyclone wind radii baseline forecasts

Abstract A skill baseline for five-day, 34-, 50-, and 64-knot (1 kt = 0.514 m s−1) tropical cyclone (TC) wind radii forecasts is described. The Markov Model CLiper (MMCL) generates a sequence of 12-h forecasts out to a forecast length limited only by the length of the forecast track and intensity. The model employs a climatology of TC size based on infrared satellite imagery, a Markov chain, and a basin-specific drift. MMCL uses the initial wind radii and initial forecast track and intensity as input. Unlike the previously developed wind radii climatology and persistence model (DRCL) that reverts to a climatological size and shape after approximately 48 h, MMCL retains more of its initial size and asymmetry and is likely more palatable for use in operational forecasting. MMCL runs operationally in the western North Pacific basin, the North Indian Ocean, and the Southern Hemisphere for the Joint Typhoon Warning Center (JTWC) in Pearl Harbor, Hawaii. This work also describes the development of Atlantic and eastern North Pacific versions of MMCL. MMCL’s formulation allows unlimited extension of forecast lead time without reverting to a generic climatological size and shape. Independent forecast comparisons between MMCL and DRCL for the 2020–2022 seasons demonstrates that MMCL’s mean absolute errors are generally smaller and biases are closer to zero in North Atlantic, and eastern North Pacific basins, and in the Southern Hemisphere. This validation includes a few example forecasts and demonstrates that MMCL can be used both as a baseline for assessing wind radii forecast skill and operational use.

Study on Downscaling Correction of Near-Surface Wind Speed Grid Forecasts in Complex Terrain

Accurate forecasting of wind speeds is a crucial aspect of providing fine-scale professional meteorological services (such as wind energy generation and transportation operations etc.). This article utilizes CMA-MESO model forecast data and CARAS-SUR_1 km ground truth grid data from January, April, July, and October 2022, employing the random forest algorithm to establish and evaluate a downscaling correction model for near-surface 1 km resolution wind-speed grid forecast in the complex terrain area of northwestern Hebei Province. The results indicate that after downscaling correction, the spatial distribution of grid forecast wind speeds in the entire complex terrain study area becomes more refined, with spatial resolution improving from 3 km to 1 km, reflecting fine-scale terrain effects. The accuracy of the corrected wind speed forecast significantly improves compared to the original model, with forecast errors showing stability in both time and space. The mean bias decreases from 2.25 m/s to 0.02 m/s, and the root mean square error (RMSE) decreases from 3.26 m/s to 0.52 m/s. Forecast errors caused by complex terrain, forecast lead time, and seasonal factors are significantly reduced. In terms of wind speed categories, the correction significantly improves forecasts for wind speeds below 8 m/s, with RMSE decreasing from 2.02 m/s to 0.59 m/s. For wind speeds above 8 m/s, there is also a good correction effect, with RMSE decreasing from 2.20 m/s to 1.65 m/s. Selecting the analysis of the Zhangjiakou strong wind process on 26 April 2022, it was found that the downscaled corrected forecast wind speed is very close to the observed wind speed at the station and the ground truth grid points. The correction effect is particularly significant in areas affected by strong winds, such as the Bashang Plateau and valleys, which has significant reference value.

ACCESS-S2 seasonal forecasts of rainfall and the SAM–rainfall relationship during the grain growing season in south-west Western Australia

South-west Western Australia (SWWA) is home to a world class grains industry that is significantly affected by periods of drought. Previous research has shown a link between the Southern Annular Mode (SAM) and rainfall in SWWA, especially during winter months. Hence, the predictability of the SAM and its relationship to SWWA rainfall can potentially improve forecasts of SWWA drought, which would provide valuable information for farmers. In this paper, focusing on the 0-month lead time forecast, we assess the bias and skill of ACCESS-S2, the Australian Bureau of Meteorology’s current operational sub-seasonal to seasonal forecasting system, in simulating seasonal rainfall for SWWA during the growing season (May–October). We then analyse the relationship between the SAM and SWWA precipitation and how well this is captured in ACCESS-S2 as well as how well ACCESS-S2 forecasts the monthly SAM index. Finally, ACCESS-S2 rainfall forecasts and the simulation of SAM are assessed for a case study of extreme drought in 2010. Our results show that forecasts tend to have greater skill in the earlier part of the season (May–July). ACCESS-S2 captures the significant inverse SAM–rainfall relationship but underestimates its strength. The model also shows overall skill in forecasting the monthly SAM index and simulating the MSLP and 850-hPa wind anomaly patterns associated with positive and negative SAM phases. However, for the 2010 drought case study, ACCESS-S2 does not indicate strong likelihoods of the upcoming dry conditions, particularly for later in the growing season, despite predicting a positive (although weaker than observed) SAM index. Although ACCESS-S2 is shown to skillfully depict the SAM–SWWA rainfall relationship and generally forecast the SAM index well, the seasonal rainfall forecasts still show limited skill. Hence it is likely that model errors unrelated to the SAM are contributing to limited skill in seasonal rainfall forecasts for SWWA, as well as the generally low seasonal-timescale predictability for the region.

A Physical‐Informed Neural Network for Improving Air‐Sea Turbulent Heat Flux Parameterization

AbstractThe parameterizations of air‐sea turbulent heat flux are one of the major bottlenecks in atmosphere‐ocean coupled model development, which play a crucial role in sea surface temperature (SST) prediction. Recently, neural networks start to be applied for the development of parameterizations of interface turbulent heat flux. However, these new parameterizations are primairily developed for specific regions and have not been tested in real atmosphere‐ocean coupled models. In this study, we propose a new air‐sea heat flux parameterization using a physical‐informed neural network (PINN) based on multiple observational data sets worldwide. Evaluated with an independent observation data set, it is shown that the PINN can significantly reduce the RMSE of latent heat flux by at least about 48.6% compared to three traditional bulk formulas. Moreover, the PINN can be flexibly updated with new observational data by transfer learning. To test the performance of the new parameterization in realistic application, we implement the PINN into a global ocean‐atmosphere coupled model and make seasonal forecasts for the first time. The PINN markedly reduce the errors of equatorial SST forecast, indicating a good performance of the PINN‐based air‐sea turbulent heat flux scheme. Noticeably, due to limited observational data, the NN‐based parameterizations tend to underestimate heat flux at high wind speeds compared with bulk formula‐based parameterizations. With more data available at extreme conditions, the PINN can be improved via transfer learning and need to be futher evaluated. This study suggests that PINN‐based air‐sea heat flux parameterization is promising to improve SST simulation.