Forecast Performance Research Articles

Does media coverage of firms' environment, social, and governance (ESG) incidents affect analyst coverage and forecasts? A risk perspective

We examine whether and how media coverage of firms' environment, social, and governance (ESG) incidents is associated with analyst coverage and forecasts. We propound that the risks of firms could either increase or decrease as a result of media-covered ESG incidents, depending on the firms' actions on the media coverage, and thus its impact on analyst coverage and forecasts would vary. Based on a sample of U.S. listed companies, we find evidence that the level of analyst coverage is negatively associated with a firm's ESG incidents covered by the media. This association is more pronounced for firms with more intense industrial product market competition, more severe ESG scandals, or coverage by less sophisticated analysts. We also find that the firms' ESG incidents covered by the media would lead to higher levels of forecast error and dispersion. Our mediation analysis further reveals that business risk and information risk tend to be higher for firms covered by the media for having been involved in ESG incidents, thereby explaining why the analysts' coverage and forecasts for these firms are adversely affected. Overall, our results highlight the importance of curbing corporate social irresponsibility and improving analyst performance in forecasting.

MEAformer: An all-MLP transformer with temporal external attention for long-term time series forecasting

Transformer-based models have significantly improved performance in Long-term Time Series Forecasting (LTSF). These models employ various self-attention mechanisms to discover long-term dependencies. However, the computational efficiency is hampered by the inherent permutation invariance of self-attention, and they primarily focus on relationships within the sequence while neglecting potential relationships between different sample sequences. This limits the ability and flexibility of self-attention in LTSF. In addition, the Transformer's decoder outputs sequences in an autoregressive manner, leading to slow inference speed and error accumulation effects, especially for LTSF. Regarding the issues with Transformer-based models for LTSF, we propose a model better suited for LTSF, named MEAformer. MEAformer adopts a fully connected Multi-Layer Perceptron (MLP) architecture consisting of two types of layers: encoder layers and MLP layers. Unlike most encoder layers in Transformer-based models, the MEAformer replaces self-attention with temporal external attention. Temporal external attention explores potential relationships between different sample sequences in the training dataset. Compared to the quadratic complexity of self-attention mechanisms, temporal external attention has efficient linear complexity. Encoder layers can be stacked multiple times to capture time-dependent relationships at different scales. Furthermore, the MEAformer replaces the intricate decoder layers of the original model with more straightforward MLP layers. This modification aims to enhance inference speed and facilitate single-pass sequence generation, effectively mitigating the problem of error accumulation effects. Regarding long-term forecasting, MEAformer achieves state-of-the-art performance on six benchmark datasets, covering five real-world domains: energy, transportation, economy, weather, and disease. Code is available at: https://github.com/huangsiyuan924/MEAformer.

Implications of WRF model resolutions on resolving rainfall variability with topography over East Africa

There is an increasing need to improve the accuracy of extreme weather forecasts for life-saving applications and in support of various socioeconomic sectors in East Africa, a region with remarkable mesoscale systems due to its complex topography defined by sharp gradients in elevation, inland water bodies, and landuse conversions. This study sought to investigate the impacts of the Weather Research and Forecasting (WRF) model spatial resolution on resolving rainfall variability with topography utilizing nested domains at 12 and 2.4 km resolutions. The model was driven by the National Centers for Environmental Prediction (NCEP)-Global Data Assimilation System (GDAS) Global Forecast System (GFS) final (FNL) reanalysis to simulate the weather patterns over East Africa from 3rd April 2018 to 30th April 2018, which were evaluated against several freely available gridded weather datasets alongside rainfall data from the Kenya Meteorological Department (KMD) stations. The reference datasets and the model outputs revealed that the highlands had more rainfall events and higher maximum daily rainfall intensity compared to the surrounding lowlands, attributed to orographic lifting enhancing convection. Rainfall was inversely proportional to altitude from 500 m to 1,100 m above sea level (ASL) for both coarse and fine resolutions. The convection-permitting setup was superior in three aspects: resolving the inverse altitude-rainfall relationship for altitudes beyond 3000 m ASL, simulating heavy rainfall events over the lowlands, and resolution of the diurnal cycle of low-level wind. Although the coarse resolution setup reasonably simulated rainfall over large mountains, only the convection-permitting configuration could accurately resolve rainfall variability over contrasting topographical features. The study notes that high-resolution modeling systems and topography-sensitive bias correction techniques are critical for improving the quality of operational weather forecasts in East Africa.

Quantitative and intelligent methods for economic forecasts (birth rate time series (Ntl))

The models proposed in the field of time series forecasting are numerous. However, accurate forecasting is still one of the main challenging problems faced by decision makers in various fields, especially in financial markets. Anticipatory and intelligent methods play a crucial role in economic forecasting, helping to increase the accuracy and reliability of forecasts. This is the main reason that many researchers have in the focus of their research the development of methods for improving the performance of forecasts. Using only a linear or non-linear method is not enough to recognize and model all features and functions of the data. In addition to quantitative techniques as well as intelligent techniques that have been developed recently as a result of technological developments, a new technique or hybridization of these two techniques behaves as a very efficient approach. For this reason, in the study it is proposed to improve the modeling and forecasting results by combining a linear A RIMA method with an intelligent non-linear method, the artificial neural network -ANN. Several ARIMA-ANN hybrid model architectures are known, but in this thesis a new A RIMA-ANN technique with improving features of the existing hybrid methods is proposed. The time series under study are the birth rate in Albania obtained by INSTAT and the Bank of Albania with monthly frequencies. Time series analysis goes through three phases: defining characteristics, modeling and forecasting. The analysis is performed with ARIMA, ANN and the proposed ARIMA-ANN method. Data mining, sentiment analysis, and natural language processing (NLP) techniques are used to analyze unstructured data sources. The models, either quantitative or intelligent, give satisfactory results, but the hybrid model shows much higher performance in all series. From the measurement indicators of the prediction performance, the hybrid method has resulted with high efficiency. The ARIMA-ANN model produced much better results than the ARIMA and ANN models. These results are a consequence of the combination of the two techniques and from which results a good recognition of the different structures of the data modes. This study contributes to the field of time series forecasts in Albania and intelligent or quantitative modeling by developing a new forecast model, that of hybridization.

Predicting the Equity Premium with Combination Forecasts: A Reappraisal

Abstract This paper reappraises the usefulness of combining individual forecasts for predicting the U.S. equity premium. For comparison, we also consider penalized regression and dimension reduction approaches. We fail to find evidence of predictive ability in recent decades, regardless of the forecasting method used. Further analysis shows that an increase in the correlation of individual forecast errors is an important factor in the declining performance of combination forecasts.()

A transformer-based method for correcting significant wave height numerical forecasting errors

Accurate significant wave height (SWH) forecasting is essential for various marine activities. While traditional numerical and mathematical-statistical methods have made progress, there is still room for improvement. This study introduces a novel transformer-based approach called the 2D-Geoformer to enhance SWH forecasting accuracy. The 2D-Geoformer combines the spatial distribution capturing capabilities of SWH numerical models with the ability of mathematical-statistical methods to identify intrinsic relationships among datasets. Using a comprehensive long time series of SWH numerical hindcast datasets as the numerical forecasting database and ERA5 reanalysis SWH datasets as the observational proxies database, with a focus on a 72-hour forecasting window, the 2D-Geoformer is designed. By training the potential connections between SWH numerical forecasting fields and forecasting errors, we can retrieve SWH forecasting errors for each numerical forecasting case. The corrected forecasting results can be obtained by subtracting the retrieved SWH forecasting errors from the original numerical forecasting fields. During long-term validation periods, this method consistently and effectively corrects numerical forecasting errors for almost every case, resulting in a significant reduction in root mean square error compared to the original numerical forecasting fields. Further analysis reveals that this method is particularly effective for numerical forecasting fields with higher errors compared to those with relatively smaller errors. This integrated approach represents a substantial advancement in SWH forecasting, with the potential to improve the accuracy of operational SWH forecasts. The 2D-Geoformer combines the strengths of numerical models and mathematical-statistical methods, enabling better capture of spatial distributions and intrinsic relationships in the data. The method's effectiveness in correcting numerical forecasting errors, particularly for cases with higher errors, highlights its potential for enhancing SWH forecasting accuracy in operational settings.

Research on Frequency Matching Correction Techniques for South China Precipitation Ensemble Forecast Based on the GRAPES Model

This study focuses on the real-time precipitation forecast products of the GRAPES_MESO regional ensemble forecast model, which is developed by the Numerical Weather Prediction Center of the China Meteorological Administration and is initialized 1–3 days in advance at 12:00 UTC. Using a national-level homogenized precipitation grid dataset from surface meteorological stations as observational data, a frequency matching method (FMM) is employed to correct precipitation forecasts for different precipitation intensity levels, including light rain, moderate rain, heavy rain, and torrential rain. Case studies and statistical tests (TS scores) are conducted to compare the forecast performance before and after correction. The results indicate that the model’s Cumulative Distribution Function (CDF) curves deviate from observations, and the longer the lead time, the more significant the error. The correction coefficients (CCs) show an increasing trend with the growth of precipitation intensity, indicating that for larger precipitation amounts and longer lead times, larger CCs are needed, highlighting the necessity of correction. Analyzing two precipitation events in South China in July 2019, the FMM results in an increase in precipitation intensity and a widening of the range of heavy precipitation. The corrected precipitation magnitudes are closer to the observations. The statistical tests using TS scores reveal that the FMM has a certain correction effect on the overall precipitation forecast in the South China region, especially for longer lead times and higher precipitation intensities, where the correction effect is more significant. The necessity of frequency matching correction becomes more apparent for heavier precipitation, and the correction effect becomes more significant with longer lead times.

GA-FCFNN: A new forecasting method combining feature selection methods and feedforward neural networks using genetic algorithms

In the modern landscape, the fusion of forecasting and computational intelligence empowers organizations to extract invaluable insights from vast datasets, facilitating informed decision-making, swift adaptation to market dynamics, and the enhancement of competitiveness, ultimately fostering innovation. Notably, forecasting has recently garnered significant attention, particularly within the realm of big data. This modern terrain presents a dual challenge: effectively distilling essential insights from complex data and optimizing data utilization. To address this challenge, we introduce a pioneering forecasting model (FM) engineered to excel in both information extraction and data utilization. Our approach commences with a formula designed to compute feature similarity, leveraging trend change data. These similarity metrics, applied to features and their relationships with labels, inform our feature selection process. This method integrates spatial and temporal correlations among features, fostering robust interconnections within the selected subset. Subsequently, we employ a feedforward neural network (FNN) to optimize feature subsets based on test error. While the FM trained with the optimal feature subset exhibits enhanced performance, it still shows some deviation from actual values in initial forecasting results (FORs). To mitigate these discrepancies, we implement a data utilization strategy. We refine the initial FORs by considering the contribution of each feature subset to them, thereby reducing disparities between forecasts and actual values. Model parameters are fine-tuned using a genetic algorithm (GA). Finally, we conduct an experimental analysis, comparing our FM to seven similar models using publicly available datasets. Our experimental results consistently demonstrate that our FM achieves high forecasting performance (FOP).

An advanced framework for net electricity consumption prediction: Incorporating novel machine learning models and optimization algorithms

In recent years, the escalating demand for electric energy has underscored the need for robust prediction models capable of accurately anticipating consumption patterns. The imperative lies in enabling utilities and policymakers to optimize resource allocation, strategically plan infrastructure development, and ensure the stability and efficiency of the power grid. This study undertakes a comprehensive comparative analysis of machine learning techniques employed in predicting net electricity consumption in Turkey. The primary goal is to augment the accuracy and performance of electricity load forecasting, thereby contributing to effective energy management and fostering sustainable development within the power sector. Two machine learning models, including CatBoost and Extreme Gradient Boosting (XGBoost), are strategically integrated with optimization algorithms such as Sparrow Search Algorithm (SSA), Phasor Particle Swarm Optimization (PPSO), and Hybrid Grey Wolf Optimization (GWO). The core analysis centers on evaluating the performance of these integrated models based on key accuracy metrics and runtime efficiency. Notably, the results underscore that the XGBoost-SSA model emerges as the superior performer, exhibiting heightened accuracy and superior performance in predicting electricity consumption. This model showcases the highest coefficient of determination (R2) value and demonstrates lower errors during the testing phase, thereby presenting a promising and effective approach for electricity consumption prediction in the specific context of Turkey.

Brain activity of professional investors signals future stock performance

A major aspiration of investors is to better forecast stock performance. Interestingly, emerging "neuroforecasting" research suggests that brain activity associated with anticipatory reward relates to market behavior and population-wide preferences, including stock price dynamics. In this study, we extend these findings to professional investors processing comprehensive real-world information on stock investment options while making predictions of long-term stock performance. Using functional MRI, we sampled investors' neural responses to investment cases and assessed whether these responses relate to future performance on the stock market. We found that our sample of investors could not successfully predict future market performance of the investment cases, confirming that stated preferences do not predict the market. Stock metrics of the investment cases were not predictive of future stock performance either. However, as investors processed case information, nucleus accumbens (NAcc) activity was higher for investment cases that ended up overperforming in the market. These findings remained robust, even when controlling for stock metrics and investors' predictions made in the scanner. Cross-validated prediction analysis indicated that NAcc activity could significantly predict future stock performance out-of-sample above chance. Our findings resonate with recent neuroforecasting studies and suggest that brain activity of professional investors may help in forecasting future stock performance.

Optimal weight support vector regression ensemble with cluster-based subsampling for electricity price forecasting

Accurate prediction of short-term electricity price is the key to obtain economic benefit and also an important index of power system planning and management. Support vector regression (SVR) based ensemble works have gained remarkable achievements in terms of high accuracy and steady performance, but they are highly dependent on data representativeness and have a high computational complexity O (k * N 3) of data samples and parameter selection. To further improve the data representativeness and reduce its computational complexity, this paper develops a new approach to forecast electricity price via optimal weighted ensemble. In the model, the cluster-based subsampling algorithm is proposed to categorize the inputs being seasonally decomposed into several groups, and representative data are drawn from each group in a certain proportion to ensure that each subset trained with SVR has the same representativeness and features. Moreover, the optimal weighted combination method is presented to assign weights to the sub-SVRs to obtain the optimal support vector regression ensemble model (OWSSVRE). The experimental results show that the improved support vector regression ensemble model with the same features and representativeness of the subset has better performance in electricity price forecasting. As a result, it is suitable to support decision making in the energy and other sectors.

T49 Style Stock Prediction

Stock market prediction has long been a pursuit of investors seeking to gain insights into the future trajectory of stock prices. Traditionally, analysts relied on fundamental analysis, technical indicators, and market sentiment to forecast stock movements. However, with the advent of machine learning (ML) and artificial intelligence (AI), predictive analytics in the stock market has seen a significant shift. ML algorithms offer the advantage of processing vast amounts of data and identifying complex patterns that might elude human analysts. In the context of stock market prediction, these algorithms can analyze historical stock prices, trading volumes, market indices, news sentiment, and various other factors to generate forecasts. One notable trend in this domain is the utilization of current stock market indices as input features for ML models. By training on historical data of these indices and their corresponding effects on individual stock prices, algorithms can learn to make predictions based on the current state of the market. One common approach is to use techniques like regression, time series analysis, or deep learning models such as recurrent neural networks (RNNs) and long short-term memory networks (LSTMs). These models can capture temporal dependencies and nonlinear relationships in the data, allowing for more accurate predictions. Additionally, ensemble methods like random forests or gradient boosting can be employed to combine the strengths of multiple models for enhanced forecasting performance. Keywords: Stock market prediction,, financial stocks, stock market indices

Moist processes in NCUM global forecasts during the boreal summer monsoon

The objective of this study is to evaluate the performance of the National Center for Medium-Range Weather Forecasting Unified model (NCUM) global version in representing the moist processes during active and break monsoon periods over the Indian region. We employed the moist static energy (MSE) equation, a process-oriented diagnostic (POD), to examine the relationship between small-scale convection and large-scale circulation patterns in the forecast for eight monsoon seasons spanning from June to September 2015 to 2022. Results are validated by utilizing the European Center for Medium-Range Weather Forecasting Reanalysis (ERA5) data products. By employing PODs, we aimed to identify the influence of various MSE terms on rainfall evolution, particularly over the Central Indian (CI) region. We also investigated the systematic errors in certain processes that contribute to rainfall biases. Additionally, the study explored the moisture-convection relationship and compared the sensitivity of free tropospheric moisture to deep convection in the NCUM forecasts against the reanalysis data. The findings indicate weaknesses in the model's sensitivity to moisture and its ability to represent the rainfall-column water vapor (CWV) relationship highlighting areas where further improvements may be necessary to enhance the model's performance in monsoon forecasting. Despite the presence of systematic biases in the model's mean state of rainfall, such as dry biases over the Bay of Bengal and wet biases over the eastern Indian Ocean (EIO), the results obtained from the MSE diagnostics aligned with the existing understanding of monsoon dynamics. Specifically, the study emphasized the significance of horizontal advection of MSE as a leading term preceding the peak active/break conditions over the CI region. Hence, these findings not only showcase the usefulness of PODs in examining the strengths and weaknesses of numerical weather prediction (NWP) models but also serve as a verification tool for assessing model skills, leading to improved understanding and development of NWP models.

Deep learning of estuary salinity dynamics is physically accurate at a fraction of hydrodynamic model computational cost

AbstractSalinity dynamics in the Delaware Bay estuary are a critical water quality concern as elevated salinity can damage infrastructure and threaten drinking water supplies. Current state‐of‐the‐art modeling approaches use hydrodynamic models, which can produce accurate results but are limited by significant computational costs. We developed a machine learning (ML) model to predict the 250 mg L−1 Cl− isochlor, also known as the “salt front,” using daily river discharge, meteorological drivers, and tidal water level data. We use the ML model to predict the location of the salt front, measured in river miles (RM) along the Delaware River, during the period 2001–2020, and we compare predictions of the ML model to the hydrodynamic Coupled Ocean–Atmosphere‐Wave‐Sediment Transport (COAWST) model. The ML model predicts the location of the salt front with greater accuracy (root mean squared error [RMSE] = 2.52 RM) than the COAWST model does (RMSE = 5.36); however, the ML model struggles to predict extreme events. Furthermore, we use functional performance and expected gradients, tools from information theory and explainable artificial intelligence, to show that the ML model learns physically realistic relationships between the salt front location and drivers (particularly discharge and tidal water level). These results demonstrate how an ML modeling approach can provide predictive and functional accuracy at a significantly reduced computational cost compared to process‐based models. In addition, these results provide support for using ML models in operational forecasting, scenario testing, management decisions, hindcasting, and resulting opportunities to understand past behavior and develop hypotheses.

Does a Scale‐Aware Convective Parameterization Scheme Improve the Simulation of Heavy Rainfall Events?

AbstractPrecipitation predictability using the non‐scale‐aware and scale‐aware convective parameterization schemes (CPSs) was investigated to assess the necessity of the CPSs within the gray‐zone. This study evaluates the performance of the Weather Research and Forecasting (WRF) model's CPS for 135 heavy rainfall events (HREs) over the Korean Peninsula for 10 years (i.e., 2011–2020). We tested the Kain–Fritsch (KF) scheme (non‐scale‐aware) and Multi‐scale Kain–Fritsch (MSKF) scheme (scale‐aware) in the WRF model. The MSKF scheme shows an overall improved performance of precipitation simulation compared to the KF scheme, but the precipitation forecast performance of CPS depends on the characteristics of HREs. When the HREs are characterized by synoptic‐scale atmospheric conditions with strong winds and large‐scale water vapor transport, the forecast performance of both CPSs is similar because a cloud microphysics scheme can explicitly resolve most of the precipitation. However, in the case of HREs with weak synoptic forcing conditions (e.g., moisture transport and winds) related to the localized and meso‐scale HREs, the MSKF scheme can improve overall simulated precipitation by increasing grid‐scale precipitation and reducing the overestimation of subgrid‐scale precipitation simulated in the KF scheme. Therefore, using the scale‐aware CPS in the gray‐zone can provide more accurate precipitation forecasts regardless of the environmental condition of the HREs.

Local identification of equatorial Kelvin waves in real‐time operational forecasts

AbstractEquatorial Kelvin waves (KWs) are associated with a variety of equatorial atmospheric phenomena, namely, tropical convection, cloud and precipitation variability, tropical cyclogenesis, the onset of monsoon season over Africa and India, and even the quasi‐biennial oscillation and the Madden–Julian oscillation. As such, if operational forecasts are to improve near the Tropics, it is important that they correctly represent KWs. This article aims at developing a novel methodology for identifying KWs in real‐time operational forecasts at specific longitudes using only the meridional structures of the solutions to the free Laplace tidal equations, known as Hough vector functions. The main advantages of this newly proposed methodology are that (a) it can identify KWs at a specific longitude and (b) very low computational cost is needed to apply it to real‐time operational forecasts. The application of the method to 2015–2017 European Centre for Medium‐range Weather Forecasts (ECMWF) analysis and operational forecast data reveals a systematic bias in the representation of KWs by the ECMWF operational forecast model.

Improvement in Forecasting Short-Term Tropical Cyclone Intensity Change and Their Rapid Intensification Using Deep Learning

Abstract Forecasting the intensity of a tropical cyclone (TC) remains challenging, particularly when it undergoes rapid changes in intensity. This study aims to develop a convolutional neural network (CNN) for 24-h forecasts of the TC intensity changes and their rapid intensifications over the western Pacific. The CNN model, the DeepTC, is trained using a unique loss function, an amplitude focal loss, to better capture large intensity changes, such as those during rapid intensification (RI) events. We showed that the DeepTC outperforms operational forecasts, with a lower mean absolute error (8.9%–10.2%) and a higher coefficient of determination (31.7%–35%). In addition, the DeepTC exhibits a substantially better skill at capturing RI events than operational forecasts. To understand the superior performance of the DeepTC in RI forecasts, we conduct an occlusion sensitivity analysis to quantify the relative importance of each predictor. Results revealed that scalar quantities such as latitude, previous intensity change, initial intensity, and vertical wind shear play critical roles in successful RI prediction. Additionally, the DeepTC utilizes the three-dimensional distribution of relative humidity to distinguish RI cases from non-RI cases, with higher dry–moist moisture gradients in the mid-to-low troposphere and steeper radial moisture gradients in the upper troposphere showed during RI events. These relationships between the identified key variables and intensity change were successfully simulated by the DeepTC, implying that the relationship is physically reasonable. Our study demonstrates that the DeepTC can be a powerful tool for improving RI understanding and enhancing the reliability of TC intensity forecasts.

Why Was the Weakening of Typhoon In-fa (2021) to the East of Taiwan Island Not Forecasted in the GFS?

Abstract Typhoon In-fa (2021) experienced a weakening process on 22–23 July in a large-scale environment favorable for tropical cyclone (TC) intensification. All operational forecasts and the Global Forecast System (GFS) forecasts predicted a continuous intensification, which deviated significantly from the observation. The analysis of the GFS analysis product shows a coalescence process of Typhoon In-fa with an intraseasonal monsoon gyre during the period, resulting in an increased outer size of In-fa and well-organized convection to the east, which prevented transporting the mass and moisture into the inner-core area of In-fa, thus leading to the weakening. Nevertheless, this essential coalescence process was not captured in the GFS forecasts due to the poor prediction of the monsoon gyre. The analysis shows that the forecasted monsoon gyre on 20–22 July had an eastward location at 72- and 96-h lead times and a weaker intensity and outer circulation at 24- and 48-h lead times, leading to the forecasted TC always moving in its north and west, in agreement with numerical simulation results that the monsoon gyre with a weaker outer circulation is not conducive to the coalescence. Thus, the deep convection to the east of In-fa preventing the inward transportation of mass and moisture did not develop in the GFS forecasts. As a result, the GFS forecasted that In-fa would continue intensifying in a favorable environment on 22–23 July. The findings of this study would prompt forecasters to pay attention to the prediction of the monsoon gyre and its influence on the TC intensity in forecast products available to them.

Prediction of Diverse Boreal Summer Intraseasonal Oscillation in the GFDL SPEAR Model

Abstract Boreal summer intraseasonal oscillation (BSISO) is a primary source of predictability for summertime weather and climate on the subseasonal-to-seasonal (S2S) time scale. Using the GFDL SPEAR S2S prediction system, we evaluate the BSISO prediction skills based on 20-yr (2000–19) hindcast experiments with initializations from May to October. It is revealed that the overall BSISO prediction skill using all hindcasts reaches out to 22 days as measured by BSISO indices before the bivariate anomalous correlation coefficient (ACC) drops below 0.5. Results also show that the northeastward-propagating canonical BSISO (CB) event has a higher prediction skill than the northward dipole BSISO (DB) event (28 vs 23 days). This is attributed to CB’s more periodic nature, resulting in its longer persistence, while DB events are more episodic accompanied by a rapid demise after reaching maximum enhanced convection over the equatorial Indian Ocean. From a forecaster’s perspective, a precursory strong Kelvin wave component in the equatorial western Pacific signifies the subsequent development of a CB event, which is likely more predictable. Investigation of individual CB events shows a large interevent spread in terms of their prediction skills. For CB, the events with weaker and fluctuating amplitude during their lifetime have relatively lower prediction skills likely linked to their weaker convection–circulation coupling. Interestingly, the prediction skills of individual CB events tend to be relatively higher and less scattered during late summer (August–October) than those in early summer (May–July), suggestive of the seasonal modulation on the evolution and predictability of BSISO. Significance Statement The advance of subseasonal-to-seasonal (S2S) prediction largely depends on dynamical models’ ability to predict some major intrinsic modes in the climate system, including the boreal summer intraseasonal oscillation (BSISO). Using a newly developed S2S prediction system, we thoroughly evaluated its performance in predicting BSISO, and revealed the skill dependence on the BSISO propagation diversity. Here we provide physical explanations of what influences the BSISO predictions and identify different precursory signals for two types of BSISO, which have important implications for operational forecasts.

Forecasting demands of blood components based on prediction models

BackgroundAn adequate blood supply is an important guarantee for saving lives and protecting health. In order to manage the blood supply more effectively when the condition of demand and supply are uncertainty, it is very important to forecast the demands of blood resources. Materials and methodsSARIMAX model and LSTM model were integrated into the prediction system of blood station. The collection and supply data of blood components was directly imported into the forecasting models to achieve automatic data update and model update. The forecasting daily demands of apheresis platelets, washing red blood cells (RBCs), suspended RBCs and plasma were recorded from January to June 2023 and compared with real data. ResultsThe prediction models had good forecasting performances. In the goodness of fit results of apheresis platelet model, the maximum value of coefficient of determination (R2) could reach 87.6%, and the minimum value of the mean absolute percentage error (MAPE) was only 0.0037. The predicted data of washing RBCs could be basically fitted, and the MAPE was 0.0121. For the prediction of suspended RBCs, the R2 was greater than 66%, and the MAPE could be 0.0372. The plasma model generated very high goodness of fit results, with R2 of over 90% and the lowest MAPE of 0.0394. ConclusionThe forecasting models, which predicts future demands of different blood components based on historical data, can help managers to overcome the challenges of blood stock control more effectively, thereby reducing blood waste and blood shortages.