Statistical Forecasting Models Research Articles

Time-series forecasting of particulate organic carbon on the Sunda Shelf: Comparative performance of the SARIMA and SARIMAX models

Conducting in-depth studies of carbon dynamics in the Southeast Asia’s Sunda Shelf waters is challenging. Particulate organic carbon (POC) is a key component that can be monitored in the region, but ocean assessment requires advanced research and POC monitoring, especially in relation to ocean dynamics and forecasts. Accurate forecasting of POC in marine environments is difficult due to the complexity of oceanic processes and the influence of various environmental factors. Statistical forecasting models such as Seasonal Autoregressive Integrated Moving Average (SARIMA) or SARIMA with Exogenous Factors (SARIMAX) are useful for this purpose. This study compares these two time-series forecasting models to assess the role of exogenous factors in predicting POC variability in Indonesian seas. The models were specified as SARIMA (3, 1, 3)x(2, 0, 0, 60) and SARIMAX (3, 1, 3)x(2, 0, 0, 60). We obtained Akaike Information Criterion (AIC) values of 1237.767 for the SARIMA model and 1207.341 for SARIMAX[exog=SST]. The SARIMA model's validation metrics were 6.94 % for MAPE, 10.80 for RMSE, and 0.65 for the correlation coefficient, outperforming the SARIMAX model. While SARIMAX incorporates additional environmental variables, SARIMA outperforms it based on MAPE, RMSE, and correlation coefficients. Our findings reveal that only sea surface temperature can significantly influence POC forecasts, thus providing a new perspective on oceanic carbon dynamics. From 2022–2030, POC levels are expected to range between 108.3 and 135.9 mg/m³, with a mean value of 120.9±5.4 mg/m³, lower than that observed from 2002 to 2022, which was 118.8±11.9 mg/m³. The highest peak in POC is predicted for the end of 2026 and 2027. The annual trend shows that the highest POC values correspond to the northwest monsoon season, with the lowest during the intermonsoon period.

Humans vs. large language models: Judgmental forecasting in an era of advanced AI

This study investigates the forecasting accuracy of human experts versus large language models (LLMs) in the retail sector, particularly during standard and promotional sales periods. Utilizing a controlled experimental setup with 123 human forecasters and five LLMs—namely, ChatGPT-4, ChatGPT3.5, Bard, Bing, and Llama2—we evaluated forecasting precision through the absolute percentage error. Our analysis centered on the effect of the following factors on forecasters’ performance: the supporting statistical model (baseline and advanced), whether the product was on promotion, and the nature of external impact. The findings indicate that LLMs do not consistently outperform humans in forecasting accuracy and that advanced statistical forecasting models do not uniformly enhance the performance of either human forecasters or LLMs. Both human and LLM forecasters exhibited increased forecasting errors, particularly during promotional periods. Our findings call for careful consideration when integrating LLMs into practical forecasting processes.

Statistical seasonal forecasting of tropical cyclone landfalls on Taiwan Island

Forecasting tropical cyclone (TC) activities has been a topic of great interest and research. Taiwan Island (TW) is one of the key regions that is highly exposed to TCs originated from the western North Pacific. Here, the authors utilize two mainstream reanalysis datasets for the period 1979–2013 and propose an effective statistical seasonal forecasting model—namely, the Sun Yat-sen University (SYSU) Model—for predicting the number of TC landfalls on TW based on the environmental factors in the preseason. The comprehensive predictor sampling and multiple linear regression show that the 850-hPa meridional wind over the west of the Antarctic Peninsula in January, the 300-hPa specific humidity over the open ocean southwest of Australia in January, the 300-hPa relative vorticity over the west of the Sea of Okhotsk in March, and the sea surface temperature in the South Indian Ocean in April, are the most significant predictors. The correlation coefficient between the modeled results and observations reaches 0.87. The model is validated by the leave-one-out and nine-fold cross-validation methods, and recent 9-yr observations (2014–2022). The Antarctic Oscillation, variabilities of the western Pacific subtropical high, Asian summer monsoon, and oceanic tunnel are the possible physical linkages or mechanisms behind the model result. The SYSU Model exhibits a 98% hit rate in 1979–2022 (43 out of 44), suggesting an operational potential in the seasonal forecasting of TC landfalls on TW.摘要本文利用1979–2013年的两个主流再分析数据集, 提出了一个中山大学 (SYSU) 热带气旋统计季节预报模型, 基于4个季前环境因子对登陆台湾岛的热带气旋数量进行预报. 模型通过了留一法, 九折交叉验证法和近9年观测数据 (2014–2022) 的验证, 模型结果与实际观测的相关系数达0.87. 南极涛动, 西太平洋副热带高压变化, 亚洲夏季风和海洋通道是模型潜在的物理联系或机制. SYSU模型在1979–2022年期间的预报准确率为98%, 表现出其业务应用价值.

A Novel Hybrid Model for Financial Forecasting Based on CEEMDAN-SE and ARIMA-CNN-LSTM

Financial time series data are characterized by non-linearity, non-stationarity, and stochastic complexity, so predicting such data presents a significant challenge. This paper proposes a novel hybrid model for financial forecasting based on CEEMDAN-SE and ARIMA- CNN-LSTM. With the help of the CEEMDAN-SE method, the original data are decomposed into several IMFs and reconstructed via sample entropy into a lower-complexity stationary high-frequency component and a low-frequency component. The high-frequency component is predicted by the ARIMA statistical forecasting model, while the low-frequency component is predicted by a neural network model combining CNN and LSTM. Compared to some classical prediction models, our algorithm exhibits superior performance in terms of three evaluation indexes, namely, RMSE, MAE, and MAPE, effectively enhancing model accuracy while reducing computational overhead.

Linear statistical forecast model based on agronomic-meteorological variables for productivity and production of paddy crop in Raipur district of Chhattisgarh

Linear statistical forecast model based on agronomic-meteorological variables for productivity and production of paddy crop in Raipur district of Chhattisgarh

The Future of Total Elbow Arthroplasty: A Statistical Forecast Model for Germany.

This study provides a statistical forecast for the development of total elbow arthroplasties (TEAs) in Germany until 2045. The authors used an autoregressive integrated moving average (ARIMA), Error-Trend-Seasonality (ETS), and Poisson model to forecast trends in total elbow arthroplasty based on demographic information and official procedure statistics. They predict a significant increase in total elbow joint replacements, with a higher prevalence among women than men. Comprehensive national data provided by the Federal Statistical Office of Germany (Statistisches Bundesamt) were used to quantify TEA's total number and incidence rates. Poisson regression, exponential smoothing with Error-Trend-Seasonality, and autoregressive integrated moving average models (ARIMA) were used to predict developments in the total number of surgeries until 2045. Overall, the number of TEAs is projected to increase continuously from 2021 to 2045. This will result in a total number of 982 (TEAs) in 2045 of mostly elderly patients above 80 years. Notably, female patients will receive TEAs 7.5 times more often than men. This is likely influenced by demographic and societal factors such as an ageing population, changes in healthcare access and utilization, and advancements in medical technology. Our projection emphasises the necessity for continuous improvements in surgical training, implant development, and rehabilitation protocols.

Automated machine learning driven model for predicting platform supply vessel freight market

Platform Supply Vessels (PSVs) play an essential role in supporting oil and gas platforms, and other offshore structures by transporting crew members, personnel, provisions, and further indispensable equipment from onshore to operational sites. PSV freight rate movements are subject to a complex array of non-linear interconnected influential factors. Reviewed literature reveals an absence of forecasting studies predicting PSV freight rates. Meanwhile, Automated Machine Learning (AutoML) frameworks have never before been employed to forecast maritime freight rates. Therefore, this study investigates factors influencing PSV time charter freight rates and explores AutoML modelling in capturing non-linearities while forecasting PSV freight rates over a 1, 3 and 6-month out-of-sample forecast horizon. A total of 43 relevant factors are included in prediction modelling, the most comprehensive number of explanatory variables in the shipping forecasting literature to date. The data consists of 188 monthly observations collected from two databases: Clarksons Shipping Intelligence Network and Offshore Intelligence Network. Time-lagged variables are utilized as data of explanatory variables are not immediately available at the time of forecasting. A total of 79 complex machine learning models are tested, and the best-performing models are Eureqa Generalized Additive Model, eXtreme Gradient Boosted Trees Regressor, and Ridge Regressor with Forecast Distance Modelling, benchmarked against the proven statistical forecasting model triple exponential smoothing. The most influential factors are historical PSV time charter freight rates, newbuilding prices, number of vessel deliveries, orderbook number, total vessel sales, and the unique variable number of active drilling rigs in the market.

Evaluating the Alaska Blocking Index as an indicator of wildfire potential in Alaska's central eastern interior

AbstractIncreased Arctic air temperatures and evaporative fluxes have coincided with more frequent and destructive high‐latitude wildfires. Arctic fires impact ecosystems and people, especially at the community‐level by degrading air quality, destroying agriculture, and threatening life and property. Central Eastern Interior (CEI) Alaska is one such region that has recently experienced the effects of wildfire activity related to warming air temperatures. To improve our ability to identify fire weather events and assess their potential for extreme outbreaks at actionable lead times relevant to fire weather forecasters and managers, new metrics and approaches need to be established and applied toward understanding the physical mechanisms underlying such wildland fire characteristics. Our study uses a new, regional atmospheric circulation metric, the Alaska Blocking Index (ABI), to describe midtropospheric air pressure around Alaska, which is subsequently related to CEI fire weather conditions at the Predictive Service Area (PSA) scale in climatological and extreme events frameworks. Of note, during years of high fire activity, Build‐Up Index (BUI) values tend to be anomalously high during the duff and drought phases across the CEI PSAs, though comparatively lower BUI values are still associated with high fire activity in the Tanana Zone‐South (AK03S) PSA. Likewise, extreme BUI values are strongly tied to high ABI values and well‐defined upper‐air ridging circulation patterns in the duff and drought periods. The statistical skill of mean daily ABI values in the 6–10 day period preceding extreme duff period BUI values is modest (τ2 > 14%) in the Upper Yukon Valley (AK02) PSA, a hotbed of wildland fire activity. Extremes in ABI and CEI BUI often occur in tandem, yielding regional predictability of upper‐air weather patterns and extremes and underlying surface weather conditions, by statistical and/or dynamical forecast models, imperative for local community and governmental organizations to effectively manage and allocate Alaska's fire weather resources.

NYSolarCast: A solar power forecasting system for New York State

Solar energy generation capacity will need to be greatly increased to meet aggressive clean energy targets by New York State (NYS), which are 70% renewable energy (RE) generation by 2030, and 100% by 2040. Because solar energy is a variable, weather-driven resource, accurate forecasts of solar energy generation both in nowcast (intra-day) and day-ahead time horizons are necessary for electric utilities and independent system operators, as they maintain grid stability and maximize RE use.Meeting this need for NYS, a gridded, open-source solar power forecasting system called NYSolarCast was developed (https://github.com/ncar/NYSolarCast_delivery). NYSolarCast makes 15-min resolution predictions of global horizontal irradiance (GHI) on a 3-km grid covering all of NYS, which are then used to estimate solar power generation both for select utility-scale PV plants (15-min resolution) and for zone-aggregated distributed PV (1-h resolution). The statewide GHI forecasts are made by applying the StatCast statistical forecasting model, which is trained on over two years of GHI observations from pyranometers at all 126 NYS Mesonet stations and gridded numerical weather prediction forecasts from both the Weather Research and Forecasting model tuned for solar applications (WRF-Solar®) and NOAA’s operational High-Resolution Rapid Refresh (HRRR) model. Forecasts are made at each NYS Mesonet site and then blended outward into the rest of the grid. This paper gives an overview of NYSolarCast performance for intra-day (0–6-h) GHI and power forecasts during a one-year period.

Impact of unexpected conditions on construction cost forecasting performance: evidence from Europe

In order to prepare proper budgets, practitioners endeavor to forecast future cost fluctuations. In recent years, the COVID-19 pandemic and the Russia-Ukraine war have caused cost overruns in most construction projects, and there has been no effort to analyze the effects of such situations on cost forecasting performances. This paper aims to investigate the impact of unexpected conditions, such as pandemics and wars, on the construction cost forecasting performances of widely used univariate time series models. For this purpose, statistical forecasting models and two widely used machine learning techniques are employed for the construction cost forecasting of nine European economies in two scenarios: a stable period (2017–2020) and a fluctuating period (2020–2023). The results showed that the success of the models varied by region and period. Unexpected global conditions cause significant errors, especially in the medium and long term. There are significant increases in forecast errors between stable and fluctuating periods. This study provides comprehensive experimental results to determine the cost-forecasting risks of construction projects under unexpected global conditions. The investigation is limited to the European economies, which are mostly developed countries with low inflation rates. Therefore, the results should also be validated in underdeveloped and developing countries.

A fusion of machine learning algorithms and traditional statistical forecasting models for analyzing American healthcare expenditure

The American healthcare system allocates considerable resources compared to peer-developed nations. However, outcomes significantly trail behind, particularly in life expectancy. This study addresses questions about the enduring trends in healthcare spending as a percentage of Gross Domestic Product (GDP), notable factors contributing to this concerning trend, and the timing to apply an emergency brake to curb this accelerating trajectory. Advanced machine learning algorithms, such as Random Forest and Support Vector Regression (SVR), in conjunction with traditional statistical forecasting methods, are used to forecast future patterns. The research underscores the importance of healthcare analytics in unraveling the intricacies of the healthcare system. The findings highlight the pressing need for effective policies to confront this mounting challenge.

Enhancing Subsurface Soil Moisture Forecasting: A Long Short-Term Memory Network Model Using Weather Data

Subsurface soil moisture is a primary determinant for root development and nutrient transportation in the soil and affects the tractability of agricultural vehicles. A statistical forecasting model, Vector AutoRegression (VAR), and a Long Short-Term Memory network (LSTM) were developed to forecast the subsurface soil moisture at a 20 cm depth using 9 years of historical weather data and subsurface soil moisture data from Fort Wayne, Indiana, USA. A time series analysis showed that the weather data and soil moisture have a stationary seasonal tendency and demonstrated that soil moisture can be forecasted from weather data. The VAR model estimates volumetric soil moisture of one-day ahead with an R2, MAE (m3m−3), MSE (m6m−6), and RMSE (m3m−3) of 0.698, 0.0561, 0.0046, and 0.0382 for 2021 corn cropping season, whereas the LSTM model using inputs of previous seven days yielded R2, MAE (m3m−3), MSE (m6m−6), and RMSE (m3m−3) of 0.998, 0.00237, 0.00002, and 0.00382, respectively as tested for cropping season of 2020 and 0.973, 0.00368, 0.00003 and 0.00577 as tested for the cropping season of 2021. The LSTM model presents a viable data-driven alternative to traditional statistical models for forecasting subsurface soil moisture.

Forecasting daily COVID-19 cases with gradient boosted regression trees and other methods: evidence from U.S. cities.

There is a vast literature on the performance of different short-term forecasting models for country specific COVID-19 cases, but much less research with respect to city level cases. This paper employs daily case counts for 25 Metropolitan Statistical Areas (MSAs) in the U.S. to evaluate the efficacy of a variety of statistical forecasting models with respect to 7 and 28-day ahead predictions. This study employed Gradient Boosted Regression Trees (GBRT), Linear Mixed Effects (LME), Susceptible, Infectious, or Recovered (SIR), and Seasonal Autoregressive Integrated Moving Average (SARIMA) models to generate daily forecasts of COVID-19 cases from November 2020 to March 2021. Consistent with other research that have employed Machine Learning (ML) based methods, we find that Median Absolute Percentage Error (MAPE) values for both 7-day ahead and 28-day ahead predictions from GBRTs are lower than corresponding values from SIR, Linear Mixed Effects (LME), and Seasonal Autoregressive Integrated Moving Average (SARIMA) specifications for the majority of MSAs during November-December 2020 and January 2021. GBRT and SARIMA models do not offer high-quality predictions for February 2021. However, SARIMA generated MAPE values for 28-day ahead predictions are slightly lower than corresponding GBRT estimates for March 2021. The results of this research demonstrate that basic ML models can lead to relatively accurate forecasts at the local level, which is important for resource allocation decisions and epidemiological surveillance by policymakers.

Statistical and deep learning models for reference evapotranspiration time series forecasting: A comparison of accuracy, complexity, and data efficiency

Reference evapotranspiration (ETo) is an essential variable in agricultural water resources management and irrigation scheduling. An accurate and reliable forecast of ETo facilitates effective decision-making in agriculture. Although numerous studies assessed various methodologies for ETo forecasting, an in-depth multi-dimensional analysis evaluating different aspects of these methodologies is missing. This study systematically evaluates the complexity, computational cost, data efficiency, and accuracy of ten models that have been used or could potentially be used for ETo forecasting. These models range from well-known statistical forecasting models like seasonal autoregressive integrated moving average (SARIMA) to state-of-the-art deep learning (DL) algorithms like temporal fusion transformer (TFT). This study categorizes monthly ETo time series from 107 weather stations across California according to their length to better understand the forecasting models' data efficiency. Moreover, two forecasting strategies (i.e., recursive and multi-input multi-output) are employed for machine learning and DL models, and forecasts are assessed for different multi-step horizons. Our findings show that statistical forecasting models like Holt-Winters' exponential smoothing perform almost as well as complex DL models. Unlike statistical models, DL models generally suffer from low data efficiency and perform well only when enough data is available. Importantly, although the computational costs of most DL models are higher than statistical methods, this is not the case for all. Considering computational cost, data efficiency, and forecasting accuracy, our findings point to the superiority of the neural basis expansion analysis for interpretable time series forecasting (N-BEATS) architecture for univariate ETo time series forecasting. Moreover, our results suggest Holt-Winters and Theta methods outperform SARIMA – the most employed statistical model for ETo forecasting in the literature – in accuracy and efficiency.

Seasonal autoregressive integrated moving average (SARIMA) time-series model for milk production forecasting in pasture-based dairy cows in the Andean highlands.

Milk production in the Andean highlands is variable over space and time. This variability is related to fluctuating environmental factors such as rainfall season which directly influence the availability of livestock feeding resources. The main aim of this study was to develop a time-series model to forecast milk production in a mountainous geographical area by analysing the dynamics of milk records thorough the year. The study was carried out in the Andean highlands, using time-series models of monthly milk records collected routinely from dairy cows maintained in a controlled experimental farm over a 9-year period (2008-2016). Several statistical forecasting models were compared. The Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and Mean Absolute Percent Error (MAPE) were used as selection criteria to compare models. A relation between monthly milk records and the season of the year was modelled using seasonal autoregressive integrated moving average (SARIMA) methods to explore temporal redundancy (trends and periodicity). According to white noise residual test (Q = 13.951 and p = 0.052), Akaike Information Criterion and MAE, MAPE, and RMSE values, the SARIMA (1, 0, 0) x (2, 0, 0)12 time-series model resulted slightly better forecasting model compared to others. In conclusion, time-series models were promising, simple and useful tools for producing reasonably reliable forecasts of milk production thorough the year in the Andean highlands. The forecasting potential of the different models were similar and they could be used indistinctly to forecast the milk production seasonal fluctuations. However, the SARIMA model performed the best good predictive capacity minimizing the prediction interval error. Thus, a useful effective strategy has been developed by using time-series models to monitor milk production and alleviate production drops due to seasonal factors in the Andean highlands.

Seasonal forecast of tropical cyclones in the Southwest Pacific Ocean

AbstractPredictions of tropical cyclone (TC) activity have been a topic of recurrent interest and research in the past. Here we utilize reanalysis datasets of sea surface temperature (SST) and mixed layer depth (MLD) to build a statistical seasonal forecasting model that produces outlooks of expected TC counts in the region of the Southwest Pacific (SWP). Nevertheless, the model applicability can be extended to other regions and basins. A novel TC predictor index is developed at the daily scale and used to obtain an objective classification of synoptic weather patterns. This classification has been performed by clustering the daily index predictor fields, previously transformed into principal components, using a K‐mean algorithm. As a result, 49 daily weather types (DWTs) are presented which inform about the mean representative features and spatial patterns of both predictor and predictand variables. Thus, statistical relationships between TC activity and nonlinear combinations of predictor variables are found to assign daily rates of expected TCs. The cluster‐based model is calibrated from 1982 to 2019 and validated by recent TC season observations, demonstrating the operational application using ensembles of long‐term predictions in the Southwest Pacific. Results have shown which synoptic types of SST and MLD are favourable to cyclogenesis and activity, with additional information related to concurrent sea level pressure and precipitation synoptic patterns, as well as seasonal and interannual climate variability.

A Statistical Forecast Model for Extratropical Cyclones Including Intensity and Precipitation Type

Abstract Winter extratropical cyclones (ETCs) are dominant features of winter weather on the east coast of North America. These storms are characterized by high winds and heavy precipitation (rain, snow, and ice). ETCs are well predicted by numerical weather prediction models (NWPs) at short- to midrange forecast lead times, but prediction on seasonal time scales is lacking. We develop a set of multiple linear regression models, using stepwise regression and cross validation, to predict the number of storms expected to affect a specific location throughout the winter storm season. Each model in the set predicts a specific storm type (e.g., snow, rain, or bomb storms). This set of models is applied in a probabilistic forecast framework that uses the probability density function of the prediction in combination with climatological mean storm activity. The resulting forecast makes statements about the likelihood of below-average, average, or above-average activity for all storms and for each of the type-specific subsets of storms. Though this forecast framework could in theory be applied anywhere, we demonstrate its skill in forecasting the characteristics of the winter storm season experienced in Halifax, Nova Scotia, Canada. Significance Statement Winter storms are a disruptive but inevitable part of life on the eastern coast of North America all the way from the Carolinas to Labrador. Knowing each fall what to expect for the upcoming winter storm season is not only a matter of public interest, but also of great public safety and financial importance. Here we develop a model that uses the state of the atmosphere over the month of September to forecast the upcoming winter storm characteristics for a specified region of interest. Our model uses a multiple linear regression approach to make skilled forecasts including probability statements about the level and type of storm activity. Forecasts can be used to inform planning for the winter ahead.

Diagnosing Tropical Cyclone Intensity Variations from the Surface Wind Field Evolution

Abstract Tropical cyclone (TC) intensity fluctuations remain a challenge for TC forecasters. Occurring through a wide range of processes, such as vortex contraction, eyewall replacements, or emission of vortex Rossby waves, they are inherently multiscale, transient, and asymmetric. In a recent study, estimates of surface wind field inner-core properties from high-resolution satellite observations were spotted as valuable for the improvement of intensity variations statistical predictability. The present study evaluates how the temporal evolution of the vortex structure, at scales ranging from O(1) km to vortexwide, further provides insights on the modulation of intensity. The study is based on a set of seven realistic TC simulations with 1-km grid spacing. The surface wind field structure is studied through an original set of descriptors that characterize the radial profile, the azimuthal asymmetries, and their spectral distribution. While radial gradients evolve concurrently with intensity, the azimuthal variability of the inner core shows a stronger connection with shorter-scale intensity modulation. The increase of high-wavenumber asymmetries distributed around the ring of maximum winds is shown to precede phases of rapid (re)intensification by 5–6 h, while the concentration of asymmetry in wavenumbers 1 and 2 leads to intensity weakening. A machine learning classification finally highlights that the classification of intensification phases (i.e., intensification or weakening) can be improved by at least 11% (thus reaching ∼75%) when accounting for the evolution of the radial wind gradient and the variance distribution among scales in the ring of maximum wind, relative to the sole use of vortex-averaged parameters. Significance Statement The purpose of this study is to relate changes in the surface wind field structure of tropical cyclones to their intensity variations. We design an original set of parameters to characterize the inner- and near-core contraction and asymmetry, and evaluate their connection with TC intensity modulations in a set of realistic high-resolution numerical simulations. The main outcome of our study is that opposite trends in high- and low-wavenumber variance tend to occur prior to intensity changes on short time scales, with more widespread and local-scale asymmetry corresponding to intensification, and vortex-scale polarized asymmetry corresponding to weakening. These diagnoses are shown to improve the classification of intensification and weakening phases. These results advocate for enhancing real-time high-resolution observations of surface wind fields under TCs, and using asymmetry distribution as predictors in statistical forecast models.

Development of a dynamical statistical analog ensemble forecast model for landfalling typhoon disasters

In this report, the development of a Dynamical Statistical Analog Ensemble Forecast model for landfalling typhoon disasters (LTDs) and some applications over coastal China are described. This model consists of the following four elements: (i) obtaining the forecast track of a target landfalling typhoon, (ii) constructing its generalized initial value (GIV), (iii) identifying its analogs based on the GIV, and (iv) assembling typhoon disasters of the analogs. Typhoon track, intensity, and landfall date are introduced in GIV at this early development stage. The pre-assessment results show that the mean threat scores of two important damage levels of LTDs reach 0.48 and 0.55, respectively. Of significance is that most of the damage occurs near the typhoon centers around the time of landfall. These results indicate the promising performance of the model in capturing the main damage characteristics of typhoon disasters, which would help coastal community mitigate damage from destructive typhoons.

Climate prediction of the seasonal sea-ice early melt onset in the Bering Sea

Based on the impact of large-scale circulation anomalies on sea-ice melting, this paper develops a statistical forecasting model for the seasonal sea-ice early melt onset (EMO) in the Bering Sea using the interannual increment prediction method. The prediction model considers three physically meaningful predictors: the January Beaufort High (P1-H500), the November sea-level pressure (P2-SLP) over eastern Siberia, and the November snow cover over the eastern European Plain (P3-Snowc). P1-H500 can influence the sea surface temperature (SST) anomaly in the Bering Sea through ocean–atmosphere interactions, and this SST anomaly can persist from January to March. Subsequently, it affects the EMO in the Bering Sea. P2-SLP exhibits a close association with the east part of the midlatitude North Pacific SST in November. The colder midlatitude North Pacific SST anomalies, which persist from November until January and February of the following year, will be accompanied by warmer SST anomalies in the Bering Sea, which result in a decreased sea-ice extent and a later-than-usual EMO. The Arctic dipole anomaly in January is one of the ways in which P3-Snowc affects the EMO in the following year. The predicted EMO shows good agreement with the observed EMO in the cross-validation test for 1981–2022, with a temporal correlation coefficient of 0.45, exceeding the 99% confidence level. The prediction accuracy of the prediction model for positive and negative abnormal years of EMO is 60% and 41%, respectively.摘要基于大尺度环流异常对海冰消融的影响过程, 本文采用年际增量预测方法研制了白令海季节性海冰早期消融开始日期(EMO)的统计预测模型. 预测模型选取了3个具有明确物理意义的预测因子: 1月波弗特高压, 前期11月东西伯利亚地区海平面气压, 以及11月东欧平原积雪覆盖率. 1月波弗特高压可以通过海气相互作用影响白令海地区海温异常, 该海温异常能够从1月持续到3月, 进而影响白令海EMO. 11月东西伯利亚地区海平面气压与11月至次年2月北太平洋中纬度东部海温密切相关. 伴随着北太平洋中纬度东部冷海温异常的出现, 白令海地区会出现暖海温异常, 进而导致白令海海冰范围减少, EMO较晚. 1月北极偶极子异常是11月东欧平原积雪覆盖率影响次年白令海EMO的桥梁之一. 1981−2022年的交叉检验结果表明: 统计模型对白令海EMO具有较好的预测能力, 预测与观测的EMO之间时间相关系数达到了0.45, 超过了99%的置信水平. 统计模型对白令海EMO正常年份和异常年份的预测准确率分别为60%和41%.