Deterministic Forecasts Research Articles

English

Uncertainty is an intrinsic characteristic of demographic processes. Th is applies even more to the future. Accurate deterministic forecasts are fundamentally impossible. Th is determines the necessity to quantify the future uncertainty. Th e purpose of this research is to develop probabilistic demographic forecasts for the metropolises of Ukraine and analyze the outcome results. For the fi rst time, probabilistic demographic forecasts have been developed for individual cities of Ukraine. Th e study was carried out using the functional data approach which incorporates wide set of demographical methods and models implemented in several packages of R programming language. Chosen methodology is based entirely on statistics and does not require introducing any additional arbitrary hypotheses. At three cases (namely for fertility in Kyiv, Lviv and Kharkiv) the default method (ARIMA) showed implausible results which could be induced by unreliable current data. In these cases were used random walk model. For Odesa the both models give similar results. It is possible that in this city the underestimation of the departed population is compensated by the underestimation of the arrived, which leads to the relevance of the current fertility rates (namely their denominators) and, consequently, the consistency of the forecast results regardless of the method. Mortality forecasts are consistent with the dynamics of mortality rates being observed and the quality of current data. Th e model captured upward life expectancy trends for Dnipro and Odesa and stagnation for other cities. Th is is also could be caused by denominator inconsistencyfor the latter ones. Computation showed that the population size of Dnipro, Lviv and Kharkiv in 2040 is most likely to be below the population number reached in early 2019. Some chances for population growth remain in Odesa and Kyiv is likely to have a larger population. Th e age distribution of the population in all cities in future looks similar. The number of people over 40 years of age has least uncertainty. At the age of 20 to 40 years, the uncertainty is much higher. Th is is a consequence of the uncertainty of youth migration during the forecast horizon of 2019–2039, because all these cities are powerful educational centres and attract students. In 2040 those who were students in 2020 will reach the age of 40 and can stay in the big city or leave. Uncertainty of the number of persons under 20 is formed from two sources: uncertainty of fertility forecasts and uncertainty of the number of reproductive cohort, i.e. those 20-40-year-olds. It is needed to review these forecasts aft er receiving the results of the closest census.

What life-course research can contribute to futures studies

Futures studies is a multidisciplinary field that integrates disciplinary information through discussion. This article integrates information from life-course research with futures studies. Life-course research explores how human lives develop over time, focusing on the past and present. Futures studies may use this information to explore possible, probable, and preferable future developments of human lives. Life-course research defines a stable social context and standardized life-courses as a requirement for futures studies under shallow uncertainty. In this situation, important experiences in a person’s life and social institutions create life-course patterns that may inform deterministic forecasts. Futures studies under medium uncertainty may be carried out when life-courses change gradually over time or when similar countries are compared. Here, personality also creates life-course patterns that may inform probabilistic forecasts. When social contexts change fundamentally or unexpectedly, then futures studies under deep uncertainty are called for. In this situation, important events, social institutions, and personality may inform foresight on future life-course patterns. Finally, futures studies under recognized ignorance may use previous life-course research for inspiration. These insights contribute to cumulative knowledge building, and they underline the opportunities and limitations of futures studies.

Prediction intervals estimation of solar generation based on gated recurrent unit and kernel density estimation

With the increasing attention to the energy crisis and global warming, solar generation has become an important way to use clean solar energy and is playing an increasingly important role. Due to the highly-variable patterns of solar generation, the estimation of prediction intervals is receiving more attention, which is conducive to the safe and stable operation of the power system. In order to further improve the performance of prediction intervals of solar generation, this paper proposes a prediction intervals estimation method for solar generation based on gated recurrent unit (GRU) neural networks and kernel density estimation (KDE). GRU, a commonly used recurrent neural networks, is utilized to obtain the deterministic forecast of solar generation. In addition, according to the characteristics of solar generation, attention mechanism is designed on the GRU prediction model to further improve the prediction performance. Then, the KDE method is used to fit the prediction errors of solar generation obtained by the deterministic forecasting method. In order to verify the effectiveness of the proposed method, we have carried out a large number of experiments on freely available datasets. The experimental results show that the proposed method outperforms competing methods and can generate high-quality prediction intervals.

On the Operational Flood Forecasting Practices Using Low-Quality Data Input of a Distributed Hydrological Model

Low-quality input data (such as sparse rainfall gauges, low spatial resolution soil type and land use maps) have limited the application of physically-based distributed hydrological models in operational practices in many data-sparse regions. It is necessary to quantify the uncertainty in the deterministic forecast results of distributed models. In this paper, the TOPographic Kinematic Approximation and Integration (TOPKAPI) distributed model was used for deterministic forecasts with low-quality input data, and then the Hydrologic Uncertainty Processor (HUP) was used to provide the probabilistic forecast results for operational practices. Results showed that the deterministic forecasts by TOPKAPI performed poorly in some flood seasons, such as the years 1997, 2001 and 2008, despite which the overall accuracy of the whole study period 1996–2008 could be acceptable and generally reproduced the hydrological behaviors of the catchment (Lushi basin, China). The HUP model can not only provide probabilistic forecasts (e.g., 90% predictive uncertainty bounds), but also provides deterministic forecasts in terms of 50% percentiles. The 50% percentiles obviously improved the forecast accuracy of selected flood events at the leading time of one hour. Besides, the HUP performance decayed with the leading time increasing (6, 12 h). This work revealed that deterministic model outputs had large uncertainties in flood forecasts, and the HUP model may provide an alternative for operational flood forecasting practices in those areas with low-quality data.

Introducing entropy-based Bayesian model averaging for streamflow forecast

Bayesian Model Averaging (BMA) is a well-known statistical post-processing approach for probabilistically merging individual forecasts. In BMA, the posterior distribution of the predictand variable is determined by implementing the law of total probability. Therefore, possessing an ensemble of independent members (mutually exclusive) with the highest information content about observation variability (collectively exhaustive) is the main inherent assumption of the original BMA method. Mutually exclusive and collectively exhaustive are two contradictory criteria. Although constructing an ensemble of members that fully satisfied these two properties is practically impossible, providing a balance between them is a key requirement for enhancing the BMA performance. Through coupling BMA with Shannon entropy of information theory, this study proposes an entropy-based selection procedure to construct an ensemble of streamflow forecasts by better addressing the aforementioned contradictory criteria prior to performing the BMA. We investigate the effects of using ensembles with the aforementioned properties by comparing the results of original BMA with the proposed entropy-based BMA (En-BMA) for short- to medium-range daily streamflow forecasts in two different watersheds. The results indicate that the En-BMA leads to better results particularly for high flow predictions. Both probabilistic and deterministic high flow forecasts are more accurate and reliable when using the En-BMA approach. However, for the average flow forecasts, there are no clear differences in the general performance of both methods. The improvements observed are more pronounced for shorter lead-times and less pronounced, but still present, for longer lead times.

Reconciling solar forecasts: Probabilistic forecasting with homoscedastic Gaussian errors on a geographical hierarchy

Hierarchical forecasting and reconciliation are new to the field of solar engineering. Previous papers in this series, namely, Yang et al. (2017a, 2017b), and Yagli et al. (2019b), discussed various reconciliation techniques for deterministic solar forecasts obtained across spatio-temporal hierarchies. This paper extends the discussion into probability space, and studies how reconciliation can affect the performance of probabilistic forecasting. More specifically, qualities of the parametric predictive distributions before and after reconciliation are compared.Four minimum-trace-based reconciliation techniques are used to reconcile day-ahead and hour-ahead forecasts generated using two datasets: (1) distributed solar power generation for 318 simulated PV systems in California, and (2) satellite-derived irradiance over Arizona. The empirical result shows that reconciliation not only improves the accuracy of point forecasts, but also leads to high-quality predictive distributions in terms of sharpness, calibration, and skill score. Moreover, such improvement is quite general, and does not seem to depend on data, hierarchy structure, nor the underlying forecasting model.

Improving Air Quality Predictions over the United States with an Analog Ensemble

AbstractAir quality forecasts produced by the National Air Quality Forecasting Capability (NAQFC) help air quality forecasters across the United States in making informed decisions to protect public health from acute air pollution episodes. However, errors in air quality forecasts limit their value in the decision-making process. This study aims to enhance the accuracy of NAQFC air quality forecasts and reliably quantify their uncertainties using a statistical–dynamical method called the analog ensemble (AnEn), which has previously been found to efficiently generate probabilistic forecasts for other applications. AnEn estimates of the probability of the true state of a predictand are based on a current deterministic numerical prediction and an archive of prior analogous predictions paired with prior observations. The method avoids the complexity and real-time computational expense of model-based ensembles and is proposed here for the first time for air quality forecasting. AnEn is applied with forecasts from the Community Multiscale Air Quality (CMAQ) model. Relative to CMAQ raw forecasts, deterministic forecasts of surface ozone (O3) and particulate matter of aerodynamic diameter smaller than 2.5 μm (PM2.5) based on AnEn’s mean have lower systemic and random errors and are overall better correlated with observations; for example, when computed across all sites and lead times, AnEn’s root-mean-square error is lower than CMAQ’s by roughly 35% and 30% for O3 and PM2.5, respectively, and AnEn improves the correlation by 50% for O3 and PM2.5. Probabilistic forecasts from AnEn are statistically consistent, reliable, and sharp, and they quantify the uncertainty of the underlying prediction.

Probabilistic solar irradiance forecasting via a deep learning‐based hybrid approach

Probabilistic solar irradiance forecasting has received widespread attention in recent years, as it provides more uncertainty information for the future photovoltaic generation. In this study, a hybrid probabilistic solar irradiance prediction method is proposed, which combines a deep recurrent neural network and residual modeling. Specifically, the long short‐term memory‐based point prediction using historical records and related features is applied to obtain deterministic forecasts. Next, these deterministic forecasts are employed as inputs to estimate the residual distributions. Furthermore, maximum likelihood estimation is utilized to compute the parameters of the residual distribution. Finally, the point prediction and residual distribution jointly generate the final probabilistic forecasting results. Compared with other deterministic and probabilistic forecasting models, the proposed method yields promising results on a publicly available dataset. © 2020 Institute of Electrical Engineers of Japan. Published by Wiley Periodicals LLC.

Impacts of Multiple Radiance Data Assimilation on the Simulation of Typhoon Chan-Hom

With the module of assimilating AMSU-A (Advanced Microwave Sounding Unit-A) and AIRS (Atmospheric Infrared Sounder) data in the WRFDA (Weather Research and Forecasting Model Data Assimilation) system, the impacts of joint assimilation of the radiance observations from two satellites on the simulation of typhoon Chan-hom (2015) are addressed. For comparison, experiments with the assimilation of solely GTS (Global Telecommunications System) data, AMSU-A data, or AIRS data are also performed. The results show that, compared to other experiments, the analysis field after assimilating multiple radiance data is closer to the observation. The simulated steering flow in its forecast field is conductive to the northeast twist of the typhoon. In addition, the simulated rainband and the FSS (fraction skill score) calculated from the experiment with assimilating multiple radiance data are better. In the deterministic forecast, better performance is obtained from the simulation with multiple radiance data in the forecast of track, MSLP (minimum sea level pressure), and MSW (maximum surface wind).

Integration of quantitative precipitation forecasts with real-time hydrology and hydraulics modeling towards probabilistic forecasting of urban flooding

As the world becomes more urbanized and heavy precipitation events increase in frequency and intensity, urban (pluvial) flooding is an emerging concern. Recent advances in hydrologic and hydraulic modeling, high-resolution quantitative precipitation forecasting, and ensemble forecasting have improved the ability to evaluate flash flooding potential in urban areas. The Probabilistic Urban Flash Flood Information Nexis (PUFFIN) app integrates these components into a tool to evaluate the probability of an urban flash flood event and to identify specific infrastructure components at risk. PUFFIN uses a combination of deterministic and probabilistic precipitation forecasts to provide analyses over a range of spatial and temporal resolutions. A case study for the City of Roanoke, Virginia in the United States demonstrates how PUFFIN can be used to assess the risk of urban flooding. However, with minimal modification, the application can be adapted for implementation in other locations.

Enhanced Wind Generation Forecast Using Robust Ensemble Learning

This letter proposes a robust ensemble learning scheme to enhance short-term prediction of wind power generation. The ensemble problem associated with pruning and combination is formulated as a worst-case robust approximation problem, taking forecast uncertainty in individual predictors into consideration. This problem is then transformed into the scaled form of the augmented Lagrangian and is solved via the alternating direction method of multipliers (ADMM). The proposed scheme can be applied to both deterministic and probabilistic forecasting. A comprehensive study is carried out to illustrate the advantage of the proposed scheme in both point and interval forecasting.

Deterministic and probabilistic wind speed forecasting with de-noising-reconstruction strategy and quantile regression based algorithm

Wind energy has become a kind of attractive alternative energy in power generation field due to its nonpolluting and renewable properties. Wind speed forecasting acts an important role in programming and operation of power systems. However, achieving high precision wind speed forecasts is still consider as an arduous and challenging issue with the randomization and transient exist in wind speed time series. For this reason, this paper proposed two novel de-noising-reconstruction-based hybrid models which consist of novel signal decomposed methods, feature selection approaches and predictors based on quantile regression and optimization algorithm to achieve more accurate short term wind speed forecasting. The developed hybrid models firstly eliminate inherent noise from the wind speed sequences via decomposed method and subsequently construct the appropriate datasets for the forecasting engines by adopting the feature selection method; finally, establish the predictors for the forecasting task. To verify the effectiveness of proposed forecasting models, 1-h and 2-h wind speed data collected from Yumen, Gansu province of China mainland is used as case studies. The computational results demonstrated that the developed hybrid models yield better performance contrast with those of other models involved in this research in terms of both wind speed deterministic and probabilistic forecasting.

A Technique for Seamless Forecast Construction and Validation from Weather to Monthly Time Scales

Abstract Seamless prediction means bridging discrete short-term weather forecasts valid at a specific time and time-averaged forecasts at longer periods. Subseasonal predictions span this time range and must contend with this transition. Seamless forecasts and seamless validation methods go hand-in-hand. Time-averaged forecasts often feature a verification window that widens in time with growing forecast leads. Ideally, a smooth transition across daily to monthly time scales would provide true seamlessness—a generalized approach is presented here to accomplish this. We discuss prior attempts to achieve this transition with individual weighting functions before presenting the two-parameter Hill equation as a general weighting function to blend discrete and time-averaged forecasts, achieving seamlessness. The Hill equation can be tuned to specify the lead time at which the discrete forecast loses dominance to time-averaged forecasts, as well as the swiftness of the transition with lead time. For this application, discrete forecasts are defined at any lead time using a Kronecker delta weighting, and any time-averaged weighting approach can be used at longer leads. Time-averaged weighting functions whose averaging window widens with lead time are used. Example applications are shown for deterministic and ensemble forecasts and validation and a variety of validation metrics, along with sensitivities to parameter choices and a discussion of caveats. This technique aims to counterbalance the natural increase in uncertainty with forecast lead. It is not meant to construct forecasts with the highest skill, but to construct forecasts with the highest utility across time scales from weather to subseasonal in a single seamless product.

Hierarchical demand forecasting benchmark for the distribution grid

We present a comparative study of different probabilistic forecasting techniques on the task of predicting the electrical load of secondary substations and cabinets located in a low voltage distribution grid, as well as their aggregated power profile. The methods are evaluated using standard KPIs for deterministic and probabilistic forecasts. We also compare the ability of different hierarchical techniques in improving the bottom level forecasters’ performances. Both the raw and cleaned datasets, including meteorological data, are made publicly available to provide a standard benchmark for evaluating forecasting algorithms for demand-side management applications.

Load probability density forecasting by transforming and combining quantile forecasts

Compared with traditional deterministic load forecasting, probabilistic load forecasting (PLF) help us understand the potential risks in the power system operation by providing more information about future uncertainties of the loads. Quantile forecasting, as a kind of non-parametric probabilistic forecasting method, has been well developed and widely used in PLF. However, the results of quantile forecasts are discrete, which contain fewer details than density forecasts which provide the most comprehensive information. This paper proposes a novel day-ahead load probability density forecasting method by transforming and combining multiple quantile forecasts. The proposed method includes two main steps: transformation and combination. In the first step, the kernel density estimation method is used to transform the individual quantile forecast into the probability density curve; in the second step, an optimization problem is established to obtain the weighted combination of different probability density forecasts. The perturbation search method is applied to determine the optimal weight of each individual forecast. We demonstrate the effectiveness and superiority of our proposed method using comprehensive case studies on the real-world load data from Guangdong province in China, ISO New England (ISO-NE) in the US and Irish smart meter data. Case studies show that the combined model is robust to kernel function selection in the transformation step and has better forecasting performance. Compared with the best individual model, the purposed combined model has an accuracy improvement of 1.54% in the Guangdong dataset and 2.9% in the ISO-NE dataset in terms of the continuous ranked probability score. The proposed combination forecasting method can be robust in high volatility scenarios.

Internal and external coupling of Gaussian mixture model and deep recurrent network for probabilistic drought forecasting

Developing an accurate drought forecasting model is significantly important for minimizing the economic and ecological damage due to drought hazards. The competitive performances of deep neuron networks in drought forecasting have been proved in the last few years. However, the uncertainty of drought events has not been effectively addressed owing to massive computational requirements involved in deep learning framework. Instead of developing a heavy ensemble probabilistic model, this study introduced an improved long short-term memory network that is aggregated with a conditional probability model for probabilistic drought forecasting. Two different aggregation strategies are carefully discussed, implemented, and evaluated. For external coupling, the conditional probability model is used as a postprocessing technique that applied on deterministic forecasting. For internal coupling, the conditional probability model is embedded into the internal structure of a long short-term memory network, making the improved long short-term memory network model probabilistic in nature. The proposed methods were demonstrated though actual drought forecasts in the three-river headwater region of China. In addition, three state-of-the-art models: autoregressive moving average, generalized linear regression, and artificial neural networks, were developed for drought forecasting served as a comparison. The forecasting abilities were evaluated using four types of verification methods with respect to the predictive mean and interval. The obtained results showed that the improved model could successfully forecast drought index as well as uncertainty information. Moreover, the internal aggregation strategy was validated as superior in the probabilistic drought forecasting in the three-river headwater region.

AeDES: a next-generation monitoring and forecasting system for environmental suitability of Aedes-borne disease transmission

Aedes-borne diseases, such as dengue and chikungunya, are responsible for more than 50 million infections worldwide every year, with an overall increase of 30-fold in the last 50 years, mainly due to city population growth, more frequent travels and ecological changes. In the United States of America, the vast majority of Aedes-borne infections are imported from endemic regions by travelers, who can become new sources of mosquito infection upon their return home if the exposed population is susceptible to the disease, and if suitable environmental conditions for the mosquitoes and the virus are present. Since the susceptibility of the human population can be determined via periodic monitoring campaigns, the environmental suitability for the presence of mosquitoes and viruses becomes one of the most important pieces of information for decision makers in the health sector. We present a next-generation monitoring and forecasting system for underline{textit{Ae}}{} textit{des}-borne diseases’ environmental suitability (AeDES) of transmission in the conterminous United States and transboundary regions, using calibrated ento-epidemiological models, climate models and temperature observations. After analyzing the seasonal predictive skill of AeDES, we briefly consider the recent Zika epidemic, and the compound effects of the current Central American dengue outbreak happening during the SARS-CoV-2 pandemic, to illustrate how a combination of tailored deterministic and probabilistic forecasts can inform key prevention and control strategies .

Texture‐based classification of high‐resolution precipitation forecasts with machine‐learning methods

AbstractObject‐based methods are commonly used for verification and postprocessing of high‐resolution precipitation forecasts. They usually detect objects based on intensity criteria only, without considering the spatial organization of rainfall, known as texture. This article evaluates the performance of several machine‐learning methods to detect “continuous” and “intermittent” rainfall patterns in the forecasts of the French convective‐scale Arome model. A sliding‐window segmentation algorithm, which applies a classification model at each grid point, is implemented. Several classifiers and input textural features are compared. Overall, intermittent precipitation is the most difficult to detect. The random forest classifier is shown to provide the best classification results independently of the predictor used, with a surprising ability to extract a relevant signal from a synthetic descriptor such as the rainfall cumulative distribution function, as well as from the large amount of unprocessed information provided by neighbouring grid points. On the other hand, the logistic regression classifier needs a texture‐oriented predictor, such as the statistics derived from the grey‐level co‐occurrence matrix, to perform well. Global insight into model behaviour is then obtained by examining the importance of input features. Finally, we show that random forests trained on Arome deterministic forecasts can be applied successfully to discriminate between precipitation textures in different Arome configuration outputs and gridded observations.

Robustness of Short-Term Wind Power Forecasting against False Data Injection Attacks

The accuracy of wind power forecasting depends a great deal on the data quality, which is so susceptible to cybersecurity attacks. In this paper, we study the cybersecurity issue of short-term wind power forecasting. We present one class of data attacks, called false data injection attacks, against wind power deterministic and probabilistic forecasting. We show that any malicious data can be injected to historical data without being discovered by one of the commonly-used anomaly detection techniques. Moreover, we testify that attackers can launch such data attacks even with limited resources. To study the impact of data attacks on the forecasting accuracy, we establish the framework of simulating false data injection attacks using the Monte Carlo method. Then, the robustness of six representative wind power forecasting models is tested. Numerical results on real-world data demonstrate that the support vector machine and k-nearest neighbors combined with kernel density estimator are the most robust deterministic and probabilistic forecasting ones among six representative models, respectively. Nevertheless, none of them can issue accurate forecasts under very strong false data attacks. This presents a serious challenge to the community of wind power forecasting. The challenge is to study robust wind power forecasting models dealing with false data attacks.

Improving subseasonal precipitation forecasts through a statistical–dynamical approach : application to the southwest tropical Pacific

Subseasonal forecasts are based on coupled general circulation models that often have a good representation of large-scale climate drivers affecting rainfall. Yet, they have more difficulty in providing accurate precipitation forecasts. This study proposes a statistical-dynamical post-processing scheme based on a bayesian framework to improve the quality of subseasonal forecasts of weekly precipitation. The method takes advantage of dynamically-forecast precipitation (calibration) and large-scale climate features (bridging) to enhance forecast skill through a statistical model. It is applied to the austral summer precipitation reforecasts in the southwest tropical Pacific, using the Meteo-France and ECMWF reforecasts in the Subseasonal-to-seasonal (S2S) database. The large-scale predictors used for bridging are climate indices related to El Nino Southern Oscillation and the Madden–Julian Oscillation, that are the major sources of predictability in the area. Skill is assessed with a Mean Square Skill Score for deterministic forecasts, while probabilistic forecasts of heavy rainfall spells are evaluated in terms of discrimination (ROC skill score) and reliability. This bayesian method leads to a significant improvement of all metrics used to assess probabilistic forecasts at all lead times (from week 1 to week 4). In the case of the Meteo-France S2S system, it also leads to strong error reduction. Further investigation shows that the calibration part of the method, using forecast precipitation as a predictor, is necessary to achieve any improvement. The bridging part, and particularly the ENSO-related information, also provides additional discrimination skill, while the MJO-related information is not really useful beyond week 2 over the region of interest.