Forecast Lead Time Research Articles

Evaluation of Surface Conditions from Operational Forecasts Using in situ Saildrone Observations in the Pacific Arctic

Abstract Observations from uncrewed surface vehicles (saildrones) in the Bering, Chukchi, and Beaufort Seas during June – September 2019 were used to evaluate initial conditions and forecasts with lead times up to 10 days produced by eight operational numerical weather prediction centers. Prediction error behaviors in pressure and wind are found to be different from those in temperature and humidity. For example, errors in surface pressure were small in short-range (<6 days) forecasts, but they grew rapidly with increasing lead time beyond 6 days. Non-weighted multi-model means outperformed all individual models approaching a 10-day forecast lead time. In contrast, errors in surface air temperature and relative humidity could be large in initial conditions and remained large through 10-day forecasts without much growth, and non-weighted multi-model means did not outperform all individual models. These results following the tracks of the mobile platforms are consistent with those at a fixed location. Large errors in initial condition of sea surface temperature (SST) resulted in part from the unusual Arctic surface warming in 2019 not captured by data assimilation systems used for model initialization. These errors in SST led to large initial and prediction errors in surface air temperature. Our results suggest that improving predictions of surface conditions over the Arctic Ocean requires enhanced in situ observations and better data assimilation capability for more accurate initial conditions as well as better model physics. Numerical predictions of Arctic atmospheric conditions may continue to suffer from large errors if they do not fully capture the large SST anomalies related to Arctic warming.

A Hybrid Multivariate Deep Learning Network for Multistep Ahead Sea Level Anomaly Forecasting

Abstract The accumulated remote sensing data of altimeters and scatterometers have provided new opportunities for ocean state forecasting and have improved our knowledge of ocean–atmosphere exchanges. Studies on multivariate, multistep, spatiotemporal sequence forecasts of sea level anomalies (SLA) for different modalities, however, remain problematic. In this paper, we present a novel hybrid and multivariate deep neural network, named HMnet3, which can be used for SLA forecasting in the South China Sea (SCS). First, a spatiotemporal sequence forecasting network is trained by an improved convolutional long short-term memory (ConvLSTM) network using a channelwise attention mechanism and multivariate data from 1993 to 2015. Then a time series forecasting network is trained by an improved long short-term memory (LSTM) network, which is realized by ensemble empirical mode decomposition (EEMD). Finally, the two networks are combined by a successive correction method to produce SLA forecasts for lead times of up to 15 days, with a special focus on the open sea and coastal regions of the SCS. During the testing period of 2016–18, the performance of HMnet3 with sea surface temperature anomaly (SSTA), wind speed anomaly (SPDA), and SLA data is much better than those of state-of-the-art dynamic and statistical (ConvLSTM, persistence, and climatology) forecast models. Stricter testbeds for trial simulation experiments with real-time datasets are investigated, where the eddy classification metrics of HMnet3 are favorable for all properties, especially for those of small-scale eddies.

Tropical cyclone track and intensity prediction skill of GFS model over NIO during 2019 & 2020

The Tropical Cyclone (TC) track prediction using different NWP models and its verification is the critical task to provide prior knowledge about the model errors, which is beneficial for giving the model guidance-based real-time cyclone warning advisories. This study has attempted to verify the Global Forecast System (GFS) model forecasted tropical cyclone track and intensity over the North Indian Ocean (NIO) for the years 2019 and 2020. GFS is one of the operational models in the India Meteorological Department (IMD), which provides the medium-range weather forecast up to 10 days. The forecasted tracks from the GFS forecast are obtained using a vortex tracker developed by Geophysical Fluid Dynamics Laboratory (GFDL). A total of 13 tropical cyclones formed over the North Indian Ocean, eight during 2019 and five in 2020 have been considered in this study. The accuracy of the model predicted tracks and intensity is verified for five days forecasts (120 h) at 6-h intervals; the track errors are verified in terms of Direct Position Error (DPE), Along Track Error (ATE) and Cross-Track Error (CTE). The annual mean DPE over NIO during 2019 (51–331 km) is lower than 2020 (82–359 km), and the DPE is less than 150 km up to 66 h during 2019 and 48 h during 2020. The positive ATE (76–332 km) indicates the predicted track movement is faster than the observed track during both years. The positive CTE values for most forecast lead times suggest that the predicted track is towards the right side of the observed track during both years. The cyclone Intensity forecast for the maximum sustained wind speed (MaxWS) and central mean sea level pressure (MSLP) are verified in terms of mean error (ME) and root mean square error (RMSE). The errors are lead time independent. However, most of the time model under-predicted the cyclone intensity during both years. Finally, there is a significant variance in track and intensity errors from the cyclone to cyclone and Bay of Bengal basin to the Arabian Sea basin.

Improved runoff forecasting performance through error predictions using a deep-learning approach

Accurate runoff prediction is critical for various fields of hydrology, agriculture, and environmental studies. Numerous hydrologic models have been developed and demonstrate good performances in runoff simulation. However, errors are inherent in forecasted runoff predictions, which can cause uncertainty in real-time flood warning systems. In order to improve the predictive performance of hydrologic modeling, this study used a deep learning approach as a post-processor to correct for errors associated with hydrologic data. The proposed model uses the long short-term memory model with sequence-to-sequence structure as a post-processor to improve runoff forecasting. Specifically, the deep learning approach was used to estimate errors in forecasted hourly runoff provided from National Water Model in Russian River basin, California, United States. Error prediction in hourly runoff with lead times between 1 and 18 h were developed using observed precipitation and errors from upstream stream gages to improve the predictive performance of National Water Model. The predictive performance of the model was evaluated using numerous statistical metrics, and results show that the long short-term memory model with sequence-to-sequence post-processor improved runoff predictions compared to standalone results from the National Water Model. Statistical values of percent bias decreased from a range of −60%–80% to −15%–10% when the post-processor model was used, and similarly root mean square errors of runoff prediction decreased from 120 cms to 20 cms. Thus, this study demonstrates the power of deep learning model to improve hydrologic modeling results, especially those with short forecasting lead times.

Intraseasonal scale ensemble forecasts of precipitation and evapotranspiration for the Madeira River basin using different physical parameterizations

Eta Regional Model of CPTEC-INPE is used to obtain intraseasonal (30-day) 8-member ensemble forecasts over the Madeira River basin for the period 2002–2012. The initial and boundary conditions are taken from Atmospheric General Circulation Global Model in six members and from Global Coupled Ocean-Atmosphere Model in two members. The intraseasonal forecasts produced by dynamic downscaling with Eta Regional model ensemble have satisfactory skill. The skill of the ensemble mean is better than the individual members up to 15-days lead time forecasts. The ensemble mean reproduces the seasonal cycle and spatial distribution of the hydrological variables. Members with the relaxation technique of Betts-Miller-Janjic produced better results. The forecasts by the members that used Kain-Fritsch scheme presented larger deviations from observations. Substantial improvements in skill are obtained through bias correction. This is the first work to attempt dynamic downscaling over the Madeira Basin in the intraseasonal time scale for a period of 10 years. The ensemble downscaled products have potential to be fed into surface hydrological models for forecasting droughts and floods and related hydrological variables over the basin.

Improvements of the Sub-Seasonal Precipitation Predicted by the BCC_S2S Forecast System Over Eastern China in Summer Using the Singular Value Decomposition Bias Correction Method

A bias correction scheme has been developed based on the singular value decomposition (SVD) analysis in this study, and it is further validated and used to improve the skill of sub-seasonal precipitation over Eastern China in summer predicted by the Beijing Climate Center sub-seasonal to seasonal forecast system (BCC_S2SFS). Results show that the BCC_S2SFS prediction skill of summer precipitation over Eastern China at the sub-seasonal scale is up to 1 days in advance and exhibits clear regional and inter-annual differences. Further adopting the bias correction scheme can significantly improve the skill of BCC_S2SFS in predicting the sub-seasonal precipitation over Eastern China in summer with different lead time especially longer than 10 days. Compared to the original prediction of BCC_S2SFS, the temporal (spatial) correlation coefficient between the bias corrected predictions and observations over Eastern China can be increased by 0.15, 0.55, and 0.56 (0.14, 0.17, and 0.19) during the forecast lead time of 0–10, 11–20, and 21–30 days, respectively. The bias correction scheme developed in this study shows large potential application prospects in the operational forecast.

Research on particle swarm optimization in LSTM neural networks for rainfall-runoff simulation

Flood forecasting is an essential non-engineering measure for flood prevention and disaster reduction. Many models have been developed to study the complex and highly random rainfall-runoff process. In recent years, artificial intelligence methods, such as the artificial neural network (ANN), have attempted to construct rainfall-runoff models. The more advanced deep learning methods of long short-term memory (LSTM) network have been proved to better predict hydrological time series. However, the selection of LSTM hyperparameters in the past mostly relied on the experience of the staff, which often led to failure to achieve the best performance. The aim of this study is to develop a method to improve flood forecast accuracy and lead time. A deep learning neural network model based on LSTM networks and particle swarm optimization (PSO) is proposed in this paper. The PSO algorithm was used to optimize the LSTM hyperparameter to improve the ability to learn data sequence features. The model focuses on the Jingle Watershed in the Fenhe River and the Lushi Watershed in the Luohe River and was used to predict flood processes using rainfall and runoff observation data from stations in the watersheds. We evaluated the performance of the model with the Nash Sutcliffe efficiency coefficient, root mean square error, and bias. The results show that the PSO-LSTM model outperforms the M-EIES, ANN, PSO-ANN, and LSTM at all stations in the watersheds. The PSO-LSTM model improves the flood forecasting accuracy at different lead times, especially for those exceeding 6 h, and has higher prediction accuracy and stability. The PSO-LSTM model could be used to improve accuracy in short-term flood forecast applications.

Evaluating the assimilation of S5P/TROPOMI near real-time SO<sub>2</sub> columns and layer height data into the CAMS integrated forecasting system (CY47R1), based on a case study of the 2019 Raikoke eruption

Abstract. The Copernicus Atmosphere Monitoring Service (CAMS), operated by the European Centre for Medium-Range Weather Forecasts on behalf of the European Commission, provides daily analyses and 5 d forecasts of atmospheric composition, including forecasts of volcanic sulfur dioxide (SO2) in near real time. CAMS currently assimilates total column SO2 products from the GOME-2 instruments on MetOp-B and MetOp-C and the TROPOMI instrument on Sentinel-5P, which give information about the location and strength of volcanic plumes. However, the operational TROPOMI and GOME-2 data do not provide any information about the height of the volcanic plumes, and therefore some prior assumptions need to be made in the CAMS data assimilation system about where to place the resulting SO2 increments in the vertical. In the current operational CAMS configuration, the SO2 increments are placed in the mid-troposphere, around 550 hPa or 5 km. While this gives good results for the majority of volcanic emissions, it will clearly be wrong for eruptions that inject SO2 at very different altitudes, in particular exceptional events where part of the SO2 plume reaches the stratosphere. A new algorithm, developed by the German Aerospace Centre (DLR) for GOME-2 and TROPOMI, optimized in the frame of the ESA-funded Sentinel-5P Innovation–SO2 Layer Height Project, and known as the Full-Physics Inverse Learning Machine (FP_ILM) algorithm, retrieves SO2 layer height from TROPOMI in near real time (NRT) in addition to the SO2 column. CAMS is testing the assimilation of these products, making use of the NRT layer height information to place the SO2 increments at a retrieved altitude. Assimilation tests with the TROPOMI SO2 layer height data for the Raikoke eruption in June 2019 show that the resulting CAMS SO2 plume heights agree better with IASI plume height data than operational CAMS runs without the TROPOMI SO2 layer height information and show that making use of the additional layer height information leads to improved SO2 forecasts. Including the layer height information leads to higher modelled total column SO2 values in better agreement with the satellite observations. However, the plume area and SO2 burden are generally also overestimated in the CAMS analysis when layer height data are used. The main reason for this overestimation is the coarse horizontal resolution used in the minimizations. By assimilating the SO2 layer height data, the CAMS system can predict the overall location of the Raikoke SO2 plume up to 5 d in advance for about 20 d after the initial eruption, which is better than with the operational CAMS configuration (without prior knowledge of the plume height) where the forecast skill is much more reduced for longer forecast lead times.

Systematic calculation of finite-time mixed singular vectors and characterization of error growth for persistent coherent atmospheric disturbances over Eurasia.

Singular vectors (SVs) have long been employed in the initialization of ensemble numerical weather prediction (NWP) in order to capture the structural organization and growth rates of those perturbations or "errors" associated with initial condition errors and instability processes of the large scale flow. Due to their (super) exponential growth rates and spatial scales, initial SVs are typically combined empirically with evolved SVs in order to generate forecast perturbations whose structures and growth rates are tuned for specified lead-times. Here, we present a systematic approach to generating finite time or "mixed" SVs (MSVs) based on a method for the calculation of covariant Lyapunov vectors and appropriate choices of the matrix cocycle. We first derive a data-driven reduced-order model to characterize persistent geopotential height anomalies over Europe and Western Asia (Eurasia) over the period 1979-present from the National Centers for Environmental Prediction v1 reanalysis. We then characterize and compare the MSVs and SVs of each persistent state over Eurasia for particular lead-times from a day to over a week. Finally, we compare the spatiotemporal properties of SVs and MSVs in an examination of the dynamics of the 2010 Russian heatwave. We show that MSVs provide a systematic approach to generate initial forecast perturbations projected onto relevant expanding directions in phase space for typical NWP forecast lead-times.

Wind Speed Forecasts of a Mesoscale Ensemble for Large-Scale Wind Farms in Northern China: Downscaling Effect of Global Model Forecasts

To facilitate wind power integration for the electric power grid operated by the Inner Mongolia Electric Power Corporation—a major electric power grid in China—a high-resolution (of 2.7 km grid intervals) mesoscale ensemble prediction system was developed that forecasts winds for 130 wind farms in the Inner Mongolia Autonomous Region. The ensemble system contains 39 forecasting members that are divided into 3 groups; each group is composed of the NCAR (National Center for Atmospheric Research) real-time four-dimensional data assimilation and forecasting model (RTFDDA) with 13 physical perturbation members, but driven by the forecasts of the GFS (Global Forecast System), GEM (Global Environmental Multiscale Model), and GEOS (Goddard Earth Observing System), respectively. The hub-height wind predictions of these three sub-ensemble groups at selected wind turbines across the region were verified against the hub-height wind measurements. The forecast performance and variations with lead time, wind regimes, and diurnal and regional changes were analyzed. The results show that the GFS group outperformed the other two groups with respect to correlation coefficient and mean absolute error. The GFS group had the most accurate forecasts in ~59% of sites, while the GEOS and GEM groups only performed the best on 34% and 2% of occasions, respectively. The wind forecasts were most accurate for wind speeds ranging from 3 to 12 m/s, but with an overestimation for low speeds and an underestimation for high speeds. The GEOS-driven members obtained the least bias error among the three groups. All members performed rather accurately in daytime, but evidently overestimated the winds during nighttime. The GFS group possessed the fewest diurnal errors, and the bias of the GEM group grew significantly during nighttime. The wind speed forecast errors of all three ensemble members increased with the forecast lead time, with the average absolute error increasing by ~0.3 m/s per day during the first 72 h of forecasts.

Improving early-season wheat yield forecasts driven by probabilistic seasonal climate forecasts

Seasonal climate forecasts (SCF) are evolving rapidly alongside improvements in climate modelling and downscaling research, and have great potential for weather-sensitive sectors, especially agriculture, by reducing weather-related risks and increasing productivity. Skilful yield forecasts at the beginning of, or before, a cropping season can provide farmers and other stakeholders in agribusiness with the necessary information for early planning and actions. Only a few yield forecast studies have a forecast lead time of four months or longer due to the problem complexity. To enable SCFs from Global Climate Models (GCMs) to be used for early-season yield forecasts, this paper uses a statistical downscaling technique, Extended Copula Post-Processing (ECPP) and the Schaake shuffle, to downscale four climate variables to generate weather-like daily data that are suitable for agricultural applications. Climate forecasts drive a process-based crop model APSIM (Agricultural Production Systems sIMulator) to simulate crop forecasts on 50 stations, well-distributed across the Australian grain zone. To focus on yield forecast skills attributable to SCF, we propose best practice management rules to predict water-limited winter wheat yield. Yield forecasts from ECPP have a significant improvement over quantile mapping downscaling and raw SCF from the Australian recent seasonal forecast model ACCESS-S1 in terms of bias, accuracy, reliability, and overall forecast skill. In addition, even at the beginning of a cropping season with a forecast lead time of four or more months, yield forecasts driven by ECPP illustrate higher skill than climatology, a benchmark for yield forecast. Early-season yield forecasts driven by SCFs provides a promising alternative to regression/machine-learning-based forecasts. Performance sensitivity and issues, and gaps on using skilful SCFs to help growers with their farming decision-making are discussed.

Application of HEC-HMS for flood forecasting in hazara catchment Pakistan, south Asia

Floods have become more severe and frequent as a result of climate change around the world, posing a hazard to public safety and economic development. This study investigates the use of distributed hydrological models in flash flood risk management in a small watershed in Hazara, Pakistan, with the goal of improving Pakistan's early warning lead time. First, the HEC-HMS model was built using geographic data and the river network's structure, then calibrated and verified using eight high rainfall events from 2013. demonstrating that the HEC-HMS model could simulate floods in the research area Second, given that rainfall and flood events have happened, this paper proposes an analysis approach for a flood forecasting and warning system, as well as criteria for sending urban-stream flash flood alerts based on rainfall, in order to provide sufficient lead time. The DEMs (digital elevation models) of the research regions were processed using HEC-Geo HMS, an ArcView GIS tool for catchment delineation, terrain pre-processing, and basin processing. The model was calibrated and verified using previously observed data. The proposed flood prediction and risk reduction methodology is nonstructural. The Hydrologic Modeling System (HEC-HMS), which provides a sufficient lead time forecast and computes the runoff/stage threshold conditions, is at the heart of the flood warning application. For flood risk assessment, data from the Pakistan Meteorological Department (PMD) is entered into a hydro-meteorological database and then into the HEC-HMS. A server-client application was utilised to visualise the real-time flood scenario and send out an early warning message. The outcomes of this study will be used to develop flood validation measures in the Hazara stream watershed to deal with potential flash floods.

Choosing between post-processing precipitation forecasts or chaining several uncertainty quantification tools in hydrological forecasting systems

Abstract. This study aims to decipher the interactions of a precipitation post-processor and several other tools for uncertainty quantification implemented in a hydrometeorological forecasting chain. We make use of four hydrometeorological forecasting systems that differ by how uncertainties are estimated and propagated. They consider the following sources of uncertainty: system A, forcing, system B, forcing and initial conditions, system C, forcing and model structure, and system D, forcing, initial conditions, and model structure. For each system's configuration, we investigate the reliability and accuracy of post-processed precipitation forecasts in order to evaluate their ability to improve streamflow forecasts for up to 7 d of forecast horizon. The evaluation is carried out across 30 catchments in the province of Quebec (Canada) and over the 2011–2016 period. Results are compared using a multicriteria approach, and the analysis is performed as a function of lead time and catchment size. The results indicate that the precipitation post-processor resulted in large improvements in the quality of forecasts with regard to the raw precipitation forecasts. This was especially the case when evaluating relative bias and reliability. However, its effectiveness in terms of improving the quality of hydrological forecasts varied according to the configuration of the forecasting system, the forecast attribute, the forecast lead time, and the catchment size. The combination of the precipitation post-processor and the quantification of uncertainty from initial conditions showed the best results. When all sources of uncertainty were quantified, the contribution of the precipitation post-processor to provide better streamflow forecasts was not remarkable, and in some cases, it even deteriorated the overall performance of the hydrometeorological forecasting system. Our study provides an in-depth investigation of how improvements brought by a precipitation post-processor to the quality of the inputs to a hydrological forecasting model can be cancelled along the forecasting chain, depending on how the hydrometeorological forecasting system is configured and on how the other sources of hydrological forecasting uncertainty (initial conditions and model structure) are considered and accounted for. This has implications for the choices users might make when designing new or enhancing existing hydrometeorological ensemble forecasting systems.

Performance of 2020 Real-Time Atlantic Hurricane Forecasts from High-Resolution Global-Nested Hurricane Models: HAFS-globalnest and GFDL T-SHiELD

Abstract The global-nested Hurricane Analysis and Forecast System (HAFS-globalnest) is one piece of NOAA’s Unified Forecast System (UFS) application for hurricanes. In this study, results are analyzed from 2020 real-time forecasts by HAFS-globalnest and a similar global-nested model, the Tropical Atlantic version of GFDL’s System for High‐resolution prediction on Earth‐to‐Local Domains (T-SHiELD). HAFS-globalnest produced the highest track forecast skill compared to several operational and experimental models, while T-SHiELD showed promising track skills as well. The intensity forecasts from HAFS-globalnest generally had a positive bias at longer lead times primarily due to the lack of ocean coupling, while T-SHiELD had a much smaller intensity bias particularly at longer forecast lead times. With the introduction of a modified planetary boundary layer scheme and an increased number of vertical levels, particularly in the boundary layer, HAFS forecasts of storm size had a smaller positive bias than occurred in the 2019 version of HAFS-globalnest. Despite track forecasts that were comparable to the operational GFS and HWRF, both HAFS-globalnest and T-SHiELD suffered from a persistent right-of-track bias in several cases at the 4–5-day forecast lead times. The reasons for this bias were related to the strength of the subtropical ridge over the western North Atlantic and are continuing to be investigated and diagnosed. A few key case studies from this very active hurricane season, including Hurricanes Laura and Delta, were examined.

Modeling digital camera monitoring count data with intermittent zeros for short-term prediction

Digital camera monitoring has revolutionised survey designs in many fields, as an important source of information. The extended sampling coverage offered by this monitoring scheme makes it preferable compared to other traditional methods of survey. However, data obtained from digital camera monitoring are often highly variable, and characterized by sparse periods of zero counts, interspersed with missing observations due to outages. In practice, missing data of relatively shorter duration are mostly observed and are often imputed using interpolation techniques, ignoring long-term trends leading to inherent estimation biases. In this study, we investigated time series forecasting methods that adequately handle intermittency and produced plausible estimates for imputation and forecasting purposes. The study utilised a yearlong digital camera monitoring data set of hourly counts of powerboat launches at three boat ramps in Western Australia. Several time series forecasting methods were evaluated and the accuracies of their point estimates of forecasts for various lead times in hours of up to one week were assessed using cross-validation techniques. Intermittent demand forecasting techniques, including Croston's method and Syntetos-Boylan Approximation (SBA) models, and count data forecasting methods including autoregressive conditional Poisson (ACP) models, integer-valued moving average (INMA) models, and integer-valued autoregressive (INAR) models were evaluated. ACP and INAR models performed better than intermittent demand forecasting techniques for short forecast horizons and provided some evidence of their sufficiency in predicting the dynamics in recreational boating activities. This result established that, in as much as intermittency may be a key feature for a given dataset, it should not override the systemic characteristics of data in the application of forecasting techniques. Our results provide plausible estimates for short-term missing data and forecasts for monitoring events, with applications in supporting proper tracking of usage of facilities, guiding resource allocations and providing insightful perspectives for management decisions.

Lead time prediction for sheeter machine production in a paper conversion industry

Lead time is a critical performance measure in any manufacturing setting Key Performance Indicator (KPI). The same is true in the paper conversion industry, which has a significant degree of product variability. Due to the great variety of their products, all industries must be able to foresee and plan ahead in order to meet client demand. With contemporary research concentrating on machine learning and simulation techniques, businesses must implement a manufacturing execution system (MES) to track data. However, without such a framework, applying machine learning and simulation approaches becomes difficult. This study introduces a novel method for forecasting lead time (special to sheeter machines used in the paper conversion sector) by combining the time required to process the reel (sheeting time) with the human (setup) elements. The method used to calculate the sheeting time takes product parameters into account, allowing for product-specific lead time forecast. As a result, a very successful 'product-specific' lead time prediction approach for small scale enterprises has been developed that enables production planning without relying on current and data-intensive prediction methods such as machine learning and simulation.

Forecasting Nodal Price Difference Between Day-Ahead and Real-Time Electricity Markets Using Long-Short Term Memory and Sequence-to-Sequence Networks

Price forecasting is at the center of decision making in electricity markets. Much research has been done in forecasting energy prices for a single market while little research has been reported on forecasting price difference between markets, which presents higher volatility and yet plays a critical role in applications such as virtual trading. To this end, this paper takes the first attempt at it and employs novel deep learning architecture with Bidirectional Long-Short Term Memory (LSTM) units and Sequence-to-Sequence (Seq2Seq) architecture to forecast nodal price difference between day-ahead and real-time markets. In addition to value prediction, these deep learning architectures are also used to develop classification models to predict the price difference bands/ranges. The proposed methods are tested using historical PJM market data, and evaluated using Root Mean Squared Error (RMSE) and other customized performance metrics. Case studies show that both deep learning methods outperform common methods including ARIMA, XGBoost and Support Vector Regression (SVR) methods. More importantly, the deep learning methods can capture the magnitude and timing of price difference spikes. Numerical results show the Seq2Seq model performs particularly well and demonstrates generalization capability to extended forecasting lead time.

Dynamic Thermal Rating Forecasting Methods: A Systematic Survey

Dynamic Thermal Rating (DTR) allows optimum electric power line rating use. It is an intelligent grid technology predicting changes in line rating due to changing physical and environmental conditions. This study performed a meta-analysis of DTR forecasting methods by classifying the methods, implementing them, and comparing their outputs for a 24hr forecast lead time. It implemented deep learning methods of Recurrent Neural Network (RNN), Ensemble Means forecasting and Convolution Neural Network (CNN). RNN uses the initial outcome of a specific neural network layer as feedback to the network to predict the layer’s outcome. Ensemble Means forecasting is a Monte-Carlo simulation process producing random, equally viable forecasting solutions. On the other hand, CNN uses unsupervised learning to predict features with minimal errors. This survey systematically implements Quantile Regression (QR), RNN, CNN and Ensemble means forecasting. Point error metrics and probabilistic error metrics of sharpness, skill, and bias were used in the methods’ evaluation. All methods tested prove to be efficient, but 50th percentile QR appears more conservative, secure and less error-prone. It achieved between 35% - 45% line capacity utilization over the Static Thermal Rating (STR). On average, judging by the error metrics of all methods, 50th percentile quantile regression proves highly reliable and provides a better conviction in our choice of DTR forecasting.

Ensemble-Based Assimilation of Satellite All-Sky Microwave Radiances Improves Intensity and Rainfall Predictions for Hurricane Harvey (2017).

Ensemble‐based data assimilation of radar observations across inner‐core regions of tropical cyclones (TCs) in tandem with satellite all‐sky infrared (IR) radiances across the TC domain improves TC track and intensity forecasts. This study further investigates potential enhancements in TC track, intensity, and rainfall forecasts via assimilation of all‐sky microwave (MW) radiances using Hurricane Harvey (2017) as an example. Assimilating Global Precipitation Measurement constellation all‐sky MW radiances in addition to GOES‐16 all‐sky IR radiances reduces the forecast errors in the TC track, rapid intensification (RI), and peak intensity compared to assimilating all‐sky IR radiances alone, including a 24‐hr increase in forecast lead‐time for RI. Assimilating all‐sky MW radiances also improves Harvey's hydrometeor fields, which leads to improved forecasts of rainfall after Harvey's landfall. This study indicates that avenues exist for producing more accurate forecasts for TCs using available yet underutilized data, leading to better warnings of and preparedness for TC‐associated hazards in the future.

Russian Rivers Streamflow Forecasting Using Hydrograph Extrapolation Method

This paper presents a method of hydrograph extrapolation, intended for simple and efficient streamflow forecasting with up to 10 days lead time. The forecast of discharges or water levels is expressed by a linear formula depending on their values on the date of the forecast release and the five previous days. Such forecast techniques were developed for more than 2700 stream gauging stations across Russia. Forecast verification has shown that this method can be successfully applied to large rivers with a smooth shape of hydrographs, while for small mountain catchments, the accuracy of the method tends to be lower. The method has been implemented into real-time continuous operations in the Hydrometcentre of Russia. In the territory of Russia, 18 regions have been identified with a single dependency of the maximum lead time of good forecasts on the area and average slope of the catchment surface for different catchments of each region; the possibilities of forecasting river streamflow by the method of hydrograph extrapolation are approximately estimated. The proposed method can be considered as a first approximation while solving the problem of forecasting river flow in conditions of a lack of meteorological information or when it is necessary to quickly develop a forecasting system for a large number of catchments.