Ranked Probability Skill Score Research Articles

Probabilistic Neural Networks for Ensemble Postprocessing

Abstract Accurate temperature forecasts are critical for various industries and sectors. We propose a probabilistic neural network (PNN), an extension of the distributional regression network (DRN), for 2-m temperature forecasts, consisting of three variants with different inputs and target variables. The first variant, standardized anomaly probabilistic neural network (SAPNN), employs a two-step approach involving standardized anomalies and global PNN modeling to effectively capture underlying features and anomalies. The second variant, PNN with geographical predictors (PNNGE), incorporates raw and static geographical predictors to enhance predictive performance. The third variant, PNN with station one-hot encoding (PNNEN), utilizes raw with station one-hot encoding predictors to represent geographical information effectively. We compare three PNN variants with two benchmarks: 1) standardized anomaly model output statistics (SAMOS) and 2) three DRN variants identical to those applied to PNN. These evaluations utilize ECMWF data from 2019 to 2020 at 6-h intervals up to 72 h over Hebei, China. Results show that SAPNN and PNNGE are better than SAMOS, while PNNEN notably exhibits a significant 14% improvement in the continuous ranked probability skill score (CRPSS). Moreover, the PNN variants exhibit comparable or superior performance to DRN regarding forecast accuracy, CRPSS, and reliability, showcasing a better-calibrated spread–error relationship. This study highlights the value of the proposed PNN variants with a distribution output in capturing nonlinear relationships within different sources of predictors and improving temperature forecast skills. Significance Statement This study aims to improve the accuracy, skills, and reliability of 2-m temperature forecasts, which are crucial in agriculture and energy management. To achieve this, we extend a popular artificial intelligence framework and explore four data sources with two schemes to systematically compare the predictive performances in making temperature forecasts. The findings of this research are vital as they offer novel ways to improve forecast skills. Imagine having a weather app that is significantly more accurate, enabling you to plan your day better. This study is about discovering innovative approaches to enhance forecast skills and reliability, which could benefit various aspects of our daily lives. One of the new methods even exhibits a 14% improvement in forecast skills.

The First Operational Version of Taiwan Central Weather Bureau’s One-Tier Global Atmosphere–Ocean Coupled Forecast System for Seasonal Prediction

Abstract The first version of the Taiwan Central Weather Bureau one-tier (TCWB1T) fully coupled global atmospheric and oceanic modeling forecast system had been developed and implemented as a routine operation for seasonal prediction at Central Weather Bureau (CWB) in 2017, with a minor revision in 2020. Based on NCEP CFSv1, the global atmospheric model in NCEP CFSv1 was replaced by CWB’s atmospheric global spectral model (GSM) and coupled with the GFDL MOM3. Several parameters have been tested and tuned in the CWB atmospheric GSM, achieving an optimal configuration with better sea surface temperature (SST) predictions for integration more than one year. Using NCEP CFSR as the initial condition, TCWB1T conducted hindcasts from 1982 to 2011 and forecasts from 2012 to 2019 to analyze its performance. The results of these hindcasts and forecasts show that the TCWB1T can make useful predictions as verified against the observations of OISST, ERSST, CFSR, and GPCP based on the methods of EOF, RMSE, anomaly correlation, ranked probability skill score (RPSS), reliability diagram (RD), and relative operating characteristics (ROCs). TCWB1T also has the same level of skill scores as NCEP CFSv2 and/or the ECMWF fifth-generation seasonal forecast system (SEAS5), based on EOF, anomaly pattern correlation, climatological bias, RMSE, temporal correlation, and anomaly correlation percentage of forecast skill. TCWB1T shows forecast skill that is better in winter than in summer. Overall, it indicates that TCWB1T can be used for seasonal ENSO predictions.

A comparison of two statistical postprocessing methods for heavy‐precipitation forecasts over India during the summer monsoon

AbstractAccurate ensemble forecasts of heavy precipitation in India are vital for many applications and essential for early warning of damaging flood events, especially during the monsoon season. In this study we investigate to what extent Quantile Mapping (QM) and Ensemble Model Output Statistics (EMOS) statistical postprocessing reduce errors in precipitation ensemble forecasts over India, in particular for heavy precipitation. Both methods are applied to day‐1 forecasts at 12‐km resolution from the 23‐member National Centre for Medium Range Weather Forecasting (NCMRWF) global ensemble prediction system (NEPS‐G). By construction, QM leads to distributions close to the observed ones, while EMOS optimizes the ensemble spread, and it is not a priori clear which is better suited for practical applications. The methods are therefore compared with respect to several key aspects of the forecasts: local distributions, ensemble spread, and skill for forecasting precipitation amounts and the exceedance of heavy precipitation thresholds. The evaluation includes rank histograms, Continuous Ranked Probability Skill Scores (CRPSS), Brier Skill Scores (BSS), reliability diagrams, and receiver operating characteristic. EMOS performs best not only with respect to correcting under‐ or overdispersive ensembles, but also in terms of forecast skill for precipitation amounts and heavy precipitation events, with positive CRPSS and BSS in most regions (both up to about 0.4 in some areas), while QM in many regions performs worse than the raw forecast. QM performs best with respect to the overall local precipitation distributions. Which aspects of the forecasts are most relevant depends to some extent on how the forecasts are used. If the main criteria are the correction of under‐ or overdispersion, forecast reliability, match between the forecasted distribution for individual days and observations (CRPSS), and the skill in forecasting heavy‐precipitation events (BSS), then EMOS is the better choice for postprocessing NEPS‐G forecasts for short lead times.

Adapting subseasonal-to-seasonal (S2S) precipitation forecast at watersheds for hydrologic ensemble streamflow forecasting with a machine learning-based post-processing approach

Accurate and reliable precipitation predictions made by dynamical forecast models could provide crucial information for human socioeconomic activities by enabling hydrologic forecasts at the Subseasonal-to-Seasonal (S2S) timescale. To utilize available S2S precipitation predictions for hydrologic forecasts, post-processing techniques have been applied to adapt the raw S2S precipitation to local watersheds. However, conventional statistical-based post-processing techniques are more focused on correcting the forecast bias, but rather limited in improving the predictive skill of available S2S precipitation forecasts. In this study, we combine the Random Forest classifiers (RF) with the Bias Correction and Spatial Disaggregation (BCSD) to adapt the 10-member ensemble precipitation forecast from the NASA Goddard Earth Observing System model version 5 (GEOS5) at 4 watersheds located in the NCEI South climate region of the United States. The adapted S2S precipitation is further applied for streamflow forecast by forcing a classical lumped hydrologic model. The performance of S2S precipitation as well as the corresponding streamflow predictions are benchmarked with the randomly resampled precipitation and the corresponding Ensemble Streamflow Prediction (ESP) framework-generated streamflow predictions. Evaluation statistics of Kling-Gupta Efficiency (KGE), Continuous Ranked Probability Skill Score (CRPSS), Reliability, Resolution, and Sharpness are employed to evaluate the predictive skill of precipitation and streamflow both deterministically and probabilistically. Our results indicate that dynamical S2S precipitation after forecast adaptation leads to consistently higher deterministic skill over ESP at all forecast lead times and across study watersheds. However, at longer forecast lead times beyond 10–15 days, S2S precipitation with a limited ensemble size does not present higher probabilistic skill than ESP. Our results shows that the joint application of RF and BCSD improves the predictive skill of the raw S2S precipitation at study watersheds in contrast to BCSD. Further, the added predictive skill of S2S precipitation brought by RF propagates into streamflow predictions, predominantly at longer forecast lead times exceeding 10 days. Overall, our results highlight the potential success of future work to apply other data-driven approaches to adapt the raw precipitation to local watersheds for more accurate and reliable streamflow forecasts at the S2S timescale.

Operational low-flow forecasting using LSTMs

This study focuses on exploring the potential of using Long Short-Term Memory networks (LSTMs) for low-flow forecasting for the Rhine River at Lobith on a daily scale with lead times up to 46 days ahead. A novel LSTM-based model architecture is designed to leverage both historical observation and forecasted meteorological data to carry out multi-step discharge time series forecasting. The feature and target selection for this deep learning (DL) model involves evaluating the use of different spatial resolutions for meteorological forcing (basin-averaged or subbasin-averaged), the impact of incorporating past discharge observations, and the use of different target variables (discharge Q or time-differenced discharge dQ). Then, the model is trained using the ERA5 dataset as meteorological forcing, and employed for operational forecast with ECMWF seasonal forecast (SEAS5) data. The forecast results are compared to a benchmark process-based model, wflow_sbm. This study also explores the flexibility of the DL model by fine-tuning the pretrained model with limited SEAS5 dataset. Key findings from feature and target selection include: (1) opting for subbasin-averaged meteorological variables significantly improves model performance compared to a basin-averaged approach. (2) Utilizing dQ as the target variable greatly boosts short-term forecast accuracy compared to using Q, with a mean absolute error (MAE) of 25 m3 s−1 and mean absolute percentage error (MAPE) of 0.02 for the first lead time, ensuring reliability and accuracy at the onset of the forecast horizon. (3) While incorporating historical discharge improves the forecasting of Q, its impact on predicting dQ is less pronounced for short lead times. In the operational forecast with SEAS5, compared to the wflow_sbm model, the DL model exhibits skill in forecasting low flows as evidenced by Continuous Ranked Probability Skill Score (CRPSS) median values of all lead times above zero, and better accuracy in forecasting drought events within short lead times. The wflow_sbm model shows higher accuracy for longer lead times. In the exploration of fine-tuning approach, the fine-tuned model generates marginal short-term enhancements in forecasting low-flow events over a non-fine-tuned model. Overall, this study contributes to advancing the field of low-flow forecasting using deep learning approach.

Can Machine Learning Models Be a Suitable Tool for Predicting Central European Cold Winter Weather on Subseasonal to Seasonal Time Scales?

Abstract Skillful weather prediction on subseasonal to seasonal time scales is crucial for many socioeconomic ventures. But forecasting, especially extremes, on these time scales is very challenging because the information from initial conditions is gradually lost. Therefore, data-driven methods are discussed as an alternative to numerical weather prediction models. Here, quantile regression forests (QRFs) and random forest classifiers (RFCs) are used for probabilistic forecasting of central European mean wintertime 2-m temperatures and cold wave days at lead times of 14, 21, and 28 days. ERA5 reanalysis meteorological predictors are used as input data for the machine learning models. For the winters of 2000/01–2019/20, the predictions are compared with a climatological ensemble obtained from E-OBS observational data. The evaluation is performed as full distribution predictions for continuous values using the continuous ranked probability skill score and as binary categorical forecasts using the Brier skill score. We find skill at lead times up to 28 days in the 20-winter mean and for individual winters. Case studies show that all used machine learning models are able to learn patterns in the data beyond climatology. A more detailed analysis using Shapley additive explanations suggests that both random forest (RF)-based models are able to learn physically known relationships in the data. This underlines that RF-based data-driven models can be a suitable tool for forecasting central European mean wintertime 2-m temperatures and the occurrence of cold wave days. Significance Statement Because of the chaotic nature of weather, it is very complicated to make predictions with traditional numerical methods 2–4 weeks in advance. Therefore, we use alternative, interpretable methods that “learn” to find statistically relevant patterns in meteorological data that can be used for forecasting central European mean surface wintertime temperatures and cold wave days. These methods are part of the so-called machine learning methods that do not rely on the traditional numerical equations anymore. We test our methods for 20 winters between 2000/01 and 2019/20 against a static weather prediction consisting of the past 30 winters. For single winters and in a mean over the 20 predicted winters, we find improved predictions up to 4 weeks in advance.

Improving Global Subseasonal to Seasonal Precipitation Forecasts Using a Support Vector Machine‐Based Method

AbstractSubseasonal to seasonal (S2s) precipitation forecasts provide great potential for hydrological forecasting at an extended range. The study proposed a support vector machine (SVM) regression‐based method to improve S2s precipitation forecasts from the European Center for Medium‐Range Weather Forecasts (ECMWF) across the globe (60°N to 60°S). Results suggested that the SVM‐based method significantly improved ECMWF daily precipitation forecasts in representing the spatiotemporal variation of precipitation with higher consistency and reduced errors when compared against observations. Furthermore, the SVM‐based method enhanced the probabilistic skill of ECMWF forecasts, providing improved ranked probability skill score (RPSS) for real‐time forecasts in 2020 (e.g., RPSSECMWF = −0.03 and RPSSrg3 = 0.08 for lead week 1). The most substantial improvement from the SVM‐based method is witnessed in regions with complex terrains where ECMWF yielded the worst skill, such as the Andes mountain range, Congo River Basin, and the Tibet Plateau. However, the SVM‐based post‐processing method did not alter the characteristics of precipitation forecasts regarding climate zone and lead time. Both ECMWF and post‐processed forecasts showed higher skill in the temperate and continental climate zones when the lead time is shorter than 2 weeks. In comparison, low‐latitude regions exhibited higher predictability when the lead time is longer than 5 weeks, which is attributed to the slow variation of boundary conditions such as the El Niño‐Southern Oscillation (ENSO).

Less is more? A novel method for identifying and evaluating non-informative tracers in sediment source mixing models

PurposeAccelerated soil erosion poses a global hazard to soil health. Understanding soil and sediment behaviour through sediment fingerprinting enables the monitoring and identification of areas with high sediment delivery. Land-use specific sediment source apportionment is increasingly determined using the Bayesian mixing model MixSIAR with compound-specific stable isotopes (CSSI). Here, we investigate CSSIs of fatty acid (FA) tracer selection with a novel method to identify and investigate the effect of non-informative tracers on model performance.MethodsTo evaluate CSSI tracer selection, mathematical mixtures were generated using source soils (n = 28) from the Rhine catchment upstream of Basel (Switzerland). Using the continuous ranked probability (CRP) skill score, MixSIAR’s performance was evaluated for 11 combinations of FAs and 15 combinations of FAs with δ15N as a mixing line offset tracer. A novel scaling and discrimination analysis (SDA) was also developed to identify tracers with non-unique mixing spaces.ResultsFA only tracer combinations overestimated pasture contributions while underestimating arable contributions. When compared to models with only FA tracers, utilizing δ15N to offset the mixing line resulted in a 28% improvement in the CRP skill score. δ15N + δ13C FA26 was the optimal tracer set resulting in a 62% model improvement relative to δ15N + all δ13C FAs. The novel SDA method demonstrated how δ13C FA tracers have a non-unique mixing space and thus behave as non-informative tracers. Importantly, the inclusion of non-informative tracers decreased model performance.ConclusionsThese results indicate that MixSIAR did not handle non-informative CSSI tracers effectively. Accordingly, it may be advantageous to remove non-informative tracers, and where feasible, all combinations and permutations of tracers should be assessed to optimize tracer selection. Application of these tracer selection steps can help improve and advance the performance of sediment fingerprinting models and ultimately aid in improving erosion mitigation and management strategies.

Catchment‐scale skill assessment of seasonal precipitation forecasts across South Korea

AbstractClimate change is expected to make droughts more frequent and severe all over the world. South Korea is no exception and is already suffering from extreme droughts such as one that prolonged from 2013 to 2015 and caused nation‐wide damage. To mitigate drought damages, better management of existing water infrastructure is essential. A promising opportunity to improve operational decisions is to make use of seasonal weather forecasts that are provided by general circulation models and can be downscaled to the catchment scale. This study hence assesses the skill of seasonal forecasts over 20 catchments in South Korea, where the largest reservoirs are located. Datasets from four weather forecasting centres (ECMWF, UK Met Office, Météo France and DWD) have been evaluated over the period 2011–2020, and their skill quantified using the Continuous Ranked Probability Skill Score (CRPSS). We analyse how skill varies across the seasons and years, and if it can be linked to catchments characteristics. In doing so, we develop a methodology and a Python package to implement it, which is freely available for future applications to other regions. As for the study case, our results showed that among the four forecasting centres, ECMWF's forecasts were the most skilful in South Korea. In particular, seasonal forecasts outperform the climatology for 2 months of lead time and are more skilful during the wet season and in dry years. Linear bias correction is found to be useful to correct systematic seasonal biases, whereas we found no significant correlation between the catchment characteristics and forecast skill. We also investigated the possibility of anticipating dry years from seasonal forecasts and/or ENSO indices but found no significant link.

A Multimodel Real-Time System for Global Probabilistic Subseasonal Forecasts of Precipitation and Temperature

Abstract A global multimodel probabilistic subseasonal forecast system for precipitation and near-surface temperature is developed based on three NOAA ensemble prediction systems that make their forecasts available publicly in real time as part of the Subseasonal Experiment (SubX). The weekly and biweekly ensemble means of precipitation and temperature of each model are individually calibrated at each grid point using extended logistic regression, prior to forming equal-weighted multimodel ensemble (MME) probabilistic forecasts. Reforecast skill of week-3–4 precipitation and temperature is assessed in terms of the cross-validated ranked probability skill score (RPSS) and reliability diagram. The multimodel reforecasts are shown to be well calibrated for both variables. Precipitation is moderately skillful over many tropical land regions, including Latin America, sub-Saharan Africa and Southeast Asia, and over subtropical South America, Africa, and Australia. Near-surface temperature skill is considerably higher than for precipitation and extends into the extratropics as well. The multimodel RPSS skill of both precipitation and temperature is shown to exceed that of any of the constituent models over Indonesia, South Asia, South America, and East Africa in all seasons. An example real-time week-3–4 global forecast for 13–26 November 2021 is illustrated and shown to bear the hallmarks of the combined influences of a moderate Madden–Julian oscillation event as well as weak–moderate ongoing La Niña event. Significance Statement This paper develops a system for forecasting of precipitation and temperatures globally over land, several weeks in advance, with a focus on biweekly averaged conditions between three and four weeks ahead. The system provides the likelihood of biweekly and weekly conditions being below, near, or above their long-term averages, as well the probability of exceeding (or not exceeding) any threshold value. Using historical data, the precipitation forecasts are demonstrated to have skill in many tropical regions, and the temperature forecasts are more widely skillful. While weather and seasonal range forecasts have become quite generally available, this is one of the first examples of a publicly available, calibrated multimodel probabilistic real-time forecasting system for the subseasonal range.

The DWD climate predictions website: Towards a seamless outlook based on subseasonal, seasonal and decadal predictions

The climate predictions website of the Deutscher Wetterdienst (DWD, https://www.dwd.de/climatepredictions) presents a consistent operational outlook for the coming weeks, months and years, focusing on the needs of German users. At global scale, subseasonal predictions from the European Centre of Medium-Range Weather Forecasts as well as seasonal and decadal predictions from the DWD are used. Statistical downscaling is applied to achieve high resolution over Germany. Lead-time dependent bias correction is performed on all time scales. Additionally, decadal predictions are recalibrated.The website offers ensemble mean and probabilistic predictions for temperature and precipitation combined with their skill (mean squared error skill score, ranked probability skill score). Two levels of complexity are offered: basic climate predictions display simple, regionally averaged information for Germany, German regions and cities as maps, time series and tables. The skill is presented as traffic light. Expert climate predictions show complex, gridded predictions for Germany (at high resolution), Europe and the world as maps and time series. The skill is displayed as the size of dots. Their color is related to the signal in the prediction.The website was developed in cooperation with users from different sectors via surveys, workshops and meetings to guarantee its understandability and usability. The users realize the potential of climate predictions, but some need advice in using probabilistic predictions and skill. Future activities will include the further development of predictions to improve skill (multi-model ensembles, teleconnections), the introduction of additional products (data provision, extremes) and the further clarification of the information (interactivity, video clips).

Probabilistic solar nowcasting based on all-sky imagers

The interest in shortest-term solar irradiance forecasts (nowcasts) increases steadily with the increase share of distributed solar power generation. Such solar irradiance nowcasts are beneficial for different stakeholders, from generation to transmission and demand, in order to tackle challenges caused by the variability of solar irradiance. In space and time highly resolved nowcasts can be obtained by all sky imager (ASI) systems, which analyze the sky conditions from sky images. Deterministic nowcasts from ASI systems are subject to significant uncertainties. Reliable uncertainty information are very helpful for any application, in order to derive practical benefit from nowcasts. Therefore, such nowcasts should be probabilistic in nature, which provide probability distributions. Meaningful indicators for the uncertainties at hand are provided by prediction intervals for distinct confidence levels derived from the probability distributions. Thus, a real time capable nonparametric probabilistic quantile nowcasting method based on deterministic ASI nowcast is developed. The method takes irradiance variabilities as main predictor of nowcast uncertainties into account. A benchmark against three distinct baseline models is conducted over an extensive data set, using a variety of recently recommended scores. Overall average continuous ranked probability skill scores (Clear-Sky Dependent Climatology as baseline) for nowcasts up to 20 min ahead of 0.72 ± 0.08 (direct normal irradiance) and 0.62 ± 0.09 (global horizontal irradiance) are reached. For a better evaluation of the actual performance of the probabilistic nowcasts, a discretization of the validation data set into eight irradiance variability conditions is performed. All scores are determined for each of these distinct conditions.

Advanced seasonal predictions for vine management based on bioclimatic indicators tailored to the wine sector

The potential increase in the adoption value of seasonal forecasts is spotlighted in this paper by introducing observation-forecast blending for wine-sector indicators over the Iberian Peninsula. The predictions of six bioclimatic indicators (temperature and precipitation based) considered highly important from the perspective of wine-sector users were prepared for each month of the growing season and combined with previous observations as the indicator period progresses. The performance of this approach was then assessed with Pearson’s correlation coefficient and Fair Ranked Probability Skill Score (FRPSS). The results show a marked increase in the skill metrics during the growing season from the early combinations for all the indicators. This progressive improvement of the forecasting skill offers the users an opportunity to ponder anticipation and confidence in their decisions and, thus, facilitate the future uptake of seasonal forecasting in their decision planning.

Daily ensemble river discharge reforecasts and real-time forecasts from the operational Global Flood Awareness System

Abstract. Operational global-scale hydrological forecasting systems are used to help manage hydrological extremes such as floods and droughts. The vast amounts of raw data that underpin forecast systems and the ability to generate information on forecast skill have, until now, not been publicly available. As part of the Global Flood Awareness System (GloFAS; https://www.globalfloods.eu/, last access: 3 December 2022) service evolution, in this paper daily ensemble river discharge reforecasts and real-time forecast datasets are made free and openly available through the Copernicus Climate Change Service (C3S) Climate Data Store (CDS). They include real-time forecast data starting on 1 January 2020 updated operationally every day and a 20-year set of reforecasts and associated metadata. This paper describes the model components and configuration used to generate the real-time river discharge forecasts and the reforecasts. An evaluation of ensemble forecast skill using the continuous ranked probability skill score (CRPSS) was also undertaken for river points around the globe. Results show that GloFAS is skilful in over 93 % of catchments in the short (1 to 3 d) and medium range (5 to 15 d) against a persistence benchmark forecast and skilful in over 80 % of catchments out to the extended range (16 to 30 d) against a climatological benchmark forecast. However, the strength of skill varies considerably by location with GloFAS found to have no or negative skill at longer lead times in broad hydroclimatic regions in tropical Africa, western coast of South America, and catchments dominated by snow and ice in high northern latitudes. Forecast skill is summarised as a new headline skill score available as a new layer on the GloFAS forecast Web Map Viewer to aid user interpretation and understanding of forecast quality.

Calibration of ECMWF SEAS5 based streamflow forecast in Seasonal hydrological forecasting for Citarum river basin, West Java, Indonesia

Study region:Citarum river basin, West Java — Indonesia. Study focus:We look at the skill of Empirical Quantile Mapping corrected ECMWF SEAS5 (SEAS5 EQM bias-corrected) based streamflow forecasts in the Citarum river basin. We focus on July to October because these are agriculturally important months in Java. We use a high-resolution hydrologic model (wflow_sbm) with data for the period 1989–2009. New hydrological insights for the region:Water users and agricultural practitioners commonly need monthly to seasonal hydrological forecasts. The forecasts should be sufficiently skillful and provide information that is relevant to the decisions makers in order to have practical value to them. We assess if skilful SEAS5 EQM bias-corrected based seasonal forecasts are available with the purpose to support rice production. In this streamflow forecast calibration, we look at different aggregation days and different lead times. For the verification, we use the Continuous Ranked Probability Skill Score (CRPSS), Brier Skill Score (BSS), and Mean Average Error (MAE). We also look at the correlation, the Root Mean Square Error (RMSE), and the Receiver operating characteristic Skill (ROCS). The LT1 and LT2 forecast show higher skills than longer lead times. Meanwhile, streamflow based on the aggregated forecast at 30 to 60 days aggregation days is more skillful than larger aggregations. In Indonesia, this study is a study that initiates using a hydrological model with inputs from a seasonal rainfall forecast.

A Bayesian Hierarchical Model Combination Framework for Real‐Time Daily Ensemble Streamflow Forecasting Across a Rainfed River Basin

AbstractThe frequent occurrence of floods during the rainy season is one of the threats in rainfed river basins, especially in river basins of India. This study implemented a Bayesian hierarchical model combination (BHMC) framework to generate skillful and reliable real‐time daily ensemble streamflow forecast and peak flow and demonstrates its utility in the Narmada River basin in Central India for the peak monsoon season (July–August). The framework incorporates information from multiple sources (e.g., deterministic hydrological forecast, meteorological forecast, and observed data) as predictors. The forecasts were validated with a leave‐1‐year‐out cross‐validation using accuracy metrics such as BIAS and Pearson correlation coefficient (R) and probabilistic metrics such as continuous ranked probability skill score, probability integral transform (PIT) plots, and the average width of the 95% confidence intervals (AWCI) plots. The results show that the BHMC framework can increase the forecast skill by 40% and reduce absolute bias by at least 28% compared to the raw deterministic forecast from a physical model, the Variable Infiltration Capacity model. In addition, PIT and AWCI show that the framework can provide sharp and reliable streamflow forecast ensembles for short lead times (1–3‐day lead time) and provide useful skills beyond up to 5‐day lead time. These will be of immense help in emergency and disaster preparedness.

Evaluation of the forecast skill of North American Multi‐Model Ensemble for monthly and seasonal precipitation forecasts over Iran

AbstractNorth American Multi‐Model Ensemble (NMME) precipitation forecast skill over Iran is evaluated using Taylor diagrams and ranked probability skill scores (RPSS) as determined over a 29‐year test period (1991–2019). The forecast skill for both monthly (October through June for lead‐times of 0.5–3.5 months) and seasonal (October–December [OND], January–March [JFM], and April–June [AMJ] for lead‐times of 1.5–3.5 months) timescales is evaluated using six NMME models as well as multi‐model ensemble means (MMM). The latest versions of these models for forecasting Iran's precipitation have not been evaluated thus far. The Global Precipitation Climatology Center (GPCC) version 2020 dataset is used to verify the models. Among individual NMME models, Geophysical Fluid Dynamics Laboratory‐Seamless System for Prediction and Earth System Research (GFDL‐SPEAR) has generally the highest forecast skill. Both Taylor diagrams and RPSS of most of the models have indicated that the highest forecast skill is found for the month of November such that the Pearson correlation for both SPEAR and MMM is statistically significant for all lead‐times. For both monthly and seasonal timescales, the temporal Pearson correlation (TPC) between the observed and forecasts of the MMM is higher than the TPC of the individual models. The spatial Pearson correlation (SPC) and normalized centred root mean square error (NCRMSE) of the SPEAR is close to MMM, but the normalized standard deviation (NSD) of the SPEAR is closer to one compared to the MMM for months from November to March and two seasons (OND and JFM seasons). The MMM precipitation forecasts are underestimated over the northern regions and Zagros mountains for JFM and OND seasons for both 1.5‐ and 2.5‐month lead‐times. The degree to which the forecast skill of MMM is dependent on the El Niño–Southern Oscillation (ENSO) connections with precipitation over Iran is examined. Significant Spearman correlations between simultaneous observed Niño3.4 index and Iran precipitation are found for OND, but not for JFM and AMJ. The MMM reproduces the observed ENSO teleconnections to the tropical Pacific in OND, consistent with forecast skill in that season. However, the MMM also produces forecast skill in JFM and AMJ when the ENSO influence is marginal, showing that ENSO is not the only source of skill in the models.

Long-Term Streamflow Prediction Using Hybrid SVR-ANN Based on Bayesian Model Averaging

The Volga River, as the primary supplier of the Caspian Sea, plays a huge role in its ecosystem sustainability. In this study, we analyze its runoff predictability for different monthly forecast horizons. Additionally, the meteorological and hydrological variables affecting runoff in each month are identified. A wide range of the potential variables was first collected and the Boruta variable preprocessing method was employed to select the important ones. Then the hybrid models were created by combining the selected variables and the data-driven models [i.e., support vector regression (SVR), artificial neural network (ANN), and multiple linear regression (MLR)]. To postprocess the predicted data, the Bayesian model averaging (BMA) method was employed using the combination of the Boruta–artificial neural network (B-ANN) and the Boruta–support vector regression (B-SVR) models. Finally, the Kling-Gupta efficiency (KGE) and the continuous ranked probability skill score (CRPSS) probabilistic evaluation criteria were applied to evaluate the hybrid models. The results showed that the streamflow of the previous steps is the most crucial variable in predicting the streamflow of all next horizons, while its significance decreases as the forecast time horizon increases. Moreover, the temperature variables have the unlike effect on the streamflow prediction and the minimum temperature for winter and spring and the maximum and the average temperatures for summer and autumn are the most effective ones. Correspondingly, BMA-B-ANN-SVR presents the best performance among the hybrid models (for example, the median of 0.8 and 0.91 for CRPSS and KGE in the first horizon, respectively); the reliability of its predicted runoff for different forecast horizons is much better than other hybrid models (B-ANN and B-SVR).

Multi-model seasonal prediction of global surface temperature based on partial regression correction method

The increased climate change is having a huge impact on the world, with the climatic change sensitive and vulnerable regions at significant risk particularly. Effective understanding and integration of climate information are essential. It helps to reduce the risks associated with adverse weather conditions and to better adapt to the impacts of climate variability and change. Using the hindcast data from Japan Meteorological Agency/Meteorological Research Institute (JMA/MRI) coupled prediction system version 2 (JMA/MRI-CPS2), National Centers for Environmental Prediction (NCEP) Climate Forecast System model version 2 (CFSv2), and Canadian Centre for Climate Modeling and Analysis (CCCma) Coupled Climate Model, versions 3 (CanCM3) seasonal prediction model systems, the performance of seasonal prediction for global surface temperature in boreal summer and winter is comprehensively evaluated and compared for 1982–2011 from the perspective of deterministic and probabilistic forecast skills in this study, and a partial regression correction (PRC) method is introduced to correct seasonal predictions. The results show high prediction skills in the tropics, particularly in the equatorial Pacific, while poor skills on land. In general, JMA/MRI-CPS2 has slightly better prediction performance than CFSv2 and CanCM3 in the tropics. CFSv2 is generally superior to JMA/MRI-CPS2 and CanCM3 in the extratropical northern hemisphere and East Asia, especially for the abnormal low winter temperature prediction in East Asia. CanCM3 shows good deterministic forecast skills in extra-tropics but performs slightly worse in probabilistic forecasting. Based on the respective strengths of each seasonal prediction model, an ensemble forecast correction with observational constraint is implemented by partial regression, and the improvement of skills in ensemble predicting has been analyzed. Compared to the simple multi-model ensemble (MME), the correction improved the global-average temporal correlation coefficient and multi-year mean anomaly correlation coefficient by about 0.1 and 0.13, respectively. The validation tests indicate that the corrected ensemble forecast has higher ranked probability skill scores than that of the MME, which is improved by more than 0.06 in the tropics. Meanwhile, when the training period is sufficiently long, it may have the potential for future seasonal temperature predictions from the perspective of stable zonal partial regression coefficients.

How Skillful Are the European Subseasonal Predictions of Wind Speed and Surface Temperature?

Abstract Subseasonal forecasts of 100-m wind speed and surface temperature, if skillful, can be beneficial to the energy sector as they can be used to plan asset availability and maintenance, assess risks of extreme events, and optimally trade power on the markets. In this study, we evaluate the skill of the European Centre for Medium-Range Weather Forecasts’ subseasonal predictions of 100-m wind speed and 2-m temperature. To the authors’ knowledge, this assessment is the first for the 100-m wind speed, which is an essential variable of practical importance to the energy sector. The assessment is carried out on both forecasts and reforecasts over European domain gridpoint wise and also by considering several spatially averaged domains, using several metrics to assess different attributes of forecast quality. We propose a novel way of synthesizing the continuous ranked probability skill score. The results show that the skill of the forecasts and reforecasts depends on the choice of the climate variable, the period of the year, and the geographical domain. Indeed, the predictions of temperature are better than those of wind speed, with enhanced skill found for both variables in the winter relative to other seasons. The results also indicate significant differences between the skill of forecasts and reforecasts, arising mainly due to the differing ensemble sizes. Overall, depending on the choice of the geographical domain and the forecast attribute, the results show skillful predictions beyond 2 weeks, and in certain cases, up to 6 weeks for both variables, thereby encouraging their implementation in operational decision-making.