Raw Forecasts Research Articles

Verifying operational intra-day solar forecasts from ECMWF and NOAA

Global horizontal irradiance (GHI) forecasting at intra-day horizons of up to 12-h ahead is vital to grid integration of solar photovoltaics, but has been fundamentally difficult for all methods that do not involve numerical weather prediction (NWP), since non-NWP methods are unable to extrapolate the data to a horizon beyond a fews hours. The European Centre for Medium-Range Weather Forecasts (ECMWF) and the National Oceanic and Atmospheric Administration (NOAA) are the most representative weather centers in Europe and America, respectively. To understand their operational impact and value to grid integration, the ECMWF’s High Resolution (HRES) model and two models from NOAA, namely, Rapid Refresh (RAP) and High-Resolution Rapid Refresh (HRRR), are validated through the Murphy–Winkler distribution-oriented verification framework, over the year 2020, at seven locations. All forecasts are retrieved at native horizontal resolutions—9 km for HRES, 13 km for RAP, and 3 km for HRRR; it depicts the “off-the-shelf” scenario if these forecasts are to be utilized by end users. Three simple linear correction methods, each being statistically optimal in its own respect, are used to post-process the raw forecasts. It was found that 1–12-h-ahead ECMWF’s HRES forecasts have a significantly lower root mean square error (14.0–33.7%) as compared to NOAA’s HRRR (19.0–53.2%) and RAP (19.2–45.9%). Even after the large biases in HRRR and RAP forecasts are removed, those post-processed versions are still inferior to the raw HRES forecasts.

Postprocessing Ensemble Weather Forecasts for Introducing Multisite and Multivariable Correlations Using Rank Shuffle and Copula Theory

Abstract Statistical methods have been widely used to postprocess ensemble weather forecasts for hydrological predictions. However, most of the statistical postprocessing methods apply to a single weather variable at a single location, thus neglecting the intersite and intervariable dependence structures of forecast variables. This study synthesized a multisite and multivariate (MSMV) postprocessing framework that extends the univariate method to the MSMV version by directly rearranging the postprocessed ensemble members (post-reordering strategy) or by rearranging the latent variables used in the univariate method (pre-reordering strategy). Based on the univariate generator-based postprocessing (GPP) method, the two reordering strategies and three dependence reconstruction methods [rank shuffle (RS), Gaussian copula (GC), and empirical copula (EC)] totaling six MSMV methods (RS-Pre, GC-Pre, EC-Pre, RS-Post, GC-Post, and EC-Post) were evaluated in postprocessing ensemble precipitation and temperature forecasts for the Xiangjiang basin in China using the 11-member ensemble forecasts from the Global Ensemble Forecasting System (GEFS). The results showed that raw GEFS forecasts tend to be biased for both the forecast ensembles and the intersite and intervariable dependencies. The univariate method can improve the univariate performance of ensemble mean and spread but misrepresent the intersite and intervariable dependence among the forecast variables. The MSMV framework can well utilize the advantages of the univariate method and also reconstruct the intersite and intervariable dependencies. Among the six methods, RS-Pre, RS-Post, GC-Post, and EC-Post perform better than the others with respect to reproducing the univariate statistics and multivariable dependences. The post-reordering strategy is recommended to combine the univariate method (i.e., GPP) and reconstruction methods.

Reconstructing climate trends adds skills to seasonal reference crop evapotranspiration forecasting

Abstract. Evapotranspiration plays an important role in the terrestrial water cycle. Reference crop evapotranspiration (ETo) has been widely used to estimate water transfer from vegetation surface to the atmosphere. Seasonal ETo forecasting provides valuable information for effective water resource management and planning. Climate forecasts from general circulation models (GCMs) have been increasingly used to produce seasonal ETo forecasts. Statistical calibration plays a critical role in correcting bias and dispersion errors in GCM-based ETo forecasts. However, time-dependent errors resulting from GCM misrepresentations of climate trends have not been explicitly corrected in ETo forecast calibrations. We hypothesize that reconstructing climate trends through statistical calibration will add extra skills to seasonal ETo forecasts. To test this hypothesis, we calibrate raw seasonal ETo forecasts constructed with climate forecasts from the European Centre for Medium-Range Weather Forecasts (ECMWF) SEAS5 model across Australia, using the recently developed Bayesian joint probability trend-aware (BJP-ti) model. Raw ETo forecasts demonstrate significant inconsistencies with observations in both magnitudes and spatial patterns of temporal trends, particularly at long lead times. The BJP-ti model effectively corrects misrepresented trends and reconstructs the observed trends in calibrated forecasts. Improving trends through statistical calibration increases the correlation coefficient between calibrated forecasts and observations (r) by up to 0.25 and improves the continuous ranked probability score (CRPS) skill score by up to 15 (%) in regions where climate trends are misrepresented by raw forecasts. Skillful ETo forecasts produced in this study could be used for streamflow forecasting, modeling of soil moisture dynamics, and irrigation water management. This investigation confirms the necessity of reconstructing climate trends in GCM-based seasonal ETo forecasting and provides an effective tool for addressing this need. We anticipate that future GCM-based seasonal ETo forecasting will benefit from correcting time-dependent errors through trend reconstruction.

Sub‐Seasonal Prediction of Drought and Streamflow Anomalies for Water Management in India

AbstractMeteorological and hydrologic prediction at short to sub‐seasonal scales is essential for reservoir operations to mitigate droughts. We examine the skills in the meteorological forecast from the Subseasonal Experiment (SubX) and Extended Range Forecast System (ERFS) for precipitation, maximum and minimum temperatures at 1, 7, 15, and 30 days lead. We bias‐corrected meteorological forecasts using the multivariate bias correction method for hydrologic prediction. The Variable Infiltration Capacity model was used to simulate total runoff and root‐zone soil moisture for India. We also developed a streamflow forecast for the five major river basins that have large reservoirs. Bias correction of meteorological forecast (precipitation, maximum and minimum temperatures) resulted in a considerable improvement in hydrologic and meteorological forecast skills. The Environmental Modeling Center (EMC) model from the SubX provides either better or equal forecast skills for the raw meteorological forecast compared to ERFS, which is an operational product in India. We examined the forecast skills of the meteorological and hydrological products for the two major droughts that occurred recently. We find that most forecast models effectively captured the onset, peak, and termination of the North Indian drought in 2015–2016 and the South Indian drought in 2016–2017 at a 30‐day lead. Bias correction of the meteorological forecast improved the streamflow forecast for the selected drought event upstream of the major reservoirs. The EMC model showed better forecast skills for the two major droughts than other forecast products. Overall, the SubX products show potential for short‐to‐sub‐seasonal scale hydrologic prediction that can assist water management in India.

Predictive performance of ensemble hydroclimatic forecasts: Verification metrics, diagnostic plots and forecast attributes

AbstractPredictive performance is one of the most important issues for practical applications of ensemble hydroclimatic forecasts. While different forecasting studies tend to use different combinations of verification metrics and diagnostic plots, this paper presents a literature survey of popular verification metrics, diagnostic plots and their corresponding forecast attributes. Twenty metrics to quantify predictive performance of ensemble, deterministic and categorical forecasts are detailed; six types of diagnostic plots to visualize the relationship between ensemble forecasts and observations are presented; and the correspondence relationships of verification metrics and diagnostic plots with ten forecast attributes are illustrated. Numerical experiments are devised for raw ensemble precipitation forecasts generated by the Climate Forecast System version 2 (CFSv2) to illustrate the verification metrics, diagnostic plots and forecast attributes. The results suggest that verification metrics that pay attention to some similar attributes can be related to one another. Meanwhile, the metrics are not necessarily in full accordance as they tend to reflect different attributes. Time series plots visualize forecasts over a period of interest and graphically illustrate forecast bias, reliability and skill. Quantile range plots provide an overall diagnosis for the association of ensemble forecasts with observations. Overall, verification metrics can be illustrated by diagnostic plots for individual grid cells and also be screened by spatial plots at regional and global scales. The survey presented in this article can serve as a roadmap for the selection of verification metrics, diagnostic plots and forecast attributes for verifying ensemble hydroclimatic forecasts.This article is categorized under: Science of Water > Hydrological Processes

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.

Introducing long‐term trends into subseasonal temperature forecasts through trend‐aware postprocessing

AbstractSkilful subseasonal forecasts are crucial for issuing early warnings of extreme weather events, such as heatwaves and floods. Operational subseasonal climate forecasts are often produced by global climate models not dissimilar to seasonal forecast models, which typically fail to reproduce observed temperature trends. In this study, we identify that the same issue exists in the subseasonal forecasting system. Subsequently, we adapt a trend‐aware forecast postprocessing method, previously developed for seasonal forecasts, to calibrate and correct the trend in subseasonal forecasts. We modify the method to embed 30‐year climate trends into the calibrated forecasts even when the available hindcast period is shorter. The use of 30‐year trends is to robustly represent long‐term climate changes and overcome the problem that trends inferred from a shorter period may be subject to large sampling variability. Calibration is applied to 20‐year ECMWF subseasonal forecasts and AWAP observations of Australian minimum and maximum temperatures with forecast horizons of up to 4 weeks. Relative to day‐of‐year climatology, raw week‐1 forecasts reproduce temperature trends of the 20‐year observations in many regions while raw week‐4 forecasts do not exhibit the 20‐year observed trends. After trend‐aware postprocessing, the behaviour of forecast trends is related to raw forecast skill regarding accuracy. Calibrated week‐1 forecasts show apparent trends consistent with the 20‐year observations, as the calibration transfers forecast skill and embeds the 20‐year observed trends into the forecasts when raw forecasts are inherently skilful. In contrast, calibrated week‐4 forecasts exhibit the 30‐year observed trends, as the calibration reverts the forecasts to the 30‐year observed climatology with trends when raw forecasts have little skill. For both weeks, the trend‐aware calibrated forecasts are more reliable, and as skilful as or more skilful than raw forecasts. The extended trend‐aware method can be applied to deliver high‐quality subseasonal forecasts and support decision‐making in a changing climate.

Improved Trend-Aware Postprocessing of GCM Seasonal Precipitation Forecasts

Abstract Climate trends have been observed over the recent decades in many parts of the world, but current global climate models (GCMs) for seasonal climate forecasting often fail to capture these trends. As a result, model forecasts may be biased above or below the trendline. In our previous research, we developed a trend-aware forecast postprocessing method to overcome this problem. The method was demonstrated to be effective for embedding observed trends into seasonal temperature forecasts. In this study, we further develop the method for postprocessing GCM seasonal precipitation forecasts. We introduce new formulation and evaluation features to cater for special characteristics of precipitation amounts, such as having a zero lower bound and highly positive skewness. We apply the improved method to calibrate ECMWF SEAS5 forecasts of seasonal precipitation for Australia. Our evaluation shows that the calibrated forecasts reproduce observed trends over the hindcast period of 36 years. In some regions where observed trends are statistically significant, forecast skill is greatly improved by embedding trends into the forecasts. In most regions, the calibrated forecasts outperform the raw forecasts in terms of bias, skill, and reliability. Wider applications of the new trend-aware postprocessing method are expected to boost user confidence in seasonal precipitation forecasts.

Postprocessing of hydrometeorological ensemble forecasts based on multisource precipitation in Ganjiang River basin, China

Flood forecasting is a challenging task with major concerns that include forecast uncertainty and lead time. This paper aims to comprehensively address the data correction issue for hydrological models by developing an integrated postprocessing framework for hydrometeorological ensemble forecasts based on three types of ensemble precipitation forecasts (ECMWF, GEFS, and CFSv2). This framework mainly comprises the Ensemble Preprocessor (EPP) and the Bayesian model averaging (BMA) scheme: (1) the EPP eliminates various deviations from the ensemble precipitation forecasts, with the canonical event model extracting useful predictive information and extending the lead time by linking the forecasts seamlessly, and (2) the BMA scheme develops more skillful and reliable probabilistic hydrological forecasts from ensemble streamflow forecasts generated by three hydrological models (XAJ, VIC and DTVGM) driven by the postprocessed ensemble precipitation forecasts. The Ganjiang River Basin was selected as a case study to examine the capacity and efficiency of our developed framework. We found that the performance of the integrated postprocessing of hydrometeorological ensemble forecasts benefitted from a reasonable design of canonical events, especially when the observed precipitation was factored into the canonical events. Streamflow forecasts with increased lead times and accuracy were obtained by employing postprocessed precipitation forecasts instead of raw forecasts as the input for the hydrological models. We also found that the BMA scheme further improved the forecast effect by generating more skillful and reliable forecasts than a single hydrological model, which was verified by a series of statisticalindices and uncertainty analysis. Our results underscore the importance of considering both input and output for hydrological models in hydrological forecasting.

Extending a joint probability modelling approach for post-processing ensemble precipitation forecasts from numerical weather prediction models

Statistical post-processing has been widely employed to correct bias and dispersion errors in raw ensemble precipitation forecasts from numerical weather prediction models. One prominent post-processing scheme is to establish a joint probability model by fitting a bivariate distribution of raw forecasts and corresponding observations. However, current joint probability models only incorporate ensemble mean as the predictor, and ensemble spread is not considered. This is a major disadvantage of joint probability models as ensemble spread can be informative for forecast uncertainty. In this paper, we propose a two-step calibration approach to combine the strengths of joint probability models and the useful information included in the ensemble spread. In the first step, we take the seasonally coherent calibration (SCC) model as an example of joint probability models to calibrate the ensemble mean. As SCC for precipitation forecasts involves transformations for data normalization and special treatments of zero values, we explore three different ways to estimate ensemble mean values when establishing the SCC model. In the second step, we re-calibrate the ensemble forecasts produced in the first step to incorporate ensemble spread information from the raw forecasts. The performance of this two-step calibration is evaluated using ensemble precipitation forecasts from the Australian Bureau of Meteorology. We find that forecasts calibrated using the two-step calibration have better skills than SCC calibrated forecasts, especially for heavy precipitation events. Strengths of joint probability models and raw ensemble spread information are well utilized in the proposed two-step calibration approach.

Convolutional neural network-based statistical post-processing of ensemble precipitation forecasts

Raw forecasts from numerical weather prediction models suffer from systematic bias and cannot be directly used in applications such as hydrological forecasting. Statistical post-processing methods can be used to remove the bias and achieve reliable ensemble forecasts. However, traditional post-processing methods generally use local raw forecasts as the only predictor, which limits their ability to extract spatial information from raw forecasts. In this paper, we develop a convolutional neural network (CNN)-based post-processing method for precipitation forecasts to make use of spatial information and atmospheric circulation variables as auxiliary predictors. The results show that the proposed CNN-based post-processing model outperforms traditional methods in forecast accuracy and reliability, especially for heavy rain. The improvements of the CNN-based model relative to a state-of-the-art joint probability model can reach 5% and 8% in terms of Brier skill score for heavy rain and rainstorms at the lead time of 1 day. The improvements can be attributed to the use of auxiliary predictors such as total column water forecasts in the proposed model. Moreover, the CNN-based model transcends a fully connected network (FCN)-based post-processing model, which illustrates the benefits of using CNNs to extract spatial information. The results illustrate the proposed CNN-based post-processing model is able to utilize spatial information and various auxiliary predictors to improve forecast skill and quantify forecast uncertainty, which is important for further applications such as hydrological forecasting.

Ensemble solar forecasting and post-processing using dropout neural network and information from neighboring satellite pixels

Ensemble weather forecasts are often found to be under-dispersed and biased. Post-processing using spatio-temporal information is, therefore, required if one wishes to improve the quality of the raw forecasts. It is on this account that the present article generates and post-processes ensemble solar forecasts using satellite-derived irradiance not only from the focal pixel but also from the neighboring pixels. The ensemble forecasting model of choice is a dropout neural network with Monte Carlo sampling, eliminating the need for training multiple models and ensuring parameter diversity in ensemble forecasting. Subsequently, ensemble forecasts are post-processed using both parametric and nonparametric post-processing techniques, such as nonhomogenous regression, generalized additive model, linear quantile regression, or quantile random forests. The proposed forecasting framework is demonstrated and verified using four years of half-hourly data, at seven locations in the United States. Continuous ranked probability skill scores as high as 66% have been obtained when comparing the proposed method to a conditional climatology reference. The content of this article may be useful to a wide range of stakeholders in the power system, including but not limited to: independent system operators, who aim at efficiently maintaining the system’s reliability; utility- and distributed-scale PV plant owners, who wish to avoid penalties for power deviation between the scheduled and real-time delivery; and forecast retailers, who can benefit from selling solar forecasts of higher quality.

System bias correction of short-term hub-height wind forecasts using the Kalman filter

Wind energy is a fluctuating source for power systems, which poses challenges to grid planning for the wind power industry. To improve the short-term wind forecasts at turbine height, the bias correction approach Kalman filter (KF) is applied to 72-h wind speed forecasts from the WRF model in Zhangbei wind farm for a period over two years. The KF approach shows a remarkable ability in improving the raw forecasts by decreasing the root-mean-square error by 16% from 3.58 to 3.01 m s−1, the mean absolute error by 14% from 2.71 to 2.34 m s−1, the bias from 0.22 to − 0.19 m s−1, and improving the correlation from 0.58 to 0.66. The KF significantly reduces random errors of the model, showing the capability to deal with the forecast errors associated with physical processes which cannot be accurately handled by the numerical model. In addition, the improvement of the bias correction is larger for wind speeds sensitive to wind power generation. So the KF approach is suitable for short-term wind power prediction.

A parsimonious post-processor for uncertainty evaluation of ensemble precipitation forecasts: an application to quantitative precipitation forecasts for civil protection purposes

Abstract In this study, we present a post-processing technique designed to assess conditional predictive uncertainty for quantitative precipitation forecasts (QPFs) that extends to the multivariate predictor case. The theoretical foundation of the mixed binary-continuous precipitation process representation for a single predictor is coupled with the univariate version of the model conditional processor (MCP) that allows considering ensemble forecasts while maintaining a parsimonious approach. The experiment set-up is based on a weather vigilance zone (WVZ) of the national warning system in southern Italy and the relative QPFs issued by the Italian Civil Protection Department. Various aspects of the quality of the probabilistic forecast from the uncertainty processor were evaluated. The results indicate that post-processed forecasts can provide improved performances in terms of accuracy and reliability, tend to correct bias and are generally less under-dispersive than raw forecasts for the investigated area. Furthermore, we explored the possibility of issuing warnings employing the full predictive distribution and moving to the use of probabilistic thresholds here identified through a receiver operating characteristic (ROC) analysis. Considering the probability of exceeding a critical rainfall value allowed successful discrimination between events and non-events for critical precipitation occurrences and proved to be a valuable approach to decision-makers and information providers.

Probabilistic fire-danger forecasting: A framework for week-two forecasts using statistical post-processing techniques and the Global ECMWF Fire Forecast System (GEFF)

AbstractWildfire guidance two weeks ahead is needed for strategic planning of fire mitigation and suppression. However, fire forecasts driven by meteorological forecasts from numerical weather prediction models inherently suffer from systematic biases. This study uses several statistical-postprocessing methods to correct these biases and increase the skill of ensemble fire forecasts over the contiguous United States 8–14 days ahead. We train and validate the post-processing models on 20 years of European Centre for Medium-range Weather Forecast (ECMWF) reforecasts and ERA5 reanalysis data for 11 meteorological variables related to fire, such as surface temperature, wind speed, relative humidity, cloud cover, and precipitation. The calibrated variables are then input to the Global ECMWF Fire Forecast (GEFF) system to produce probabilistic forecasts of daily fire-indicators which characterize the relationships between fuels, weather, and topography. Skill scores show that the post-processed forecasts overall have greater positive skill at Days 8–14 relative to raw and climatological forecasts. It is shown that the post-processed forecasts are more reliable at predicting above- and below-normal probabilities of various fire indicators than the raw forecasts and that the greatest skill for Days 8–14 is achieved by aggregating forecast days together.

Machine learning for precipitation forecasts post-processing — Multi-model comparison and experimental investigation

AbstractObtaining high-quality quantitative precipitation forecasts is a key precondition for hydrological forecast systems. Due to multisource uncertainties (e.g., initial conditions, model structures and parameters), raw forecasts are subject to systematic biases; hence, statistical post-processing is often required to reduce these errors before the forecasts can proceed to hydrological applications. Machine learning (ML) algorithms are canonical statistical models, and they are diverse in type and variation. It is important to verify and compare their performance in the same scenario (e.g., precipitation post-processing). In this paper, we conduct a large-scale comparison study for the major ML models with diverse model structures and regularization strategies as post-processors for improving the quality of precipitation forecasts. Specifically, we compare the efficiency and effectiveness of 21 ML algorithms on solving this task. Daily reforecast precipitation with lead times up to 8 days from the Global Ensemble Forecast System and corresponding observations are employed to determine the usability of different models in the Yalong River basin in China. The performance of each model is validated by a group of carefully designed experiments and statistical metrics. The results reveal that improvements in model structures are more effective than regularization strategies. Among these algorithms, the optimized extra-trees regressor exhibit the best performance, effectively reduce overestimation and achieve the best skill in forecasting precipitation. Eleven ensemble members and a 2-day time window can be used as predictors to obtain the best model performance. The systematic experiments and findings also offer useful guidelines for other related studies.

Comparing and Combining Deterministic Surface Temperature Postprocessing Methods over the United States

AbstractCommon methods for the postprocessing of deterministic 2-m temperature (T2m) forecasts over the United States were evaluated from +12- to +120-h lead. Forecast data were extracted from the Global Ensemble Forecast System (GEFS) v12 reforecast dataset and thinned to a ½° grid. Analyzed data from the European Centre/Copernicus reanalysis (ERA5) were used for training and validation. Data from the 2000–18 period were used for training, and 2019 forecasts were validated. The postprocessing methods compared were the raw forecast guidance, a decaying-average bias correction (DAV), quantile mapping (QM), a univariate model output statistics (uMOS) algorithm, and a multivariate (mvMOS) algorithm. The mvMOS algorithm used the raw forecast temperature, the DAV adjustment, and the QM adjustment as predictors. Forecasts from all the postprocessing methods reduced the root-mean-square error (RMSE) and bias relative to the raw guidance. QM produced forecasts with slightly higher error than DAV. DAV estimates were the most consistent from day to day. The uMOS and mvMOS algorithms produced statistically significant lower RMSEs than DAV at forecast leads longer than 1 day, with mvMOS exhibiting the lowest error. Taylor diagrams showed that the MOS methods reduced the variability of the forecasts while improving forecast-analyzed correlations. QM and DAV modified the distribution of forecasts to more closely exhibit those of the analyzed data. A main conclusion is that the judicious statistical combination of guidance from multiple postprocessing methods is capable of producing forecasts with improved error statistics relative to any one individual technique. As each method applied here is algorithmically relatively simple, this suggests that operational deterministic postprocessing combining multiple correction methods could produce improved T2m guidance.

Calibrated ensemble forecasts of the height of new snow using quantile regression forests and ensemble model output statistics

Abstract. Height of new snow (HN) forecasts help to prevent critical failures of infrastructures in mountain areas, e.g. transport networks and ski resorts. The French national meteorological service, Météo-France, operates a probabilistic forecasting system based on ensemble meteorological forecasts and a detailed snowpack model to provide ensembles of HN forecasts. These forecasts are, however, biased and underdispersed. As for many weather variables, post-processing methods can be used to alleviate these drawbacks and obtain meaningful 1 to 4 d HN forecasts. In this paper, we compare the skill of two post-processing methods. The first approach is an ensemble model output statistics (EMOS) method, which can be described as a nonhomogeneous regression with a censored shifted Gamma distribution. The second approach is based on quantile regression forests, using different meteorological and snow predictors. Both approaches are evaluated using a 22 year reforecast. Thanks to a larger number of predictors, the quantile regression forest is shown to be a powerful alternative to EMOS for the post-processing of HN ensemble forecasts. The gain of performance is large in all situations but is particularly marked when raw forecasts completely miss the snow event. This type of situation happens when the rain–snow transition elevation is overestimated by the raw forecasts (rain instead of snow in the raw forecasts) or when there is no precipitation in the forecast. In that case, quantile regression forests improve the predictions using the other weather predictors (wind, temperature, and specific humidity).

Improvements in subseasonal forecasts of rainfall extremes by statistical postprocessing methods

End users of seasonal rainfall forecasts demand not only skilful forecasts of rainfall totals but also skilful forecasts of rainfall extremes. For better forecasts of rainfall extremes, the copula-based postprocessing (CPP) method, which was originally designed for forecasting rainfall totals, is modified with a hybrid probability distribution to model low-to-medium and heavy rainfall separately and to allow the forecast of extreme rainfall events that have never occurred in observed records. A case study for 17 rainfall stations in Queensland, Australia is carried out to test the forecast performance of the modified CPP to postprocess the raw Australian Community Climate and Earth-System Simulator (ACCESS) Seasonal model (ACCESS-S) seasonal rainfall forecast at a daily scale. The modified CPP improves the overall skill of forecasting 12 rainfall indices from the raw forecast and outperforms two quantile mapping based methods in most cases. The use of the hybrid distribution leads to more promising forecast skill for heavy and very heavy rainfall related indices. The forecast skill decreases with longer lead times and the modified CPP leads to neutral forecasts (i.e. forecasts with skill similar to climatology forecasts) for most rainfall indices beyond 0-month lead time. The skill improvement has been found in all selected climate regions and initialisation dates (from the 1st day of each month), though more substantial improvement is observed in the rainfall stations within the tropical zone where the raw forecast is particularly unskilful.

A five-parameter Gamma-Gaussian model to calibrate monthly and seasonal GCM precipitation forecasts

Calibration is necessary for improving raw forecasts generated by global climate models (GCMs) to fully utilize potential benefits of the forecasts in practical applications. Based on quantile mapping (QM), this paper proposes a five-parameter Gamma-Gaussian model to calibrate monthly and seasonal GCM precipitation forecasts. While QM directly maps forecasts to observations without accounting for the dependency relationship, the Gamma-Gaussian model employs the Gamma distribution to normalize precipitation forecasts and observations using normal quantile transform (NQT) and then formulates a bivariate Gaussian distribution to characterize the dependency relationship. A case study is devised to calibrate global precipitation forecasts generated by the Climate Forecast System version 2 (CFSv2) using both QM and Gamma-Gaussian models. The results show that both QM and Gamma-Gaussian models can effectively correct biases. While QM can improve forecast reliability to some degree by reducing biases, reliability is not always satisfactory. The Gamma-Gaussian model outperforms QM because it can ensure forecast reliability and coherence. To facilitate the selection of marginal distributions for the purpose of calibrating GCM precipitation forecasts, six alternative distributions, i.e., Gamma, lognormal, generalized extreme value, generalized logistic, Pearson type III and Kappa distributions, are employed to characterize the marginal distribution of forecasts (observations) in NQT. It is observed that the Gamma distribution is overall the most suitable and that the alternative distributions tend to fit sample-specific noises and get penalized under cross validation. Overall, the Gamma-Gaussian model can serve as an effective tool to calibrate raw GCM forecasts for hydrological modeling and water resources management.