Ensemble forecast systems are an integral part of the National Weather Service's (NWS) program for producing skillful, reliable deterministic and probabilistic forecasts, but statistical postprocessing is often required to address known deficiencies. Here, we test improvements to the ensemble postprocessing procedure currently implemented at the NWS Meteorological Development Laboratory (MDL) that generates a deterministic quantitative precipitation forecast from multimodel ensemble means. The existing approach uses recent forecast and analysis data to estimate cumulative distributions of precipitation and then applies a quantile mapping between the mean forecast and analysis. In the current work, we use a revised version of this approach that, in addition, utilizes ensemble mean inputs improved by a coalescence procedure based on the feature alignment technique (FAT). This is a variational technique for determining the displacements needed to adjust precipitation features of individual ensemble members toward their positions in the raw ensemble mean. The performance of the combined coalescence-quantile mapping algorithm was evaluated over a 2-yr period of forecasts from version 12 of the NCEP Global Ensemble Forecast System (GEFSv12) after extensive testing of configurable parameters to ensure the physical reasonableness of the precipitation fields. The coalescence procedure by itself was shown to correct some of the raw ensemble mean deficiencies related to position errors in the individual ensemble forecasts, especially as lead times increase. The combined approach resulted in improved forecasts as measured by fractional bias and equitable threat scores. Implementation of the techniques described here for operational guidance is being considered by the National Oceanic and Atmospheric Administration (NOAA) and is applicable to other forecast models.

Postprocessing is a critical step in attaining calibrated and reliable probabilistic forecast output from numerical weather prediction models. A novel deep learning framework is proposed to postprocess 20 years of 7-and 14-day precipitation accumulation reforecasts from the Global Ensemble Forecast System at subseasonal time scales (week 1, week 2, and combined weeks 3-4 forecasts) over the contiguous United States. The network builds upon previous studies and is a combination of three parallel-trained components suitable for subseasonal prediction. The first is a ResUnet architecture which learns nonlinear relationships between binned observed precipitation and input images of weather and geographical variables. The second conditions the network to the month-of-year via a feature-wise linear modulation (FiLM) layer. The third helps the network learn when to revert the forecast to that of climatology. The RUFCO (named for its components ResUnet, FiLM, and Climatological-Offramp) forecasts are compared against raw and climatological forecasts as well as those from a state-of-the-art distributional regression postprocessing model, "censored, shifted gamma distribution (CSGD)," and a simple bias-corrected model. At week 1, every method exhibited a competitive advantage over climatological forecasts. At week 2, RUFCO generated forecasts with statistically significant improvement over climatology at 82%-94% of the domain, beating CSGD's coverage of 76%-90% of the grid points. At week 3, RUFCO's skillful coverage was 65%-85%, while CSGDs dropped to only 12%-37%. At the longer lead times, RUFCO achieved the highest domain-averaged skill scores across seasons. However, the network tends to "smooth" forecast skill, making it less competitive with CSGD in limited areas with strongly spatially varying biases.

In the postprocessing of ensemble forecasts of weather variables, it is standard practice to first calibrate the forecasts in a univariate setting, before reconstructing multivariate ensembles that have a correct covariability in space, time, and across variables, via so-called “reordering” methods. Within this framework though, postprocessors cannot fully extract the skill of the raw forecast that may exist at larger scales. A multi-temporal-scale modulation mechanism for precipitation is here presented, which aims at improving the forecasts over different accumulation periods, and which can be coupled with any univariate calibration and multivariate reordering techniques. The idea, originally known under the term “canonical events,” has been implemented for more than a decade in the Meteorological Ensemble Forecast Processor (MEFP), a component of the U.S. National Weather Service’s (NWS) Hydrologic Ensemble Forecast Service (HEFS), although users were left with material in the gray literature. This paper proposes a formal description of the mechanism and studies its intrinsic connection with the multivariate reordering process. The verification of modulated and unmodulated forecasts, when coupled with two popular methods for reordering, the Schaake shuffle and ensemble copula coupling (ECC), is performed on 11 Californian basins, on both precipitation and streamflow. Results demonstrate the clear benefit of the multi-temporal-scale modulation, in particular on multiday total streamflow. However, the relative gain depends on the method used for reordering, with more benefits expected when this latter method is not able to reconstruct an adequate temporal structure on the calibrated precipitation forecasts.

© 2023 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses). Corresponding author: Jerald A. Brotzge, jerald.brotzge@wku.edu

This article describes proposed revised methods for the statistical postprocessing of precipitation amount intended for the NOAA's National Blend of Models using the Global Ensemble Forecast System version 12 data (GEFSv12). The procedure updates the previously established procedure of quantile mapping, weighting of sorted members, and dressing of the ensemble. The revised method leverages the long reforecast training dataset that has become available to improve quantile mapping of GEFSv12 data by eliminating the use of supplemental locations, that is, training data from other grid points. It establishes improved definitions of cumulative distributions through a spline-fitting approach. It provides updated algorithms for the weighting of sorted members based on closest-member histogram statistics, and it establishes an objective method for the dressing of the quantile-mapped, weighted ensemble. Verification statistics and case studies are provided in the accompanying article (Part II).

Tropical cyclones are extreme events with enormous and devastating consequences to life, property, and our economies. As a result, large-scale efforts have been devoted to improving tropical cyclone forecasts with lead times ranging from a few days to months. More recently, subseasonal forecasts (e.g., 2-6-week lead time) of tropical cyclones have received greater attention. Here, we study whether bias-corrected, subseasonal tropical cyclone reforecasts of the GEFS and ECMWF models are skillful in the Atlantic basin. We focus on the peak hurricane season, July-November, us-ing the reforecast years 2000-19. Model reforecasts of accumulated cyclone energy (ACE) are produced, and validated, for lead times of 1-2 and 3-4 weeks. Week-1-2 forecasts are substantially more skillful than a 31-day moving-window climato-logy, while week-3-4 forecasts still exhibit positive skill throughout much of the hurricane season. Furthermore, the skill of the combination of the two models is found to be an improvement with respect to either individual model. In addition to the GEFS and ECMWF model reforecasts, we develop a statistical modeling framework that solely relies on daily sea sur-face temperatures. The reforecasts of ACE from this statistical model are capable of producing better skill than the GEFS or ECMWF model, individually, and it can be leveraged to further enhance the model combination reforecast skill for the 3-4-week lead time.

The water resources of the western United States have enormous agricultural and municipal demands. At the same time, droughts like the one enveloping the West in the summer of 2021 have disrupted supply of this strained and precious resource. Historically, seasonal forecasts of cool-season (November-March) precipitation from dynamical models such as North American Multi-Model Ensemble (NMME) and the Seasonal Forecasting System 5 (SEAS5) from the European Centre for Medium-RangeWeather Forecasts have lacked sufficient skill to aid inWestern stakeholders' and water managers' decision-making. Here, we propose a new empirical-statistical framework to improve cool-season precipitation forecasts across the contiguous United States (CONUS). This newly developed framework is called the Statistical Climate Ensemble Forecast (SCEF) model. The SCEF framework applies a principal component regression model to predictors and predictands that have undergone dimensionality reduction, where the predictors are large-scale meteorological variables that have been prefiltered in space. The forecasts of the SCEF model captures 12.0% of the total CONUS-wide standardized observed variance over the period 1982/83-2019/20, whereas NMME captures 7.2%. Over the more recent period 2000/01-2019/20, the SCEF, NMME, and SEAS5 models respectively capture 11.8%, 4.0%, and 4.1% of the total CONUS-wide standardized observed variance. An important finding is that much of the improved skill in the SCEF, with respect to models such as NMME and SEAS5, can be attributed to better forecasts across most of the western United States.

NOAA has created a global reanalysis dataset, intended primarily for initialization of reforecasts for its Global Ensemble Forecast System, version 12 (GEFSv12), which provides ensemble forecasts out to 135-days lead time. The reanalysis covers the period 2000-19. It assimilates most of the observations that were assimilated into the operational data assimilation system used for initializing global predictions. These include a variety of conventional data, infrared and microwave radiances, global positioning system radio occultations, and more. The reanalysis quality is generally superior to that from NOAA's previous-generation Climate Forecast System Reanalysis (CFSR), demonstrated in the fit of shortterm forecasts to the observations and in the skill of 5-day deterministic forecasts initialized from CFSR versus GEFSv12. Skills of reforecasts initialized from the new reanalyses are similar but slightly lower than skills initialized from a preoperational version of the real-time data assimilation system conducted at the higher, operational resolution. Control member reanalysis data on vertical pressure levels are made publicly available.

For the newly implemented Global Ensemble Forecast System, version 12 (GEFSv12), a 31-yr (1989-2019) ensemble reforecast dataset has been generated at the National Centers for Environmental Prediction (NCEP). The reforecast system is based on NCEP's Global Forecast System, version 15.1, and GEFSv12, which uses the Finite Volume 3 dynamical core. The resolution of the forecast system is similar to 25 km with 64 vertical hybrid levels. The Climate Forecast System (CFS) reanalysis and GEFSv12 reanalysis serve as initial conditions for the Phase 1 (1989-99) and Phase 2 (2000-19) reforecasts, respectively. The perturbations were produced using breeding vectors and ensemble transforms with a rescaling technique for Phase 1 and ensemble Kalman filter 6-h forecasts for Phase 2. The reforecasts were initialized at 0000 (0300) UTC once per day out to 16 days with 5 ensemble members for Phase 1 (Phase 2), except on Wednesdays when the integrations were extended to 35 days with 11 members. The reforecast dataset was produced on NOAA's Weather and Climate Operational Supercomputing System at NCEP. This study summarizes the configuration and dataset of the GEFSv12 reforecast and presents some preliminary evaluations of 500-hPa geopotential height, tropical storm track, precipitation, 2-m temperature, and MJO forecasts. The results were also compared with GEFSv10 or GEFS Subseasonal Experiment reforecasts. In addition to supporting calibration and validation for the National Water Center, NCEP Climate Prediction Center, and other National Weather Service stakeholders, this high-resolution subseasonal dataset also serves as a useful tool for the broader research community in different applications.

This paper utilizes statistical and statistical-dynamical methodologies to select, from the full observational record, a minimal subset of dates that would provide representative sampling of local precipitation distributions across the contiguous United States (CONUS). The CONUS region is characterized by a great diversity of precipitation-producing systems, mechanisms, and large-scale meteorological patterns (LSMPs), which can provide favorable environment for local precipitation extremes. This diversity is unlikely to be adequately captured in methodologies that rely on grossly reducing the dimensionality of the data-by representing it in terms of a few patterns evolving in time-and thus requires data thinning techniques based on high-dimensional dynamical or statistical data modeling. We have built a novel high-dimensional empirical model of temperature and precipitation capable of producing statistically accurate surrogate realizations of the observed 1979-99 (training period) evolution of these fields. This model also provides skillful hindcasts of precipitation over the 2000-20 (validation) period. We devised a subsampling strategy based on the relative entropy of the empirical model's precipitation (ensemble) forecasts over CONUS and demonstrated that it generates a set of dates that captures a majority of high-impact precipitation events, while substantially reducing a heavy-precipitation bias inherent in an alternative methodology based on the direct identification of large precipitation events in the Global Ensemble Forecast System (GEFS), version 12 reforecasts. The impacts of data thinning on the accuracy of precipitation statistical postprocessing, as well as on the calibration and validation of the Hydrologic Ensemble Forecast Service (HEFS) reforecasts are yet to be established. Significance StatementHigh-impact weather events are usually associated with extreme precipitation, which is notoriously difficult to predict even using highly resolved state-of-the-art numerical weather prediction models based on first physical principles. The same is true for statistical models that use past data to anticipate the future behavior likely to stem from an observed initial state. Here we use both types of models to identify the occurrences of the states, over the historical climate record, which are likely to lead to extreme precipitation events. We show that the overall statistics of precipitation over the contiguous United States is encapsulated in a greatly reduced set of such states, which could substantially alleviate the computational burden associated with testing of hydrological forecast models used for decision support.

Conventional statistical postprocessing techniques offer limited ability to improve the skills of probabilistic guidance for heavy precipitation. This paper introduces two artificial neural network (ANN)-based, geographically aware, and computationally efficient postprocessing schemes, namely, the ANN-multiclass (ANN-Mclass) and the ANN-censored, shifted gamma distribution (ANN-CSGD). Both schemes are implemented to postprocess Global Ensemble Forecast System (GEFS) forecasts to produce probabilistic quantitative precipitation forecasts (PQPFs) over the contiguous United States (CONUS) using a short (60 days), rolling training window. The performances of these schemes are assessed through a set of hindcast experiments, wherein postprocessed 24-h PQPFs from the two ANN schemes were compared against those produced using the benchmark quantile mapping algorithm for lead times ranging from 1 to 8 days. Outcomes of the hindcast experiments show that ANN schemes overall outperform the benchmark as well as the raw forecast over the CONUS in predicting probability of precipitation over a range of thresholds. The relative performance varies among geographic regions, with the two ANN schemes broadly improving upon quantile mapping over the central, south, and southeast, and slightly underperforming along the Pacific coast where skills of raw forecasts are the highest. Between the two schemes, the hybrid ANN-CSGD outperforms at higher rainfall thresholds (i.e., >50 mm day(-1)), though the outperformance comes at a slight expense of sharpness and spatial specificity. Collectively, these results confirm the ability of the ANN algorithms to produce skillful PQPFs with a limited training window and point to the prowess of the hybrid scheme for calibrating PQPFs for rare-to-extreme rainfall events.

Characteristics of the European Centre for Medium-Range Weather Forecast’s (ECMWF’s) 0000 UTC diagnosed 2-m temperatures ( T 2m ) from 4D-Var and global ensemble forecasts initial conditions were examined in 2018 over the contiguous United States at 1/2° grid spacing. These were compared against independently generated, upscaled high-resolution T 2m analyses that were created with a somewhat novel data assimilation methodology, an extension of classical optimal interpolation (OI) to surface data analysis. The analysis used a high-resolution, spatially detailed climatological background and was statistically unbiased. Differences of the ECMWF 4D-Var T 2m initial states from the upscaled OI reference were decomposed into a systematic component and a residual component. The systematic component was determined by applying a temporal smoothing to the time series of differences between the ECMWF T 2m analyses and the OI analyses. Systematic errors at 0000 UTC were commonly 1 K or more and larger in the mountainous western United States, with the ECMWF analyses cooler than the reference. The residual error is regarded as random in character and should be statistically consistent with the spread of the ensemble of initial conditions after inclusion of OI analysis uncertainty. This analysis uncertainty was large in the western United States, complicating interpretation. There were some areas suggestive of an overspread initial ensemble, with others underspread. Assimilation of more observations in the reference OI analysis would reduce analysis uncertainty, facilitating more conclusive determination of initial-condition ensemble spread characteristics.

Wildfire 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.

Common methods for the postprocessing of deterministic 2-m temperature (T-2m) 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 1/2 degrees 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 T-2m guidance.

NCEP has implemented the first version of the Finite Volume Cubed Sphere (FV3) dynamic core based Global Forecast System (GFS v15) into operations in June 2019, replacing the spectral model-based GFS. This is the first instantiation of NOAA's Unified Forecast System (UFS), which is being built as a comprehensive coupled Earth system model using modern tools and software infrastructure (e.g., NEMS, NUOPC, and ESMF) to support research and operations. Advancements in model physics and data assimilation are in development using CCPP and JEDI frameworks. Testing and evaluation of UFS are facilitated through the development of unified workflow and METplus capabilities. All these initiatives involve significant engagement with the research community, with emphasis on more efficient and streamlined transition of research advances to operations (R2O). The next major operational upgrade towards UFS is for the Global Ensemble Forecast System (GEFSv12) planned for implementation later this year. Compared to the currently operational GEFS, the next GEFS version 12 incorporates the following advances: the same FV3 based global model and UFS infrastructure as in GFS, higher resolution (~25km), increased membership (31), GFSv15 physics, advanced stochastic physics perturbations (SKEB and SPPT), and 2-tiered SST approach using SST anomalies from CFSv2 as input. For the first time, GEFSv12 will provide ensemble based operational weather predictions for sub-seasonal scales with daily 00z forecasts going out to 35 days. GEFSv12 also comes with 20-year reanalysis, 30-year reforecasts and 3-year retrospective forecasts to support stakeholder needs for calibration and validation. In addition, GEFSv12 absorbs the global wave ensembles and aerosol capabilities (control member only) through one-way coupling, taking major steps towards building a unified system and simplifying NCEP’s production suite. This presentation describes the design and development of GEFSv12 and discusses results from the evaluation of the retrospective and reforecast experiments. Significant improvements were noted in both deterministic and probabilistic forecast metrics for several variables including 500 hPa geopotential height anomaly correction, 850 hPa temperature and winds, near surface variables, precipitation, tropical cyclone tracks and intensity, and modes of variability including MJO and NAO. Substantial improvements were also noted in the performance of wave ensemble and aerosol predictions. This presentation also describes NOAA’s efforts towards accelerating further development of fully coupled UFS consisting of six component models of the Earth system: the FV3 dynamical core for the atmosphere, MOM6 for the ocean, Noah MP for the land surface, GOCART for aerosols, CICE5/CICE6 for sea ice and WW3 for ocean surface waves. Combined with data assimilation advances, an ambitious goal of unifying both the high-resolution deterministic (GFSv17) and probabilistic (GEFSv13) predictions for global medium range and sub-seasonal time scales is planned to significantly advance the global prediction capabilities at NCEP.

Forecast skill of numerical weather prediction (NWP) models for precipitation accumulations over California is rather limited at subseasonal time scales, and the low signal-to-noise ratio makes it challenging to extract information that provides reliable probabilistic forecasts. A statistical postprocessing framework is proposed that uses an artificial neural network (ANN) to establish relationships between NWP ensemble forecast and gridded observed 7-day precipitation accumulations, and to model the increase or decrease of the probabilities for different precipitation categories relative to their climatological frequencies. Adding predictors with geographic information and location-specific normalization of forecast information permits the use of a single ANN for the entire forecast domain and thus reduces the risk of overfitting. In addition, a convolutional neural network (CNN) framework is proposed that extends the basic ANN and takes images of large-scale predictors as inputs that inform local increase or decrease of precipitation probabilities relative to climatology. Both methods are demonstrated with ECMWF ensemble reforecasts over California for lead times up to 4 weeks. They compare favorably with a state-of-the-art postprocessing technique developed for medium-range ensemble precipitation forecasts, and their forecast skill relative to climatology is positive everywhere within the domain. The magnitude of skill, however, is low for week-3 and week-4, and suggests that additional sources of predictability need to be explored.

Spatial variability of precipitation is analyzed to characterize to what extent precipitation observed at a single location is representative of precipitation over a larger area. Characterization of precipitation representativeness is made in probabilistic terms using a parametric approach, namely, by fitting a censored shifted gamma distribution to observation measurements. Parameters are estimated and analyzed for independent precipitation datasets, among which one is based on high-density gauge measurements. The results of this analysis serve as a basis for accounting for representativeness error in an ensemble verification process. Uncertainty associated with the scale mismatch between forecast and observation is accounted for by applying a perturbed-ensemble approach before the computation of scores. Verification results reveal a large impact of representativeness error on precipitation forecast reliability and skill estimates. The parametric model and estimated coefficients presented in this study could be used directly for forecast postprocessing to partly compensate for the limitation of any modeling system in terms of precipitation subgrid-scale variability.

NCEP has implemented the first version of the Finite Volume Cubed Sphere (FV3) dynamic core based Global Forecast System (GFS v15) into operations in June 2019, replacing the spectral model-based GFS. This is the first instantiation of NOAA's Unified Forecast System (UFS), which is being built as a comprehensive coupled Earth system model using modern tools and software infrastructure (e.g., NEMS, NUOPC, and ESMF) to support research and operations. Advancements in model physics and data assimilation are in development using CCPP and JEDI frameworks. Testing and evaluation of UFS are facilitated through the development of unified workflow and METplus capabilities. All these initiatives involve significant engagement with the research community, with emphasis on more efficient and streamlined transition of research advances to operations (R2O). The next major operational upgrade towards UFS is for the Global Ensemble Forecast System (GEFSv12) planned for implementation later this year. Compared to the currently operational GEFS, the next GEFS version 12 incorporates the following advances: the same FV3 based global model and UFS infrastructure as in GFS, higher resolution (~25km), increased membership (31), GFSv15 physics, advanced stochastic physics perturbations (SKEB and SPPT), and 2-tiered SST approach using SST anomalies from CFSv2 as input. For the first time, GEFSv12 will provide ensemble based operational weather predictions for sub-seasonal scales with daily 00z forecasts going out to 35 days. GEFSv12 also comes with 20-year reanalysis, 30-year reforecasts and 3-year retrospective forecasts to support stakeholder needs for calibration and validation. In addition, GEFSv12 absorbs the global wave ensembles and aerosol capabilities (control member only) through one-way coupling, taking major steps towards building a unified system and simplifying NCEP’s production suite. This presentation describes the design and development of GEFSv12 and discusses results from the evaluation of the retrospective and reforecast experiments. Significant improvements were noted in both deterministic and probabilistic forecast metrics for several variables including 500 hPa geopotential height anomaly correction, 850 hPa temperature and winds, near surface variables, precipitation, tropical cyclone tracks and intensity, and modes of variability including MJO and NAO. Substantial improvements were also noted in the performance of wave ensemble and aerosol predictions. This presentation also describes NOAA’s efforts towards accelerating further development of fully coupled UFS consisting of six component models of the Earth system: the FV3 dynamical core for the atmosphere, MOM6 for the ocean, Noah MP for the land surface, GOCART for aerosols, CICE5/CICE6 for sea ice and WW3 for ocean surface waves. Combined with data assimilation advances, an ambitious goal of unifying both the high-resolution deterministic (GFSv17) and probabilistic (GEFSv13) predictions for global medium range and sub-seasonal time scales is planned to significantly advance the global prediction capabilities at NCEP.