Precipitation Forecasts Research Articles

Evaluation of precipitation forecasting methods and an advanced lightweight model

Abstract Precipitation forecasting is crucial for warning systems and disaster management. This study focuses on deep learning-based methods and categorizes them into three categories: Recurrent Neural Network (RNN-RNN-RNN), Convolutional Neural Network (CNN-CNN-CNN), and CNN-RNN-CNN methods. Then, we conduct a comprehensive evaluation of typical methods in these three categories using the SEVIR precipitation dataset. The results show that RNN-RNN-RNN suffers from instability in long-term forecasts due to error accumulation, CNN-CNN-CNN struggles to capture temporal signals but produces relatively stable forecasts, and CNN-RNN-CNN significantly increases model complexity and inherits the drawbacks of RNN, leading to worse forecasts. Here, we propose an advanced lightweight precipitation forecasting model (ALPF) based on CNN. Experimental results demonstrate that ALPF can effectively forecast spatial-temporal features, maintaining CNN’s feature extraction capabilities while avoiding error accumulation in RNN’s propagation. ALPF achieves long-term stable precipitation forecasts and can better capture large precipitation amounts.

Short-Term Precipitation Radar Echo Extrapolation Method Based on the MS-DD3D-RSTN Network and STLoss Function.

Short-term precipitation forecasting is essential for agriculture, transportation, urban management, and tourism. The radar echo extrapolation method is widely used in precipitation forecasting. To address issues like forecast degradation, insufficient capture of spatiotemporal dependencies, and low accuracy in radar echo extrapolation, we propose a new model: MS-DD3D-RSTN. This model employs spatiotemporal convolutional blocks (STCBs) as spatiotemporal feature extractors and uses the spatial-temporal loss (STLoss) function to learn intra-frame and inter-frame changes for end-to-end training, thereby capturing the spatiotemporal dependencies in radar echo signals. Experiments on the Sichuan dataset and the HKO-7 dataset show that the proposed model outperforms advanced models in terms of CSI and POD evaluation metrics. For 2 h forecasts with 20 dBZ and 30 dBZ reflectivity thresholds, the CSI metrics reached 0.538, 0.386, 0.485, and 0.198, respectively, representing the best levels among existing methods. The experiments demonstrate that the MS-DD3D-RSTN model enhances the ability to capture spatiotemporal dependencies, mitigates forecast degradation, and further improves radar echo prediction performance.

Development of Multiscale EnKF within GSI and Its Applications to Multiple Convective Storm Cases with Radar Reflectivity Data Assimilation Using the FV3 Limited-Area Model

Abstract To improve the representation of all relevant scales in initial conditions for large-domain convection-allowing models, a new multiscale ensemble Kalman filter (MEnKF) algorithm is developed and implemented within the Gridpoint Statistical Interpolation analysis system (GSI) data assimilation framework coupled with the Finite-Volume Cubed-Sphere Dynamical Core (FV3) limited-area model. The algorithm utilizes ensemble background error covariances filtered to match the observations assimilated. This is realized in a sequential manner. 1) When assimilating coarse-resolution observations such as radiosondes, ensemble background perturbations are filtered to remove scales smaller than those the observations can represent, along with relatively large horizontal localization radii to ensure low-noise and balanced analysis increments. 2) The resulting ensemble analyses from the first step then serve as the background to assimilate denser observations such as radar data with smaller localization radii. Several passes can be taken to assimilate all observations. In this paper, vertically increasing horizontal filter scales are used when assimilating rawinsonde and surface observations together, while radar data are assimilated in the second step. The algorithm is evaluated through six convective storm cases during May 2021, with cycled assimilation of either conventional data only or with additional radar reflectivity followed by 24-h ensemble forecasts. Overall, positive impacts of the MEnKF on forecasts are obtained regardless of reflectivity data; its advantage over the single-scale EnKF is most significant in surface humidity and temperature forecasts up to at least 12 h. More accurate hourly precipitation forecasts with MEnKF can last up to 24 h for light rain. Furthermore, MEnKF forecasts higher ensemble probabilities for the observed hazardous events.

Evaluating machine learning models for precipitation prediction in Casablanca City

Accurate precipitation forecasting is a vital task for many domains, such as agriculture, water management, flood prevention, and crop yield estimation. The use of machine learning (ML) approaches has improved precipitation forecasting accuracy, exhibiting promising results in capturing the intricate connections between various meteorological variables and precipitation patterns. However, given the vast array of available ML models, a comparative analysis is imperative for identifying the most effective models for precipitation prediction. This study aims to examine the capacities of ML algorithms to forecast precipitation based on weather data for the city of Casablanca, Morocco, which faces challenges in water management and climate change adaptation. Eight different ML models’ performances are compared: linear regression, polynomial regression, K-nearest neighbors (KNN), support vector machine (SVM), decision tree (DT), random forest (RF), XGBoost, and an ensemble learning model. These models are evaluated based on their mean absolute error (MAE), mean squared error (MSE), and R-squared (R2 ) value to determine their effectiveness. The study showcases the potential of ML models in predicting precipitation by utilizing meteorological parameters such as temperature, humidity, wind speed, and pressure.

Improving inland precipitation forecast in China through data assimilation of microwave temperature sounding data from a three‐orbit constellation

AbstractMicrowave temperature sounders onboard polar‐orbiting satellites can provide global observation data twice a day, supplying a large amount of temperature information for global data assimilation and serving as a crucial instrument to improve operational numerical forecasts. However, regional numerical forecasts are still subject to a lack of polar‐orbiting satellite data within regional model domains, and even multiple polar‐orbiting satellites may simultaneously miss measurements. Establishing a three‐orbit observation system of polar‐orbiting satellites is crucial to improve the spatiotemporal coverage of polar‐orbiting satellite data. In this study, we investigate the impact of assimilating microwave temperature sounding data from a three‐orbit constellation on precipitation forecasts in inland China based on the data from the US afternoon‐orbit satellite NOAA‐19, the European morning‐orbit satellite Meteorological Operational satellite‐A and the Chinese early‐morning‐orbit satellite Fengyun‐3E (FY‐3E) launched recently. The research results indicate that there are data gaps at 0600 and 1800 UTC in the East Asian region only for the morning‐orbit and afternoon‐orbit satellite observations. The FY‐3E satellite can provide additional microwave temperature sounding observations over the eastern region of China, thus partially compensating for the gap in polar‐orbiting satellite data in China. Moreover, the additional assimilation of the FY‐3E data can further improve numerical forecasts, effectively adjusting the spatial structure and eastward movement of the weather system, thereby considerably increasing the prediction accuracy of rainfall location and intensity. Rolling‐prediction results show that the data from the three‐orbit constellation provide a stable and notable improvement in precipitation forecasts in inland China, especially for forecasts longer than nine hours and amounts of rainfall below 10 mm. These research findings provide valuable insights for optimizing the assimilation application of polar‐orbiting satellite data in regional numerical forecasts.

A machine learning model that outperforms conventional global subseasonal forecast models

Skillful subseasonal forecasts are crucial for various sectors of society but pose a grand scientific challenge. Recently, machine learning-based weather forecasting models outperform the most successful numerical weather predictions generated by the European Centre for Medium-Range Weather Forecasts (ECMWF), but have not yet surpassed conventional models at subseasonal timescales. This paper introduces FuXi Subseasonal-to-Seasonal (FuXi-S2S), a machine learning model that provides global daily mean forecasts up to 42 days, encompassing five upper-air atmospheric variables at 13 pressure levels and 11 surface variables. FuXi-S2S, trained on 72 years of daily statistics from ECMWF ERA5 reanalysis data, outperforms the ECMWF’s state-of-the-art Subseasonal-to-Seasonal model in ensemble mean and ensemble forecasts for total precipitation and outgoing longwave radiation, notably enhancing global precipitation forecast. The improved performance of FuXi-S2S can be primarily attributed to its superior capability to capture forecast uncertainty and accurately predict the Madden-Julian Oscillation (MJO), extending the skillful MJO prediction from 30 days to 36 days. Moreover, FuXi-S2S not only captures realistic teleconnections associated with the MJO but also emerges as a valuable tool for discovering precursor signals, offering researchers insights and potentially establishing a new paradigm in Earth system science research.

Improving the Short-Range Precipitation Forecast of Numerical Weather Prediction through a Deep Learning-Based Mask Approach

Improving the Short-Range Precipitation Forecast of Numerical Weather Prediction through a Deep Learning-Based Mask Approach

Reinterpreting ENSO's Role in Modulating Impactful Precipitation Events in California

AbstractWater years (WY) 2017 and 2023 were anomalously wet for California, each alleviating multiyear drought. In both cases, this was unexpected given La Niña conditions, with most seasonal forecasts favoring drier‐than‐normal winters. We analyze over seven decades of precipitation and snow records along with mid‐tropospheric circulation to identify recurring weather patterns driving California precipitation and Sierra Nevada snowpack. Tropical forcing by ENSO causes subtle but important differences in these wet weather patterns, which largely drives the canonical seasonal ENSO‐precipitation relationship. However, the seasonal frequency of these weather patterns is not strongly modulated by ENSO and remains a primary source of uncertainty for seasonal forecasting. Seasonal frequency of ENSO‐independent weather patterns was a major cause of anomalous precipitation in WY2017, record‐setting snow in WY2023, and differences in precipitation outcome during recent El Niño winters 1983, 1998, and 2016. Improved understanding of recurrent atmospheric weather patterns could help to improve seasonal precipitation forecasts.

Effects of Assimilating Ground-Based Microwave Radiometer and FY-3D MWTS-2/MWHS-2 Data in Precipitation Forecasting

This study investigates the impacts of the joint assimilation of ground-based microwave radiometer (MWR) and FY-3D microwave sounder (MWTS-2/MWHS-2) observations on the analyses and forecasts for precipitation forecast. Based on the weather research and forecasting data assimilation (WRFDA) system, four experiments are conducted in this study, concerning a heavy precipitation event in Beijing on 2 July 2021, and 10-day batch experiments were also conducted. The key study findings include the following: (1) Both ground-based microwave radiometer and MWTS-2/MWHS-2 data contribute to improvements in the initial fields of the model, leading to appropriate adjustments in the thermal structure of the model. (2) The forecast fields of the experiments assimilating ground-based microwave radiometer and MWTS-2/MWHS-2 data show temperature and humidity performances closer to the true fields compared with the control experiment. (3) Separate assimilation of two types of microwave radiometer data can improve precipitation forecasts, while joint assimilation provides the most accurate forecasts among all the experiments. In the single-case, compared with the control experiment, the individual and combined assimilation of MWR and MWTS-2/MWHS-2 improves the six-hour cumulative precipitation threat score (TS) at the 25 mm level by 57.1%, 28.9%, and 38.2%, respectively. The combined assimilation also improves the scores at the 50 mm level by 54.4%, whereas individual assimilations show a decrease in performance. In the batch experiments, the MWR_FY experiment’s TS of 24 h precipitation forecast improves 28.5% at 10 mm and 330% at 25 mm based on the CTRL.

Contrasting the Effects of X-Band Phased Array Radar and S-Band Doppler Radar Data Assimilation on Rainstorm Forecasting in the Pearl River Delta

Contrasting the X-band phased array radar (XPAR) with the conventional S-Band dual-polarization mechanical scanning radar (SMSR), the XPAR offers superior temporal and spatial resolution, enabling a more refined depiction of the internal dynamics within convective systems. While both SMSR and XPAR data are extensively used in monitoring and alerting for severe convective weather, their comparative application in numerical weather prediction through data assimilation remains a relatively unexplored area. This study harnesses the Weather Research and Forecasting Model (WRF) and its data assimilation system (WRFDA) to integrate radial velocity and reflectivity from the Guangzhou SMSR and nine XPARs across Guangdong Province. Utilizing a three-dimensional variational approach at a 1 km convective-scale grid, the assimilated data are applied to forecast a rainstorm event in the Pearl River Delta (PRD) on 6 June 2022. Through a comparative analysis of the results from assimilating SMSR and XPAR data, it was observed that the assimilation of SMSR data led to more extensive adjustments in the lower- and middle-level wind fields compared to XPAR data assimilation. This resulted in an enlarged convergence area at lower levels, prompting an overdevelopment of convective systems and an excessive concentration of internal hydrometeor particles, which in turn led to spurious precipitation forecasts. However, the sequential assimilation of both SMSR and XPAR data effectively reduced the excessive adjustments in the wind fields that were evident when only SMSR data were used. This approach diminished the generation of false echoes and enhanced the precision of quantitative precipitation forecasts. Additionally, the lower spectral width of XPAR data indicates its superior detection accuracy. Assimilating XPAR data alone yields more reasonable adjustments to the low- to middle-level wind fields, leading to the formation of small-to-medium-scale horizontal convergence lines in the lower levels of the analysis field. This enhancement significantly improves the model’s forecasts of composite reflectivity and radar echoes, aligning them more closely with actual observations. Consequently, the Threat Score (TS) and Equitable Threat Score (ETS) for heavy-rain forecasts (>10 mm/h) over the next 5 h are markedly enhanced. This study underscores the necessity of incorporating XPAR data assimilation in numerical weather prediction practices and lays the groundwork for the future joint assimilation of SMSR and XPAR data.

Assessing Multi-Scale Atmospheric Circulation Patterns for Improvements in Sub-Seasonal Precipitation Predictability in the Northern Great Plains

This study leverages the relationships between the Great Plains low-level jet (GP-LLJ) and the circumglobal teleconnection (CGT) to assess the enhancement of 30-day rainfall forecast in the Northern Great Plains (NGP). The assessment of 30-day simulated precipitation using the Climate Forecast System (CFS) is contrasted with the North American Regional Reanalysis, searching for sources of precipitation predictability associated with extended wet and drought events. We analyze the 30-day sources of precipitation predictability using (1) the characterization of dominant statistical modes of variability of 900 mb winds associated with the GP-LLJ, (2) the large-scale atmospheric patterns based on 200 mb geopotential height (HGT), and (3) the use of GP-LLJ and CGT conditional probability distributions using a continuous correlation threshold approach to identify when and where the forecast of NGP precipitation occurs. Two factors contributing to the predictability of precipitation in the NGP are documented. We found that the association between GP-LLJ and CGT occurs at two different scales—the interdiurnal and the sub-seasonal, respectively. The CFS reforecast suggests that the ability to forecast sub-seasonal precipitation improves in response to the enhanced simulation of the GP-LLJ and CGT. Using these modes of climate variability could improve predictive frameworks for water resources management, governance, and water supply for agriculture.

Assessment of Subseasonal-to-Seasonal (S2S) Precipitation Forecast Skill for Reservoir Operation in the Yaque Del Norte River, Dominican Republic

Operational forecasters desire information about how their reservoir and riverine systems will evolve over monthly to seasonal timescales. Seasonal traces of hydrometeorological variables at a daily or sub-daily resolution are needed to drive hydrological models at this timescale. Operationally available models such as the Climate Forecast System (CFS) provide seasonal precipitation forecasts, but their coarse spatial scale requires further processing for use in local or regional hydrologic models. We focus on three methods to generate such forecasts: (1) a bias-adjustment method, in which the CFS forecasts are bias-corrected by ground-based observations; (2) a weather generator (WG) method, in which historical precipitation data, conditioned on an index of the El Niño–Southern Oscillation, are used to generate synthetic daily precipitation time series; and (3) a historical analog method, in which the CFS forecasts are used to condition the selection of historical satellite-based mean areal precipitation (MAP) traces. The Yaque del Norte River basin in the Dominican Republic is presented herein as a case study, using an independent dataset of rainfall and reservoir inflows to assess the relative performance of the methods. The methods showed seasonal variations in skill, with the MAP historical analog method having the strongest overall performance, but the CFS and WG methods also exhibited strong performance during certain seasons. These results indicate that the strengths of each method may be combined to produce an ensemble forecast product.

Probabilistic assessment of postfire debris-flow inundation in response to forecast rainfall

Abstract. Communities downstream of burned steep lands face increases in debris-flow hazards due to fire effects on soil and vegetation. Rapid postfire hazard assessments have traditionally focused on quantifying spatial variations in debris-flow likelihood and volume in response to design rainstorms. However, a methodology that provides estimates of debris-flow inundation downstream of burned areas based on forecast rainfall would provide decision-makers with information that directly addresses the potential for downstream impacts. We introduce a framework that integrates a 24 h lead-time ensemble precipitation forecast with debris-flow likelihood, volume, and runout models to produce probabilistic maps of debris-flow inundation. We applied this framework to simulate debris-flow inundation associated with the 9 January 2018 debris-flow event in Montecito, California, USA. When the observed debris-flow volumes were used to drive the probabilistic forecast model, analysis of the simulated inundation probabilities demonstrates that the model is both reliable and sharp. In the fully predictive model, however, in which debris-flow likelihood and volume were computed from the atmospheric model ensemble's predictions of peak 15 min rainfall intensity, I15, the model generally under-forecasted the inundation area. The observed peak I15 lies in the upper tail of the atmospheric model ensemble spread; thus a large fraction of ensemble members forecast lower I15 than observed. Using these I15 values as input to the inundation model resulted in lower-than-observed flow volumes which translated into under-forecasting of the inundation area. Even so, approximately 94 % of the observed inundated area was forecast to have an inundation probability greater than 1 %, demonstrating that the observed extent of inundation was generally captured within the range of outcomes predicted by the model. Sensitivity analyses indicate that debris-flow volume and two parameters associated with debris-flow mobility exert significant influence on inundation predictions, but reducing uncertainty in postfire debris-flow volume predictions will have the largest impact on reducing inundation outcome uncertainty. This study represents a first step toward a near-real-time hazard assessment product that includes probabilistic estimates of debris-flow inundation and provides guidance for future improvements to this and similar model frameworks by identifying key sources of uncertainty.

Comparative analysis of precipitable water vapour data in the Tarim Basin, China

Atmospheric water vapour is an important contributor to global energy and water cycles and its detection is necessary for precipitation forecasting and climate change research. To explore the applicability of reanalysis data for determining precipitable water vapour (PWV) content in the Tarim Basin, China, this study compared the reliability of two reanalysis PWV products, ERA-5 (PWERA) and MERRA-2 (PWMER), using data from four ground-based global navigation satellite system stations representing basin ecosystems (Oasis, Gobi Desert, Central Desert, and Alpine; dataset: PWGNSS) as reference values. We conducted correlation analysis between PWGNSS and both reanalysis PWV products at different time scales to verify the products’ accuracy. The results showed that the reanalysis PWV products were mostly overestimated at the Desert station, while they were mostly underestimated at the Oasis station. The applicability of PWERA and PWMER varied by season and location, with better applicability at the Oasis station from spring to autumn. The applicability of reanalysis PWV products was lower during precipitation periods than during non-precipitation, and varied by location. During non-precipitation situation, the PWMER had better applicability at the Oasis station and at the Central Desert station when the PWGNSS exceeded 25 mm. Meanwhile, PWERA had better applicability in precipitation situation at the Gobi Desert station from April to June and at Oasis station from August to September when PWGNSS was above 30 mm, at the Central Desert station from May to August, and when the PWGNSS exceeded 30 mm.

Aerosol impacts on summer precipitation forecast over the North China Plain by using Thompson aerosol aware scheme in WRF: Process analysis and response sensitivity to aerosol concentrations during a Heavy Rainfall Event in Beijing

Aerosol-Cloud-Interactions (ACIs) are possibly important factors affecting precipitation while are usually not considered in numerical weather prediction systems (NWP) due to the complexities and uncertainties involved by considering aerosol information. This research focused on the use of the Thompson aerosol-aware scheme in WRF (Weather Research and Forecasting model). This paper summarized the process analysis and response sensitivity to aerosol concentrations during a heavy rainfall event in Beijing. The July 16, 2018 rainfall event -a typical precipitation case involving both warm and cold-phase clouds was thoroughly analyzed, including cloud fraction, downward shortwave radiation, water vapor consumption, cloud microphysical processes, net latent heat etc., which verified that the aerosol-cloud interaction paths were accurately and adequately considered in Thompson aerosol-ware scheme. By tracking key parameter changes, it is confirmed that the CCN/IN activation and relative roles in changing precipitation at different stages are reasonable in Thompson aerosol-aerosol scheme. According to the rainfall characteristics, the precipitation was divided into two phases. And the difference in the aerosol effect on precipitation in the two precipitation phases is mainly due to the complexity of cold rain process and hypothesis. Sensitivity simulations also revealed that the current concentrations of background aerosol in Beijing are actually high enough and further ten-times of aerosol concentration increase didn't bring significant additional precipitation response.

Quantifying and reducing flood forecast uncertainty by the CHUP-BMA method

Abstract. The Bayesian model averaging (BMA), hydrological uncertainty processor (HUP), and HUP-BMA methods have been widely used to quantify flood forecast uncertainty. This study proposes the copula-based hydrological uncertainty processor BMA (CHUP-BMA) method by introducing a copula-based HUP in the framework of BMA to bypass the need for a normal quantile transformation of the HUP-BMA method. The proposed ensemble forecast scheme consists of eight members (two forecast precipitation inputs; two advanced long short-term memory, LSTM, models; and two objective functions used to calibrate parameters) and is applied to the interval basin between the Xiangjiaba and Three Gorges Reservoir (TGR) dam sites. The ensemble forecast performance of the HUP-BMA and CHUP-BMA methods is explored in the 6–168 h forecast horizons. The TGR inflow forecasting results show that the two methods can improve the forecast accuracy over the selected member with the best forecast accuracy and that the CHUP-BMA performs much better than the HUP-BMA. Compared with the HUP-BMA method, the forecast interval width and continuous ranked probability score metrics of the CHUP-BMA method are reduced by a maximum of 28.42 % and 17.86 % within all forecast horizons, respectively. The probability forecast of the CHUP-BMA method has better reliability and sharpness and is more suitable for flood ensemble forecasts, providing reliable risk information for flood control decision-making.

Skilful probabilistic medium‐range precipitation and temperature forecasts over Vietnam for the development of a future dengue early warning system

AbstractDengue fever is a source of substantial health burden in Vietnam. Given the well‐established influence of temperature and precipitation on vector biology and disease transmission, predictions of meteorological variables, such as those issued by ECMWF as a world‐leading provider of global ensemble forecasts, are likely to be valuable model inputs to a future dengue early warning system. In the absence of established verification at municipal and regional scales, this study assesses the skill of rainy season (May–October) ensemble precipitation and 2‐m temperature retrospective forecasts over North and South Vietnam initialized for dates during the period 2001–2020, evaluated against the ERA5 reanalysis for the same period. Forecasts are found to be significantly skilful compared with both climatology and persistence for lead times up to 10 days, including for cumulative precipitation values considered against independent rain gauge data. Rank histograms demonstrate that ensembles generally avoid excessive bias and consistently positive CRPSS values indicate substantial skill for temperature and cumulative precipitation forecasts for all spatial scales considered, despite differences in rainy season characteristics between North and South Vietnam. This forecast reliability demonstrates that meteorological input data based on ECMWF ensemble forecasts would add appreciably more value to the development of a future dengue early warning system compared to reference forecasts like climatology or persistence. These results raise hope for further exploration of predictive skill for relevant meteorological variables, particularly focused on their downscaling to produce district‐level epidemiological forecasts for urban areas where dengue is most prevalent.

Impacts of Moisture Advection Scheme on Precipitation in the Steep Topography Region between the Tibetan Plateau and the Sichuan Basin

Abstract Water vapor transport is a crucial process in modeling and can contribute to errors in precipitation forecasts. To investigate the sensitivity of precipitation to the moisture advection scheme, this study introduced the two-step shape-preserving advection scheme (TSPAS), which has been proven to improve precipitation simulation over steep topography at lower resolutions, into the Southwest Center Weather Research and Forecast (WRF)-based Intelligent Numerical Grid Forecast System (SWC-WINGS) at a convection-permitting resolution. According to experiments conducted throughout the summer of 2021, the precipitation over the eastern slope of the Tibetan Plateau (TP) is highly sensitive to the moisture advection scheme. TSPAS successfully improved precipitation over the eastern slope of the TP, especially for torrential rainfall. The fractions skill score (FSS) is improved by 0.075 (27.78%) for daily precipitation with a threshold of 100 mm. Compared with the experiment with the original WRF advection scheme, the TSPAS reduced the overestimation of precipitation in the topographic region and excessive water vapor transport in a low-level atmosphere. To understand the precipitation improvement contributed by the advection scheme, additional experiments were conducted for a particular precipitation process from two approaches: switching advection schemes during the rainfall evolution and updating the variables related to moisture advection individually. Results demonstrate that the precipitation improvement is mainly contributed by the moisture advection scheme before the precipitation. Among the different variables, the combination of wind and water vapor was the most influential factor causing the precipitation improvement under the TSPAS.

Deep Learning of a 200-Member Ensemble with a Limited Historical Training to Improve the Prediction of Extreme Precipitation Events

Abstract This study introduces a deep learning (DL) scheme to generate reliable and skillful probabilistic quantitative precipitation forecasts (PQPFs) in a postprocessing framework. Enhanced machine learning model architecture and training mechanisms are proposed to improve the reliability and skill of PQPFs while permitting computationally efficient model fitting using a short training dataset. The methodology is applied to postprocessing of 24-h accumulated PQPFs from an ensemble forecast system recently introduced by the Center for Western Weather and Water Extremes (CW3E) and for lead times from 1 to 6 days. The ensemble system was designed based on a high-resolution version of the Weather Research and Forecasting (WRF) Model, named West-WRF, to produce a 200-member ensemble in near–real time (NRT) over the western United States during the boreal cool seasons to support Forecast-Informdayed Reservoir Operations (FIRO) and studies of prediction of heavy-to-extreme events. Postprocessed PQPFs are compared with those from the raw West-WRF ensemble, the operational Global Ensemble Forecast System version 12 (GEFSv12), and the ensemble from the European Centre for Medium-Range Weather Forecasts (ECMWF). As an additional baseline, we provide PQPF verification metrics from a recently developed neural network postprocessing scheme. The results demonstrate that the skill of postprocessed forecasts significantly outperforms PQPFs and deterministic forecasts from raw ensembles and the recently developed algorithm. The resulting PQPFs broadly improve upon the reliability and skill of baselines in predicting heavy-to-extreme precipitation (e.g., >75 mm) across all lead times while maintaining the spatial structure of the high-resolution raw ensemble.

Improving Flood Forecasting Skill by Combining Ensemble Precipitation Forecasts and Multiple Hydrological Models in a Mountainous Basin

Ensemble precipitation forecasts (EPFs) derived from single numerical weather predictions (NWPs) often miss extreme events, and individual hydrological models (HMs) often fail to accurately capture all types of flows, including flood peaks. To address these shortcomings, this study introduced four “EPF + HM” schemes for ensemble flood forecasting (EFF) by combining two EPFs and two HMs. A generator-based post-processing (GPP) method was applied to correct biases and under-dispersion within the raw EPF data. The effectiveness of these schemes in delivering high-quality flood forecasts was assessed using both deterministic and probabilistic metrics. The results indicate that, once post-processed by GPP, all proposed schemes show improvements in both deterministic and probabilistic performances, with skillful flood forecasts for 1–7 lead days. The deterioration in forecast performance with extended lead times is also lessened. Notably, the results indicate that uncertainty within hydrological models has a more pronounced impact on capturing flood peaks than uncertainty in precipitation inputs. This study recommends combining individual EPF with multiple hydrological models for reliable flood forecasting. In conclusion, effective flood forecasting necessitates employing post-processing techniques to correct EPFs and accounting for the uncertainty inherent in hydrological models, rather than relying solely on the uncertainty of the input data.