Forecast Experiments Research Articles

Combining Machine Learning and Numerical Simulation for High-Resolution PM2.5 Concentration Forecast.

Forecasting ambient PM2.5 concentrations with spatiotemporal coverage is key to alerting decision makers of pollution episodes and preventing detrimental public exposure, especially in regions with limited ground air monitoring stations. The existing methods rely on either chemical transport models (CTMs) to forecast spatial distribution of PM2.5 with nontrivial uncertainty or statistical algorithms to forecast PM2.5 concentration time series at air monitoring locations without continuous spatial coverage. In this study, we developed a PM2.5 forecast framework by combining the robust Random Forest algorithm with a publicly accessible global CTM forecast product, NASA's Goddard Earth Observing System "Composition Forecasting" (GEOS-CF), providing spatiotemporally continuous PM2.5 concentration forecasts for the next 5 days at a 1 km spatial resolution. Our forecast experiment was conducted for a region in Central China including the populous and polluted Fenwei Plain. The forecast for the next 2 days had an overall validation R2 of 0.76 and 0.64, respectively; the R2 was around 0.5 for the following 3 forecast days. Spatial cross-validation showed similar validation metrics. Our forecast model, with a validation normalized mean bias close to 0, substantially reduced the large biases in GEOS-CF. The proposed framework requires minimal computational resources compared to running CTMs at urban scales, enabling near-real-time PM2.5 forecast in resource-restricted environments.

Optimization and Evaluation of SO2 Emissions Based on WRF-Chem and 3DVAR Data Assimilation

Emission inventories are important for modeling studies and policy-making, but the traditional “bottom-up” emission inventories are often outdated with a time lag, mainly due to the lack of accurate and timely statistics. In this study, we developed a “top-down” approach to optimize the emission inventory of sulfur dioxide (SO2) using the Weather Research and Forecasting model coupled with Chemistry (WRF-Chem) and a three-dimensional variational (3DVAR) system. The observed hourly surface SO2 concentrations from the China National Environmental Monitoring Center were assimilated and used to estimate the gridded concentration forecast errors of WRF-Chem. The concentration forecast errors were then converted to the emission errors by assuming a linear response from SO2 emission to concentration by grids. To eliminate the effects of modelling errors from aspects other than emissions, a strict data-screening process was conducted. Using the Multi-Resolution Emission Inventory for China (MEIC) 2010 as the a priori emission, the emission inventory for October 2015 over Mainland China was optimized. Two forecast experiments were conducted to evaluate the performance of the SO2 forecast by using the a priori (control experiment) and optimized emissions (optimized emission experiment). The results showed that the forecasts with optimized emissions typically outperformed the forecasts with 2010 a priori emissions in terms of the accuracy of the spatial and temporal distributions. Compared with the control experiment, the bias and root-mean-squared error (RMSE) of the optimized emission experiment decreased by 71.2% and 25.9%, and the correlation coefficients increased by 50.0%. The improvements in Southern China were more significant than those in Northern China. For the Sichuan Basin, Yangtze River Delta, and Pearl River Delta, the bias and RMSEs decreased by 76.4–94.2% and 29.0–45.7%, respectively, and the correlation coefficients increased by 23.5–53.4%. This SO2 emission optimization methodology is computationally cost-effective.

Household appliance usage recommendation based on demand forecasting and multi­objective optimization

Accelerated population growth in the 21st century and increased demand for energy sources, associated with climate change, have resulted in two main challenges: the search for sustainable energy sources and the need to find more efficient ways to use extant sustainable sources. The forecasting module provides an estimate of the future usage of these appliances and it is the source of the recommended module’s suggestion. Time Series Forecasting techniques, such as Autoregressive Integrated Moving Average, Long­Short Term Memory (LSTM), Gated Recurrent Units, Echo State Networks (ESN), and Support Vector Regression, were tested for the predictive module. Multi­objective optimization techniques such as Non­Sorted Genetic Algorithm II (NSGA II), Multi­Objective Particle Swarm Optimization (MOPSO), Speed constrained Multi-­objective Particle Swarm Optimization (SMOPSO), and Strength Pareto Evolutionary Algorithm two (SPEA2), for example, were tested for the Recommendation Module. The Forecasting and Recommendation module experiments were performed independently. In the Forecasting Module, the results and statistical tests revealed LSTM as the best­ suited technique for forecasting the loads of the majority of the appliances tested (in this case seven) in terms of root mean square error. In the experiments performed for the recommendation module, NSGA II showed a higher overall performance compared to other metrics in terms of hyper volume of the Pareto Front generated. This work presents the potential of using both Predictive Models and Multi­Objective Optimization Techniques combined to reduce energy usage in household environments.

A Convolutional Transformer Model for Multivariate Time Series Prediction

This paper proposes a multivariate time series prediction framework based on a transformer model consisting of convolutional neural networks (CNNs). The proposed model has a structure that extracts temporal features of input data through a CNN and interprets the correlation between variables through an attention mechanism. This framework solves the problem of the inability to simultaneously analyze the temporal features of the input data and the correlation between variables, which is a limitation of the forecasting models presented in existing studies. We designed a forecasting experiment using several time series datasets with various data characteristics to precisely evaluate the proposed model. In addition, comparative experiments were performed between the proposed model and several predictive models proposed in recent studies. Furthermore, we conducted ablation studies on the extent to which the proposed CNN structure in the prediction model affects the forecasting results by substituting a specific layer of the model. The results of the experiments showed that the proposed predictive model exhibited good performance in predicting time series data with a clear cycle and high correlation between variables, and improved the accuracy by approximately 3% to 5% compared with that of previous studies’ time series prediction models.

A Probabilistic Forecast Experiment and Its Verification for Relative Humidity and Leaf Wetness with Downscaling Data of One-Week Ensemble Prediction System of the Japan Meteorological Agency

A Probabilistic Forecast Experiment and Its Verification for Relative Humidity and Leaf Wetness with Downscaling Data of One-Week Ensemble Prediction System of the Japan Meteorological Agency

A No-Code Machine Learning System for Next Day Equity and Crypto Forecasting Experiments

A No-Code Machine Learning System for Next Day Equity and Crypto Forecasting Experiments

Ensemble Forecasting Experiments Using the Breeding of Growing Modes with Perturbed Land Variables

Although land surface influences atmospheric processes significantly, insufficient studies have been conducted on the ensemble forecasts using the breeding of growing modes (BGM) with perturbed land surface variables. To investigate the practicability of perturbed land variables for ensemble forecasting, we used the ARWv3 mesoscale model to generate ensembles for an event of 24 h heavy rainfall with perturbed atmospheric and land variables by the BGM method. Results show that both atmospheric and land variables can generate initial perturbations with BGM, except that they differ in time and saturation characteristics, e.g., saturation is generally achieved in approximately 30 h with a growth rate of ~1.30 for atmospheric variables versus 102 h and growth rate of 1.02 for land variables. With the increase in precipitation, the importance of the perturbations of land variables also increases as compared to those of atmospheric variables. Moreover, the influence of the perturbations of land variables on simulated precipitation is still relatively large, although smaller than that of atmospheric variables, e.g., the spreads of perturbed atmospheric and land subsets were 7.3 and 3.8 mm, respectively. The benefits of perturbed initialisation can also be observed in terms of probability forecast. All findings indicate that the BGM method with perturbed land variables has the potential to ensemble forecasts for precipitation.

Assessment of Storm‐Scale Real Time Assimilation of GOES‐16 GLM Lightning‐Derived Water Vapor Mass on Short Term Precipitation Forecasts During the 2020 Spring Forecast Experiment

AbstractThis study assesses the impact of assimilating pseudo‐observations for water vapor mass derived from the Geostationary Operational Environmental Satellites GOES‐16 lightning data on short‐term quantitative precipitation forecasts (QPFs) over the contiguous United States (CONUS), with an emphasis given to regions characterized by an overall poor radar coverage. The GOES‐16/17 provide nearly uniform and high temporal resolution total lightning observations over most of the Americas. To leverage this information for convective scale weather forecasts, a three‐dimensional variational data assimilation (DA) package developed by the National Severe Storm Laboratory was employed to assimilate these data. To mimic operational settings, the Weather Research and Forecasting Model configuration used in NOAA Global Systems Laboratory's High‐Resolution Rapid Refresh Model version 4 was utilized. During the 2020 NOAA Hazardous Weather Testbed Spring Forecasting Experiment, four experiments were run in real time during a 29‐day period over CONUS and surrounding territories to assess the added value of GOES‐16 lightning over conventional radar data. Overall, the lightning DA (LDA) showed benefit in improving QPFs up to 6 hr, with the best improvements seen during the first 3 hr. Owing to notably larger lightning activity over the eastern CONUS, the most noticeable impacts from the LDA were seen there. Case‐by‐case analysis revealed that positive impacts from the LDA were seen over areas characterized by both good (eastern two thirds of CONUS) and poor radar coverage (western one third of CONUS). Highlighting the inherent difficulties in developing an optimal observation operator based on moisture for LDA applications, the GLM‐based DA experiments systematically overestimated rainfall over the eastern two thirds of CONUS while underestimating it over the western one third of CONUS.

Cross-temporal forecast reconciliation: Optimal combination method and heuristic alternatives

Forecast reconciliation is a post-forecasting process aimed to improve the quality of the base forecasts for a system of hierarchical/grouped time series. Cross-sectional and temporal hierarchies have been considered in the literature, but generally, these two features have not been fully considered together. The paper presents two new results by adopting a notation that simultaneously deals with both forecast reconciliation dimensions. (i) The closed-form expression of the optimal (in the least squares sense) point forecasts fulfilling both contemporaneous and temporal constraints. (ii) An iterative procedure that produces cross-temporally reconciled forecasts by alternating forecast reconciliation along one single dimension (either cross-sectional or temporal) at each iteration step. The feasibility of the proposed procedures, along with first evaluations of their performance as compared to the most performing ‘single dimension’ (either cross-sectional or temporal) forecast reconciliation procedures, is studied through a forecasting experiment on the 95 quarterly time series of the Australian Gross Domestic Product from Income and Expenditure sides. For this dataset, the new procedures, in addition to providing fully coherent forecasts in both cross-sectional and temporal dimensions, improve the forecast accuracy of the state-of-the-art point forecast reconciliation techniques.

A comparison of combined data assimilation and machine learning methods for offline and online model error correction

Recent studies have shown that it is possible to combine machine learning methods with data assimilation to reconstruct a dynamical system using only sparse and noisy observations of that system. The same approach can be used to correct the error of a knowledge-based model. The resulting surrogate model is hybrid, with a statistical part supplementing a physical part. In practice, the correction can be added as an integrated term (i.e. in the model resolvent) or directly inside the tendencies of the physical model. The resolvent correction is easy to implement. The tendency correction is more technical, in particular it requires the adjoint of the physical model, but also more flexible. We use the two-scale Lorenz model to compare the two methods. The accuracy in long-range forecast experiments is somewhat similar between the surrogate models using the resolvent correction and the tendency correction. By contrast, the surrogate models using the tendency correction significantly outperform the surrogate models using the resolvent correction in data assimilation experiments. Finally, we show that the tendency correction opens the possibility to make online model error correction, i.e. improving the model progressively as new observations become available. The resulting algorithm can be seen as a new formulation of weak-constraint 4D-Var. We compare online and offline learning using the same framework with the two-scale Lorenz system, and show that with online learning, it is possible to extract all the information from sparse and noisy observations.

Evaluation of an experimental Warn‐on‐Forecast 3DVAR analysis and forecast system on quasi‐real‐time short‐term forecasts of high‐impact weather events

AbstractThe purpose of this study is to demonstrate the capability of an experimental, weather‐adaptive, high‐resolution, deterministic Warn‐on‐Forecast (WoF) analysis and forecast system (WoF3DVAR‐AFS) for predicting high‐impact severe weather events that occurred during the Hazardous Weather Testbed 2019 Spring Forecast Experiments. WoF3DVAR‐AFS uses a three‐dimensional variational (3DVAR) method as its core data assimilation system and the Advanced Research Version of the Weather Research and Forecasting (WRF‐ARW) model as its forward model. Surface measurements provided in meteorological aviation reports and the Oklahoma Mesonet, Doppler radar data, and spaceborne total lightning observations provided by the Geostationary Lightning Mapper are assimilated at 15‐min frequency over a target domain determined by the “Day 1” Convective Outlook product from the Storm Prediction Center. The chief goal of this system is to complement probabilistic forecasts generated by ensemble analysis and forecast systems, such as the experimental Warn‐on‐Forecast System (WoFS) with a higher‐resolution deterministic member to aid forecasters' decision‐making. We performed both qualitative and quantitative evaluations on 0–6 hr forecasts launched hourly from 1900 to 0300 UTC the next day for each of the 12 cases. Aggregated subjective forecast evaluation metrics from each individual case, as well as detailed comparison against available verification datasets, suggest that the forecasts are generally skillful in terms of composite reflectivity fields, quantitative precipitation forecasts, and the strength and location of rotation tracks and damaging winds. This study presents initial efforts to assess the performance of WoF3DVAR‐AFS and provides possible directions for further improvements, including the development of a weather‐adaptive, dual‐resolution analysis and forecast system hybrid with an ensemble system, such as the experimental Warn‐on‐Forecast system.

Lagged ensembles in sub‐seasonal predictions

AbstractThe benefits of using lagged ensembles for S2S predictions have been evaluated using the ECMWF sub‐seasonal to seasonal (S2S) forecasts. The ECMWF S2S forecasts are currently produced twice a week from a 51‐member ensemble. An experiment where the ECMWF forecasts are produced daily instead of twice weekly has been performed to identify the minimum lagged‐ensemble configuration which will produce probabilistic forecasts at least as skilful as the current configuration. Results suggest that the extended‐range prediction of the Madden–Julian Oscillation (MJO) is not deteriorated with the daily lagged ensemble approach, but there is a slight degradation during the first 10 days. In the northern Extratropics, the probabilistic skill scores are improved after week 2 when using lagged ensembles, but only if the daily ensemble size exceeds 20 members. The lagged ensemble approach is more beneficial in the Tropics than in the northern Extratropics, showing even benefits in forecast week 1 over the Tropics, despite the 24 hr difference between two consecutive forecasts. As expected, the daily lagged ensemble method is more beneficial for longer lead times and the optimal lagged ensemble window increases from 2 days for week 2 to 4 days for week 4. An idealized numerical simulation, based on a Monte Carlo approach, confirms the benefits of lagged ensembles. The results obtained are generally consistent with the ECMWF forecast experiment, suggesting that such a simulation could be used as a cheap alternative to address the benefit of lagged versus burst ensembles.

Seasonal prediction skills in the CAMS-CSM climate forecast system

The seasonal prediction skills in the CAMS-CSM (the acronym stands for the Chinese Academy of Meteorological Sciences Climate System Model) climate forecast system is evaluated with a set of retrospective forecast experiments during the period of 1981–2019. The CAMS-CSM, which has been registered for the sixth phase of the coupled model intercomparison project (CMIP6), is an atmosphere–ocean–land–sea ice fully coupled general circulation model. The assimilation scheme used in the forecast system is the 3-dimentional nudging, including both the atmospheric and oceanic components. The analyses mainly focus on the seasonal predictable skill of sea surface temperature, 2-m air temperature, and precipitation anomalies. The analyses revealed that the model shows a good prediction skill for the SST anomalies, especially in the tropical Pacific, in association with El Niño-Southern Oscillation (ENSO) events. The anomaly correlation coefficient (ACC) score for ENSO can reach 0.75 at 6-month lead time. Furthermore, the extreme warm/cold Indian Ocean dipole (IOD) events are successfully predicted at 3- and even 6-month lead times. The whole ACC of IOD events between the observation and the prediction can reach 0.51 at 2-month lead time. There are reliable seasonal prediction skills for 2-m air temperature anomalies over most of the Northern Hemisphere, where the correlation is mainly above 0.4 at 2-month lead time, especially over the East Asia, North America and South America. However, the seasonal prediction for precipitation still faces a big challenge. The source of precipitation predictability over the East Asia can be partly related to strong ENSO events. Additionally, the anomalous anticyclone over the western North Pacific (WPAC) which connects the ENSO events and the East Asian summer monsoon (EASM) can be well predicted at 6-month lead time.

Scale-Dependent Value of QPF for Real-Time Streamflow Forecasting

AbstractIncorporating rainfall forecasts into a real-time streamflow forecasting system extends the forecast lead time. Since quantitative precipitation forecasts (QPFs) are subject to substantial uncertainties, questions arise on the trade-off between the time horizon of the QPF and the accuracy of the streamflow forecasts. This study explores the problem systematically, exploring the uncertainties associated with QPFs and their hydrologic predictability. The focus is on scale dependence of the trade-off between the QPF time horizon, basin-scale, space-time scale of the QPF, and streamflow forecasting accuracy. To address this question, the study first performs a comprehensive independent evaluation of the QPFs at 140 U.S. Geological Survey (USGS) monitored basins with a wide range of spatial scales (~10 – 40,000 km2) over the state of Iowa in the Midwestern United States. The study uses High-Resolution Rapid Refresh (HRRR) and Global Forecasting System (GFS) QPFs for short and medium-range forecasts, respectively. Using Multi-Radar Multi-Sensor (MRMS) quantitative precipitation estimate (QPE) as a reference, the results show that the rainfall-to-rainfall QPF errors are scale-dependent. The results from the hydrologic forecasting experiment show that both QPFs illustrate clear value for real-time streamflow forecasting at longer lead times in the short- to medium-range relative to the no-rain streamflow forecast. The value of QPFs for streamflow forecasting is particularly apparent for basin sizes below 1,000 km2. The space-time scale, or reference time (tr) (ratio of forecast lead time to basin travel time) ~ 1 depicts the largest streamflow forecasting skill with a systematic decrease in forecasting accuracy for tr > 1.

New Observation Operators for Cloud Liquid/Ice Water Path From ABI and Their Impact on Assimilation and Hurricane Forecasts

AbstractThe cloud liquid/ice water path (LWP/IWP) can provide valuable information about the cloud and precipitation. For the first time, the Gridpoint Statistical Interpolation (GSI) system is applied to examine the impact of the high spatial and temporal resolution LWP and IWP products derived from the Advanced Baseline Imager (ABI) on the new generation geostationary environmental satellites (GOES‐R series) in numerical weather prediction (NWP). To extend the GSI's capability to assimilate LWP and IWP products, a pair of new observation operators of LWP and IWP is developed to avoid the requirement for hydrometeor control variables. The new observation operators enable the direct use of hydrometeor background information from the NWP model and allow the vertical propagation of the water vapor increment according to the Jacobian distributions for the new LWP and IWP operators. The new LWP and IWP observation operators are evaluated by the single observation tests and the cycling assimilation and forecast experiments of Hurricane Irma (2017) and Hurricane Maria (2017). The results from single observation tests show that the assimilations of pseudo‐LWP and pseudo‐IWP can achieve positive water vapor increments, which also spread vertically according to the Jacobian distributions. The impact studies on two hurricanes demonstrate that the ABI LWP and IWP measurements can be successfully assimilated into the NWP model by the new observation operators, which enhances the water vapor analysis and further improves other atmospheric state variables in the forecast fields. The improved environmental fields lead to the overall forecast improvement for the two hurricanes, with the most notable track improvement.

Increased impact of the Tibetan Plateau Spring Snow Cover to the Meiyu Rainfall over the Yangtze River Valley after 1990s

AbstractThis study investigated the increased impact of the spring (March-April-May) snow cover extent (SCE) over the western Tibetan Plateau (TP) (SSTP) to the Meiyu rainfall (June-July, JJ) over the Yangtze River valley (YRV) (MRYRV) after 1990s. The correlation between the MRYRV and SSTP is significantly increased from 1970-1992 (P1) to 1993-2015 (P2). In P1, the MRYRV-related SSTP anomalies locate southwest TP which cause a perturbation near the SWJ core and favor an eastward propagation in the form of a wave train. The wave train results in a southward shift of the SWJ over the ocean south of Japan in JJ and exerts a limited effect on the MRYRV. Differently, in P2, the MRYRV-related anomalous SSTP causes an anomalous cooling temperature and upper-level cyclonic system centered over the northwestern TP. The cyclonic system develops and extends eastward to the downstream region with time and reaches coastal East Asia in JJ. The anomalous westerly winds along its south flank cause an enhanced subtropical westerly jet (SWJ) which is accompanied by an anomalous lower-level air convergence and ascent motion near the YRV region, favoring enhanced MRYRV. In addition, the forecast experiments performed with empirical regression models illustrate that the prediction skill of the MRYRV variation is clearly increased in P2 with the additional forecast factor of the SSTP.

Using Machine Learning to Generate Storm-Scale Probabilistic Guidance of Severe Weather Hazards in the Warn-on-Forecast System

AbstractA primary goal of the National Oceanic and Atmospheric Administration Warn-on-Forecast (WoF) project is to provide rapidly updating probabilistic guidance to human forecasters for short-term (e.g., 0–3 h) severe weather forecasts. Postprocessing is required to maximize the usefulness of probabilistic guidance from an ensemble of convection-allowing model forecasts. Machine learning (ML) models have become popular methods for postprocessing severe weather guidance since they can leverage numerous variables to discover useful patterns in complex datasets. In this study, we develop and evaluate a series of ML models to produce calibrated, probabilistic severe weather guidance from WoF System (WoFS) output. Our dataset includes WoFS ensemble forecasts available every 5 min out to 150 min of lead time from the 2017–19 NOAA Hazardous Weather Testbed Spring Forecasting Experiments (81 dates). Using a novel ensemble storm-track identification method, we extracted three sets of predictors from the WoFS forecasts: intrastorm state variables, near-storm environment variables, and morphological attributes of the ensemble storm tracks. We then trained random forests, gradient-boosted trees, and logistic regression algorithms to predict which WoFS 30-min ensemble storm tracks will overlap a tornado, severe hail, and/or severe wind report. To provide rigorous baselines against which to evaluate the skill of the ML models, we extracted the ensemble probabilities of hazard-relevant WoFS variables exceeding tuned thresholds from each ensemble storm track. The three ML algorithms discriminated well for all three hazards and produced more reliable probabilities than the baseline predictions. Overall, the results suggest that ML-based postprocessing of dynamical ensemble output can improve short-term, storm-scale severe weather probabilistic guidance.

Testing the seismic quiescence hypothesis through retrospective trials of alarm-based earthquake prediction in the Kurile\u2013Japan subduction zone

We make trial binary forecasts for the Kurile–Japan subduction zone for the period 1988–2014 by hypothesizing that seismic quiescence (i.e., the absence of earthquakes of M ≥ 5 for a minimum period of Tq) is a precursor of a large (7.5 ≤ Mw < 8.5) earthquake in the coming period Ta within a radius R of the quiescence. We evaluate the receiver-operating-characteristic diagram constructed using a range of forecast models specified by (Tq, R, Ta). A forecast experiment targeting eight large earthquakes in the studied spacetime suggests that the risk of a large earthquake is modestly (probability gain G ~ 2) but significantly (p-value less than 5%) heightened for several years following a long quiescent period of Tq ≥ 9 years, within several tens of kilometers of the quiescence. We then attempt cross-validation, where we use half the data for training [i.e., optimization of (Tq, R, Ta)] and the remaining half for evaluation. With only four target earthquakes available for evaluation of the forecasts in each of the learning and evaluation periods, our forecast scheme did not pass the cross-validation test (with a criterion that the p-value is less than 5%). Hence, we cannot formally deny the possibility that our positive results for the overall period are a ghost arising from over-fitting. However, through detailed comparison of optimal models in the overall test with those in the cross-validation tests, we argue that severe over-fitting is unlikely involved for the modest G of ~ 2 obtained in the overall test. There is thus a reasonable chance that the presently tested type of quiescence will pass the cross-validation test when more target earthquakes become available in the near future. In the meantime, we find that G improves to ~ 5 when target earthquakes are limited to 8 ≤ Mw < 8.5, though we cannot say anything about the possible involvement of over-fitting because we have only three such very large target earthquakes.

Influence of a portable near-surface observing network on experimental ensemble forecasts of deep convection hazards during VORTEX-SE

AbstractEnsemble forecasts are generated with and without the assimilation of near-surface observations from a portable, mesoscale network of StickNet platforms during the Verification and Origins of Rotation in Tornadoes EXperiment – Southeast (VORTEX-SE). Four VORTEX-SE intensive observing periods are selected to evaluate the impact of StickNet observations on forecasts and predictability of deep convection within the southeast United States. StickNet observations are assimilated with an experimental version of the HighResolution RapidRefresh Ensemble (HRRRE) in one experiment, and withheld in a control forecast experiment. Overall, StickNet observations are found to effectively reduce mesoscale analysis and forecast errors of temperature and dewpoint. Differences in ensemble analyses between the two parallel experiments are maximized near the StickNet array and then either propagate away with the mean low-level flow through the forecast period or remain quasi-stationary, reducing local analysis biases. Forecast errors of temperature and dewpoint exhibit periods of improvement and degradation relative to the control forecast, and error increases are largely driven on the storm scale. Convection predictability, measured through subjective evaluation and objective verification of forecast updraft helicity, is driven more by when forecasts are initialized (i.e., more data assimilation cycles with conventional observations) rather than the inclusion of StickNet observations in data assimilation. It is hypothesized that the full impact of assimilating these data is not realized in part due to poor sampling of forecast sensitive regions by the StickNet platforms, as identified through ensemble sensitivity analysis.

Ensemble prediction using a new dataset of ECMWF initial states – OpenEnsemble 1.0

Abstract. Ensemble prediction is an indispensable tool in modern numerical weather prediction (NWP). Due to its complex data flow, global medium-range ensemble prediction has almost exclusively been carried out by operational weather agencies to date. Thus, it has been very hard for academia to contribute to this important branch of NWP research using realistic weather models. In order to open ensemble prediction research up to the wider research community, we have recreated all 50+1 operational IFS ensemble initial states for OpenIFS CY43R3. The dataset (OpenEnsemble 1.0) is available for use under a Creative Commons licence and is downloadable from an https server. The dataset covers 1 year (December 2016 to November 2017) twice daily. Downloads in three model resolutions (TL159, TL399, and TL639) are available to cover different research needs. An open-source workflow manager, called OpenEPS, is presented here and used to launch ensemble forecast experiments from the perturbed initial conditions. The deterministic and probabilistic forecast skill of OpenIFS (cycle 40R1) using this new set of initial states is comprehensively evaluated. In addition, we present a case study of Typhoon Damrey from year 2017 to illustrate the new potential of being able to run ensemble forecasts outside of major global weather forecasting centres.