Toward the Improvement of High-Impact Probabilistic Forecasts with a Sensitivity-Based Convective-Scale Ensemble Subsetting Technique

2020 ◽  
Vol 148 (12) ◽  
pp. 4995-5014
Author(s):  
Austin Coleman ◽  
Brian Ancell
Keyword(s):  
Degrees Of Freedom ◽  
Forecast Accuracy ◽  
High Sensitivity ◽  
Lead Times ◽  
Forecast Errors ◽  
Ensemble Forecasts ◽  
Skill Scores ◽  
Probabilistic Forecasts ◽  
Convective Scale ◽  
Model Initialization

AbstractEnsemble sensitivity analysis (ESA) is a useful and computationally inexpensive tool for analyzing how features in the flow at early forecast times affect different relevant forecast features later in the forecast. Given the frequency of observations measured between model initialization times that remain unused, ensemble sensitivity may be used to increase predictability and forecast accuracy through an objective ensemble subsetting technique. This technique identifies ensemble members with the smallest errors in regions of high sensitivity to produce a smaller, more accurate ensemble subset. Ensemble subsets can significantly reduce synoptic-scale forecast errors, but applying this strategy to convective-scale forecasts presents additional challenges. Objective verification of the sensitivity-based ensemble subsetting technique is conducted for ensemble forecasts of 2–5-km updraft helicity (UH) and simulated reflectivity. Many degrees of freedom are varied to identify the lead times, subset sizes, forecast thresholds, and atmospheric predictors that provide most forecast benefit. Subsets vastly reduce error of UH forecasts in an idealized framework but tend to degrade fractions skill scores and reliability in a real-world framework. Results reveal this discrepancy is a result of verifying probabilistic UH forecasts with storm-report-based observations, which effectively dampens technique performance. The potential of ensemble subsetting and likely other postprocessing techniques is limited by tuning UH forecasts to predict severe reports. Additional diagnostic ideas to improve postprocessing tool optimization for convection-allowing models are discussed.

scholarly journals Characterizing the Atmospheric Conditions Leading to Large Error Growth in Volcanic Ash Cloud Forecasts

2018 ◽  
Vol 57 (4) ◽  
pp. 1011-1019 ◽  
Author(s):  
H. F. Dacre ◽  
N. J. Harvey
Keyword(s):  
Flow Separation ◽  
Volcanic Ash ◽  
Dispersion Model ◽  
Forecast Accuracy ◽  
Large Error ◽  
Ensemble Member ◽  
Forecast Errors ◽  
Atmospheric Conditions ◽  
Wind Fields ◽  
Ensemble Forecasts

ABSTRACTVolcanic ash poses an ongoing risk to safety in the airspace worldwide. The accuracy with which volcanic ash dispersion can be forecast depends on the conditions of the atmosphere into which it is emitted. In this study, meteorological ensemble forecasts are used to drive a volcanic ash transport and dispersion model for the 2010 Eyjafjallajökull eruption in Iceland. From analysis of these simulations, the authors determine why the skill of deterministic-meteorological forecasts decreases with increasing ash residence time and identify the atmospheric conditions in which this drop in skill occurs most rapidly. Large forecast errors are more likely when ash particles encounter regions of large horizontal flow separation in the atmosphere. Nearby ash particle trajectories can rapidly diverge, leading to a reduction in the forecast accuracy of deterministic forecasts that do not represent variability in wind fields at the synoptic scale. The flow‐separation diagnostic identifies where and why large ensemble spread may occur. This diagnostic can be used to alert forecasters to situations in which the ensemble mean is not representative of the individual ensemble‐member volcanic ash distributions. Knowledge of potential ensemble outliers can be used to assess confidence in the forecast and to avoid potentially dangerous situations in which forecasts fail to predict harmful levels of volcanic ash.

scholarly journals Intrinsic Predictability of the 20 May 2013 Tornadic Thunderstorm Event in Oklahoma at Storm Scales

2016 ◽  
Vol 144 (4) ◽  
pp. 1273-1298 ◽  
Author(s):  
Yunji Zhang ◽  
Fuqing Zhang ◽  
David J. Stensrud ◽  
Zhiyong Meng
Keyword(s):  
Weather Prediction ◽  
Numerical Weather Prediction Model ◽  
Lead Times ◽  
Forecast Errors ◽  
Ensemble Forecasts ◽  
Positive Feedbacks ◽  
Convective Storms ◽  
Cold Pools ◽  
Initial Errors ◽  
Environmental Instability

Abstract Using a high-resolution convection-allowing numerical weather prediction model, this study seeks to explore the intrinsic predictability of the severe tornadic thunderstorm event on 20 May 2013 in Oklahoma from its preinitiation environment to initiation, upscale organization, and interaction with other convective storms. This is accomplished through ensemble forecasts perturbed with minute initial condition uncertainties that were beyond detection capabilities of any current observational platforms. It was found that these small perturbations, too small to modify the initial mesoscale environmental instability and moisture fields, will be propagated and evolved via turbulence within the PBL and rapidly amplified in moist convective processes through positive feedbacks associated with updrafts, phase transitions of water species, and cold pools, thus greatly affecting the appearance, organization, and development of thunderstorms. The forecast errors remain nearly unchanged even when the initial perturbations (errors) were reduced by as much as 90%, which strongly suggests an inherently limited predictability for this thunderstorm event for lead times as short as 3–6 h. Further scale decomposition reveals rapid error growth and saturation in meso-γ scales (regardless of the magnitude of initial errors) and subsequent upscale growth into meso-β scales.

Influence of the Madden–Julian Oscillation on Forecasts of Extreme Precipitation in the Contiguous United States

2011 ◽  
Vol 139 (2) ◽  
pp. 332-350 ◽  
Author(s):  
Charles Jones ◽  
Jon Gottschalck ◽  
Leila M. V. Carvalho ◽  
Wayne Higgins
Keyword(s):  
United States ◽  
Extreme Precipitation ◽  
Western Hemisphere ◽  
Boreal Winter ◽  
Lead Times ◽  
Madden Julian Oscillation ◽  
Extreme Precipitation Events ◽  
Skill Scores ◽  
Probabilistic Forecasts ◽  
Precipitation Events

Abstract Extreme precipitation events are among the most devastating weather phenomena since they are frequently accompanied by loss of life and property. This study uses reforecasts of the NCEP Climate Forecast System (CFS.v1) to evaluate the skill of nonprobabilistic and probabilistic forecasts of extreme precipitation in the contiguous United States (CONUS) during boreal winter for lead times up to two weeks. The CFS model realistically simulates the spatial patterns of extreme precipitation events over the CONUS, although the magnitudes of the extremes in the model are much larger than in the observations. Heidke skill scores (HSS) for forecasts of extreme precipitation at the 75th and 90th percentiles showed that the CFS model has good skill at week 1 and modest skill at week 2. Forecast skill is usually higher when the Madden–Julian oscillation (MJO) is active and has enhanced convection occurring over the Western Hemisphere, Africa, and/or the western Indian Ocean than in quiescent periods. HSS greater than 0.1 extends to lead times of up to two weeks in these situations. Approximately 10%–30% of the CONUS has HSS greater than 0.1 at lead times of 1–14 days when the MJO is active. Probabilistic forecasts for extreme precipitation events at the 75th percentile show improvements over climatology of 0%–40% at 1-day lead and 0%–5% at 7-day leads. The CFS has better skill in forecasting severe extremes (i.e., events exceeding the 90th percentile) at longer leads than moderate extremes (75th percentile). Improvements over climatology between 10% and 30% at leads of 3 days are observed over several areas across the CONUS—especially in California and in the Midwest.

scholarly journals Assessing the Ensemble Predictability of Precipitation Forecasts for the January 2015 and 2016 East Coast Winter Storms

2017 ◽  
Vol 32 (3) ◽  
pp. 1057-1078 ◽  
Author(s):  
Steven J. Greybush ◽  
Seth Saslo ◽  
Richard Grumm
Keyword(s):  
East Coast ◽  
Initial Conditions ◽  
Wave Pattern ◽  
Forecast Errors ◽  
Ensemble Forecasts ◽  
Winter Storms ◽  
Position Errors ◽  
Skill Scores ◽  
Forecast Lead Time ◽  
East Coast Winter Storms

Abstract The ensemble predictability of the January 2015 and 2016 East Coast winter storms is assessed, with model precipitation forecasts verified against observational datasets. Skill scores and reliability diagrams indicate that the large ensemble spread produced by operational forecasts was warranted given the actual forecast errors imposed by practical predictability limits. For the 2015 storm, uncertainties along the western edge’s sharp precipitation gradient are linked to position errors of the coastal low, which are traced to the positioning of the preceding 500-hPa wave pattern using the ensemble sensitivity technique. Predictability horizon diagrams indicate the forecast lead time in terms of initial detection, emergence of a signal, and convergence of solutions for an event. For the 2016 storm, the synoptic setup was detected at least 6 days in advance by global ensembles, whereas the predictability of mesoscale features is limited to hours. Convection-permitting WRF ensemble forecasts downscaled from the GEFS resolve mesoscale snowbands and demonstrate sensitivity to synoptic and mesoscale ensemble perturbations, as evidenced by changes in location and timing. Several perturbation techniques are compared, with stochastic techniques [the stochastic kinetic energy backscatter scheme (SKEBS) and stochastically perturbed parameterization tendency (SPPT)] and multiphysics configurations improving performance of both the ensemble mean and spread over the baseline initial conditions/boundary conditions (IC/BC) perturbation run. This study demonstrates the importance of ensembles and convective-allowing models for forecasting and decision support for east coast winter storms.

scholarly journals Storm-Scale Data Assimilation and Ensemble Forecasts for the 27 April 2011 Severe Weather Outbreak in Alabama

2015 ◽  
Vol 143 (8) ◽  
pp. 3044-3066 ◽  
Author(s):  
Nusrat Yussouf ◽  
David C. Dowell ◽  
Louis J. Wicker ◽  
Kent H. Knopfmeier ◽  
Dustan M. Wheatley
Keyword(s):  
Short Range ◽  
Type System ◽  
Severe Weather ◽  
Forecast Errors ◽  
Model Errors ◽  
Ensemble Forecasts ◽  
Low Level ◽  
Design Storm ◽  
Probabilistic Forecasts ◽  
Arw Model

Abstract As part of NOAA’s Warn-on-Forecast (WoF) initiative, a multiscale ensemble-based assimilation and prediction system is developed using the WRF-ARW model and DART assimilation software. To evaluate the capabilities of the system, retrospective short-range probabilistic storm-scale (convection allowing) ensemble analyses and forecasts are produced for the 27 April 2011 Alabama severe weather outbreak. Results indicate that the storm-scale ensembles are able to analyze the observed storms with strong low-level rotation at approximately the correct locations and to retain the supercell structures during the 0–1-h forecasts with reasonable accuracy. The system predicts the low-level mesocyclones of significant isolated tornadic supercells that align well with the locations of radar-derived rotation. For cases with multiple interacting storms in close proximity, the system tends to produce more variability in mesocyclone forecasts from one initialization time to the next until the observations show the dominance of one of the cells. The short-range ensemble probabilistic forecasts obtained from this continuous 5-min storm-scale 6-h-long update system demonstrate the potential of a frequently updated, high-resolution NWP system that could be used to extend severe weather warning lead times. This study also demonstrates the challenges associated with developing a WoF-type system. The results motivate future work to reduce model errors associated with storm motion and spurious cells, and to design storm-scale ensembles that better represent typical 1-h forecast errors.

Ensemble weather forecast of precipitation with a stochastic weather generator based on analogues circulation

2020 ◽  
Author(s):  
Meriem Krouma ◽  
Pascal Yiou ◽  
Céline Déandréis ◽  
Soulivanh Thao
Keyword(s):  
Model Development ◽  
Weather Generator ◽  
Lead Times ◽  
The European Union ◽  
Ensemble Forecasts ◽  
Analogue Model ◽  
Data Set ◽  
Stochastic Weather Generator ◽  
Horizon 2020 ◽  
Skill Scores

<p><strong>Abstract</strong></p><p>The aim of this study is to assess the skills of a stochastic weather generator (SWG) to forecast precipitation in Europe. The SWG is based on the random sampling of circulation analogues, which is a simple form of machine learning simulation. The SWG was developed and tested by Yiou and Déandréis (2019) to forecast daily average temperature and the NAO index. Ensemble forecasts with lead times from 5 to 80 days were evaluated with CRPSS scores against climatology and persistence forecasts. Reasonable scores were obtained up to 20 days.  In this study, we adapt the parameters of the analogue SWG to optimize the simulation of European precipitations. We then analyze the performance of this SWG for lead times of 2 to 20 days, with the forecast skill scores used by Yiou and Déandréis (2019). To achieve this objective, the SWG will use ECA&D precipitation data (Haylock. 2002), and the analogues of circulation will be computed from sea-level pressure (SLP) or geopotential heights (Z500) from the NCEP reanalysis. This provides 100-member ensemble forecasts on a daily time increment. We will evaluate the seasonal dependence of the forecast skills of precipitation and the conditional dependence to weather regimes. Comparisons with “real” medium range forecasts from the ECMWF will be performed.</p><p><strong>References</strong></p><p>Yiou, P., and Céline D.. Stochastic ensemble climate forecast with an analogue model. Geoscientific Model Development 12, 2 (2019): 723‑34.</p><p>Haylock, M. R. et al.. A European daily high-resolution gridded data set of surface temperature and precipitation for 1950-2006. J. Geophys. Res. - Atmospheres 113, D20 (2008): doi:10.1029/2008JD010201.</p><p> </p><p><strong>A</strong><strong>cknowledge</strong></p><p>This project has received funding from the European Union’s Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie grant agreement No 813844.</p>

scholarly journals Statistical Postprocessing of Ensemble Precipitation Forecasts by Fitting Censored, Shifted Gamma Distributions*

2015 ◽  
Vol 143 (11) ◽  
pp. 4578-4596 ◽  
Author(s):  
Michael Scheuerer ◽  
Thomas M. Hamill
Keyword(s):  
Probability Distributions ◽  
Precipitable Water ◽  
Training Dataset ◽  
Ensemble Forecasts ◽  
Gamma Distributions ◽  
Predictive Probability ◽  
Skill Scores ◽  
Probabilistic Forecasts ◽  
The Mean ◽  
Set Up

Abstract A parametric statistical postprocessing method is presented that transforms raw (and frequently biased) ensemble forecasts from the Global Ensemble Forecast System (GEFS) into reliable predictive probability distributions for precipitation accumulations. Exploratory analysis based on 12 years of reforecast data and ⅛° climatology-calibrated precipitation analyses shows that censored, shifted gamma distributions can well approximate the conditional distribution of observed precipitation accumulations given the ensemble forecasts. A nonhomogeneous regression model is set up to link the parameters of this distribution to ensemble statistics that summarize the mean and spread of predicted precipitation amounts within a certain neighborhood of the location of interest, and in addition the predicted mean of precipitable water. The proposed method is demonstrated with precipitation reforecasts over the conterminous United States using common metrics such as Brier skill scores and reliability diagrams. It yields probabilistic forecasts that are reliable, highly skillful, and sharper than the previously demonstrated analog procedure. In situations with limited predictability, increasing the size of the neighborhood within which ensemble forecasts are considered as predictors can further improve forecast skill. It is found, however, that even a parametric postprocessing approach crucially relies on the availability of a sufficiently large training dataset.

Implications of Ensemble Quantitative Precipitation Forecast Errors on Distributed Streamflow Forecasting

2010 ◽  
Vol 11 (1) ◽  
pp. 69-86 ◽  
Author(s):  
Giuseppe Mascaro ◽  
Enrique R. Vivoni ◽  
Roberto Deidda
Keyword(s):  
Hydrological Model ◽  
Initial Conditions ◽  
Forecast Accuracy ◽  
Precipitation Forecast ◽  
Forecast Errors ◽  
Distributed Hydrological Model ◽  
Streamflow Forecasting ◽  
Ensemble Forecasts ◽  
Wide Range ◽  
Operational Flow

Abstract Evaluating the propagation of errors associated with ensemble quantitative precipitation forecasts (QPFs) into the ensemble streamflow response is important to reduce uncertainty in operational flow forecasting. In this paper, a multifractal rainfall downscaling model is coupled with a fully distributed hydrological model to create, under controlled conditions, an extensive set of synthetic hydrometeorological events, assumed as observations. Subsequently, for each event, flood hindcasts are simulated by the hydrological model using three ensembles of QPFs—one reliable and the other two affected by different kinds of precipitation forecast errors—generated by the downscaling model. Two verification tools based on the verification rank histogram and the continuous ranked probability score are then used to evaluate the characteristics of the correspondent three sets of ensemble streamflow forecasts. Analyses indicate that the best forecast accuracy of the ensemble streamflows is obtained when the reliable ensemble QPFs are used. In addition, results underline (i) the importance of hindcasting to create an adequate set of data that span a wide range of hydrometeorological conditions and (ii) the sensitivity of the ensemble streamflow verification to the effects of basin initial conditions and the properties of the ensemble precipitation distributions. This study provides a contribution to the field of operational flow forecasting by highlighting a series of requirements and challenges that should be considered when hydrologic ensemble forecasts are evaluated.

scholarly journals A Comparison of Precipitation Forecast Skill between Small Convection-Allowing and Large Convection-Parameterizing Ensembles

2009 ◽  
Vol 24 (4) ◽  
pp. 1121-1140 ◽  
Author(s):  
Adam J. Clark ◽  
William A. Gallus ◽  
Ming Xue ◽  
Fanyou Kong
Keyword(s):  
Wrf Model ◽  
Precipitation Forecast ◽  
Convective Precipitation ◽  
Weather Research And Forecasting ◽  
Lead Times ◽  
Grid Spacing ◽  
Ensemble Forecasts ◽  
Probabilistic Forecasts ◽  
Central United States ◽  
Statistical Consistency

Abstract An experiment has been designed to evaluate and compare precipitation forecasts from a 5-member, 4-km grid-spacing (ENS4) and a 15-member, 20-km grid-spacing (ENS20) Weather Research and Forecasting (WRF) model ensemble, which cover a similar domain over the central United States. The ensemble forecasts are initialized at 2100 UTC on 23 different dates and cover forecast lead times up to 33 h. Previous work has demonstrated that simulations using convection-allowing resolution (CAR; dx ∼ 4 km) have a better representation of the spatial and temporal statistical properties of convective precipitation than coarser models using convective parameterizations. In addition, higher resolution should lead to greater ensemble spread as smaller scales of motion are resolved. Thus, CAR ensembles should provide more accurate and reliable probabilistic forecasts than parameterized-convection resolution (PCR) ensembles. Computation of various precipitation skill metrics for probabilistic and deterministic forecasts reveals that ENS4 generally provides more accurate precipitation forecasts than ENS20, with the differences tending to be statistically significant for precipitation thresholds above 0.25 in. at forecast lead times of 9–21 h (0600–1800 UTC) for all accumulation intervals analyzed (1, 3, and 6 h). In addition, an analysis of rank histograms and statistical consistency reveals that faster error growth in ENS4 eventually leads to more reliable precipitation forecasts in ENS4 than in ENS20. For the cases examined, these results imply that the skill gained by increasing to CAR outweighs the skill lost by decreasing the ensemble size. Thus, when computational capabilities become available, it will be highly desirable to increase the ensemble resolution from PCR to CAR, even if the size of the ensemble has to be reduced.

Sampling errors on convective scales: What can we learn from a 1000-member ensemble?

2020 ◽  
Author(s):  
Tobias Necker ◽  
Martin Weissmann ◽  
Yvonne Ruckstuhl ◽  
Juan Ruiz ◽  
Takemasa Miyoshi ◽  
...  
Keyword(s):  
Degrees Of Freedom ◽  
Initial Conditions ◽  
Sampling Error ◽  
Correction Method ◽  
Sampling Errors ◽  
Atmospheric Models ◽  
Ensemble Forecasts ◽  
Advantages And Disadvantages ◽  
Convective Scale ◽  
High Impact Weather

<p>Current regional forecasting systems particularly aim at the forecast of convective events and related hazards. Most weather centers apply high-resolution ensemble forecasts that resolve convection explicitly but can only afford a limited ensemble size of less than 100 members. Given that the degrees of freedom of atmospheric models are several magnitudes higher implies sampling errors. Sampling errors and fast error growth on convective scales in turn lead to a low predictability. Consequently, improving initial conditions and subsequent forecasts requires a better understanding of error correlations in both space and time.<br>For this purpose, we conducted the first convective-scale 1000-member ensemble simulation over central Europe. Several 1000-member ensemble forecasts are investigated during a high impact weather period in summer 2016 using ensemble sensitivity analysis. Spatial and spatiotemporal correlations are used to quantify sampling errors on convective scales. Correlations of the 1000-member ensemble forecast serve as truth to assess the performance of different localization approaches. Those approaches include a standard distance-based localization technique and a statistical sampling error correction method as proposed by Anderson (2012). Our study highlights advantages and disadvantages of existing methods and emphasises the need of different localization approaches for different scales and variables. Several results are published in Necker et al (2020a) and (2020b).</p>

Sign in / Sign up

Export Citation Format

Share Document