Webinars - Weeks 3-4/S2S Webinar Series
4/1/2024
Atmospheric and Surface Dynamics During the Winter 2021 Warm Spells in the Southern Great Plains
Presenter: Taylor Grace, Dr. Kathy Pegion, and Dr. Jeffrey Basara
School of Meteorology
The University of Oklahoma
Norman, OK
Continual extreme heat is adversely affecting numerous global regions, imposing significant
socioeconomic burdens on sectors such as public health, agriculture, and water resources.
Notably, increasing temperatures due to climate change are not exclusive to boreal summer,
with significant warming signals during boreal winter. Extreme warm temperatures during boreal
winter are becoming more frequent (e.g., January 2023 in Europe; South America in July and
August 2023). While the ramifications of winter warm spells mirror those of summer heat waves,
the distinctive characteristics and underlying mechanisms governing these events remain
inadequately understood. Therefore, this study examines the subseasonal-to-seasonal (S2S)
characteristics of extreme warm boreal winter temperatures across the Southern Great Plains of the United States during December of 2021. Employing a relative threshold heat wave definition spanning from 1950 to 2022, we identify two distinct winter warm spells during December 2021: the period spanning from November 29 to December 17, 2021, and another occurring between December 22 and December 31, 2021. Temperature anomalies observed during these events ranged from +3 ̊C to surpassing +10 ̊C. A time-lagged composite analysis, utilizing ERA5 reanalysis, identifies critical atmospheric, oceanic, and surface conditions during the December 2021 case study. Surprisingly, the conventional presence of a co-located blocking high, often associated with extreme heat events, was conspicuously absent during the examined period. The first winter warm spell played a key role in initiating land-atmospheric feedbacks, ultimately contributing to the development of the subsequent warm spell across the region. In addition, warm air advection emerged as a non-significant factor in augmenting surface temperatures during these extreme events. One method to increase the predictability of these extreme warm temperatures during the boreal winter is to understand key characteristics associated with these extreme events, underscoring the main goal of this case study.
Future Changes In Seasonal Climate Predictability
Presenter: Dr. Dillon Amaya, PhD.
Physical Sciences Laboratory, NOAA, Boulder, CO
2:30-2:55 PM
Seasonal forecasts provide critical decision support tools for managing important socioeconomically-relevant resources. As the result of continued model development, the skill of such tools has improved over the years. However, further advancements are hampered by the climate’s “potential predictability”, a potential upper limit for how accurately we can predict different climate parameters. Recent studies have shown that potential predictability and actual forecast skill have varied throughout the historical record, primarily as a result of natural decadal variability. In this study, we explore whether potential predictability will change in the future as a distinct response to anthropogenic climate change. We quantify the potential predictability limits of the El Niño-Southern Oscillation (ENSO) as well as global surface temperature, precipitation, and upper atmospheric circulation anomalies from 1921-2100 by applying a perfect model framework to five coupled model large ensembles. We find that the sign, magnitude, and timing of predictability changes are highly model dependent, with some producing a robust increase or decrease in potential predictability by 2100, and others producing no significant change. While there is large inter-model uncertainty in future predictability changes, a common physical relationship emerges that allows us to anticipate how real-world predictability may change in the coming decades. In particular, predictability changes in each model are strongly linked to their projected change in ENSO amplitude, with a 10% change in Nino3.4 standard deviation giving rise to a 14% change in globally averaged forecast skill. Therefore, historical forecast skill relationships that depend on ENSO and its teleconnections may be altered as the climate continues to change.
3/4/2024
Applied climate Services: Managing Risk for Food Production, Fire Mitigation, and Energy Production in Guatemala
Presenter: Dr. Diego Pons, PhD.
University of Denver, Denver, CO
2:00-2:25 PM
The provision of Climate Services is an approach to climate risk management that has the potential to reduce certain aspects of climate vulnerability in countries prone to climatic extremes. Successful climate services provided in an adequate institutional environment have the potential to increase adaptive capacity and to reduce sensitivities in communities exposed to drought or extreme rainfall events. In this context, adequate climate service provision can inform standard operation procedures for relevant institutions and stakeholders on the ground in charge of managing human, natural, and financial capital to reduce or mitigate the impact of climatic events. In this paper we review three case studies that provide evidence on the potential usability of climate services at the institutional level in Guatemala. First, we demonstrate how climate service provision can help leading development institutions to aid communities manage the climate risk associated with drought and food production in Guatemala’s eastern Dry Corridor area. The second case study shows the usability of seasonal climate forecast systems to anticipate wildfire risks in the lowlands of Guatemala by the National Forest Institute. Finally, the third case study shows how climate service provision can help hydroelectrical power plants manage risks associated with a decrease in seasonal precipitation affecting energy production as well an increase in precipitation that could potentially damage communities allocated downhill from the dams. Our results preliminary suggest that climate service provision has the potential to reduce climate impacts in vulnerable communities if the institutional environment is adequate, and the standard operation procedures are in well-articulated, in place and funded.
Tropical and Midlatitude S2S Prediction using UFS and Machine Learning
Presenter: Drs. Eric D. Maloney, Elizabeth Barnes, Jack Cahill, Zaibeth Carlo Frontera, Yu-Cian Tsai, PhD.
Department of Atmospheric Science
Colorado State University
Ft. Collins, CO
2:30-2:55 PM
This talk will present results from several studies that address subseasonal prediction of North American precipitation and Western Hemisphere tropical cyclones, including use of Unified Forecast System (UFS) and machine learning (ML). First, we present results that use ML to predict errors in subseasonal North American precipitation forecasts in UFS hindcasts. Specifically, by looking at when/where there are errors in the UFS, neural networks can be used to understand what atmospheric conditions helped produce these errors via explainability methods. Our ‘Errors of Opportunity’ approach identifies phase 4 of the Madden-Julian oscillation (MJO) and phases 1 and 2 of the BSISO as significant factors in aiding UFS error prediction across different regions and seasons. For example, we see high accuracy for underestimates of geopotential heights in the Pacific Northwest during Spring stemming from the UFS’s failure to accurately forecast teleconnection patterns.
We also demonstrate that several prototypes of the UFS produce common subseasonal prediction errors over the tropical east Pacific and Atlantic, affecting the conditions that modulate tropical cyclones in these basins. In particular, when the UFS is initiated in MJO phases with a strong dipole of convection across the Maritime Continent, prominent subseasonal UFS forecast errors result in the Western Hemisphere. These errors occur because the UFS maintains an MJO that propagates eastward too slowly, and maintains its strength too long after initialization. An analysis of tropical cyclone genesis potential suggests that substantial errors in prediction of TC genesis result from these forecast biases. Logistic regression (LR) and neural network (NN) models utilizing ENSO and MJO indices and other local environmental information are also used to predict east Pacific and Atlantic cyclogenesis, and demonstrate enhanced forecasting skill relative to climatology. Overall, the NN model shows superior performance compared to the LR model, retaining skill out to three weeks lead time for the east Pacific, and out to four weeks for the Atlantic basin.
2/5/2024
California Investments in Forecasting Development
Presenter: Dr. Michael L Anderson, PhD., PE
State Climatologist, California Department of Water Resources
2:00-2:25 PM
California experiences extreme variability in precipitation outcomes from year to year and within
a year. California also experiences a Mediterranean climate where on average 90% of the
annual precipitation falls from October through April and on average 50% of the annual
precipitation falls in the 90-day window from December through February. Atmospheric rivers
have been identified as key drivers in wet or dry water year outcomes and in flood impacts.
California has spent a decade making investments in forecast capabilities at a range of time
scales to inform resource management and emergency response to extreme events. This talk
will walk through the decade of extremes as motivation for investment and the state of the
collaborative space developed to support improved forecast capabilities.
On the seasonal predictability and forecast skill of the North Pacific western boundary current system and adjacent marginal seas
Presenter: Dr. Youngji Joh, PhD.
Atmospheric and Oceanic Sciences, Princeton University, Princeton, NJ
2:30-2:55 PM
Ocean Western Boundary Current (WBC) systems play an integral role in the climate system by
transporting heat, salt, mass, and biogeochemical elements from the tropics to the mid-latitude.
Over the recent several decades, global WBC systems have experienced exceptional rapid
warming and increasing ocean temperature episodes, including marine heatwave events. This
talk will describe the seasonal variability, predictability, and forecast skill of the major WBC
system and adjacent marginal seas over the Northwestern Pacific, focusing on the sea surface
temperature (SST) variability of Kuroshio-Oyashio Extension (KOE) region and East/Japan Sea
(EJS). Using the observational reanalysis and Geophysical Fluid Dynamics Laboratory (GFDL)
Seamless system for Prediction and EArth system Research (SPEAR) seasonal forecast
system, we show how the summer season KOE and EJS SST variability can be linked to large-
scale air-sea coupled processes and thus can be predictable 8-9 months in advance. We
highlight the role of basin-scale air-sea interactions during the previous winter, which can induce
a spatial evolution of both the significant atmospheric (e.g., local anti-cyclonic circulations) and
oceanic advection (e.g., Ekman transport) anomalies. This talk discusses potential sources of
seasonal predictability of the North Pacific climate and related initialization strategies for
improved forecasts.
Slides and recording are currently unavailable for this presentation
11/6/2023
Calibrated probabilistic sub-seasonal forecasting for Pakistan’s monsoon rainfall during 2022
Presenter: Dr. Bohar Singh Columbia Climate School International Research Institute for Climate & Society
2:00-2:25 PM ET
A Polar Low Genesis Potential Index and Its Application to Subseasonal Prediction
Presenter: Dr. Zhuo Wang Department of Atmospheric Sciences, University of Illinois
2:25-2:55 PM ET
10/2/2023
Update on the Development of the MJO-Teleconnections Diagnostics Package
Presenter: Cristiana Stan, George Mason University, Fairfax, VA, USA
2:00-2:25 PM ET
4/3/2023
Strongly Coupled Land-Atmosphere Data Assimilation and Its Influence on Near-surface Weather Forecasting
Presenter: Dr. Zhaoxia Pu, University of Utah
2:00-2:25 PM
Near-surface weather forecasts are critical for protecting life and human activities. However, they remain a challenging problem in modern numerical weather prediction (NWP) due to difficulties in surface data assimilation and uncertainties in representing complicated land–atmosphere interactions in numerical models. This seminar will summarize recent developments from the author’s research team to understand and develop effective data assimilation methods that enhance near-surface weather forecasts. Results from several recent journal publications are summarized and presented to introduce strongly coupled land–atmosphere data assimilation in the context of land–atmosphere interaction. The first part of the work evaluated the association between near-surface variables and soil moisture with observations, coupled land–atmosphere model, and data assimilation systems. Results indicated a strong coupling between soil moisture and the low-level atmosphere, especially the atmospheric boundary layer. Then, the weakly and strongly coupled land–atmosphere data assimilation methods were compared regarding their influence on the prediction of near-surface atmospheric conditions. Results showed that strongly coupled land–atmosphere data assimilation, with simultaneous corrections to the land and atmospheric conditions, outperformed weakly coupled data assimilation. Finally strongly coupled land–atmosphere data assimilation in an ensemble Kalman filter data assimilation system was implemented with an NWP model. Its positive impacts on predicting both atmosphere and land states were demonstrated. Near the end of the seminar, the recent efforts in developing strongly coupled land–atmosphere data assimilation with NOAA UFS and JEDI data assimilation system as well as their potential impacts on medium-range weather forecasting and S2S prediction will be briefly discussed.
3/6/2023
What caused the quad-states tornado outbreak on 10-11 December 2021?
Presenter: Dr. Dongmin Kim, CIMAS, University of Miami, Miami, FL
2:30-2:55 PM
US tornado activity increases sharply in boreal spring, reaching its peak around April-May-June, and then decreases rapidly thereafter. However, the occurrence of intense tornado outbreaks is not limited to spring and early summer. On December 10 and 11, 2021, a series of unusual and destructive winter tornadoes developed in Arkansas, and Missouri, then swept across Illinois, Tennessee, and Kentucky - the quad-state tornado outbreak, resulting in 71 confirmed tornadoes, 89 fatalities, 672 injuries, and at least $3.9 billion in property damages. This study uses observational data and numerical model simulations to investigate large-scale atmosphere-ocean conditions that led to (or contributed to) the quad-state tornado outbreak. Our analysis indicates that the quad-state tornado outbreak was closely linked to an exceptionally strong negative Pacific North American (PNA) pattern, which developed in early December 2020 and persisted throughout December 10-11. Enhanced by the 2021-22 La Niña condition, the negative PNA pattern produced a strong and persisting atmospheric ridge over the central and eastern US. This atmospheric ridge shifted the mid-latitude jet stream northward and produced southerly wind anomalies, thus increasing the flux of warm and moist air from the Gulf of Mexico (GoM) to the central and eastern US. The poleward-shifted jet stream also warmed the GoM, and thus further enhanced the flux of warm and moist air to the US. These large-scale environments increased atmospheric instability and tornado activity across the Mississippi-Ohio Valley. Further analysis indicates that such conditions may develop if a negative PNA persists for at least six days, as in the case of the quad-state tornado outbreak. This study suggests that the persistence of the negative PNA phase is one of the key factors that may improve the subseasonal predictability of winter season US tornado activity.
2/6/2023
Improving Atmospheric Models by Accounting for Chaotic Physics
Presenter: Dr. Prashant D. Sardeshmukh, CIRES, University of Colorado, Boulder, CO
2:00-2:25 PM
It is well known that randomly perturbing an atmospheric model’s diabatic tendencies at each model timestep can increase its probabilistic forecast skill, mainly by increasing the spread of ensemble forecasts and making it more consistent with the errors of ensemble-mean forecasts. Less obvious and less well established is that such perturbations can also reduce the errors of the ensemble-mean forecasts and improve the model’s mean climate, variability, and sensitivity to forcing. In a recent study, we found clear reductions in ensemble-mean forecast errors in large ensembles of 15-day forecasts made with NOAA’s Global Forecast System (GFS) model. The nearly ubiquitous reductions around the globe, obtained throughout the forecast range, were interpreted as arising in effect from modifications to the model’s deterministic evolution operator by stochastic noise-induced drifts. Such drifts are associated mainly with reductions of mechanical and thermal damping by chaotic boundary-layer and cloud-radiative processes. We also found similar noise-induced impacts on the mean climate and variability of NOAA’s newer Fv3GFS model (specifically, the “MRW/S2S Application” of the Unified Forecast System).
Our results are consistent with many previous studies performed with models ranging from simple stochastically forced models to comprehensive global weather and climate models. They suggest that the parameterized diabatic tendencies in most current global atmospheric models are likely not sufficiently chaotic and this deficiency could be partly remedied by specifying additional stochastic terms. Using some empirical guidance in such specifications may be unavoidable, given the generally intractable complexity of the diabatic interactions.
Neural network-based methods to post-process probabilistic weeks 3-4 precipitation accumulation forecasts
Presenter: Dr. Rochelle Worsnop, CIRES, University of Colorado, Boulder, CO
2:30-2:55 PM
Statistical post-processing is a method to correct systematic biases and ensemble dispersion errors inherent in raw forecasts output from numerical weather prediction models. For this Joint Technology Transfer Initiative (JTTI) project, we develop data-driven machine learning algorithms to identify and learn statistical relationships between 20-years of past model-observation pairs to calibrate probabilistic weeks 3–4 precipitation accumulation forecasts. This approach uses the latest version of the Global Ensemble Forecast System (GEFSv12) reforecasts and gridded Parameter-elevation Relationships on Independent Slopes Model (PRISM) precipitation observations to train and cross-validate the algorithms. We first show results from an artificial neural network (ANN) framework that is able to pool together grid points across the contiguous United States to increase the amount of data to train the models and reduce overfitting. We then discuss our current methods to use a convolutional neural network (CNN) to learn nonlinear relationships in 2D maps of weather variables that represent the large-scale moisture and pressure patterns across the U.S. These post-processing algorithms are compared against a more traditional statistical approach that fits the data to a censored, shifted gamma distribution to calibrate the GEFSv12 forecasts. Comparisons are shown in the form of reliability diagrams and continuous ranked probability skill scores.
11/7/2022
Applying Machine Learning to Improve Subseasonal to Seasonal (S2S) Forecasts
Presenter: Dr. Judah Cohen, Atmospheric and Environmental Research, Lexington, MA
2:00-2:25 PM
The discipline of seasonal climate prediction began as an exercise in simple statistical techniques. However, today the large government forecast centers almost exclusively rely on complex fully coupled dynamical forecast systems for their subseasonal to seasonal (S2S) predictions while statistical techniques are mostly neglected and those techniques still in use have not been updated in decades. In our funded WPO S2S project we have been applying machine learning (ML) to increase the accuracy of S2S predictions using large ensembles of global climate models (GCMs). One technique that we have been applying is the adaptive bias correction (ABC) method that combines state-of-the-art dynamical forecasts with observations using ML. For example, when applied to the the European Centre for Medium-Range Weather Forecasts (ECMWF), ABC improves temperature forecasting skill by 60-90% and precipitation forecasting skill by 40-69% in the contiguous U.S. We couple these performance improvements with a practical work-flow, based on Cohort Shapley, for explaining ABC skill gains and identifying higher-skill windows of opportunity based on specific climate conditions.
In a second application of ML, we are developing online learning algorithms for aggregating probabilistic forecasts into one optimal forecast with the smallest possible loss given the available historical information. Online learning algorithms are a family of ML algorithms that produce time-varying, adaptive ensembles by learning a weight vector for combining the set of model ensemble members at each forecast iteration. As real-time weather outcomes are observed, the algorithm adjusts the weights of each model according to its past performance, upweighting ensemble members that have historically performed well and downweighing those that have performed poorly according to a skill loss function. To explore the performance of no-regret online learning for probabilistic weather forecasting, we evaluated the rank probability skill score (RPSS) of precipitation and temperature SubX model forecasts with 3–4-week lead time over the United States made by our online learning algorithms and compared them to equally weighted mean ensemble forecasts or a uniform ensemble.
Bayesian Joint Probability (BJP) Calibration of Subseasonal Model Forecasts
Presenter: Dr. Dan Collins, NOAA CPC
2:30-2:55 PM
Subseasonal (Week 3-4) forecasts of temperature and precipitation at the Climate Prediction Center (CPC) rely on multiple tools, including dynamical model data, calibrated model data, and various statistical tools. While dynamical model forecasts are helpful for informing outlooks, dynamical climate models can be systematically biased and lack reliability of probabilities in temperature and precipitation forecasts. Statistical calibration of models can improve bias and reliability beyond the potential gains of advances in dynamical models, and without the need to make changes to the base model code. One such calibration method, Bayesian Joint Probability (BJP) has been in use experimentally at CPC for the subseasonal timescale to calibrate temperature forecasts. BJP calibration has been shown to improve skill and reliability of 2-meter temperature over raw dynamical models in a test environment using three Subseasonal Prediction Experiment (SubX) models, CFSv2, GEFSv12, and the NOAA Earth System Research Laboratories (ESRL-FIMv2) model; and to improve the skill and reliability of seasonal precipitation predictions when applied to the North American Multi-Model Ensemble (NMME); as well as to improve skill over other calibration methods such as Extended Linear Regression (ELR) and Probability Anomaly Correlation (PAC). While BJP calibration is currently used in an experimental setting for subseasonal temperature, given the demonstrated success of the methodology upcoming work aims to transition BJP calibration to operational temperature forecasts from CFSv2, ECMWF, GEFSv12, and, when available UFS subseasonal data, as well as to extend BJP to subseasonal precipitation. An overview of the skill of these calibrated forecasts and the products generated from these tools will be provided.
10/3/2022
Title: Strongly Coupled Data Assimilation with a Linear Inverse Model
Presenter: Dr. Greg Hakim, University of Washington
2:00-2:15 PM
Abstract: Strongly coupled data assimilation (SCDA), such as using atmospheric observations to update ocean analyses, can potentially use observations more effectively than approaches that limit the influence of observations to the domain in which they were taken. To work well, SCDA requires accurate estimates of the cross-domain covariances, which is extremely computationally demanding, leading to weakly-coupled approximations. Here we break the computational bottleneck using a linear inverse model (LIM) to emulate the coupled atmosphere—ocean dynamics, with a goal of improving week 3/4 forecasts. The LIM operates in a reduced dimensional space, which promotes rapid prototyping, including a wide range of data assimilation approaches. The LIM also serves as a reference benchmark for nonlinear approaches, such as neural networks.
We show experiments using a Kalman filter, with a LIM trained on 25 years of Climate Forecast System Reanalysis gridded data, to evaluate SCDA relative to uncoupled and weakly coupled approaches during on an independent 7-year period at daily resolution. SCDA Sea-surface temperature (SST) analysis errors are reduced over 20\% in global-mean mean-squared error, and 40% in local regions, relative to a control experiment where only SST observations are assimilated with an SST LIM. The analysis improvements improve SST forecast skill for leads of at least 50 days. In contrast, extratropical Northern Hemisphere 2m air temperature forecast errors increase for coupled data assimilation in these experiments, despite reduction during the training period, indicating that 25 years of daily training data is not sufficient to fully capture the coupled dynamics. Experiments with larger sample sizes, and an ensemble approach using JRA reanalysis data will be briefly discussed.
Title: Ensemble Predictability of Week 3/4 Precipitation and Temperature over the United States via Cluster Analysis of the Large-Scale Circulation
Presenter: Dr. David Straus, George Mason University
2:20-2:45 PM
Abstract: Forecasting the Week 3/4 period presents many challenges, resulting in a need for improvements to forecast skill. At this time spanned from initial conditions, numerical models struggle to present skillful forecasts of temperature, precipitation, and associated extremes. Nor does this period fully extend into the boundary-dependent climate time scale. One approach to improve Week 3-4 forecasts is to utilize the better predicted, large-scale circulation to make forecasts of temperature and precipitation anomalies, using the association between the preferred patterns of geopotential height (hereafter regimes) and surface weather obtained from reanalysis products. The functionality of regime classification has been well documented (Amini and Straus 2018, Riddle et al. 2013, Dawson and Palmer 2015). This study explores the utility of k-means cluster analysis of geopotential heights to identify regimes and using the forecasted regimes to make skillful predictions of temperature and precipitation in the Week 3/4 period.
9/12/2022
A diagnostic toolbox: assessing the representation of stratosphere-troposphere coupling in the Global Ensemble Forecast System (GEFSv12)
Presenter: Dillon Elsbury, CIRES and Laura Ciasto, NOAA CPC
2:00-2:25 PM
Dynamical coupling between the stratosphere and troposphere is an important source of sub-seasonal (two weeks - two months) predictive skill for tropospheric temperature and precipitation. However, taking advantage of this source of predictive skill is difficult in practice because forecast systems have stratospheric temperature and circulation biases arising from coarse vertical resolution, low model tops, and poor process representation (e.g., the Quasi-Biennial Oscillation, QBO). In this seminar, we share aspects of the diagnostic toolbox we have developed to analyze the representation of stratosphere-troposphere coupling, and its application to the NOAA Global Ensemble Forecast Prediction System version 12 (GEFSv12) forecast system.
NOAA’s GEFSv12 includes a 20-year reanalysis and 30-years of reforecasts, with reforecasts spanning 35 days (5 weeks) from the time of initialization. We use our new diagnostic toolbox to document the temperature and wind biases present in the GEFSv12 stratosphere, evaluate coupling between the stratospheric polar vortex and the mid-latitude surface circulation, provide an assessment of the representation of the QBO, and include a preliminary analysis of coupling between the QBO and the Madden-Julian Oscillation (MJO). Going forward, the toolbox will be implemented at the Climate Prediction Center in order to allow future evaluations of the stratosphere in NOAA Unified Forecast System prototypes, hindcasts, and retrospective forecasts.
Captured QBO-MJO Connection in a Subseasonal Prediction System
Presenter: Kai Huang, PhD Candidate, George Mason University
2:30-2:55 PM
The Quasi Biennial Oscillation (QBO) leaves dominant impacts on the Madden-Julian Oscillation (MJO) activity in observations with a stronger and more eastward-propagating MJO in QBO easterly (QBOE) winters than in QBO westerly (QBOW) winters. However, such QBO-MJO connection is poorly represented in the uninitialized simulations by current general circulation models (GCMs). For the first time, this paper applies a stratospheric zonal-mean nudging in a subseasonal prediction system to capture the QBO-MJO connection. Two strong MJO cases in a QBO neutral winter are selected for the experiment. The QBO temperature and zonal wind anomalies are nudged separately as well as together for the hindcasts of two selected MJO cases. Results show that only by nudging the QBO temperature anomalies while leaving the zonal wind free, the prediction system could capture the observed QBO-MJO connection. The tropopause instability is found to be positively correlated to the MJO amplitude in the hindcasts, but it alone could not fully explain the captured QBO-MJO connection. Instead, the freely-evolving zonal wind anomalies as a result of the thermal wind balance to the nudged QBO temperature are likely to be the key for such capturing.
8/8/2022
The Development of UFS Coupled GEFS for Weather and Subseasonal Forecasts
Presenter: Dr. Yuejian Zhu
In collaboration with Unified Forecast System (UFS) Research to Operation / Subseasonal to Seasonal (R2O/S2S) and Medium Range Weather (MRW) applications, focus of the WPO/CTB funded project to improve NCEP Global Ensemble Forecast System (GEFS) for weather and subseasonal prediction from current operational GEFS. The GEFS v12 was implemented in NCEP operation in September 2020, which is the 1st UFS application with coupling to Wave Watch 3 (WW3) ensembles with 25km uniform horizontal resolution; 31 members and out to 35 days to cover subseasonal prediction. The operational GEFS demonstrated improved forecast capability and excellent performance for weather and subseasonal time scales including MJO predictions, surface temperature and precipitation through retrospective forecasts and 31 years reforecasts before it was implemented.
In this presentation, quantifying forecast uncertainty is the main focus to discuss through the integration of a fully coupled UFS and various stochastic physical perturbation schemes to initialize a coupled global ensemble forecast experiments toward GEFS v13 implementation. The experiments are closely configured to the development of the UFS coupled prototype experiments. For this study, a UFS prototype version 5 (P5) based coupled GEFS experiment has been carried out with the optimum atmospheric model perturbations. The experiments run a 2-year period (initialized once per week at 00UTC time) with 10 perturbed and 1 unperturbed members, out to 35 days. A full evaluation of the experiment will be presented in terms of various evaluation metrics.
6/6/2022
Direct assessment of the surface impacts of the January 2021 sudden stratospheric warming with S2S ensemble forecasts
Presenter: Dr. Nicholas Davis, NCAR
Subseasonal weather prediction can reduce economic disruption and loss of life, especially during “windows of opportunity” when noteworthy events in the Earth system are followed by characteristic weather patterns. Sudden stratospheric warmings (SSWs), breakdowns of the winter stratospheric polar vortex, are one such event. They often precede warm temperatures in Northern Canada and cold, stormy weather throughout Europe and the United States - including the most recent SSW on January 5th, 2021. Here we assess the drivers of surface weather in the weeks following the SSW through initial condition “scrambling” experiments using real-time CESM2(WACCM6) Earth system prediction framework ensemble forecasts. We find that the SSW itself had a limited impact, and that stratospheric polar vortex stretching and wave reflection had no discernible contribution to the record cold in North America in February. Instead, the tropospheric circulation and bidirectional coupling between the troposphere and stratosphere were dominant contributors to variability.
IRI SubX-based Real-Time Subseasonal Precipitation and Temperature Forecasts
Presenter: Dr. Andrew W. Robertson, IRI, Columbia Climate School, Columbia University
This talk will present IRI’s real-time global subseasonal forecasts for precipitation and temperature, which have been issued every Friday since 2018. The output of three SubX models is first calibrated for each model separately using extended logistic regression, and the resulting probabilities averaged over the 3 models to form a multi-model mean, for weekly and biweekly averages of precipitation and near-surface temperature up to 4 weeks ahead. Forecasts are issued as the probability of three tercile categories, above/below median, as well as the full forecast PDF, plotted versus the climatological one, allowing "forecasts of opportunity" when predictable shifts in the distribution occur associated with sources of S2S predictability. The talk will include an assessment of skill using both hindcasts and real time forecasts.
5/3/2022
3/7/2022
A Deep Learning Filter for Intraseasonal Variability of the Tropics
Presenter: Rama Sesha Sridhar Mantripragada, George Mason University
One of the challenges faced by the post-processing of subseasonal to seasonal (S2S) and seasonal forecasts is the ability to isolate the intraseasonal variability (20-100 days) in the forecast anomalies. If the length of the forecast permits applying conventional methods for time series filtering (e.g., band-pass, running averaging), these methods require a backward extrapolation of the forecast. The extrapolation step may introduce artifactual or spurious features. In most cases the length of the forecast does not permit applying conventional filtering methods. In this study, a new method based on a Convolutional Neural Network (CNN) model is developed to extract the intraseasonal signal in the S2S forecast of tropical wind stress. The power spectrum of the signal predicted by the 1D CCN is similar to the power spectrum of the band-pass filtered signal. When applied to the tropical belt, the spatial pattern of the wind stress predicted by CNN shows only small differences from the band-pass filtered field.
Developing and Verifying a Subseasonal Outlook Tool for Extratropical Storminess
Presenters: Dr. Edmund Chang, Stony Brook University and Yutong Pan, NOAA/NWS/CPC
Extratropical cyclones give rise to much of the high impact weather in the mid- to high- latitudes during the cool season, including heavy precipitation and strong winds. Thus it is important for many stakeholders to be warned of approaching periods of increased or decreased potential of storm activities. While individual cyclone tracks can be predicted out to about a week or so, from week 2 on, statistics summarizing cyclone activity, or storminess, are more useful. Storminess can be defined based on Lagrangian cyclone tracking or by Eulerian variance statistics. In this talk, we will describe the storminess indices that we developed, as well as some forecast evaluation statistics using hindcast data. We will also describe a web page that has been developed to display subseasonal storminess forecasts based on GEFSv12 and CFSv2 operational forecast products.
2/7/2022
Application of large-scale precipitation tracking (LPT) to real-time MJO monitoring and forecasts
Presenters: Dr. Chidong Zhang, NOAA/OAR/PMEL & Dr. Wanqiu Wang, NOAA/NWS/CPC
The Madden-Julian Oscillation (MJO) is the dominant mode of intraseasonal variability in the tropics. One of the major characteristics of the MJO is the large regions of enhanced and suppressed tropical rainfall. Realistically capturing rainfall variations is important for the representation of the MJO in its monitoring and forecasts. Under the support of the Weather Program Office/Climate Testbed program, a Large-scale Precipitation Tracking (LPT) tool was recently developed using Climate Prediction Center (CPC) CMORPH precipitation analysis and Climate Forecast System (CFS) forecasts. In this presentation, we describe the application of the LPT to real-time monitoring and forecast at the CPC. The presentation will focus on the following three aspects: (1) Configuration of the LPT package for real-time monitoring and forecasts in CPC computer system environment; (2) Comparison of LPT with other MJO indices including the Real-time Multivariate MJO (RMM) index and the OLR MJO Index (OMI); and (3) Evaluation of forecast skill of LPT and RMM. Areas for further enhancement of the LPT tool for real-time application will also be discussed.
12/6/2021
A seasonal probabilistic forecast system for U.S. regional precipitation based on the tropical Pacific and Atlantic SSTAs
Presenter: Dr. Dongmin Kim, NOAA/AOML
The U.S. regional precipitation is modulated by the tropical Pacific sea surface temperature anomalies (SSTAs) in boreal winter, and tropical Atlantic SSTAs in boreal summer via their atmospheric teleconnections. The tropical SSTAs in dynamical seasonal forecast models are well predicted up to 6-month lead time. However, the prediction skill for boreal summer U.S. regional precipitation is generally very low. In order to improve the seasonal prediction skill for boreal summer U.S. regional precipitation, here we explore inter-basin SSTA contrast between the tropical Pacific and Atlantic as a potential predictor since it can directly modulate U.S. summer and fall precipitation via a Gill-type response driven by atmospheric convection over the Caribbean Sea (Kim et al., 2020). To test whether this new index could serve as a potential predictor for U.S. precipitation, we develop and evaluate a hybrid probabilistic forecast model for U.S. precipitation based on tropical Pacific and Atlantic SSTAs. We first evaluate seasonal forecast skills of the tropical Pacific and Atlantic SSTAs in the North American Multi-Model Ensemble (NMME). The tropical Pacific SSTAs predicted by NMME have fairly good forecast skills at lead times of up to five months for all initialized months (i.e., anomaly correlation > 0.6), whereas skillful forecasts for the tropical Atlantic SSTAs are limited to three months. The hybrid forecast model is developed by using multiple regression coefficients applied to the NMME-predicted tropical Pacific and Atlantic SSTAs, and using the ensemble spreads. In boreal summer (July-September), the hybrid forecast has better forecast skill for 3-month lead U.S. precipitation than the NMME dynamical forecast, while in boreal winter (December-February) the prediction skills are comparable. We further explore a 3-month lead probabilistic forecast skill for U.S. precipitation and carry out a cross-validation test. The hybrid model successfully predicts extreme precipitation events in the U.S. while the NMME models failed to capture some of them. Our results suggest that the probabilistic hybrid prediction model is a valuable forecasting tool, especially for U.S. summer and fall precipitation.
The development, evaluation and applications of CPC Week 2-4 excessive heat forecast tools and services
Presenter: Jon Gottschalck, NOAA/CPC
The CPC releases a Week-2 U.S. Hazards Outlook that highlights potential hazardous events related to temperature, precipitation and wind - including extreme heat events. The talk will outline the objectives of this project which includes applied research and development for both the Week-2 and Week 3-4 forecast periods targeting extreme heat events. Quantitative evaluation as well as qualitative review of some case studies will be presented from both dynamical post processing, statistical and hybrid (statistical-dynamical) developed methods to gain understanding of forecast skill for both the Week-2 and Week 3-4 periods. The status of both experimental and operational products and services and their increasing role in climate IDSS opportunities will be presented. The primary benefit and goal of the project for stakeholders is continued advancement of forecast lead time, where and when possible, for decision making ahead of hazardous extreme heat episodes. These extreme heat periods affect nearly all aspects of daily life across many sectors that includes the health, agriculture and energy sectors, among others. The project work maps well to the key target areas outlined in the “Weather Research and Forecasting Innovation Act of 2017” – namely (1) extending outlooks of extremes further into the subseasonal time period and (2) targeting the extreme heat problem.
11/1/2021
Improving S2S precipitation forecasts in UFS through Tropical Nudging and Explainable Machine Learning
Presenter: Dr. Eric D. Maloney, Colorado State University
Boreal-wintertime hindcasts in the Unified Forecast System with the tropics nudged toward reanalysis improve United States (U.S.) West Coast precipitation forecasts at Weeks 3-4 lead times when compared to those without nudging. To diagnose the origin of these improvements, a multivariate k-means clustering method is used to group hindcasts into subsets by their initial conditions. One cluster characterized by an initially strong Aleutian Low demonstrates larger improvements at Weeks 3-4 with nudging compared to others. The greater improvements with nudging for this cluster are related to the model error in simulating the interaction between the Aleutian Low and the teleconnection patterns associated with the Madden-Julian oscillation (MJO) and El Nino-Southern Oscillation (ENSO). Improving forecasts of tropical intraseasonal precipitation, especially during early MJO phases under non-cold ENSO, may be important for producing better Weeks 3-4 precipitation forecasts for the U.S. West Coast.
We will also present future plans to combine state-of-the-art developments in machine learning with process-based diagnostics of the tropical budget to understand and correct precipitation biases in coupled UFS hindcasts. In particular, we will diagnose how model biases and errors in tropical variability (e.g. MJO) and associated teleconnections to midlatitudes lead to errors in U.S. precipitation on S2S timescales. We will also present methods to reduce these errors via post-processing on a forecast-by-forecast basis.
Predicting Indian Monsoon onset in S2S scale: A Machine Learning Framework
Presenter: Dr. Nachiketa Acharya, Pennsylvania State University
Monsoon Advance of the southwest monsoon over the Indian mainland is marked by monsoon onset over Kerala (MOK) as it is the beginning of the rainy season for the country. MOK is associated with many changes in the large-scale and local atmospheric variables. Although the climatological date of MOK is 1st June; it varies by a few days from year to year with a standard deviation. As the arrival of the monsoon is crucial for farmers to plan their crop strategy during the season, a reliable prediction of MOK is very crucial. Since 2005, India Meteorological Department (IMD) has been issuing operational forecasts for MOK. This forecasting system is based on a Principal Component Regression (PCR) between large-scale circulation features in the Asia-Pacific region, local pre-monsoon rainfall peak in, and the MOK date. We aim to improve upon the accuracy of MOK predictions to explore sophisticated machine learning (ML) based prediction models. Two families of ML models: decision trees that involve bagging (Random Forest) and boosting (XGBoost), and neural networks based on extreme learning machine (ELM), have been explored in this study and have been compared to the benchmark PCR model. All the prediction models were implemented in a leave-one-out manner from 1975 to 2020 and their skill is assessed using a set of statistical skill metrics. Results suggest that the ELM methods are out performed compared to all other models.
10/4/2021
Using Simple, Explainable Neural Networks to Predict the Madden-Julian Oscillation
Presenter: Dr. Zane Martin, CSU
The tropical Madden-Julian oscillation (MJO) is among the most important sources of S2S predictability on Earth. Despite the importance of MJO prediction, few studies have utilized machine learning (ML) techniques to predict or understand the MJO. As ML methods have proven useful in predicting other climate oscillations like ENSO, it is important to characterize and benchmark ML modeling for the MJO. Here we present a simple ML framework for real-time MJO prediction using shallow artificial neural networks (ANNs). These ANNs make skillful MJO predictions out to ~17 days in October-March and ~10 days in April-September, and efficiently capture aspects of MJO predictability found in more complex, computationally-expensive models. Varying model input and applying ANN explainability techniques further reveal sources and regions important for ANN prediction skill. This simple ML framework can be more broadly adapted and applied to predict and understand other climate oscillations or phenomena.
The CPC Global Tropics Hazards Outlook: Background, Current Operational Products and Work to Transition to a Probabilistic Format Targeting the Week 2-3 Period
Presenter: Jon Gottschalck, NOAA/NWS/CPC
The Climate Prediction Center (CPC) issues the Global Tropics Hazards Outlook (GTH) once per week (Tuesday), although an update is provided on Friday for the northern Pacific and Atlantic Ocean basins during hurricane season. The product highlights anticipated areas of potentially hazardous (1) above- and below-normal weekly total rainfall and mean temperature and (2) regions where tropical cyclogenesis is favored and at times reduced compared to climatology for an upcoming two week period. The talk will outline and review the purpose, physical basis, forecast tools and stakeholders of the product as time permits. Also presented will be the status and review of completed work as part of a funded project to convert the current outlook from a more deterministic forecast to a probabilistic format and shifts the forecast period to target the subseasonal Week 2-3 time scale. This will include review of the methods used in developing the necessary statistical and dynamical model based forecast tools to support this product transition as well as a report out on the historical (i.e., from reforecasts) and realtime (where available) forecast skill for these new guidance products. The new format will be presented along with the necessary planning and outreach required to make this transition. The outlook transition is aimed to improve the information and expand the content contained in the current product as well as to extend the forecast lead time for extreme events -- consistent with the goals of the "Weather Research and Forecasting Innovation Act" of 2017.
9/13/2021
Developing Experimental S2S Sea Ice Predictions with a UFS-based System
Presenter: Dr. Wanqiu Wang, NOAA/NWS/CPC
Sea ice predictions at subseasonal to seasonal (S2S) time scales have become important products for stakeholders. For example, the NWS Alaska Region requires sea-ice forecasts for the next few weeks to seasons. The Climate Prediction Center (CPC) has been providing sea ice predictions for week-2 to 9-month target periods based on an experimental sea ice prediction system (CFSm5) consisting of the Climate Forecast System (CFS) atmospheric component and the Geophysical Fluid Dynamics Laboratory Modular Ocean Model version 5 (MOM5). Sea ice in CFSm5 is initialized from a MOM5-based CPC sea ice initialization system (CSIS). Sea ice forecasts from CFSm5 are significantly better than that from the operational CFS. The NWS Alaska Region uses these CPC sea ice predictions to provide guidance to the DOI, USCG and other partners. CPC’s sea ice predictions are also regularly used by Alaska Center for Climate Assessment and Policy (ACCAP) in Alaska Region Climate Outlooks.
The recent successful development and improvement of the coupled Unified Forecast System (UFS) by the Dynamics and Coupled Modeling Group of the Environmental Modeling Center (EMC) provided an opportunity for CPC to upgrade the CFSm5 to a UFS-based model for the S2S sea ice predictions. In this talk, we report our progress in the use of UFS in sea ice predictions. The final goal is to provide improved real-time week-3/4 and seasonal sea ice outlooks. We will present two major efforts with the UFS: (1) Experiments to adjust cloud parameterizations to reduce model errors in sea surface temperature and sea ice coverage and (2) An evaluation of sea ice predictions based on hindcasts completed with the UFS and comparisons with operational CFS, CFSm5, and observations. The potential of using a multi-model ensemble based on UFS, CFS, and CFSm5 will also be discussed.
Evaluating the Potential of a Blocking Predictor in a Hybridized Dynamical-statistical Model for Improved Week 3-4 Temperature and Precipitation Outlooks
Presenter: Dr. Cory Baggett, NOAA/NWS/CPC
Atmospheric blocking manifests as large, quasi-stationary anticyclones at high latitudes that reverse the climatological westerlies in the mid-latitudes and block the storm track. Blocking has long been recognized as an important phenomenon that impacts both upstream and downstream temperature and precipitation anomalies due to the prolonged, several week timescale that blocking and its impacts can persist. For example, extended drought and heat in California have been linked to blocking over the northeastern Pacific Ocean, while these same blocks can lead to atmospheric river landfalls in Alaska. Similarly, extreme cold conditions over the eastern United States have been linked to blocking over the western Atlantic. Thus, a greater understanding of blocking and its impacts has the potential to improve subseasonal outlooks, including NOAA’s Climate Prediction Center’s (CPC) Week 3-4 temperature and precipitation outlooks.
In this work, we investigate the extent to which Week 3-4 prediction skill can be improved by adding blocking as a predictor to CPC’s statistical suite of tools. In particular, we focus on improving CPC’s multiple linear regression model (MLR), whose present version uses as predictors the current state of ENSO and the MJO at forecast initialization along with the long-term trend to forecast above or below normal outlooks of Week 3-4 temperature and precipitation. We test several formulations of blocking as a predictor, including traditional indices based on the latitudinal reversal of geopotential height gradients along with more common indices such as the North Atlantic Oscillation (NAO) and the Pacific-North American pattern (PNA) that are known to be associated with blocking. Furthermore, we analyze the ability of the dynamical models, including the GEFS (version 12), to forecast blocking at extended leads with the purpose of using forecasted blocking predictors in a hybrid-statistical-dynamical approach. In this analysis, we found that the GEFS can skillfully forecast blocking predictors with correlations exceeding 0.5 at leads extending to Days +12 to 14.
Here, we verify temperature and precipitation forecasts by the MLR for initializations during November-April over an independent verification period of 2011-2021. While improved precipitation skill scores were not attainable, temperature skill scores exhibit notable improvements. It is found that using Day +14 forecasted values of the NAO from the GEFS as a predictor increases the Heidke Skill Score (HSS) of the MLR for temperature over CONUS from 10.6 to 15.0 but decreases skill scores over Alaska. Similarly, using Day +12 forecasted values of the PNA increases the HSS over Alaska from 17.2 to 25.7 but decreases skill scores over CONUS. Thus, a merged-MLR that uses the NAO to forecast for CONUS and the PNA to forecast for Alaska increases the HSS averaged over the entire domain from 11.7 to 16.8, an improvement of 44%. Furthermore, when only considering forecast initializations with an amplified PNA on Day 0, so-called forecasts of opportunity, the HSS increases by 27%, from 21.5 to 27.4. Likewise, when the NAO is amplified on Day 0 the HSS increases from 11.9 to 23.8, an improvement of 100%. From these results, we conclude that the addition of forecasted, blocking-related predictors adds value to CPC’s statistical models, which in turn offers the potential to improve CPC’s official Week 3-4 outlooks.
08/02/2021
Tropical Dynamics Diagnostics for Numerical Weather Prediction
Presenter: Dr. Maria Gehne, NOAA/OAR/PSL & CIRES/CU Boulder
Precipitation and moisture variability in the Tropics covers a vast collection of scales from a few to several thousands of kilometers and from hours to weeks. Current operational numerical weather prediction (NWP) models struggle with representing the full range of scales and phenomena in the Tropics. It has been shown that in particular skill in the representation of larger scale phenomena such as convectively coupled equatorial waves (CCEWs) can be important for reducing error propagation from the Tropics to the Mid-latitudes at longer lead times. Novel diagnostics are therefore needed to assess NWP forecast skill of large spatial, long time scale phenomena such as CCEWs as well as to assess the representation of physical processes that are important for CCEW initiation and maintenance.
Here we apply diagnostics from a tropical variability diagnostics toolbox currently under development at NOAA PSL to model output from two recent versions of the Unified Forecast System (UFS), operational V15 forecasts and experimental retrospective V16 forecasts from December 2019 through March 2020. The diagnostics include space-time coherence spectra to identify preferred scales of dynamics-precipitation coupling, pattern correlation of precipitation Hovmoeller diagrams to assess model skill in zonal propagation and an EOF based CCEW skill assessment. Two more diagnostics that look at the coevolution of moisture and convection in the Tropics can be used by model developers to assess whether changes in the parameterizations lead to improvements of this relationship. These diagnostics are in the process of being added to the Model Diagnostics Task Force (MDTF) code repository and the upcoming METplus python release (METcalcpy and METplotpy).
Results show that the V16 forecasts are more realistic in reproducing the statistical relationship between precipitation and column moisture, and slightly more skillful in their coherence between precipitation and model dynamics at CCEW scales. However, this does not necessarily translate to a significant improvement in traditional precipitation skill scores such as ETS or MSESS. This highlights the utility of these process and phenomena focused diagnostics as the goal is to allow for better understanding of NWP model performance regarding coupling between moist convective processes and synoptic to planetary scale tropical circulations.
METplus Verification and Diagnostics Framework for S2S
Presenter: Tara Jensen, NCAR/RAL
Verification and validation activities are critical for the success of modeling and prediction efforts at organizations around the world. Having reproducible results via a consistent framework is equally important for model developers and users alike. The Model Evaluation Tools (MET) was developed over a decade ago and expanded to the METplus framework with a view towards providing a consistent platform delivering reproducible results.
The METplus system is an umbrella verification, validation and diagnostic tool for use by thousands of users from both US and international organizations. These tools are designed to be highly flexible to allow for quick adaption to meet additional evaluation and diagnostic needs. A suite of python wrappers have been implemented to facilitate a quick set-up and implementation of the system, and to enhance the pre-existing plotting capabilities. Recently, several organizations within the National Oceanic and Atmospheric Adminstration (NOAA), the United States Department of Defense (DOD), and international partnerships such as Unified Model (UM) Partnership led by the Met Office have adopted the tools for their use both operationally and for research purposes. Many of these organizations are also now contributing to METplus development, leading to a more robust and dynamic framework for the entire earth system modeling community to use.This presentation will provide an overview of METplus and how it can be used for evaluation and diagnosis of weeks 3-4 subseasonal prediction.
06/07/2021
Development of Fully Coupled UFS-based Seasonal-to-Subseasonal Prototypes
Presenter: Dr. Avichal Mehra, NOAA/NWS/EMC
This presentation will discuss the scientific formulation of the seasonal-to-subseasonal UFS model prototypes, which are being developed as we work towards the operational implementation of GFS v17 and GEFS v13 in FY2024. The current prototype is a four-way coupled atmosphere-ocean-ice-wave model. The atmosphere model uses the Finite Volume Cubed Sphere (FV3) dynamical core and the Common Community Physics Package (CCPP) for the physics driver, with GFS physics and GFDL microphysics. The other model components are the Modular Ocean Model (MOM6), the Los Alamos Sea Ice Model (CICE) and the WAVEWATCH III (WW3) wave model. Recent developments and prototype experiments will be discussed along with salient results and improvements. A brief overview of future development plans will also be discussed.
Open Community Development Using the UFS-Weather-Model
Presenter: Dr. Arun Chawla, NOAA/NWS/EMC
The UFS weather model is a fully coupled weather model that serves the needs of several applications ranging from short range Limited Area weather models all the way to fully coupled global systems for sub seasonal to seasonal ranges. In this talk we will give an overview on the GitHub repository, the code management and regression test systems that have been put in place for developers to use and contribute to the development of the ufs-weather-model.
05/03/2021
Subseasonal Earth System Prediction Framework with CESM2 and its Application to the January 2021 Sudden Stratospheric Warming
Presenter: Dr. Jadwiga (Yaga) Richter, NCAR
A framework to enable Earth system predictability research on the subseasonal timescale is developed with the Community Earth System Model, version 2 (CESM2) using two model configurations that differ in their atmospheric components. One configuration uses the Community Atmosphere Model, version 6 (CAM6) with its top near 40 km, referred to as CESM2(CAM6). The other employs the Whole Atmosphere Community Climate Model, version 6 (WACCM6) whose top extends to ~ 140 km in the vertical and it includes fully interactive tropospheric and stratospheric chemistry (CESM2(WACCM6)). Both configurations were used to carry out subseasonal reforecasts for the time period 1999 to 2020 following the Subseasonal Experiment’s (SubX) protocol. Weekly real-time forecasts with CESM2(WACCM6) contribute to the multi-model mean ensemble forecast used to issue the NOAA weeks 3-4 outlooks. We discuss here the system design and basic skill of the two systems. In addition, we discuss experiments with the CESM2(WACCM6) system that were designed to elucidate the contribution of the stratosphere to the surface impacts in the month following the January 2021 sudden stratospheric warming.
Understanding Coupled vs Uncoupled GEFS Subseasonal Forecast Skill Differences
Presenter: Dr. Xin-Zhong Liang, UMD
This talk compares subseasonal skills between GEFS coupled prototype-5 model forecasts (CPL) at a horizontal resolution of 50km (C192) and its uncoupled atmosphere-only reforecasts (RFC) of 25km (C384) for two years (October 2017 – Sep 2019) and attempts to understand the physical processes that may explain their differences. CPL’s tropical skill (per the average 500hPa height anomaly correlation) is significantly better than RFC throughout 1-4 weeks. Correspondingly, CPL improves over RFC the MJO skill at all lead times, with the 50% anomaly correlation reaching 23-day for RMM1 and 31-day for RMM2. As compared to the ERA5 reanalysis, CPL largely reduces RMS errors in surface latent heat flux and outgoing longwave radiation over tropical Indian to western Pacific Oceans, and hence tends to enhance the predictive skill for convection activities over the MJO key region. In addition, CPL produces warm SST biases, while RFC have the cold biases. Their differences increase slightly with lead times and reach ~0.35 ºC averaged over global oceans between 50oN-50oS. They are identified with the smaller biases of downward shortwave radiation in CPL than RFC, by ~2 W/m 2 on average between 50ºN-50ºS. These results seem to indicate significant changes in cloud-radiation interaction from RFC to CPL. While a comprehensive diagnostic analysis is in progress, this talk will present a preliminary understanding of how air-sea interaction with cloud-radiation feedback affect subseasonal forecast skills.
04/05/2021
Accelerating Progress in Subseasonal to Seasonal Prediction Capabilities by Improving Subgrid-Scale Parameterizations in the UFS
Presenter: Dr. Benjamin Green, NOAA/OAR/GSL - CIRES/CU
This ongoing work tests the impact of incorporating three new parameterizations of atmospheric subgrid-scale physical processes into the existing physics suite used for NOAA’s Unified Forecast System (UFS) – specifically, the coupled atmosphere (FV3), ocean (MOM6), sea ice (CICE6), and wave (WWIII) model – in order to advance subseasonal to seasonal (S2S) prediction capabilities over the U.S. and globally. Specifically, the current operational parameterizations of scale-aware convection, planetary boundary layer, and cloud microphysics developed at NOAA’s Environmental Modeling Center are swapped out one-at-a-time and replaced (temporarily, for experimental purposes) with schemes developed at least in part by NOAA’s Global Systems Laboratory. These physics tests have been done within the coupled UFS (“C-UFS”) under the Common Community Physics Package (CCPP) framework. Through the one-at-a-time approach to test different physics schemes, deficiencies in the existing operational physics can be exposed (and corrected).
Currently, this work has completed 3 additional sets of 35-d hindcasts initialized twice per month from 1 April 2011 through 15 March 2018 (for 168 hindcasts per set), and analyses comparing each of these three additional sets to the control experiment (operational physics) is underway. So far, none of the experiments has been found to be uniformly better (or worse) than any of the others, in terms of biases (including for precipitation and 2-m temperature) and skill scores (including for an index of the Madden-Julian Oscillation).
So far, the most impactful findings are that (i) in some instances, running hindcasts at shorter timescales (~1 week) may be sufficient to reveal biases that exist at subseasonal timescales; and (ii) some bias patterns are not impacted by physics changes. In the near future, more evaluation of biases in the upper-level circulation – important because of the impact on teleconnections – will take place.
Seamless S2S Prediction at GFDL: Coupled and Convective-Scale Prediction
Presenter: Dr. Lucas Harris, NOAA/OAR/GFDL
GFDL has developed a seamless modeling system within the UFS for "minutes-to-millennia" simulation. Strengths have been on weather timescales (out to 10 days) and on seasonal-decadal-centennial simulation (2 months and beyond). The seamlessness of the system allows us to then leverage findings on both ends to the intermediate subseasonal-to-seasonal timescales. Two models are most useful for S2S prediction: the SPEAR coupled-climate system and the nonhydrostatic SHiELD weather prediction system. A series of results are demonstrated for each system. Prototypes of SPEAR showed 27 days of MJO prediction skill, and skillful predictions of Northern Hemisphere wintertime temperature anomalies out to week five. A subseasonal configuration of SHiELD shows a greatly improved diurnal cycle of land precipitation, and similar MJO prediction skill. The prediction skill of the MJO is greatly extended by using SHiELD variable-resolution capabilities to zoom-in to 4-km resolution over the Maritime Continent. We also show promising results for explicit subseasonal prediction of US severe weather outbreaks. Prospects for further development will be discussed.
03/01/2021
Subseasonal-to-Seasonal (S2S) Prediction of Atmospheric Rivers, Ridging Events, and Precipitation over the Western U.S. to Benefit Water Management
Presenter: Dr. Michael DeFlorio, University of California San Diego
Water resource managers are in need of more skillful Subseasonal-to-Seasonal (S2S) forecasts of atmospheric rivers (ARs) and precipitation over the western United States. Demand for improved S2S forecast skill in this region is fundamentally driven by its considerable interannual variability of precipitation, along with a rapidly growing population, an increasing desire to optimize resource management in a warming climate, and other socioeconomic and policymaking considerations. In this talk, specific decisions and actions made by water resource managers that would be affected by improved S2S forecasts of ARs, ridging, and precipitation will be identified. In addition, an overview of key research and experimental forecast product development efforts at CW3E to meet these stakeholder needs will be presented.
Seasonal Prediction with Ocean Eddy Resolving Models
Presenter: Dr. Ben Kirtman, University of Miami - CIMAS
Increasingly, high resolution observations and coupled model experiments with eddy-resolving oceans indicate that western boundary currents (WBCs) are regions of strong ocean-atmosphere interactions that are critical components of the climatic mean state and variability. The high SSTs and strong SST gradients couple with the atmosphere to pump moisture into the marine boundary layer, accelerate winds, sharpen SST fronts, and introduce significant decadal climate variability that affects the frequency and intensity of extreme events (e.g., heat waves, cold spells, droughts, floods, extreme winds) at remote locations. This talk describes a set in retrospective sub-seasonal to seasonal prediction, predictability and simulation experiments using an ocean eddy resolving version of CESM. We described how the mesoscale features in the ocean feedback onto representation of the large-scale mean state, the simulation of sub-seasonal to seasonal variability, the local imprint of large scale-climatic features (associated extreme events) and retrospective forecast skill. Similar numerical experiments with an eddy parameterized version will be used for comparison.
02/01/2021
MJO Prediction Skill and Eddy Meridional Mixing of Moisture: Insights From a Model with Superparameterized Physics
Presenter: Dr. Stefan Tulich, NOAA/OAR/ESRL/PSL - CIRES
There is a growing appreciation of the importance of eddy meridional mixing of moisture towards MJO simulation and prediction. However, it remains unclear as to what determines the strength of such mixing. Here it is proposed that a key factor is the efficiency of warm rain production in convecting regions through its effects on the strength of convection coupling to synoptic-scale rotational eddy circulations. Evidence to support this idea is documented in a series of superparameterized hindcast simulations of the MJO, with results showing a strong relationship between MJO prediction skill, meridional eddy moisture transports, and the efficiency of warm rain production, where the latter is modulated through changes in microphysical parameter settings. These findings may help to explain why MJO fidelity in models is often found to be associated with bulk metrics of precipitation-moisture coupling.
A Toolbox for Verification and Validation of Stratosphere-Troposphere Coupling Processes in Current and Future NOAA S2S Forecast Systems
Presenter: Dr. Zachary Lawrence, NOAA/OAR/ESRL/PSL - CIRES
Two-way coupling between the stratosphere and troposphere can be an important source of predictive skill in the troposphere on subseasonal-to-seasonal (S2S) timescales. However, S2S forecast models often have significant temperature and circulation biases in the stratosphere that can negatively affect the representation of stratospheric variability and stratosphere-troposphere coupling. It is thus important to evaluate and document models’ representations of the stratosphere and stratosphere-troposphere coupling processes in order to properly leverage such information for S2S forecasts, and improve future versions of S2S models. Here we will discuss our ongoing development of a comprehensive toolbox for verifying and validating stratosphere-troposphere coupling processes for NOAA S2S models, and the preliminary application of it to the recently released set of 35-day GEFSv12 hindcasts. We will also discuss our plans to transition this work to operations so that stratospheric information is regularly monitored for useful applications to weeks 3-4 forecasts, and regularly validated with future versions of NOAA models.
01/04/2021
Skillful Subseasonal Prediction of the United States Extreme Warm Days and Standardized Precipitation Index in Boreal Summer
Presenter: Dr. Douglas Miller, University of Illinois
Skillful subseasonal prediction of extreme heat and precipitation greatly benefits multiple sectors, including water management, public health, and agriculture, in mitigating the impact of extreme events. A statistical model is developed to predict the weekly frequency of extreme warm days and 14-day standardized precipitation index (SPI) during boreal summer in the United States (US). We use a leading principal component of US soil moisture and an index based on the North Pacific sea surface temperature (SST) as predictors. The model outperforms the CFSv2 at weeks 3-4 in the eastern US. It is found that the North Pacific SST anomalies are excited by a persistent Circumglobal Teleconnection (CGT) pattern which allows the SST anomalies to persist several weeks. Extreme values of the SST index are shown to occur simultaneously with a wave train spanning from the Pacific to the eastern US, which leads to an increased frequency of blocking and extreme temperature occurrence. Extreme dry soil moisture conditions persist into week 4 and are associated with an increase in sensible heat flux which may help maintain the overlying anticyclone.
MJO-ENSO Interaction in an Energetic Framework
Presenter: Dr. Cristiana Stan, George Mason University
The relationship between wind power and ENSO energetics is expanded and applied to study the impact of MJO on ENSO. In this framework, the MJO wind stress exerts a conditional influence on the oceanic Kelvin wave activity. The necessary condition for MJO-El Niño interactions is a coherent phasing between the westerly MJO winds and downwelling Kelvin waves. The composite of El Niño events for which this condition is satisfied shows a greater magnitude of mean perturbation wind, buoyancy power, and available potential energy than the composite of El Nino events in which the MJO wind power is out of phase with the oceanic Kelvin waves. The ENSO events affected by the MJO wind stress show an earlier onset of a flattened, El-Niño like state of thermocline and greater sea surface temperature (SST) anomalies at the peak of the event. In contrast, El-Niño events not influenced by the MJO winds show a later onset, and are dominated by a transient-like thermocline along with periods of upwelling Kelvin wave activity. The MJO wind power, measured by the covariability of MJO‐related wind stress and oceanic Kelvin wave activity, is combined with the SST anomalies, and two new indices are proposed. The MJO‐Kelvin wave‐ENSO (MaKE) index is proposed as an ENSO predictor and is shown to slightly outperform Niño 3.4 when applied to observational data sets and to greatly outperform Niño 3.4 when applied to CFSv2 reforecasts of the years 1980–2014. The MJO‐Kelvin wave Influence (MaKI) index is proposed to predict MJO influence on developing El Niño events. This index performs reasonably well when applied to observations. The forecast skill of MaKI in the CFSv2 reforecasts suggests that this model does not predict the observed MJO‐ENSO relationship as measured by this index.
11/02/2020
The Role of Air-sea Coupling in SubX
Presenter: Dr. Ben Kirtman, University of Miami - CIMAS
This talk examines the role of air-sea coupling in the SubX forecasts from two perspectives. First, we examine how current dynamical prediction systems forecast individual MJO events that occurred during the Dynamics of the MJO (DYNAMO) field campaign and the effects of ocean-atmospheric coupling on the forecasts of these events. Previous studies have shown the apparent need for active coupling in models to produce a realistic MJO, while others only need atmospheric dynamics. Past work on the DYNAMO MJO events has demonstrated that the October event (MJO1) did not need ocean-atmospheric coupling (an “uncoupled” event) for maintenance and propagation and the November event (MJO2) needed ocean-atmospheric coupling (a “coupled” event). We group the SubX models into a coupled ensemble and an uncoupled ensemble to examine the effects of ocean-coupling on the forecasts of these two events. Second, we examine the evolution of local air-sea feedbacks in mid--latitude region western boundary currents, and how these feedbacks are affected by resolution.
Sources of Tropical Subseasonal Skill in CFSv2
Presenter: Dr. Carl Schreck, North Carolina State University
This study applies Fourier filtering to a combination of rainfall estimates from TRMM and forecasts from the CFSv2. The combined data are filtered for low-frequency (LF, ≥120 days) variability, the MJO, and convectively coupled equatorial waves. The filtering provides insight into the sources of skill for the CFSv2. The LF filter, which encapsulates persistent anomalies generally corresponding with SSTs, has the largest contribution to forecast skill beyond week 2. Variability within the equatorial Pacific is dominated by its response to ENSO, such that both the unfiltered and the LF-filtered forecasts are skillful over the Pacific through the entire 45-day CFSv2 forecast. In fact, the LF forecasts in that region are more skillful than the unfiltered forecasts or any combination of the filters. Verifying filtered against unfiltered observations shows that subseasonal variability has very little opportunity to contribute to skill over the equatorial Pacific. Any subseasonal variability produced by the model is actually detracting from the skill there. The MJO primarily contributes to CFSv2 skill over the Indian Ocean, particularly during March–May and MJO phases 2–5. However, the model misses opportunities for the MJO to contribute to skill in other regions. Convectively coupled equatorial Rossby waves contribute to skill over the Indian Ocean during December–February and the Atlantic Ocean during September¬–November. Convectively coupled Kelvin waves show limited potential skill for predicting weekly averaged rainfall anomalies since they explain a relatively small percent of the observed variability.
10/05/2020
Hybrid Prediction of Weekly Tornado Activity out to Week 3: Utilizing Weather Regimes
Presenter: Dr. Zhuo Wang, University of Illinois
Tornadoes are one of the high-impact weather phenomena that can induce life loss and property damage. Here, we investigate the relationship between large-scale weather regimes and tornado occurrence in boreal spring. Results show that weather regimes strongly modulate the probability of tornado occurrence in the United States due to changes in shear and convective available potential energy, and that persisting weather regimes (lasting 3 days) contribute to greater than 70% of outbreak days (days with 10 tornadoes) . A hybrid model based on the weather regime frequency predicted by a numerical model is developed to predict above/below normal weekly tornado activity and has skill better than climatology out to week 3. The hybrid model can be applied to real-time forecasting and aide in mitigation of severe weather events.
Predictability of the UFS over CONUS at Subseasonal Time Scale during Boreal Summer
Presenter: Dr. V. Krishnamurthy, George Mason University
The predictability of the Unified Forecast System (UFS) Coupled Model Prototype 2, developed by the National Centers for Environmental Prediction, is assessed over the continental United States (CONUS) for the boreal summer. The retrospective forecasts of low-level horizontal wind, precipitation and 2m temperature for 2011–2017 are examined to determine the predictability at subseasonal time scale. Using a data-adaptive method, the leading modes of variability are obtained and identified to be related to El Niño-Southern Oscillation (ENSO), intraseasonal oscillation (ISO) and warming trend. The sources of better predictability are identified by examining the forecast errors and correlations in the weekly averages of the leading modes of variability. During the boreal summer, the ISO followed by the trend in UFS are found to provide better predictability in weeks 1–4 compared to the ENSO mode and the total anomaly. The western part of the CONUS seems to have better predictability on weekly time scale in all the three modes.