AccuWeather has a global long-range forecast team that studies current and projected tele-connections and climate relationships and creates a set of analog years that closely match observed and climate model output weather trends. Forecast ideas are submitted by each team member and a consensus is then reached.

The Atlantic Support Vector Regression (ASVR) model is a collaboration between researchers from the University of Oklahoma, the Commonwealth Science and Industrial Research Organisation (CSIRO), and the University of Technology Sydney. Seasonal forecasts of the number of named storms are based on a machine learning technique, Support Vector Regression, using monthly sea surface temperature anomaly (SSTA) data during April and May as the basis for the initial predictor selection. The first experimental forecast was issued for the 2017 hurricane season. Work is underway to extend the model forecast to statistics, other than just the seasonal number of named storms, to include the Accumulated Cyclone Energy (ACE) and the number of hurricanes.

Climate Analytics (CA) issues an Atlantic basin seasonal tropical cyclone forecast early April that is updated early June. The forecast is based on robust precursors of low vertical wind shear and high sea surface temperatures during the season. The robust precursors are identified by applying causal effect networks based on a methodology developed by Runge et al. A simple linear model is used to predict the July-October accumulated cyclone energy (ACE). More details on the methodology are provided in Pfleiderer et al.

Coastal Carolina University (CCU) has issued North Atlantic Ocean Basin seasonal Tropical Cyclone and Hurricane outlooks since 2013. The model system predicts family genesis numbers along with the probability of hurricane land-falls on U.S. coastlines. The forecasts utilize historical data to develop statistical and empirical blended mathematical models, which have demonstrated significant levels of skill in the prediction of hurricane seasons back to 1950. An interactively coupled model suite also forecasts the track, intensity and coastal surge, flood and inundation of any incoming hurricane.

Colorado State University (CSU) has issued Atlantic basin seasonal hurricane forecasts since 1984, primarily using statistical techniques. These forecasts utilize historical data to develop statistical models which show significant levels of skill in the prediction of past hurricane seasons. Their models also utilize currently-available dynamical model forecasts for large-scale atmosphere/ocean conditions as additional predictions into CSU's seasonal hurricane outlooks.

The Oceans and Climate Lab at Colorado University (CU) Boulder issues predictions of named storms based on the spatial pattern and intensity of outgoing longwave radiation (OLR) over Africa. The scientific basis and prediction methods used are published in the peer-reviewed journal Geophysical Research Letters (Karnauskas 2006; Karnauskas and Li 2016).

Commodity Weather Group's (CommodityWx) forecasting technique involves parsing out similar ENSO evolution situations through peak hurricane season (August-October). Those cases are further refined to reflect sea surface temperatures in the main development region in the Atlantic.

The Cuban Institute of Meteorology (InsMet) developed its first statistical seasonal forecast model in 1995, and the issuance of these forecasts began in 1996. This methodology is based on the solution of a linear regression equation set, and the evaluation of a discriminant function, according to climatological and analogical criteria. This scheme considers ENSO as the fundamental modulator of TC activity in the North Atlantic and also considers the atmospheric circulation during March and April in the North Atlantic and the SSTs in the tropical Atlantic and the Caribbean Sea.

Data Transmission Network (DTN)’s Atlantic hurricane season forecast incorporates at a minimum, factors such as the El Niño Southern Oscillation evolution in the equatorial Pacific Ocean, sea surface temperature anomaly patterns in the months leading up to hurricane season and those projected during the season, the Arctic Oscillation, and the Atlantic Multi-decadal Oscillation. Previous seasonal records with similar patterns in place are used as a proxy to determine whether a below-, near-, or above-normal season is expected. Finally, the forecast is further refined using various seasonal model guidance to view individual contributions from sea level pressure, wind shear and precipitation anomalies to determination the expected frequency of systems and intensity spectrum.

Earth Networks has provided annual Atlantic hurricane forecasts since 2007. Developed in April and issued in early May, our forecasts employ sophisticated statistical models in combination with historical analogs chosen by our team of meteorologists. Our statistical models are built using several key atmospheric and oceanic predictors that have shown skill in forecasting seasonal Accumulated Cyclone Energy, the number of named storms, hurricanes, and major hurricanes. Beginning in 2021, forecasts will also include United States seasonal hurricane landfall probabilities and projections broken down by the Gulf of Mexico, Florida, and the East Coast regions.

European Centre for Medium-Range Weather Forecasts (ECMWF) dynamical forecast based on System 5.

Global Weather Oscillations (GWO) utilizes proprietary ClimatePulse technology that tracks specific physical mechanisms which in turn cause forcing of the earth’s climate, atmosphere, and oceanic processes – including ENSO events. The historical ClimatePulse cycles are predictable into the future and are correlated with historical hurricane landfalls and then tracked from the late 1800's to present - and into the future. GWO issues Atlantic Basin hurricane landfall predictions in January.

The National Oceanic and Atmospheric Administration (NOAA) has been issuing Atlantic hurricane season outlooks since August 1998. These outlooks are based on predictions of the main climate factors known to influence seasonal Atlantic hurricane activity. Prediction tools involve statistical regression and climate-based analogues, dynamical model predictions from the NOAA Climate Forecast System (CFS) and the NOAA Geophysical Fluid Dynamics Lab (GFDL) models FLOR-FA and HI-FLOR, and statistical/ dynamical hybrid forecast tools which are also based on the CFS.

The forecast from the North American Multi-Model Ensemble (NMME) is based on a multiple regression relationship between the observed Atlantic hurricane season activity and two predicted circulation variables. The two predictors used in this model are the mean forecasted August to October vertical wind shear over the main development region and the preseason observed North Atlantic sea surface temperature.

The interdisciplinary group of North Carolina State University (NCSU) Department of Statistics and Department of Marine, Earth and Atmospheric Sciences has been releasing annual hurricane prediction forecasts since 2005. The seasonal outlook is made by using a Poisson regression model, along with variable selection techniques. The model is trained and validated using historical data. The main predictors in the model include January-February measures of sea surface temperature in the Atlantic and forecasted ENSO index measures.

Penn State University's (PSU) Earth System Science Center has been issuing annual seasonal tropical cyclone forecasts for the North Atlantic basin since 2007. Seasonal forecasts are made by using a multivariate Poisson regression model, trained on corrected historical tropical cyclone counts and climate predictors, and published in the Journal of Geophysical Research in 2012. The primary variables considered by the statistical model in these forecasts include sea surface temperatures in the main development region and anticipated El Nino conditions.

Radiant Solutions (Maxar) has been issuing North Atlantic tropical seasonal forecasts to its energy and agriculture customers since the early 2000s. The forecast system we leverage is a combination of inputs from physical/dynamical weather and climate models as well as inputs from statistical pattern recognition/matching algorithms. The forecast system is dynamically tuned to capture signals around the world that have proven to show skill as leading indicators to tropical activity (number of events and magnitude of events). These signals originate from time lagged indicators of observed weather as well as indicators from the expected patterns sourced from the physical/dynamical models. The statistical algorithms then weight these various inputs based on historical performance over a variety of categories to ultimately produce the blended output that is then used for the annual Official MDA North Atlantic Tropical Seasonal Forecast.

reask forecasts tropical cyclone activity globally using a Bayesian hierarchical model with machine learning based climate predictors. Forecasts for all 4 Northern hemisphere basins are released in early June with the Southern hemisphere basins following in October. In an effort to communicate model uncertainty, reask's forecasts always provide the full distribution of projected cyclone activity.

Seoul National University (SNU) forecasts are based on the weighted sum of tropical cyclones and hurricanes computed for four distinct regions of the Atlantic basin as estimated from an empirial model applied to NCEP CFSv2 operational seasonal forecast output for the months of August-October. The predictors for this model are the seasonal averaged sea surface temperature, zonal wind at 200 and 850 hPa, magnitude of vertical wind shear and relative vorticity at 850 hPa. These final seasonal predictions are calculated by adding the recent 10-yr (2007–2016) climatology of genesis frequencies in June, July and November to our ASO predictions.

Servicio Meteorológico Nacional de México (SMN) - Description not available.

Description coming soon.

StormGeo has been issuing hurricane seasonal outlooks starting in 2007 after devastating 2005 Atlantic tropical cyclone season. We use a blend of both statistical and dynamical models that incorporate current and projected trends of surface and upper air pressure anomalies associated with forecast water temperature anomalies over the Pacific and Atlantic Oceans. A key part of this process is to identify analog seasons that best represent current and projected atmospheric pressure trends over the Atlantic Basin and apply a weighting factor for each analog in order to derive the seasonal forecast.

The Weather Company (WSI-TWC) has been issuing seasonal tropical forecasts for the North Atlantic basin since 2006, using an optimized blend of dynamical models and proprietary statistical models. The statistical models have been built using historical values of various relevant earth-atmosphere indices along with observed historical activity levels in the tropical Atlantic. The skill of both the dynamical and statistical models has been evaluated and weights are assigned to each model in producing the final blended forecast.

Tropical Storm Risk (TSR), based at University College London in the UK, has issued seasonal outlooks for North Atlantic hurricane activity since December 1998. TSR predicts basin-wide tropical cyclone (TC) activity (numbers of storms of different strengths and the ACE index), U.S. landfalling TC activity, and Caribbean Lesser Antilles landfalling TC activity. The TSR prediction models are statistical in nature and are underpinned by atmospheric and oceanic predictors that have sound physical links to contemporaneous TC activity.

The UK Met Office (UKMO) has been issuing seasonal forecasts of Atlantic tropical storm activity since 2007. Forecasts are created using the Met Office global seasonal forecasting system, GloSea. This system uses current observations of the ocean, land and atmosphere and simulates these over the next 7 months to provide a forecast of tropical storm activity. Multiple forecasts are created and combined to produce a best-estimate forecast as well as a forecast range.

The University of Arizona (UA) forecasts the mean number and a range of hurricanes based on a statistical method relying on observational measurements from March to May over the Atlantic and Pacific. The predictors considered include the April–May multivariate ENSO index (MEI) conditioned upon the Atlantic multidecadal oscillation (AMO) index, the average zonal pseudo–wind stress across the North Atlantic in May and the average March–May tropical Atlantic sea surface temperature.

The University of Missouri (MU) began issuing seasonal tropical cyclone outlooks for the Atlantic Ocean Basin in 2018. In-house projections began in 2013. The techniques used are based on a statistical climatology and trends derived from the historical data, as well as teleconnections. Our group also uses current information available about large-scale atmospheric and oceanic conditions, and seasonal forecasts of these states issued using dynamic and statistical models.

Weather 2020, LLC has been issuing seasonal outlooks since 2017. The technology and methodology is based on the cyclicality of weather patterns, following the Cycling Pattern Hypothesis (CPH), also known as the Lezak's Recurring Cycle (Lezak et al. 2019). The methodology allows accurate predictions of when and where tropical systems will likely develop and where they will make landfall. It also provides insights on how active tropical seasons will become months before hurricane season traditionally begins.

Weatherbell issues three seasonal forecasts (March, May, August) by combining analog years with model output. Beside the standard metrics of hurricane activity, we also quantify the seasonal forecasts using a power and impact scale, which we feel is more representative than the usual Saffir-Simpson scale. Seasonal forecasts are made available to a general audience, but customers of Weatherbell are given exclusive access one week prior to the public release.

WeatherTiger's seasonal hurricane forecasts are created using a proprietary statistical modeling engine, TigerTracks. This ensemble scheme is trained using a temporally and spatially expansive set of tropical cyclone, atmospheric, and oceanic historical observations, yielding probabilistic guidance for both measures of overall activity and seasonal landfall risk. TigerTracks employs rigorous checks against overfitting, with uncertainty ranges for all predictands derived from the out-of-sample statistical ensemble.

WeatherWorks has issued seasonal forecasts for the North Atlantic Basin's Hurricane season since 1992. WeatherWorks' forecast process uses a hands-on, blended approach, with significant emphasis on analog years. The analogs are carefully chosen primarily based on their fit to current and anticipated states of various global teleconnections as well as Sea Surface Temperature patterns.

268Weather obtains their forecasts with the use of the Climate Predictability Tool (CPT) version 15.5.10, 2017 by Simon J. Mason and Michael K. Tippett. The software was view in canonical correlation analysis (CCA) mode. Input explanatory (X) files used were NOAA NCDC ERSSTv4 mean SSTs for: March 1971-2018; January to March 1971-2018 and NOAA NCEP EMC CFSv2 ensemble mean SSTs for June to November 1982-2018. The X domain used was 20°S to 30°N and 140°E to 20°W. The response (Y) variables were ACE values, named storms, hurricanes and major hurricanes for the Atlantic Basin (including the Caribbean Sea and the Gulf of Mexico) for the period 1971 to 2017.