Subseasonal to Seasonal (S2S) Experimental Forecasts

A multi-institutional collaboration sponsored by California DWR

Click here to view the latest CW3E S2S Outlook

– Archive of previous S2S Outlooks

Subseasonal Forecasts (Weeks 1-6)

Dynamical Model Atmospheric River Activity Forecasts

A multi-model experimental forecast for AR activity (defined as the # of AR days per week) at week-1, 2, and 3 lead time is shown below for the NCEP dynamical model.

This product was developed in collaboration with NASA’s Jet Propulsion Laboratory.

Weeks 1-2: Shading indicates the odds of AR activity for each day. ARs are defined using the Guan and Waliser (2015) algorithm and probability is calculated by the number of ensemble members predicting an AR at each grid point at 00 UTC on the given forecast day. Click on a panel to open in a new tab, click on title to open seven day panel plot.

Week 3: The top row shows the forecast number of AR days during week-3; the middle row shows the climatological values of AR activity in each model’s hindcast record for the week-3 verification period; the bottom row shows the departure of the AR activity forecast for that same verification period (top panel forecast minus middle panel climatology). For this row, blue values represent higher than average AR activity predicted during week-3; red values represent lower than average AR activity predicted during week-3. Grey rectangles surround grid cells where >75% of forecast ensemble members agree on the sign of the AR activity anomaly with respect to climatology. These regions can be interpreted as having higher confidence in their prediction of week-3 AR activity. The hindcast skill assessment associated with the NCEP, ECMWF, and ECCC hindcast systems is described in DeFlorio et al. 2019b.

Week 1
Week 2
Week 3

Dynamical Model S2S Ridging Forecasts

An atmospheric ridge is defined as elongated area of relatively high atmospheric pressure (American Meteorological Society; 2020). Persistent ridging events play an important role in determining if and where precipitation falls. Ridging forecasts are shown below for lead-times of weeks 1 and 2, weeks 3 and 4 and weeks 5 and 6. These forecasts are for three different ‘ridge types’ that have been shown in recent research [Gibson et al. 2020a; more methodology information here] to strongly influence atmospheric river and precipitation likelihood across the US West Coast. The North-ridge type is typically associated with widespread dry conditions across the entire US West. The South and West-ridge types are typically associated with dry conditions across Southern California and the Colorado River Basin but can also increase the likelihood of wet conditions across the Pacific Northwest. The hindcast skill assessment associated with this outlook is provided in Gibson et al. 2020b.

This product was developed in collaboration with NASA’s Jet Propulsion Laboratory.

Interpretation: In these plots, the left panel shows the occurrence frequency of each ridge type (bars) compared to climatology (horizontal line) for each of the model ensemble members. If over 50% of the ensemble members predict more ridging than expected (for this time of year) then the right panel maps are displayed indicating the likelihood of wetter or drier conditions based on how these ridge types typically influence precipitation.

Forecast Period:  

Sub-seasonal Forecast of North American Weather Regimes

Forecast from IRI.

The graphic shows forecasts of large scale meteorological patterns, issued every day during winter (along x-axis) by the NCEP CFSv2 model. The four large scale “weather regimes” are the West Coast Ridge (reds), Pacific Ridge (blues), Pacific Trough (greens), and Greenland High (yellows), each of which leads to distinct precipitation and temperature conditions over the West. Darker colors indicate more confident forecasts, which run diagonally up the chart from 0 to 45 days ahead. This product was developed with support from NOAA’s Next Generation Global Prediction System (NGGPS) program and the California Department of Water Resources. Please see https://wiki.iri.columbia.edu/index.php?n=Climate.S2S-WRs for the precipitation and temperature signatures of the regimes or Robertson, Vigaud, Yuan & Tippett (2020) for additional details.


Seasonal Forecasts (Beyond week 6)

Seasonal Precipitation Forecast based on a Machine Learning Approach

This CW3E forecast product predicts likely seasonal precipitation anomaly patterns for January-February-March over the western U.S. based on machine learning models and dynamical ensemble predictions. Four different machine learning models (Random Forest, XGBoost, LSTM, and neural networks) were pretrained based on learning relationships between remote ocean and atmospheric patterns and characteristic seasonal precipitation anomalies over the Western United States. Instead of training on observations, the machine learning models were here trained on a large climate model ensemble, which enabled more robust relationships to be established due to the very large sample size in the training phase (Gibson et al. 2021). Seasonal forecast skill is also enhanced by focusing on four characteristic large-scale precipitation patterns across the Western United States. Our historical skill assessment found that the probability of a correct forecast improves when the Machine learning models show stronger consensus for a particular pattern. As such, the ensemble agreement (or lack of) provides useful guidance on forecast uncertainty.

This product was developed in collaboration with NASA’s Jet Propulsion Laboratory.

Top row: Four precipitation anomaly patterns identified in Gibson et al. 2021. Percentages represent the number of ensemble models predicting that pattern for January – March 2022 with the red outline and plot size representing the most predicted pattern.

Middle row: Three-month precipitation anomaly pattern forecasts from the four CW3E developed machine learning models valid for January – March 2022.

Bottom row: Three-month precipitation anomaly forecasts from five North American Multi-Model Ensemble (NMME) models valid for January – March 2022. The associated precipitation pattern for each forecast is noted in the bottom right of each image.

Possible Precipitation Patterns

CW3E Machine Learning Models

North American Multi-Model Ensemble

For additional products and more information visit the Odds of Reaching 100% of Normal Water Year Precipitation webpage.


All products displayed on this page are considered experimental.