Anti-Seasonal Outlook, Seasonal Outlook: Part II

How are NOAA’s seasonal climate outlooks developed?

The Climate Prediction Center (CPC), a branch of NOAA, issues the operational seasonal climate and drought outlooks for the US.  Figure 1 shows the official US seasonal drought outlook that is valid through the end of January, 2015.  This map shows the likelihood of drought development, removal, or persistence.  Pretty sad looking for much of CA and NV.  However, these forecasts must be taken with a grain of salt, as skill beyond ~two weeks remains quite low.  Remember, these are climate outlooks, and not weather forecasts.  The CPC is not predicting individual storm events, but rather departures from mean climate states.  So, how are these maps developed?  Well, there is one dynamical model (Climate Forecast System Version 2 (CFSv2); based on atmospheric physics and weather dynamics), and a variety of statistical methods that use historical conditions and patterns to predict the future (If you are interested in the technical details, follow this link: http://www.cpc.ncep.noaa.gov/products/predictions/90day/tools.html).  Statistical models can be just as reliable as dynamical models at long lead times because the atmosphere is a nonlinear, chaotic system that is not perfectly predictable by any means.  A limited set of variables are used for seasonal prediction, and typically include temperature, precipitation, soil moisture, and sea surface temperature.

Figure 1. CPC US seasonal drought outlook. Source: http://www.cpc.ncep.noaa.gov/products/expert_assessment/sdo_summary.html

Experimental Seasonal Forecasts

Multiple dynamical models (as opposed to just CFSv2) can also be used to reduce the uncertainty of a single model.  The average forecast is taken from several different models, and is termed an ensemble forecast.  Studies have found the ensemble approach to typically be more accurate than one model, but skill at seasonal lead times remains low.  The North American Multi-Model Ensemble (NMME) is an experimental project that provides individual forecasts and an ensemble forecast for eight different models, including CFSv2.  Figure 2 shows the November average precipitation anomaly (mm/d) from the NMME ensemble and three members of the ensemble.  Red indicates abnormally dry and green indicates abnormally wet. If you just look at the ensemble, NMME shows slightly above normal precipitation for CA.  But look at how contrasting the three individual members are!  CFSv2 is extremely dry, CMC2 is extremely wet, and the NASA model splits the state of CA about 50/50, take your pick!  The ensemble basically smooth’s out the contrast found amongst the eight members.

Figure 2. November 2014 precipitation anomaly forecast from NMME ensemble and 3 individual 
members.  For all ensemble members and more forecast see: http://www.cpc.ncep.noaa.gov/products/NMME/monanom.shtml.

Come back to this post at the end of November and see which model wins!  There are an awful lot of resources out there to look at seasonal predictions, but none of them are reliable at this point.  One thing that can help seasonal prediction skill is if a model is initialized (start date of the model run) during an ENSO event (El Niño or La Niña).  Unfortunately, we are ENSO neutral right now, so the current seasonal predictions may end up being quite poor.