Abstract ID: 177
A Linear Inverse Model for Improved Model Guidance of CPC’s Week 3-4 Operational Temperature Outlooks
Lead Author: Matthew Newman
NOAA/Physical Sciences Laboratory, United States of America
Keywords: operational prediction, Weeks 3-4, empirical models, S2S predictability, forecasts of opportunity
Abstract: We present a recently developed Linear Inverse Model (LIM) that makes year-round predictions of weekly 2m air temperature (T2m) anomalies for North America at leads of 3-4 weeks. This LIM was specifically designed to provide improved model guidance for NOAA’s Climate Prediction Center (CPC) Week 3-4 Temperature outlooks, including a priori identification of forecasts of opportunity. The model is constructed from 7-day running mean anomalies of North American T2m, Northern Hemisphere tropospheric geopotential height, and tropical heating, all obtained from the JRA-55 reanalysis for the years 1979-2016. For the cold season, the state vector additionally includes 7-day running mean anomalies of Northern Hemisphere stratospheric geopotential height, whereas for the warm season it includes 7-day running mean anomalies of North American root zone soil moisture. Its cross-validated hindcast skill is comparable to operational models at NWS and ECMWF, obtained from s2sprediction.net.
Working with CPC, we have applied this version of the LIM as a probabilistic and deterministic forecast system focused on continental United States (CONUS) T2m prediction, although the LIM also yields forecasts of all its state variables. The LIM forecast system runs in near realtime, initialized with JRA-55 reanalysis data made available to NCAR RDA three days behind. We have evaluated skill of the LIM retrospective forecasts (starting in 2017, after the LIM training period) using both deterministic and probabilistic measures. Notably, LIM reliability (for both two-category forecasts, as issued by CPC, and three-category forecasts) is excellent, which gives confidence in its probabilistic predictions. Mean Heidke skill since 2017 is comparable to official CPC forecasts, with RPSS skill generally better. Additionally, for some forecast events the LIM was considerably more skillful (including the notable North American cold air outbreak in February 2021), suggesting that its model guidance provides new predictive information that could aid forecasters in the future. Lessons learned (or still to be learned) from transitioning the LIM to an operational environment will be discussed, including: (1) How to best leverage for forecasters’ use the LIM’s ability to identify forecasts of opportunity by predicting its own skill, which is particularly important for exploiting the relatively small amount of skill available on the Weeks 3-4 time scale; and (2) Decomposing the forecast into different predictable components — specifically, eigenmodes of the LIM operator representing key dynamical processes including ENSO and stratospheric sudden warmings – that allow forecasters to use physical understanding in support of operational predictions.
Co-authors:
John Albers (University of Colorado/CIRES and NOAA/PSL)
Hui Wang (NOAA/CPC)
Maria Gehne (University of Colorado/CIRES and NOAA/PSL)
