Subseasonal-to-seasonal forecast skill in the California Current System and its connection to coastal Kelvin waves
The following summarizes work published in JGR: Oceans
Major Results:
The ECWMF S2S forecasting model skillfully predicts sea level along the US west coast from Weeks 3-7.
The presence of coastally trapped waves at initialization enhances overall forecast skill.
We have developed a coastal Kelvin wave index to capture the location and amplitude of these waves as they propagate up the North American coastline in real-time.
The California Current System (CCS) is home to a diverse range of marine resources, and the management of these resources could benefit from accurate forecasts of coastal ocean conditions on subseasonal-to-seasonal (S2S) timescales. For example, rapid changes in coastal sea surface height (SSH) can lead to changes in species distribution that may increase the risk of ship strikes and unwanted bycatch.
Figure 1 Forecast skill of coastal SSH in the South, Central, and North CCS. Red stipples indicate significant skill over persistence.
In the South CCS, the forecast model has significant skill above persistence at leads above ~15-45 days when initialized in the summer/fall (Fig. 1a)
In the North CCS, the forecast skill is overall lower, however, there is dynamical skill at shorter leads (~5-10 days). This is due to stronger weather noise at higher latitudes, which reduces day-to-day SSH persistence.
Figure 2 Sea surface height anomalies during the development of the 1997-1998 El Niño. Animation depicts a Kelvin wave as it travels from the equatorial Pacific up the coastline of North America. Data is from the GLORYS ocean reanlaysis.
It is often useful to consider physical processes that may provide opportunities for enhanced forecast skill.
In the CCS, one such physical process may be related to coastally trapped waves (potentially originating from the equatorial Pacific). See the animation in Figure 2 for an example.
We test this hypothesis by creating a coastal Kelvin wave (C-KW) index based on EOFs of SSH anom. along the N. American coastline.
The C-KW index credibly captures the intensity and location of propagating coastally trapped waves from 1993-2019. See our paper for more details.
We are developing real-time version of the C-KW index for operational use. Check back for more details soon!
Using the C-KW index, we composite dynamical forecast skill based on the intensity of coastal Kelvin wave conditions at forecast initialization (i.e., either Weak, Strong, or Extreme Kelvin wave conditions).
We find that dynamical forecast skill explains 30-40% more SSH variance in Weeks 3-7 than the corresponding persistence forecast.
Additionally, we find that forecasts initialized with significant coastal KW conditions explain a 6-10% more variance than forecasts without Kelvin wave activity.
Results are consistent for the Central and North CCS.
Figure 3 Weekly forecast skill composited on the intensity of coastal Kelvin wave conditions at forecast initialization. Solid lines = dynamical forecasts. Dashed lines = persistence forecasts. Filled circles = dynamical forecast skill is significantly better than persistence.