The Role of Deep Mesoscale Eddies in Ensemble Forecast Performance

TL;DR

This study shows that deep ocean dynamics substantially influence SSH forecast skill in the Gulf of Mexico's Loop Current system during the LCE Thor separation. By analyzing two 92-day ensemble forecasts with a 32-member EF and comparing deep-field initial conditions against CPIES observations and CMEMS SSH, the authors demonstrate that smaller initial uncertainty in the deep field, represented by $η_{ref}$, yields lower upper-ocean SSH RMSE weeks into the forecast. They develop a simple yet robust best/worst member ranking based on SSH RMSE and show that deep eddy structure, particularly deep cyclones and anticyclones around the Mississippi Fan and Deep Southeast Channel, strongly guides LCE positioning and detachment timing. The findings advocate for assimilating deep observations to constrain deep initial fields, improving forecast performance, and suggest future OSSE studies to generalize this result to other boundary-current regimes.

Abstract

Present forecasting efforts rely on assimilation of observational data captured in the upper ocean (< 1000 m depth). These observations constrain the upper ocean and minimally influence the deep ocean. Nevertheless, development of the full water column circulation critically depends upon the dynamical interactions between upper and deep fields. Forecasts demonstrate that the initialization of the deep field is influential for the development and evolution of the surface in the forecast. Deep initial conditions that better agree with observations have lower upper ocean uncertainty as the forecast progresses. Here, best and worst ensemble members in two 92-day forecasts are identified and contrasted in order to determine how the deep ocean differs between these groups. The forecasts cover the duration of the Loop Current Eddy Thor separation event, which coincides with available deep observations in the Gulf. Model member performance is assessed by comparing surface variables against verifying analysis and satellite altimeter data during the forecast time-period. Deep cyclonic and anticyclonic features are reviewed, and compared against deep observations, indicating subtle differences in locations of deep eddies at relevant times. These results highlight both the importance of deep circulation in the dynamics of the Loop Current system and more broadly motivate efforts to assimilate deep observations to better constrain the deep initial fields and improve surface predictions.

Figures (11)

• Figure 1: The Gulf of Mexico with the region of interest (red box), which ranges from $22^\circ$N to $28^\circ$N, $90^\circ$W to $83^\circ$W. The Understanding Gulf Ocean Systems CPIES array (black, yellow-filled circles), and names of key features are included. Bathymetry is contoured every $500$ meters.
• Figure 2: Scatter plot of RMSE of $\eta_{ref}$ from different forecasts at Week 0 compared against the observations from the CPIES array, plot against RMSE of SSH for each forecast's respective a) Week 1, b) Week 2, c) Week 3, d) Week 4, e) Week 5, and f) Week 6. Deep mean fields for $\eta_{ref}$ are shown in Fig. \ref{['fig:IC_vs_wk6']} for the filled symbols.
• Figure 3: Deep ensemble mean $\eta_{ref}$ fields for EF20191028 (top row) and EF20191125 (bottom row), exemplifying a "better" initial condition and a "poor" initial condition, respectively. Model fields are contrasted against the CPIES observations at Week 0 (left columns) and Week 6 (right columns). The purple lines represent the 17-cm contour for the forecast, and the black lines at Week 6 represent the 17-cm contour for the verifying analysis.
• Figure 4: Schematic flow chart for member selection.
• Figure 5: Weekly RMSE of SSH for EF20191028 compared against a) the verifying analysis and b) the CMEMS satellite product. Only the best (blue, solid lines) and worst (orange, solid lines) members are shown. The range (gray, shaded-region) is bounded by the weekly maximum and minimum values of RMSE. The mean of RMSE values (black, solid line) and one standard deviation above and below this mean (black, dashed lines) are provided.