Mesoscale SST–wind stress coupling in the Peru–Chile current system: Which mechanisms drive its seasonal variability?

Abstract

Satellite observations and a high-resolution regional ocean–atmosphere coupled model are used to study the air/sea interactions at the oceanic mesoscale in the Peru–Chile upwelling current system. Coupling between mesoscale sea surface temperature (SST) and wind stress (WS) intensity is evidenced and characterized by correlations and regression coefficients. Both the model and the observations display similar spatial and seasonal variability of the coupling characteristics that are stronger off Peru than off Northern Chile, in relation with stronger wind mean speed and steadiness. The coupling is also more intense during winter than during summer in both regions. It is shown that WS intensity anomalies due to SST anomalies are mainly forced by mixing coefficient anomalies and partially compensated by wind shear anomalies. A momentum balance analysis shows that wind speed anomalies are created by stress shear anomalies. Near-surface pressure gradient anomalies have a negligible contribution because of the back-pressure effect related to the air temperature inversion. As mixing coefficients are mainly unchanged between summer and winter, the stronger coupling in winter is due to the enhanced large-scale wind shear that enables a more efficient action of the turbulent stress perturbations. This mechanism is robust as it does not depend on the choice of planetary boundary layer parameterization.

Appendix: Double time averaging of the momentum balance

The mechanisms driving the feedback of the SST anomalies on the wind speed are investigated. Monthly mean wind speed anomalies $\langle V \rangle '$, proportional to the monthly mean SST anomalies $\langle SST \rangle '$, are observed in our simulation (with primes marking the mesoscale anomalies and $\langle \rangle$ the temporal average). We want to identify the dominant mechanism that creates $\langle V \rangle '$. In this appendix, we explain why a simple time averaging of a momentum balance does not explain the mean wind speed. Then we present the double time averaging that should be used. It is similar to the one included in the NEMO code (Madec 2008).

A simple time-average of 1D momentum balance:

$$\partial _{t} {V}=\sum \limits _{F_{n}\in \{Forces\}}F_{n}$$

(14)

relates the forcing time average to the difference between the final and initial wind speed but not to the average wind speed $\langle V \rangle$, which is the variable of interest:

$$\sum \limits _{F_{n}\in \{Forces\}} \langle F_{n}\rangle =\langle \partial_{t} {V} \rangle = \left\langle \frac{{\Delta } V}{{\Delta } t} \right\rangle =\frac{V(0)-V(N_{step})}{{\Delta } t}$$

(15)

with V(p), the wind speed p time steps after the beginning of the month, $N_{step}$ the number of time steps during July, ${\Delta } t$ the time step duration, and ${\Delta } V=V(p)-V(p-1)$ the wind speed difference between 2 time steps.

The monthly mean wind speed is $\langle V \rangle =\frac{1}{N_{step}+1}\sum \nolimits _{p=0}^{N_{step}}V(p)$ and V(p) can be expressed using the initial conditions V(0): $V(p)=V(0)+ \sum \nolimits _{k=1}^{p}{\Delta } V$, so, we obtain:

$$\begin{aligned} \langle V \rangle&=\frac{1}{N_{step}+1}\sum \limits _{p=0}^{N_{step}}(V(0)+ \sum \limits _{k=1}^{p}{\Delta } V) \nonumber \\&=V(0)+\frac{1}{N_{step}+1}\sum \limits _{p=0}^{N_{step}}(\sum \limits _{k=1}^{p}{\Delta } V) \end{aligned}$$

(16)

We introduce a new metric $\lceil F \rceil$, the double time averaging of a quantity F, defined as:

$$\begin{aligned} \lceil F \rceil =\frac{1}{N_{step}+1}\sum \limits _{p=0}^{N_{step}}(\sum \limits _{k=1}^{p}F) \end{aligned}$$

(17)

(16) can be written $\langle V \rangle =V(0)+\lceil {\Delta } V \rceil$, i.e.

$$\begin{aligned} \frac{\langle V\rangle - V(0)}{{\Delta } t}=\lceil \frac{{\Delta } V}{{\Delta } t} \rceil \end{aligned}$$

(18)

$\lceil \,\rceil$ is a linear operator, so, using (14), we obtain:

$$\begin{aligned} \frac{\langle V\rangle - V(0)}{{\Delta } t}=\lceil \frac{{\Delta } V}{{\Delta } t} \rceil =\lceil \partial _{t}V \rceil =\sum \limits _{F_{n}\in \{Forces\}} \lceil F_{n}'\rceil \end{aligned}$$

(19)

The left-hand side represents the mean temporal variation around the initial state V(0). The relative contribution of $\lceil F_{n} \rceil$ indicates the dominant mechanisms.