Weakening of the Senegalo–Mauritanian upwelling system under climate change

Abstract

Upwelling processes bring nutrient-rich waters from the deep ocean to the surface. Areas of upwelling are often associated with high productivity, offering great economic value in terms of fisheries. The sensitivity of spring/summer-time coastal upwelling systems to climate change has recently received a lot of attention. Several studies have suggested that their intensity may increase in the future while other authors have shown decreasing intensity in their equatorward portions. Yet, recent observations do not show robust evidence of this intensification. The Senegalo-Mauritanian upwelling system (SMUS) located at the southern edge of the north Atlantic system (12°N–20°N) and most active in winter/spring has been largely excluded from these studies. Here, the seasonal cycle of the SMUS and its response to climate change is investigated in the database of the Coupled Models Inter comparison Project Phase 5 (CMIP5). Upwelling magnitude and surface signature are characterized by several sea surface temperature and wind stress indices. We highlight the ability of the climate models to reproduce the system, as well as their biases. The simulations suggest that the intensity of the SMUS winter/spring upwelling will moderately decrease in the future, primarily because of a reduction of the wind forcing linked to a northward shift of Azores anticyclone and a more regional modulation of the low pressures found over Northwest Africa. The implications of such an upwelling reduction on the ecosystems and local communities exploiting them remains very uncertain.

Appendices

Appendix A: Wind stress estimation from wind speed data

In principle, the drag coefficient C_d depends on both the atmospheric and the oceanic state and it is variable. Large and Pond (1982) proposed the following scaling of C_d according to the wind speed V:

$$10^{3} C_{d} = \, 0.49 + 0.065\;V\;{\text{for}}\;10 \le V < 25\;{\text{m}}\;{\text{s}}^{ - 1} ,$$

(6)

$$10^{3} C_{d} = \, 1.14\;{\text{for}}\;3 \le \;V < 10\;{\text{m}}\;{\text{s}}^{ - 1} ,$$

(7)

$$10^{3} C_{d} = 0.62 + 1.56\;V^{ - 1} \;{\text{for}}\;V < 3\;{\text{m}}\;{\text{s}}^{ - 1} .$$

(8)

Gill (1982, p. 29) proposed another scaling, based on results of Smith (1980) for large wind speeds:

$$C_{d} = 0.0011\;{\text{for}}\;V > 3\;{\text{m}}\;{\text{s}}^{ - 1} ,$$

(9)

$$10^{3} C_{d} = \, 0.061 + 0.063\;V\;{\text{for}}\;6 < V < 22\;{\text{m}}\;{\text{s}}^{ - 1} .$$

(10)

The NOAA-TM-NMF S-SWFSC-231 rapport (1996) suggests to use the value of C_d = 0.0026 with monthly mean data while Santos et al. (2012) used C_d = 0.0014 for their study of the Moroccan upwelling zone (22°N–33°N).

Over the SMUS region, the maximum wind speed is around 7 m s⁻¹. We compared the meridional component of the wind stress computed online in the IPSL-CM5A-LR climate model to the wind stress computed offline using the meridional wind speed component from the same simulations with the drag coefficient values from Santos et al. (2012) and the two methods [Eqs. (9), (10)] drag coefficient listed above (figure not shown). We have chosen to test only the meridional wind component because it is the strongest one in the SMUS region and most directly associated to the upwelling intensity. We found that the scaling proposed in Gill (1982) underestimate the meridional wind stress amplitude north of roughly 15°N, in particular in summer north of 20°N, as well as in winter between 12°N and 20°N which the season and the latitude band of the SMUS. The C_d = 0.0014 scaling as used by Santos et al. (2012) yields the closest values to the observations. In all simulations where only the wind speed components are provided, the wind stress was thus computed using this latter scaling.

Appendix B: model upwelling from vertical velocities

The net estimation of upwelling must in principle be consistent with the upward ocean mass transport diagnosed in the models (whenever available). However, this latter field is noisy (as recently emphasized by Oyarzún and Brierley 2018) and the depth at which the typical upwelling vertical velocity should be considered for comparison with upwelling indices is a difficult parameter to choose. Here, this depth has been chosen equal to the mixed layer depth averaged over the upwelling season (November–May) and over the SMUS region. In the models for which UI_total could not be computed (see last column in Table 1) we choose the depth where U_w is maximum.