Orographic shaping of US West Coast wind profiles during the upwelling season

Abstract

Spatial and temporal variability of nearshore winds in eastern boundary current systems is affected by orography, coastline shape, and air-sea interaction. These lead to a weakening of the wind close to the coast: the so-called wind drop-off. In this study, regional atmospheric simulations over the US West Coast are used to demonstrate monthly characteristics of the wind drop-off and assess the mechanisms controlling it. Using a long-term simulation, we show the wind drop-off has spatial and seasonal variability in both its offshore extent and intensity. The offshore extent varies from around 10 to 80 km from the coast and the wind reduction from 10 to 80 %. We show that when the mountain orography is combined with the coastline shape of a cape, it has the biggest influence on wind drop-off. The primary associated processes are the orographically-induced vortex stretching and the surface drag related to turbulent momentum flux divergence that has an enhanced drag coefficient over land. Orographically-induced tilting/twisting can also be locally significant in the vicinity of capes. The land-sea drag difference acts as a barrier to encroachment of the wind onto the land through turbulent momentum flux divergence. It turns the wind parallel to the shore and slightly reduces it close to the coast. Another minor factor is the sharp coastal sea surface temperature front associated with upwelling. This can weaken the surface wind in the coastal strip by shallowing the marine boundary layer and decoupling it from the overlying troposphere.

1 Introduction

The Californian current system (CCS) is an eastern boundary upwelling system (EBUS). It is among the most productive marine ecosystems in the world, supporting some of the world’s major fisheries (e.g., FAO 2009). A cross-shore pressure gradient between the North Pacific High and the Continental Thermal Low pressure systems (Huyer 1983) leads to predominantly alongshore winds favoring upwelling circulation. Wind-driven upwelling is responsible for relatively low temperatures and high productivity along the coast and extends offshore for hundreds of km. The CCS upwelling varies on seasonal time scales, with a favorable season during spring and summer (e.g., Marchesiello et al. 2003). As with the other EBUSs (i.e., Humboldt, Benguela, and Canary), equatorward winds drive alongshore currents, Ekman-induced upwelling and offshore surface flow, cross-shore exchange, and productivity. However, the spatial and temporal structure of the upwelling depends on the geographical structure of the coastal winds. For example, wind expansion fans (Winant et al. 1988) generated by numerous capes and mountain ranges can lead to spatial variations in upwelling.

In the EBUS, they are weaker winds within the nearshore drop-off zone (Fig. 1) (e.g., Capet et al. 2004; Dorman et al. 2006; Perlin et al. 2007, 2011; Renault et al. 2012; Desbiolles et al. 2014). This coastal transition zone for the wind may extend to around 100 km offshore (Boé et al. 2011). Various different processes can produce wind drop-off: sharp changes in surface drag and the atmospheric boundary layer at the land-sea interface, coastal orography (Edwards et al. 2001), the cape effect (Perlin et al. 2011), and SST-wind coupling (Chelton et al. 2007; Jin et al. 2009). Recently Perlin and Skyllinstad (2013) showed the influence of mountain flow dynamics and orography on the coastal wind on the lee side of an idealized cape. The influence of the lee trough on the coastal wind and the associated vorticity generation has been studied previously (e.g., Smolarkiewicz and Rotunno 1989; Rotunno and Smolarkiewicz 1995; Epifanio and Rotunno 2005). These studies show both that synoptic conditions and marine boundary layer (MBL) height appear to be important in the formation of orographic lee side effects. Additionally, in an idealized study Smolarkiewicz and Rotunno (1989) suggest a tilting of the baroclinically produced horizontal vorticity induces a vertical vorticity.

Fig. 1

Upper panels WRF domain configuration illustrated by the wind speed as measured by QuikSCAT (left) and simulated by WRF (right). The contour lines represent the mean wind speed intensity and the arrows the mean wind direction, while the red contour line represents the domain boundaries. Bottom panel Example of wind drop-off from QuikSCAT (black line) and WRF (red line) for the month of June 2000 between 40°N and 42°N. The month of June 2000 is characterized by a well-defined wind event around 40°N. The red dot highlights the detected drop-off length (Sect. 2.3.1). There is a good agreement between QuikSCAT and WRF, and one can note their good correspondence in terms of wind drop-off length ($L_d=75\,\hbox {km})$ between them, although QuikSCAT is not able to measure the wind in a coastal strip of 30 km width

Weaker coastal winds induce Ekman pumping through the associated wind curl; and so it could be a key to the forcing of coastal upwelling, in conjunction with upwelling due to the wind right at the coast, which induces a surface velocity divergence and offshore Ekman transport. Variations in the relative strength of Ekman pumping and coastal divergence modify the upwelling structure in the coastal current system (e.g., Marchesiello et al. 2003; Renault et al. 2012), and they also lead to variations in the biogeochemical response by changing the horizontal and vertical advection of nutrients and the eddy-induced nutrient quenching (Gruber et al. 2011; Renault et al. 2014). Finally, as suggested by Renault et al. 2014, the integrated upwelling index (i.e. the so-called Bakun upwelling index, Bakun and Nelson 1991) lacks an explicit oceanographic process. An upwelling index should take into account, among other things, the coastal wind shape characteristics. However, the wind drop-off characteristics and causal mechanisms are not well known. As reported by Renault et al. (2009) and illustrated in Fig. 1, the QuikSCAT near-coastal blind zone barely allows observational monitoring of the wind drop-off. Additionally, as illustrated in Renault et al. (2014), even recent global reanalysis products (e.g., the NOAA Climate Forecast System Reanalysis, CFSR; Saha et al. 2010) does not realistically represent the wind drop-off. When used to force a regional oceanic model, these products can induce unrealistic surfacing of the coastal undercurrent and misrepresent the eddy activity and biogeochemical response.

In this study a regional configuration of the Weather Research and Forecast (WRF) model (Skamarock et al. 2008) is implemented over the US West Coast, encompassing the CCS. The main objective is to document the processes driving the spatial and temporal distribution of wind drop-off. The primary outstanding questions are:

• (1) What is the spatial distribution of the wind drop-off in the CCS? A simple view of the weakening of the wind close to the coast may lead to the assumption of a constant wind drop-off length ($L_d$) and intensity. However, some regions may have a larger or more pronounced wind drop-off due to higher coastal mountain ranges and coastline wandering.

• (2) Is there a seasonal variability in the drop-off? The wind intensity exhibits seasonal variability and so may the drop-off profile.

• (3) What are the mechanisms driving the wind drop-off characteristics? Orography, coastline, drag, and SST can play a role in determining the weakening of the wind. Chelton et al. (2007) suggest an important SST feedback on the wind curl, which in addition could modify the wind and wind curl distribution close to the coast. It is not well known which of these mechanisms dominates. This information can be important in designing numerical experiments to simulate an upwelling region. For example, does the model have a high enough spatial resolution to resolve accurately the orography? Is coupling necessary to simulate the SST-wind curl feedbacks and therefore the upwelling?

The paper is organized as follows: Sect. 2 describes the model, methodology, and different experiments carried out. In Sect. 3 the simulated wind is briefly evaluated with respect to the observations, and the main wind drop-off characteristic as well as the underlying mechanisms are assessed. Finally, Sect. 4 evaluates the importance of the mechanisms discussed in Sect. 3 using sensitivity experiments. The results are then discussed in Sect. 5, followed by conclusions.

2 Model and in situ data

2.1 Model configuration

WRF (version 3.5, Skamarock et al. 2008) is implemented in a configuration with two nested grids. The largest domain covers the North American West Coast (154.3°W–105°W, 15°S–60°N) with a horizontal resolution of 18 km (not shown); the inner domain covers the US West Coast (144.0°W–113.7°W–22.9°N–50.9°N), with a horizontal resolution of 6 km (Fig. 1). The coarser grid (WRF18) reproduces the large-scale synoptic features that force the local dynamics in the second grid, each using an one-way offline nesting with six-hourly updates of the boundary conditions. The coarser grid simulation (WRF18) was first run independently. It is initialized with the climate forecast system reanalysis (CFSR) reanalysis (≈40 km spatial resolution; Saha et al. 2010) for 28th December 1994 and integrated for 13 years with time-dependent boundary conditions interpolated from the same six-hourly reanalysis. Forty vertical levels are used, with half of them in the lowest 1.5 km. SST forcing is derived from the Ostia daily product (Stark et al. 2007) that has a spatial resolution of 5 km. In order to take into account the diurnal, the CFSR SST diurnal cycle is first estimated as the 6-h anomalies with respect to the daily mean CFST SST at each grid point over the WRF grid. It is then added to the Ostia SST. The nested domain (WRF6) was initialized from the coarse solution WRF18 on 30th December 1994 and integrated 13 years (in the following analyses, only the WRF6 solution is used.).

A full set of parameterization schemes is included in WRF. The model configuration was setup with the following parameterizations: the WRF Single-Moment 6-class microphysics scheme (Hong and Lim 2006); the Kain-Fritch cumulus parameterization (Kain and Fritsch 1990); the rapid radiative transfer model (RRTM) for longwave radiation, based on the work by Mlawer et al. (1997); the Dudhia (1989) scheme for shortwave radiation; the Noah land surface model (Skamarock et al. 2008); and the MYNN2.5 planetary boundary layer (PBL) scheme (Nakanishi and Niino 2006).¹

2.2 The experiments

The main experiment (the control run, CR) is run over the period 1995–2007 and used to estimate a monthly mean of the wind drop-off characteristics. To assess which mechanisms control the wind drop-off, the vorticity budget described in Sect. 2 is applied to the six-hourly outputs, and then a monthly mean is derived.

Table 1 Sensitivity experiments

Sensitivity experiments (Table 1) are done over the month of June 2000 to assess the individual effects of the coastline (STRAIGHT), orography (NOTOPO), land-sea drag difference (NODRAG), and SST front (FRONT). Figure 2 shows an along-coast orography index ($H_{index}$) for CR, STRAIGHT and NOTOPO as well as a coastline meandering index ($M_{index}$). $H_{index}$ is estimated as the mean orographic height in meters over the strip of land within 100 km of the coast. $M_{index}$ is expressed in longitude degrees and is estimated using a high-pass along-shore filter (with 30 km half-width) applied to the coastal longitude. We can clearly distinguish the main capes (e.g., at 41°N and 43°N) and bays (e.g., 42°N) in $M_{index}$. The STRAIGHT coastline is represented and compared to the realistic coastline in Fig. 2c. In all the sensitivity experiments, the land and sea grid points have been updated accordingly. A short spin up of two days is enough to reach a model equilibrium. In FRONT, contrary to the other sensitivity experiments where the SST is spatially uniform (using the daily SST averaged over the whole domain), the SST is characterized by an uniform along-coast SST front defined by a cross-shore difference of 3° and a spatial extent of 25 km. For each experiment a vorticity budget is diagnosed on the six-hourly output to assess which mechanisms drive the wind drop-off.

2.3 Analysis methodology

In this study, the winter, spring, summer and fall seasons correspond to the months (January–March), (April– June), (July–September), and (October–December), respectively.

2.3.1 Wind drop-off scale estimation

The wind drop-off characteristics are estimated using monthly mean values of the simulated wind. The drop-off length ($L_d$) is computed using a wind curl threshold value. As illustrated in Fig. 1c, $L_d$ is estimated by detecting where the cross-shore wind curl is >3 × 10⁻⁵ s⁻¹ (see the dashed line in Fig. 4). The percentage of along-shore wind reduction at the coast ($W_r$) is estimated as follows:

$$\begin{aligned} W_r = \frac{V(L_d) - V(coast)}{V(L_d)} \times 100 \end{aligned}$$

(1)

where V represents the along-shore wind speed at a height of 10 m above the surface. We also tried to use a threshold based on the along-shore wind gradient. It turns out the curl threshold gives the more accurate and stable estimation. This threshold appears to detect $L_d$ relatively well during the upwelling season. If the wind intensity (and therefore the wind curl) is too weak, the wind drop-off characteristics are not important for the upwelling.

Fig. 2

The orographic index ($H_{index}$, a) and coastline meandering index ($M_{index}$, b). The main capes and bays are indicated (c). The STRAIGHT configuration. The color fields represent the orogaphy, the solid (dashed) lines represent the STRAIGHT (CR) coastline. NOTOPO, NODRAG, and FRONT have the same coastline as STRAIGHT

2.3.2 Vorticity budget analysis

The Ekman pumping is generated by the vertical component of the relative vorticity (i.e., the vertical component of the vorticity vector):

$$\begin{aligned} \xi = \frac{\partial {v}}{\partial {x}} - \frac{\partial {u}}{\partial {y}} \,. \end{aligned}$$

(2)

For diagnosing the mechanisms that drive the wind drop-off, a vorticity budget analysis is performed as follows:

$$\begin{aligned} \dfrac{D}{Dt}\left( \xi + f\right)= & {} \overbrace{-\left( \xi + f\right) \left( \frac{\partial {u}}{\partial {x}} + \frac{\partial {v}}{\partial {y}}\right) }^ {\text{ Vortex } \text{ stretching }} \overbrace{ - \left( \frac{\partial {w}}{\partial {x}}\frac{\partial {v}}{\partial {z}} - \frac{\partial {w}}{\partial {y}}\frac{\partial {u}}{\partial {z}}\right) }^{\text{ Tilting/Twisting }}\nonumber \\&\overbrace{ + \frac{1}{\rho ^2}\left( \frac{\partial {\rho }}{\partial {x}}\frac{\partial {P}}{\partial {y}} -\frac{\partial {\rho }}{\partial {y}}\frac{\partial {P}}{\partial {x}}\right) }^{\text {Solenoidal}} \overbrace{+ \left( \frac{\partial {Fr_x}}{\partial {x}} - \frac{\partial {Fr_y}}{\partial {y}}\right) }^{\text {Turb. flux divergence}} \,, \end{aligned}$$

(3)

where ${u,v,w}$ are the 3D components of the wind, $\rho$ is the air density, $P$ is the air pressure, and $Fr_x$ and $Fr_y$ are the vertical eddy mixing terms: $Fr_{x} = \frac{\partial {\bar{u'w'}}}{\partial {z}}$ and $Fr_{y} = \frac{\partial {\bar{v'w'}}}{\partial {z}}$. The first term on the right-hand side (RHS) is the vortex stretching term ($Vs$). The second term is the tilting/twisting ($TiTw$) term. The third term corresponds to the solenoidal term, and the last term is the turbulent flux divergence ($TFD$). In our results the solenoidal term is negligible compared to the other terms, so it will not be shown. The budget is estimated and integrated over the first 50 m above the ocean surface at each 6 h of model output (when we instead integrate over the first 400 m, our results are not significantly different).

2.4 Wind speed from in situ data

The hourly wind data from the US National Data Buoy Center (NDBC) (http://www.ndbc.noaa.gov) over the California shelf are used to validate the nearshore wind of the atmospheric simulation. The position and features of each buoy used are described by Dorman and Winant (1995) and are summarized in Table 2. Note, the anemometer height correction has not been applied on the data used.

Table 2 Buoys main characteristics

The NDBC wind velocity accuracy requirements are the greater of $1\, \mathrm{m s}^{-1}$ or 10 % for speed (Beardsley et al. 1997). Gilhousen (1987) conducted field trials and found high correlation between buoys moored within 5 km of each other, with speed standard deviations of 0.6–0.8 ms⁻¹.

3 Seasonal variability

3.1 Model evaluation

We first present an evaluation of the simulated wind with respect to the satellite-derived wind (i.e., QuikSCAT) and some coastal buoys from the National Buy Data Center. Then, we focus on the seasonal-mean features of the coastal wind profiles.

The WRF 10 m wind speed and direction compare fairly well with the daily Ifremer QuikSCAT data descbribed by Bentamy and Fillon (2012) over the period July 1999 to December 2007, both in amplitude and direction. Note the QuikSCAT statistics errors used bellow are derived from this study. The mean correlation between zonal (meridional) mean wind speed from WRF and QuikSCAT is 0.85 (0.86); the significance levels are $\sigma >95\,\%$. The mean bias is 0.4 ms⁻¹, the same order of magnitude as the QuikSCAT mean bias (0.6 ms⁻¹). Because QuikSCAT winds are equivalent neutral winds and are compared to the actual winds from the model, this difference cannot necessarily be interpreted as a model bias. The variability is also similar: the WRF (QuikSCAT) standard deviations are around 3.2 ms⁻¹ (3.6 ms⁻¹) and 2.6 ms⁻¹ (2.9 ms⁻¹) for zonal winds and meridional winds. The RMS differences are 1.5 and 1.6 ms⁻¹ for the zonal and meridional wind speeds, respectively. Note the RMS difference has the same order of magnitude as the QuikSCAT wind error (up to 2 ms⁻¹).

The mean seasonal wind speed intensity from the CR is plotted in Fig. 3 with superimposed arrows showing the mean wind magnitude and direction. The spatial pattern of the mean seasonal wind is well captured by the model. It realistically reproduces both the seasonality and the main gradients of the wind. The upwelling season that occurs in spring and summer is marked by a well defined alongshore wind speed (up to 9 ms⁻¹ in spring and summer). As expected, the main discrepancies between QuikSCAT and WRF occur close to the coast where QuikSCAT is less reliable (e.g. see Fig. 1 due to land contamination in the radar antenna sidelobe-antenna effects. The radar backscatter from land is much brighter than that from calm ocean water, inducing significant contamination of the backscatter measurements. The wind curl (Fig. 4) is also in fair agreement with the literature and the observations. It is characterized by a strong wind curl (up to $5.5 \times 10^{-4}\,\hbox {ms}^{-1}$) along the coast, which shows some geographical variations in terms of spatial extension (from ≈10 to 100 km) and intensity. This variation will be discussed in Sect. 3.2. Note that QuikSCAT, as shown by Renault et al. (2009) and illustrated in Fig. 1, is only partially able to monitor this coastal strip of large wind curl. This is due to QuikSCAT’s blind zone (around 25 km, Bentamy and Fillon 2012) and its spatial resolution (25 km) that , when combined, induce a curl blind zone of about 50 km. However, in Fig. 1, we can note the good correspondence between QuikSCAT and WRF in terms of $L_d$ and offshore wind.

To validate the model in the nearshore zone, the WRF simulation is compared to buoy data (Fig. 5). The observed variability of the nearshore wind speed during the upwelling season is reasonably well reproduced ($r >0.7$, significance $\sigma > 95\,\%$, Fig. 5). The mean bias of the model compared to buoys is smaller (0.2 ms⁻¹) than the difference with QuikSCAT computed previously over the entire domain. Finally, the variability of the alongshore wind is also fairly well represented. The RMS of the simulated wind is comparable to that estimated from the buoys (2.6 vs. 2.7 ms⁻¹), and the RMS difference is <1.75 ms⁻¹. The same statistics were also computed over the whole year and only for the spring and summer seasons, with very similar results. To conclude, despite some small biases, the results in this section show that the atmospheric model represents the main characteristics of the wind reasonably, in particular during the upwelling season. This conclusion even holds nearshore. In the following section, the monthly mean wind drop-off characteristics are assessed as well as the mechanisms leading to the wind reduction toward the coast.

Fig. 3

Seasonal-mean wind from the CR, estimated over the period 1995–2007. The colors (arrows) represent the wind speed intensity (direction). During the spring and summer the wind blows southward along the coast. a Winter, b Spring, c Summer, d Fall

Fig. 4

Mean seasonal wind curl from the CR, estimated over the period 1995–2007. The black dashed line represents the 0.3 × 10⁻⁴ s⁻¹ value used as a threshold to determine the wind drop-off spatial length ($L_d$, see Fig. 2). The alongshore wind curl peaks in spring and summer. a Winter, b Spring, c Summer, d Fall

Fig. 5

Wind speed comparisons between the National Data Buoy Center buoys (see Sect. 2.4) and WRF during the upwelling season (Spring and Summer), the considered period for each buoy is indicated in Table 2. On the left is the correlation at each buoy location between the WRF 10 m wind speed and the observations. The name of each buoy is also indicated. On the right, the black and gray lines represent, respectively, the mean WRF 10 m wind speed and the measured 10 m wind speed at each buoy location during the upwelling season. The mean bias is <0.2 ms⁻¹. There is a good agreement between the WRF coastal wind and the buoys. Similar results are obtained when considering the full year

3.2 Seasonal wind drop-off and mechanisms

The main wind drop-off characteristic are estimated with the methods described in Sect. 2.3.1 on the six-hourly model outputs. Then monthly means are computed.

Typical wind drop-off characteristics are shown in Fig. 6 as a function of month and latitude. We can note the remarkable similarity in all three colored patterns and with the larger intensity maxima. Consistent with Fig. 4, the upwelling season is characterized by a larger wind curl. $L_d$ varies from ≈10 to ~80 km depending on the time of year with a maximum in August, when the southward wind is greatest (up to 8 ms⁻¹; Fig. 6). Note that $L_d$ is larger where the orography is high or dowstream (e.g., 36°N, 38.5°N, 40°N and 42.5°N; Fig. 6a). This suggests the wind drop-off is significantly affected by orography and by the cape effects and coastline meandering. The alongshore wind can be reduced by >8 ms⁻¹ at these locations. Interestingly, north of 43°N the wind blows southeast and is controlled by the atmospheric anticyclone position, so both alongshore wind and wind curl are weak. This is consistent with a weak upwelling in the $CU$ index (Bakun and Nelson 1991). We can estimate rough mean values for $L_d$ during spring and summer of 40 and 60 km, respectively. The corresponding alongshore wind speed reductions are respectively 4 and 5 ms⁻¹. During fall and winter, $L_d$ is not meaningful because the southward wind is weak or non-existent.

Fig. 6

Alongshore wind drop-off climatology as function of month. a The solid and dashed lines represent the CR $H_{index}$ and $M_{index}$, respectively. b Wind curl ($\xi$) integrated over a 100 km offshore band. c Wind drop-off spatial length ($L_d$, km) and d alongshore wind speed decrease. In b, c, and d, the black contour lines represent the mean alongshore wind speed ($m s^{-1}$) at an offshore distance equal to $L_d$. Note the remarkable similarity in all three colored patterns, and also with the larger intensity maxima

Fig. 7

Seasonal alongshore vorticity budget (see Eq. 3) integrated over a coastal strip 100 km wide. As in Fig. 6, the black dashed contours represent the alongshore wind speed at a distance equal to $L_d$. a The solid and dashed lines represent the CR $H_{index}$ and $M_{index}$, respectively, b vortex stretching ($Vs$), c turbulent flux divergence ($TFD$), and d tilting/twisting ($TiTw$). $Vs$ mostly depends on the orography and wind intensity, whereas $TFD$ and $TiTw$ variations are associated with the presence of capes and bays

We now investigate the mechanisms that control the spatial distribution of coastal wind. To that end, a vorticity analysis (Sect. 2) is conducted for each six-hourly model output. Figure 7 depicts the derived monthly climatology. $Vs$ and $TFD$ are the main contributors in the vorticity budget, although $TiTW$ can be significant in the vicinity of the major capes. Not surprisingly, both have a similar seasonality as the wind has with larger values during the upwelling season. This suggests the wind has to be large enough to activate these mechanisms. The spatial distribution of each term during the upwelling season is shown in Fig. 8a, b, c, as well as the wind speed and direction and the orography. High $Vs$ is associated with a high coastal orography, whereas, $TFD$ and $TiTW$ peaks are likely related to coastline meandering and/or cape effect (mainly around 36°N, 38.5°N, 40°N and 42.5°N). Capes and bays, combined with high orography enhance $L_d$ and the wind reduction. North of 42°N, the offshore wind blows southeastward and, close to the coast, turn southward. The presence of the land edge and orography likely acts as a wind barrier, inducing a large and uniform $TFD$ while reducing the zonal component of the wind. This is partly related to the sharp change of the drag coefficient between sea and land.

To summarize, the wind drop-off characteristics have a similar seasonal variability as the wind intensity. The larger the wind intensity, the larger $L_d$ and wind curl. Therefore, larger $L_d$ and wind curl occur during spring and summer when the wind intensity is largest and are associated with strong $Vs$ and $TFD$. Based on these results, our main working hypothesis is that orography combined with coastline meandering is the main player in generating the wind drop-off through $Vs$ and locally $TiTw$ and $TFD$. The drag change across the coast and SST front effects may act through $TiTW$ and $TFD$. In Sect. 4, sensitivity experiments are carried out to verify this hypothesis.

Fig. 8

Spatial distribution of the vorticity budget (Eq. 3) (top panels) during the upwelling season (spring and summer) and (bottom panels) during June 2000. The thin contour lines represent the orography, the color fields the term of the vorticity budget, the thick black contours the wind intensity and the arrows its direction. Again, $Vs$ appears to depend on orography and wind intensity, whereas $TFD$ and $TiTw$ anomalies are associated with the presence of capes and bays. Note the similarity between the upwelling season and June 2000. A conspicuous difference, however, is that $Vs$ is larger in June 2000

4 Sensitivity experiments

Using a series of simulations covering the month of June 2000 (Table 1), we focus on the sensitivity of the wind drop-off to the coastline, orography, land drag and SST front. June 2000 was characterized by a well defined and steady wind (Fig. 8d), which makes it a representative period for a good case study.

The main findings of this section, demonstrated below, are:

• The orography, through $Vs$ is the main driver of the wind drop-off. The higher the orography is, the wider and larger $L_d$ and wind reduction are. The highest coastal orography induces a 30 % reduction of the wind over a $L_d$ of 20-25 km.

• The presence of capes enhances the orographic effect by acting through $TFD$ and $TiTW$. A significant cape can double the orography effect. In that case, $L_d$ can reach $60$ km and the coastal wind can be reduced up to 75 %.

• Drag differences across the coast play a role in determining the wind direction. Through $TFD$, this effect generally induces a $L_d$ of 15 km, but the wind reduction is rather small.

• An upwelling SST front can induce a $L_d$ around 30 km (comparable to the width of the front) and a coastal wind reduction of about $10\,\%$. It shallows the MBL and induces additional TFD. In the coastal region, such SST interaction appears to be as a second-order mechanism, although it can exhibit a significant wind coupling further offshore (Chelton et al. 2007).

4.1 Wind drop-off during June 2000 in the CR

The month of June 2000 is characterized by a tongue of steady southward winds in excess of 8 ms⁻¹ extending away from the coast, with a wind core maximum situated at 40°N (Fig. 8d). This month has a slightly larger wind (by 1.5 ms⁻¹) than the climatology estimated in Sect. 3.2. As shown by Beardsley et al. (1987) and Renault et al. (2012), such enhanced southward wind is associated with the synoptic variability. The large-scale circulation brings higher momentum and colder air from the north, raises the inversion, and decreases the stability near the surface. To illustrate this, Fig. 9 depicts the mean sea level pressure derived from CFSR for June 2000 and the climatology for the same month. During June 2000 a high pressure anomaly is located off the coast, increasing the along-coast pressure gradient from south to north. Such a poleward-oriented pressure gradient cannot be balanced by the Coriolis force because the presence of coastal orography (up to 1000 m above sea level in this model configuration) precludes the development of zonal (cross-shore) flow in the lower troposphere. The pressure gradient then accelerates the along-shore flow as a southerly wind until turbulent mixing closes the force balance.

The simulated wind clearly weakens toward the coast, in particular between 38°N and 42°N where the wind intensity and the orography are large (Figs. 1, 8d). Consistent with the climatology (Sect. 3.2), the primary mechanisms generating the wind drop-off are $Vs$ where $H_{index}$ is large and $TFD$ and $TiTw$ where $M_{index}$ is large, i.e., in the presence of capes or bays (Figs. 8d–f,  10). North of 42°N, the wind blows southeast and then turns alongshore with $TFD$ acting like a barrier. As expected, the vorticity budget for the month of June 2000 is similar to that during the mean upwelling season. However, due to a stronger wind speed, $Vs$ is larger, inducing a wider $L_d$ (Fig. 10).

The wind drop-off characteristics for this month are depicted in Fig. 10. $L_d$ has the same geographical variation as the climatology, but it is larger between 38°N and 42°N where the wind intensity is greatest. $L_d$ extends from 10 up to 90 km. The coastal alongshore wind can be reduced up to $8\,\hbox {ms}^{-1}\,(W_r$ of $75\,\%$) with respect to the offshore wind, inducing a larger wind curl than that of the climatology (not shown). We can note the correspondence between the $H_{index}$, $M_{index}$ and $Vs$, $TFS$ and $TiTw$ (Figs. 8d−f, 10). In Fig. 8e, f, in good agreement with Rotunno and Smolarkiewicz (1995) and Perlin and Skyllinstad (2013), capes and bays (highlighted respectively by large negative or positive values of $M_{index}$) induce large $TiTw$ and $TFD$ and hence large values of $L_d$ and $W_r$ (e.g., 41°N, 43°N). High coastal mountain ranges (large $H_{index}$), induce large values of $L_d$ and $W_r$ through $Vs$ (e.g., between 38°N and 42°N). When high coastal orography is combined with a cape, $Vs$ increases even more, inducing a larger and stronger wind drop-off.

These results highlight the dependency of the wind drop-off characteristics on the wind speed intensity, the orography, and the coastline meandering. Compared to the climatology, the larger wind intensity during June 2000 induces a larger $Vs$, whereas the $TiTw$ and $TFD$ remain similar. The orography and coastline meandering appear to be the main players in generating the wind drop-off. This hypothesis is verified in the Sects. 4.2 and  4.3.

Fig. 9

Large-scale atmospheric conditions from the CFSR reanalysis for June 2000 and a climatological month of June (estimated over the period 1995–2007) illustrated by the mean sea level pressure (MSLP). The black (gray) contour represents the MSLP during June 2000 (Climatological June), whereas the color fields represent the June 2000 MSLP anomalies with respect to the climatological mean. There is a high pressure anomaly close to the coast that induces a stronger along-coast pressure gradient, generating positive southward wind anomalies

Fig. 10

Alongshore wind drop-off characteristics during June 2000 from CR. a Integrated orography index ($H_{index}$, solid line) and coastline index ($M_{index}$, dashed line), b wind curl ($\xi$) integrated over a 100 km offshore band, c drop-off spatial length ($L_d$), and d alongshore (at $L_d$ km offshore) and coastal wind speed intensity (black line and dashed line respectively). Note the correspondence between $H_{index}$, $M_{index}$ and wind curl. Large $H_{index}$ and $M_{index}$ induce large $L_d$ and $\xi$

4.2 Coastline, orography, and sea-land drag difference

In this section we focus on contributions of the coastline, orography, and sea-land differences in the drag coefficient differences to the wind drop-off through $Vs$, $TFD$ and $TiTw$. The June 2000 experiment was re-run but using an idealized coastline, realistic or idealized orography, and no land drag (i.e., experiments STRAIGHT, NOTOPO and NODRAG, respectively; Table 1). The broad characteristics of the surface circulation remain similar to those of CR. In particular, around 40°N, in all the experiments, a core of intense wind (up to 8 ms⁻¹) is simulated.

STRAIGHT is characterized by a realistic orography and a straight coastline, except around 41$^0$N where a cape is present. By comparing STRAIGHT to the CR, the role of coastline meandering, especially the presence of capes is clear: they enhance the orographic effect. Figure 11 depicts the mean drop-off characteristics from STRAIGHT (black line). This figure can be directly compared to its equivalent for the CR (Fig. 10). As shown before, in the north of the domain, the CR has two main capes (43°N and 41°N). The presence of those capes is reflected in larger values of $L_d$ that presents larger values. North of 41°N in STRAIGHT, $L_d$ and $W_r$ do not peak and then have homogeneous values of 25 km and 30 %, respectively. Around the cape at 41°N, $L_d$ and $W_r$ peak, reaching values of 60 km and 75 %, respectively.

North of 41°N, the alongshore-integrated vorticity budget (Fig. 11) shows that the main components are $TFD$ and $Vs$. The orography combined with the sea-land drag difference acts as a barrier. It turns the wind alongshore and reduces it (30 %). Below the cape, where the wind is alongshore and greatest (8 ms⁻¹), the combination of large orography (large $H_{index}$; Fig. 11) and the cape generate large values of $Vs$, explaining the larger values of $L_d$. Additionally, consistent with Rotunno and Smolarkiewicz 1995, the lee trough influences the coastal wind and the vorticity through $TiTw$ and $TFD$ in the vicinity of the cape.

To confirm the importance of the orography in generating the wind drop-off, we can compare NOTOPO to STRAIGHT. NOTOPO has the same coastline as STRAIGHT but without orography close to the coast ($H_{index} = 0$; Figs. 2, 11). Without orography, there is no more $Vs$ or $TiTw$ (Fig. 11), whereas $TFD$ is reduced to homogeneous values. As a result, $L_d$ is reduced to a steady value between 15 and 25 km (in the presence of a cape), and, overall, $W_r$ is rather small (<10 %). This confirms that orography is the main influence on the wind drop-off, and the coastline meandering can enhance its effect. The remaining drop-off is likely due to the sharp change of drag between sea and land.

To confirm this final point, NODRAG can be compared to NOTOPO. It has similar characteristics but has the same prescribed roughness over sea and land. In NODRAG there is no more wind drop-off (Fig. 12). By comparing NODRAG to NOTOPO, we can deduce that the land-sea drag difference acts as a barrier through $TFD$ (Fig. 11). Therefore, the offshore wind in NODRAG can penetrate the land (not shown). Overall, this confirms that the remaining $L_d$ and $W_r$ in NOTOPO is induced by the drag difference. This effect generates an $L_d$ of 15 km (that can be increased through cape effects) and a weak $W_r$ (<10 %).

Fig. 11

Alongshore wind drop-off characteristics and alongshore integrated vorticity budget (Eq. 3) over a coastal strip of 100 km from STRAIGHT (black lines) and NOTOPO (red lines). a Integrated orography index ($H_{index}$, solid lines) and coastline index ($M_{index}$, dashed lines) (n.b., STRAIGHT and NOTOPO have the same coastline. b Wind curl ($\xi$) integrated over a 100 km offshore band. c Drop-off length ($L_d$). d Alongshore offshore (at $L_d$ km offshore) and coastal wind speed intensity (respectively, solid line and dashed line). e vortex stretching ($Vs$), f turbulent flux divergence ($TFD$), and g tilting/twisting ($TiTw$). The orography plays a primary role in generating the wind drop-off: in the absence of orography (NOTOPO), $Vs$ and $TiTw$ tend toward zero, explaining the wind drop-off $L_d$ and intensity reduction

Fig. 12

As in Fig. 11 but here for the alongshore wind drop-off characteristics and alongshore-integrated vorticity budget (Eq. 3) over a coastal strip 100 km wide from NODRAG (black lines) and FRONT (red lines). The NODRAG experiment does not have a significant wind curl ($\xi$) and drop-off length ($L_d$) due to the absence of coastline meandering ($M_{index} = 0$), coastal orography ($H_{index} = 0$), or land-sea drag differences. In (d) for NOTOPO, because $L_d$ is equal to zero, the alongshore speed (at $L_d$ km offshore) and the coastal wind speed intensity are equal. The SST feedback to the atmosphere, through generation of of $Vs$ and $TFD$ induces a wind drop-off with $L_d$ up to 40 km and slight reduction of the wind

4.3 SST coastal front

The effect of an upwelling cold SST front is assessed by adding an SST front of 3 °C over a coastal band strip of $25$ km (FRONT experiment; Table 1). In this last experiment, in addition to the straight coastline, flat coastal orography, and the absence of drag difference between sea and land, a sharp SST front along the coast is imposed as a bottom boundary condition to WRF. The mean circulation remains the same with respect to NODRAG, and a core of intense wind is still present around 40°N. This simulation allows us to isolate the SST feedback due to an upwelling front, because it lacks other wind drop-off sources (i.e., no drag, no orography, and no coastline meandering except a cape around 40°N).

As expected, a sharp coastal SST front does not affect the mean general circulation (not shown). However, as shown in Fig. 13, the colder coastal SST stabilizes and shallows down the MBL (estimated by the MYNN2.5 PBL scheme in WRF, Nakanishi and Niino 2006) to 100 m close to the coast, inducing a wind drop-off through $TFD$ (Fig. 12). $L_d$ can reach up to 40 km while the wind weakens to around 1 ms⁻¹ (≈15 % reduction; Fig. 12). Note that some $Vs$ is also induced in FRONT, which is likely due to the stratification change in the lower atmosphere.

In this experiment the sharp SST front is highly idealized, and it is uniformly strong everywhere along the coast, which are obviously not the case in a realistic configuration. Therefore, the weakening of the wind estimated in this section can be considered as an upper limit of the SST effect on the coastal wind. Under realistic conditions, with a less pronounced SST front, the impact should be minor, i.e., secondary compared to the orographic effect.

Fig. 13

SST feedback effects on the alongshore marine boundary layer height (MBLH) integrated over a coastal strip 100 km wide. The black and red lines represent the NODRAG and FRONT MBLH, respectively. A cold sharp coastal SST front induces a shallowing of the marine boundary layer

5 Discussion and conclusion

Using a long-term atmospheric simulation and sensitivity experiments, we assess wind profile characteristics off the US West Coast. The simulated winds are in good agreement with both satellite measurements of offshore wind and buoy measurements of relatively nearshore wind. The high quality of the simulation allows us to determine the wind drop-off characteristics and causes.

The wind drop-off over the CCS is mainly influenced by orography combined with coastline meandering. The coastal orography induces a weakening of the wind close to the coast through a vortex stretching. Coastline meandering enlarges the orographic effect through vortex stretching, turbulent flux divergence, and tilting-twisting. Combined with orography, it induces a wind reduction up to 65 % over a wide coastal strip (60 km in the ideal experiment, 80 km in the realistic case), in good agreement with Boé et al. (2011). The sea-land drag coefficient difference mainly acts as a barrier that turns the wind alongshore. It induces a weak wind drop-off over a coastal strip of about 15 km, but the wind reduction is rather small (less than 10 %). A strong SST front can affect the coastal wind profile. By stabilizing the air column, the MBL shallows and decouples from the wind aloft, inducing a weakening of the surface coastal wind. The wind drop-off occurs over a strip slightly larger than the SST front (40 km in that case), and the wind reduction is about 15 %. However, a large SST upwelling front is needed to induce a significant additional wind drop-off. Therefore, the SST effect can be considered as secondary compared to the orography. Rotunno and Smolarkiewicz (1995) and Epifanio and Rotunno (2005) show that the MBL can modulate the formation of orographic lee-side effects. Therefore, the shallowing of the MBL by the SST feedback could also change the orographic effect. In addition to the single-change sensitivity experiments in Table 1 and Sect. 4, another experiment was made using a realistic orography and an idealized SST front. The results show that the SST effect, by shallowing the MBL, generates additional turbulent flux divergence. It enlarges $L_d$ if it is smaller than 40 km. However, since the various effects are not linear, the possible interaction between the MBL height and the orographic effect is hard to separate from the SST feedback itself.

Due to the non-uniform spatial distribution of the mechanisms shaping the wind distribution, the wind drop-off is not geographically constant. It varies from ≈15 to 80 km. It is characterized by a seasonal variability that depends on the wind intensity and the SST upwelling front. An implication of this study is that when simulating the atmospheric circulation of an upwelling region, the atmospheric model needs to resolve well the orography and the coastline to represent accurately the vortex stretching impact on the wind and wind curl variability. This is an important point when forcing an oceanic model since the coastal wind shapes will strongly influence the mesoscale eddy activity and, subsequently, the net primary production (Renault et al. 2014). An additional simulation has been made using a 2 km horizontal resolution for the month of June 2000. As in Renault et al. (2012), the simulated wind drop-off length shows a convergence with the atmospheric model resolution. This illustrates the robustness of the results and shows that the simulated wind drop-off has a clear physical origin (the orography) rather than an apparently excessive dependence on numerical resolution (demonstrated in Capet et al. 2004). This is also consistant with other numerical studies (e.g., Boé et al. 2011; Dong et al. 2011) that show the good quality of WRF solutions using a similar configuration.

The CCS is characterized by a southward coastal current that, in a monthly average, can be as large as 0.4 ms⁻¹. To assess the current feedback impact on the wind drop-off, an additional experiment was done by assuming a coastal surface current of 0.4 ms⁻¹ over a coastal strip of 30 km. With a bulk formula for the surface stress, WRF calculates the frictional velocity and the wind stress by subtracting the surface current from the 10 m wind. The results in terms of Marine Boundary Layer Height, wind drop-off characteristics, and wind stress were not significantly changed (less than 2 % difference). However, the current-stress feedback may have some highly local influences through some (sub)mesoscale features (e.g., eddies and filaments) and even during an acceleration of the surface current at time scales shorter than a month. A fully coupled model is needed to assess such effects further

Further errors may arise from the absence of coupling processes. Uncoupled simulations may result in biases in the estimation of the MBLH. The atmospheric model has been forced by the Ostia SST (Stark et al. 2007), which has a relatively high spatial resolution of 5 km. However, cloud cover over the upwelling season is problematic for obtaining high-resolution data. Therefore, the effective resolution of this product may be similar to that of the microwave satellite products (25 km). Because the oceanic response to atmospheric forcing is driven by atmospheric stability conditions that have a large dependency on the air-sea temperature difference, a smoothed temperature may induce MBL shallowing close to the coast that is also too smooth. Additionally, the SST feedback, as suggested by Chelton et al. (2007) and Castelao (2012), depends on the steadiness of the wind. The wind drop-off representation (width and magnitude) may thus have some degree of inaccuracy. However, because we show that under a steady wind condition and an artificial sharp SST front, the wind could weaken up to 15 %, this is likely to be a secondary effect compared to the orography and coastline effects (65 %). Offshore, as shown by Chelton et al. (2007), the SST gradients significantly structure the wind, but the focus of this paper is not offshore. A fully coupled model would take us a step further in accounting for more realistic wind-SST interaction and current-stress interactions through mesoscale eddies.

In this study, we show what are the characteristics of the wind drop-off and what influences it. However, it is still unclear how the coast wind drop-off works. Further work will aim to develop an idealized dynamical study, as for example Rotunno and Smolarkiewicz (1995). It could help to understand precisely the mechanisms generating the vortex stretching. For example, the presence of orography induces more subduction of cold air into the MBL, inducing a shallowing of it and then a weakening of the wind by vortex stretching.

The EBUSs (US West Coast, Peru-Chile, Benguela, and Canary) have common characteristics but also present differences. The origin of those differences are not well known. Some of them can obviously be due to a different wind regime (e.g., the coastal jet off central Chile); however, their wind drop-off characteristics may also explain differences in term of mean circulation and primary production (Gruber et al. 2011; Renault et al. 2014). For instance, the Peru-Chile EBUS is marked by the presence of the Andes. A higher orography should induce a large vortex stretching and then a large wind drop-off length. However, Renault et al. (2012) show that during a well defined atmospheric coastal jet event (10 ms⁻¹), the wind drop-off extends over a length of about 30 km, which is smaller than that of the US West Coast. This could be explained by the straight Chilean coastline: there is no cape that enlarges the orographic effect, and the Andes induce a sharper drop-off (smaller $L_d$) than the US West Coast mountains. Studies of the wind drop-off characteristics and origin using a similar methodology should be done to gain a better understanding of EBUS coastal dynamics.