Sea level interannual variability along the west coast

0 downloads 0 Views 4MB Size Report
Take-home message(s): The amplitude of sea level seasonal cycle is much larger than the. 127 interannual variability along the WCI. The sea level interannualΒ ...
1 2

Geophysical Research Letters

3

Supporting Information for

4

Sea level interannual variability along the west coast of India

5 6 7 8 9 10

I. Suresh1, J. Vialard2, M. Lengaigne2,3, T. Izumo2,3, V. Parvathi1, and P.M. Muraleedharan1 1

CSIR-National Institute of Oceanography (CSIR-NIO), Goa, India

2

LOCEAN-IPSL, Sorbonne UniversitΓ© (UPMC, Univ Paris 06)-CNRS-IRD-MNHN, Paris, France

3

Indo-French Cell for Water Sciences, IISc-NIO-IITM–IRD Joint International Laboratory, CSIR-NIO, Goa, India

11 12

Contents of this file

13

1. Details of the linear, continuously stratified ocean model

14

2. Details of the sensitivity experiments

15

3. Figures in support of the main text

16

a. Figure S1 shows the amplitude of seasonal and interannual sea level variability in the

17

Indian Ocean

18

b. Figure S2 shows that the interannual sea level variability 1st Empirical Orthogonal

19

Function (EOF1) is dominated by the influence of the Indian Ocean Dipole (IOD)

20

c. Figure S3 complements the main text Figure 2 by showing the patterns of sea level

21

anomalies and wind-stress forcing associated with an IOD event for August, October,

22

December, and February.

23

d. Figure S4 provides theoretical travel times of first baroclinic mode Rossby waves in

24 25

the Bay of Bengal (BoB) e. Figure S5 provides the process decomposition of west coast of India (WCI) signals

26

separately for positive and negative IODs, allowing to discuss their asymmetrical

27

influence on the WCI

28 29

f.

Figure S6 completes the main text Figure 3 by showing typical IOD signals (and their process decomposition) in other relevant regions

30

1

31

1. The linear, continuously-stratified ocean model

32

We use the same linear, continuously stratified (LCS) model as that used to study the

33

West Coast of India (WCI) intraseasonal variability and seasonal cycle in Suresh et al. (2013;

34

2016). This model is a modified version of the one presented in McCreary et al. (1996). The

35

equations of motion are linearized about a state of rest with a realistic background

36

stratification derived from the averaged Indian Ocean (15oS-15oN, 40oE-100oE) World Ocean

37

Atlas (Locarnini et al., 2010) climatological potential density profile, and the ocean bottom is

38

assumed flat. With these conditions, the solutions are represented as a sum of vertical normal

39

modes, which are eigenfunctions, πœ“! (𝑧), satisfying the equation,

40

subject to the boundary conditions, πœ“!" βˆ’π· = πœ“!" 0 = 0, where 𝑁! (𝑧) is the Brunt–

41

VΓ€isΓ€lΓ€ frequency and D (= 4000 m) is the ocean depth. The eigenvalue, 𝑐! , of the above

42

equation represents the characteristic Kelvin wave speed of the n-th mode. We use the first 5

43

baroclinic modes, the characteristic speeds of which are 252 cm s-1, 155 cm s-1, 88 cm s-1, 69

44

cm s-1, and 53 cm s-1, but a very similar solution is obtained when using 20 modes instead, as

45

in Suresh et al. (2013; 2016).

! !!!

πœ“!"

!

!

= βˆ’ ! ! πœ“! (𝑧), !

46

The response of each mode is obtained on a regular, 0.25Β°-resolution grid over a domain

47

that covers the Indian Ocean from 30Β°S to 30Β°N and from 30Β°E to 110Β°E, and with a coastline

48

determined by the 200-m isobaths (Smith & Sandwell, 1997), a good proxy for the

49

continental shelf break. Vertical mixing has the same form as McCreary et al. (1996), but

50

with a 7-year dissipation time scale for the first baroclinic mode. The horizontal mixing

51

coefficient is set to 5000 m2.s-1. Wind is introduced into the ocean as a body force with the

52

same profile (constant down to 50 m and ramped to zero at 100 m) as in McCreary et al.

53

(1996).

54

The model is forced with daily TropFlux wind-stress anomalies (long-term mean

55

removed) during 1979-2013 (available from http://www.incois.gov.in/tropflux/) (Praveen

56

Kumar et al., 2013). We refer to this solution as the control (CTL) experiment.

57 58

2. Details of sensitivity experiments

59

The processes that influence the sea level along the WCI are discussed in detail in the

60

main text (refer to Section 3.2). The contribution of each process is isolated using set of

61

experiments with dampers and/or special boundary condition applied in specific regions. 2

62

The dampers are constructed as in McCreary et al. (1996) from a Newtonian damping

63

which allows the waves to decay over a region of interest with an e-folding scale of 1.5 Ξ”x,

64

where Ξ”x is the grid size. To avoid distortions in the solutions, the Newtonian damping

65

coefficients are ramped linearly to zero within 1o from the edges of the damping region. As

66

demonstrated in Suresh et al. (2016), our results are not very sensitive to the specific choice

67

of these dampers or ramp. The solution obtained with such a damper will thus be free from

68

the effect of waves emanating from this region. The dampers in our experiments are designed

69

such that their edges match and the ramping coefficients add up to 1 there. That is, the sum of

70

the various process solutions is exactly equal to the CTL experiment, owing to the linearity of

71

the model.

72 73

As in McCreary et al. (1996), we also use the following special boundary condition in some sensitivity experiments: 𝒖𝒏 = 𝒏 βˆ™ 𝒗𝒏 = βˆ’π’. π’ŒΓ—

74

𝑭𝒏 𝒇

𝑣! = (π’ŒΓ—π’) βˆ™ 𝒗𝒏 = 0, 75

where n is a unit vector perpendicular to the coast, k is an upward-directed unit vector, and f

76

is the Coriolis parameter. 𝒗𝒏 = 𝑒! , 𝑣! π‘€π‘–π‘‘β„Ž 𝑒! , 𝑣! the horizontal components of velocity

77

perpendicular and parallel to the coast. 𝑭𝒏 = 𝐹! , 𝐺! π‘€β„Žπ‘’π‘Ÿπ‘’ 𝐹! and 𝐺! are given by

78

𝐹! = 𝜏 ! 𝑍! /(𝜌𝐻! ) and 𝐺! = 𝜏 ! 𝑍! /(𝜌𝐻! ) . 𝜏 ! and 𝜏 ! are the zonal and meridional wind

79

stresses, and 𝐻! =

80

vertical profile, Ξ–(𝑧), used here follows that of McCreary et al. (1996), which is constant

81

down to 50 m and ramped to zero at 100 m. The second equation in the above represents no

82

slip condition. The effect this boundary condition is to filter out alongshore forcing of coastal

83

Kelvin waves, by allowing the Ekman flow to pass through the boundaries and hence

84

generating no convergence and sea level anomaly at the coast.

! πœ“! !! !

𝑑𝑧 and 𝑍! =

! 𝑍 !!

𝑧 πœ“! 𝑧 𝑑𝑧 are the coupling factors. The

85

In this paper, we carry out a series of sensitivity experiments to separate out the

86

processes mentioned in Section 3.2. These experiments are essentially the same ones

87

presented in Suresh et al. (2016).

88

The equatorial forcing process (EQ) is isolated using a damper over the entire

89

equatorial region between 4oS and 4oN (light blue box on Fig. 2c in the main text; experiment

90

EXP1), with a 1o ramp at the northern and southern edges to minimize distortion of the signal

91

near the edges. The solution obtained with this damper has no signals emanating from the 3

92

equatorial region, i.e. no equatorial forcing. CTL-EXP1 thus yields northern Indian Ocean

93

sea-level variations resulting from equatorial forcing.

94

The BoB forcing process (BB) is isolated in an experiment with both the equatorial

95

and BoB dampers (light blue and green boxes on Fig. 2cd; EXP2). EXP2 will thus be free

96

from both EQ and BB signals. EXP1-EXP2 hence yields the BB contribution to sea-level

97

variability. We further isolate the contribution of the BoB alongshore forcing (ABB) by

98

applying the special boundary condition described above within the BoB (green box on Fig.

99

2d; EXP2’). CTL-EXP2’ hence isolates the effect of ABB forcing. The BoB interior Ekman

100

pumping contribution (IBB) is further obtained by subtracting ABB from BB solution.

101

The ST process, which includes the alongshore forcing (AST) near the southern tip of

102

India and Sri Lanka and the Ekman pumping due to wind-stress curl off the east coast of Sri

103

Lanka (IST), is separated out by using an additional damper near the southern tip of India,

104

extending up to 90oE (purple box on Fig. 2b; EXP3; i.e. EXP3 has dampers in the EQ, BB &

105

ST regions). The difference EXP2-EXP3 thus yields the effect of ST (both alongshore

106

forcing and Ekman pumping east of Sri Lanka). We further isolate the contribution of AST in

107

the ST process in an experiment EXP3’ in which we impose the special boundary condition

108

described above within the purple box. CTL-EXP3’ hence isolates the effect of alongshore

109

forcing. The IST contribution is further obtained by subtracting this AST contribution from

110

the ST process.

111

The effect of local alongshore forcing along the WCI is finally obtained by applying

112

the special boundary condition along the entire WCI, in addition to all the above dampers

113

(EXP4). EXP3-EXP4 thus isolates the effect of WCI winds. At the same time, EXP4 does

114

not have the effect of equatorial, BoB and southern tip of India forcing and is also free from

115

the effect of WCI alongshore winds. It has only the effect of interior wind forcing in the

116

Arabian Sea. Both the above processes constitute the AS process.

117

The linearity of our model ensures that EQ+BB+ST+AS=CTL, except for numerical

118

errors, which are orders of magnitude smaller than any of the components. As noted above,

119

the BB, ST and AS contributions can further be decomposed into interior (I) and alongshore

120

(A) contributions, e.g. ST=IST+AST.

121

4

122

123 124

Figure S1: Standard deviation of the observed sea level a) seasonal cycle and b) interannual

125

anomalies.

126 127

Take-home message(s): The amplitude of sea level seasonal cycle is much larger than the

128

interannual variability along the WCI. The sea level interannual variability is much larger in

129

the BoB than along the WCI. This probably explains why previous studies discussed the

130

seasonal cycle on the WCI (e.g. Suresh et al. 2016; because of its high amplitude) or the

131

interannual variability in the Bay of Bengal (e.g. Aparna et al. 2012; because it has a high

132

amplitude) but not the interannual variability on the WCI (because of its relatively weaker

133

amplitude). While the WCI SLA interannual variability is weak, it has strong societal

134

impacts. Parvathi et al. (2017) indeed showed that even small perturbations in the WCI

135

thermocline/oxycline depths during SON can potentially trigger/suppress anoxic conditions,

136

due to the upwelling-driven shallow thermocline/oxycline depths at this time of the year.

137

Those anoxic conditions have strong societal impacts, owing to their adverse influence on

138

ecosystems and fisheries. This emphasizes the need for better understanding of the WCI

139

interannual sea level variability (see introduction section in the text for more details)

140

5

141 142 143

Figure S2: Normalized first principal component (PC1, black) of the northern Indian Ocean

144

interannual monthly sea-level anomalies from the CTL experiment, along with the normalized

145

Dipole Mode index (DMI, red) and multivariate ENSO Index (MEI, blue).

146 147

Take-home message(s): PC1 has a higher correlation with DMI Index during the IOD peak

148

(SON; 0.95) than with MEI index during ENSO peak (NDJ; 0.81). In line with the previous

149

literature on the subject (see the main text), this indicates that the EOF1 captures the intrinsic

150

sea-level variability associated with the IOD, rather than that associated with the ENSO.

151

6

152 153

Figure S3: Maps of (left column) CTL SLA (cm) and wind-stress (N.m-2) vectors and (right

154

column) wind-stress curl (N.m-3; shading) and alongshore wind stresses (N.m-2; shading along

155

the coastline) lead-lag regression onto the normalized SON-averaged DMI during (a, b)

156

August; (c, d) October; (e, f) December; and (g, h) February. Red boxes marked on panel a

157

will be used in Figure S5.

158 159

Take-home message(s): Typical sequence of wind forcing and sea-level response associated

160

with a positive IOD event. Note that the west coast of India signals shift from positive in

161

boreal fall to negative in winter. 7

162 163

Figure S4: a) Sketch showing the pathways of coastal Kelvin and mid-latitude Rossby waves

164

travelling from one given latitude on the eastern rim to the same latitude on the western rim of

165

the BoB. Also shown, as a function of latitude, are b) Rossby wave phase speed, computed

166

with a typical first baroclinic wave speed of 2.5 m.s-1, c) distance across BoB basin, d) Rossby

167

wave travel time to cross the basin.

168 169

Take-home message(s): A first-baroclinic mode coastal Kelvin wave typically takes ~20

170

days (with 2.5 m.s-1) to travel around the BoB coastal waveguide (~4300 km), while Rossby

171

waves typically need 2 to 6 months, depending on the latitude, to cross the BoB basin.

8

172

173 174 175

Figure S5: Regression of WCI CTL interannual sea-level anomalies in SON to the

176

normalized SON-averaged DMI during a typical positive (red; i.e. regression only for DMI >

177

0) and negative (blue; i.e. regression only for DMI < 0) IOD events, and its decomposition

178

into the ST, EQ, BB, and AS processes.

179 180

Take-home message(s): For a given IOD amplitude, the WCI sea-level response is larger for

181

positive IODs, due to stronger ST and BB processes, which can be linked to zonal equatorial

182

wind anomalies that extend more northward during positive IOD than their negative

183

counterpart.

184

9

185 186 187

Figure S6: Same as Figure 3 in the main text, but for coastal locations on the a) eastern BoB;

188

b) east coast of India; c) southeastern Arabian Sea, indicated by red frames in Figure S3a.

189 190

Take-home message(s): Sea-level signals are almost entirely associated with equatorial

191

remote forcing in the eastern BoB. Local BoB forcing starts contributing (but rather weakly)

192

along the east coast of India, where EQ is still the most dominant process. The contributing

193

processes to the sea level interannual variability in the southeastern Arabian Sea are very

194

similar to those along the WCI (Fig. 3 in the main text), i.e., the ST and EQ, but with a

195

slightly stronger contribution of Arabian Sea local forcing.

196 197

10