Quick viewing(Text Mode)

Flow Topography Interactions Around Sri Lanka and the Maldives

Flow Topography Interactions around and the

Danielle Su Li Wan BSc (Hons), BMarSc

This thesis is presented for the degree of Doctor of Philosophy of The University of Western Oceans Graduate School & UWA Oceans Institute

2020 This page has been intentionally left blank.

Thesis Declaration

I, Danielle Su, certify that

This thesis has been substantially accomplished during enrolment in this degree.

This thesis does not contain material which has been submitted for the award of any other degree or diploma in my name, in any university or other tertiary institution.

In the future, no part of this thesis will be used in a submission in my name, for any other degree or diploma in any university or other tertiary institution without the prior approval of The University of Western Australia and where applicable, any partner institution responsible for the joint-award of this degree.

This thesis does not contain any material previously published or written by another person, except where due reference has been made in the text and, where relevant, in the Authorship Declaration that follows.

This thesis does not violate or infringe any copyright, trademark, patent, or other rights whatsoever of any person.

This thesis does not contain work that I have published, nor work under review for publication.

Signature:

Date: 8th January 2020

i Abstract

Ocean currents in the Northern (NIO) interact with the Maldives Island chain and Sri Lanka to create an Island Mass Effect (IME). Here, three-dimensional flow patterns, generated in the lee of the islands enhance oceanic primary productivity. In the NIO, reversing currents with strong temporal and spatial variability, provide a unique system for a detailed study of the IME. Although the IME in the NIO is evident from remote sensing data, there has yet to be an investigation of IME flow structure and variability across seasonal and inter-annual time scales. In this study, a high-resolution numerical model was configured to examine monsoon forcing and Indian Ocean Dipole (IOD) Mode on IME development around Sri Lanka and the Maldives. The model was validated using field and remotely sensed measurements. Results indicated that interaction between the monsoon currents and topography generated island wakes in the lee of the Maldives Island chain and around Sri Lanka. The monsoon currents also changed the location of the island wakes except in the southern section of the Maldives

(south of 4°N) where equatorial currents dominated. Seasonal transport pathways were identified through particle tracking experiments and proved that the Maldives and Sri Lanka were connected during the south-west monsoon. In contrast, there was minimal connectivity during the north-east monsoon. Negative IOD events were found to intensify island wakes through depth and energy budget analyses confirmed that wind forcing had a dominant role in kinetic energy generation around the islands. Collectively, this research contributes to a growing body of work to offer further insight on oceanographic processes in a poorly understood part of the Indian Ocean basin.

ii Table of Contents

Thesis Declaration i Abstract ii Table of Contents iii List of Figures vi List of Tables xiv Acknowledgements xv Authorship Declaration xviii

Chapter 1. Introduction ...... 1 1.1 Overview ...... 1 1.2 Objectives and Approach ...... 6 1.3 Thesis Structure ...... 8

Chapter 2. Background ...... 13 2..1 General Theory ...... 13 2.1.1 The Island Mass Effect ...... 13 2.1.2 Island Wakes ...... 15 2.2 The Northern Indian Ocean ...... 18 2.2.1 The ...... 18 2.2.2 Climate Modes ...... 28 2.2.3 Study Region ...... 32 2.2.4 Sri Lanka ...... 36 2.2.5 The Maldives Archipelago ...... 38 2.3 Numerical Models ...... 40 2.3.1 The Regional Ocean Modelling System ...... 41 2.4 Summary ...... 44

Chapter 3. Monsoonal Variability of the Island Mass Effect around 47 Sri Lanka and The Maldives ...... 3.1 Summary ...... 47 3.2 Introduction ...... 48 3.3 Methodology ...... 53 3.3.1 Model Setup...... 53 3.3.2 Validation Data Sets ...... 63 3.4 Model Evaluation ...... 64 3.5 Results ...... 72 3.5.1 Seasonal mean circulation ...... 72 3.5.2 Intraseasonal circulation variability ...... 78

iii 3.5.3 Daily circulation variability ...... 86 3.6 Discussion ...... 92 3.6.1 Seasonal mean circulation ...... 92 3.6.2 Intra-seasonal circulation variability ...... 93 3.7 Concluding Remarks ...... 94

Chapter 4. Ocean connectivity pathways between Sri Lanka and 99 The Maldives ...... 4.1 Summary ...... 99 4.2 Introduction ...... 100 4.3 Methodology ...... 105 4.3.1 Numerical Models ...... 105 4.3.2 Experimental Setup ...... 107 4.4 Results ...... 111 4.4.1 SOURCE Experiment ...... 111 4.4.2 ADVECTION Experiment ...... 116 4.4.3 DEPTH Experiment ...... 119 4.5 Discussion ...... 126 4.5.1 SOURCE Experiment ...... 127 4.5.2 ADVECTION Experiment ...... 128 4.5.3 DEPTH Experiment ...... 130 4.6 Concluding Remarks ...... 131

Chapter 5. Indian Ocean Dipole Influence on Island Wake 136 Development ...... 5.1 Summary ...... 136 5.2 Introduction ...... 137 5.3 Methodology ...... 142 5.3.1 Numerical Model ...... 142 5.3.2 Non-dimensional parameters for island wakes ...... 143 5.3.3 Data Sets ...... 149 5.3.4 Eddy Kinetic Energy Budgets ...... 151 5.4 Results ...... 153 5.4.1 Inter-annual circulation variability ...... 153 5.4.2 IOD Influence on island wake structures ...... 158 5.4.3 Okubo-Weiss Parameter and Climate Indices ...... 161 5.4.4 Energy Budget Analysis ...... 162 5.5 Discussion ...... 175 5.6 Concluding Remarks ...... 177

Chapter 6. Synthesis ...... 180 6.1 Summary ...... 180 iv 6.2 Findings and Conclusions ...... 182 6.3 Implications ...... 184 6.4 Recommendations for future work ...... 185

Bibliography...... 188

v List of Figures

1.1 10-year averaged wind speed and surface currents in the northern Indian Ocean obtained from ECMWF and HYCOM respectively. (a), (c) Northeast monsoon winds and currents; (b), (d) Southwest monsoon winds and currents. SMC – South Monsoon Current; NMC – North Monsoon Current. The magnitude of the current speeds is represented by the coloured arrows...... 3

1.2 (a) Global map with study region identified in red (b) Bathymetry of study region in the Indian Ocean...... 5

1.3 General overview of the proposed thesis chapters and schematic of the temporal scales and forcing addressed in each chapter...... 9

2.1 Concept diagram of several mechanisms that may generate the IME (Gove et al. 2016). These include current-bathymetric interactions that can drive vertical transport of water masses via upwelling, downstream mixing and eddies, and internal waves; island-associated inputs, such as submarine groundwater discharge and outflow from rivers...... 15

2.2 Monsoon wind stress fields from the 1990–1998 National Centres for Environmental Prediction (NCEP) (Kalnay et al., 1996) climatology (vectors) and depths of 20°C isotherm (Z20) from Simple Ocean Data Assimilation (SODA) (mean for 1992–2001, color shaded) for (a) January, (b) June, (c) August, and (d) November (Schott et al., 2009)...... 19

2.3 Schematic representation of the large-scale circulation of the Indian Ocean during the (a) winter monsoon (Northeast Monsoon) and the (b) summer monsoon (Southwest Monsoon) from Schott et al. (2009). Currents identified are the (SEC), South Equatorial Countercurrent (SECC), Northeast and Southeast Current (NEMC and SEMC), East African Coastal Current (EACC), (SC), Southern Gyre (SG) and Great Whirl (GW) and associated upwelling wedges (green shades), Southwest and Northeast Monsoon Currents (SMC and NMC), South Java Current (SJC), East Gyral Current (EGC), and (LC). Subsurface currents are shown in purple. Depth contours shown are for 1000 m and 3000 m (grey). Red vectors (Me) indicate directions of meridional Ekman transports. ITF indicates (from Schott et al.,2009)...... 25

2.4 Evolution of rainfall during ISMR from 1951 to 2014 (from Zheng et al. 2016). Strong (red) and weak (blue) ISMRs are identified by the departure of JJAS rainfall of each year from the JJAS rainfall climatology computed over the period 1951–2014, whose departure values are larger than +10%, smaller than 10%, and within 10% and 10% of the seasonal climatology, respectively. The dashed line denotes a value of 110% (90%) 27 vi of seasonal climatology (i.e., the mean climatology is 937.7 mm), which is 1031 mm (844 mm). The years of strong and weak ISMRs are denoted by the numbers over the bars......

2.5 Seasonal climatology of SeaWiFS chlorophyll (Jan, Apr, Aug, Oct). The climatology was created using monthly Level 3 Standard Mapped Image data from September 1997 through January 2002, which were obtained from the Goddard DAAC (http:// daac.gsfc.nasa.gov/) (Wiggert, et al., 2006) ...... 28

2.6 Seasonality of Indian Ocean Climate Modes from (Schott et al., 2009) . . 29

2.7 Schematic of the Indian Ocean Dipole Mode and the resulting changes in convection pathways and equatorial thermocline depth at (a) neutral phase (b) positive phase (c) negative phase (Bureau of Meteorology, Australia) ...... 32

2.8 Schematic of the major recirculation features in the proposed study area during the NEM (left panel) and SWM (right panel). SD represents the recirculation feature known as the Sri Lanka Dome (de Vos et al., 2013) 35

2.9 The seasonal climatology of remotely sensed SST and currents (left column of panels), salinity (center column) and chlorophyll (right column) for a region of the IO encompassing the RAMA (Research Moored Array for African–Asian–Australian Monsoon Analysis and Prediction) mooring at the equator, 80.5°E (magenta dot) (Strutton et al., 2015) ...... 36

2.10 Monthly composite images of chlorophyll-a derived from SeaWiFS during 1998-2002 showing the evolution of the chlorophyll bloom around the southern tip of and Sri Lanka during the summer monsoon (Hood et al., 2015) ...... 38

2.11 MODIS derived chlorophyll plume (left panel) around the Maldives and SST (right panel) during the NEM Sasamal (2006) ...... 40

3.1 Daily snapshots of surface chlorophyll-a concentrations around Sri Lanka

and the Maldives during the Northeast Monsoon (a) – (d) and the

Southwest Monsoon (e) – (h). (d) and (h) displays multiple von Karman

vortex streets swirling downstream from the Huvadhoo Atoll (Figure 3.2),

Maldives during the Northeast Monsoon and the Southwest Monsoon 50 respectively. Images adapted from EOSDIS Worldview ......

3.2 Bathymetry map of (a) Maldives and (b) Sri Lanka ...... 53

3.3 Model domain showing bathymetry and locations of model validation points. The observational data used for model validation are denoted in blue circles for selected tide gauge sites (Table 3.2). The locations of 56 vii ARGO profile data used to compare model results are represented in yellow shaded boxes [A1-A8]. The RAMA mooring, M1 is denoted by the pink triangle and a summary of the time periods and variables used for validation is summarized in Table 3.2......

3.4 Gridded map for grid domain values of (a) rx0 and (b) rx1. The recommended values are rx0 <0.25 and 5

3.5 Vertical grid in the SNIO-ROMS where the top panel is the zoomed inset for the top 200m and the lower panel shows the vertical resolution over the bathymetry to a depth of 5000m. The red line indicates where the thermocline, Tc is set to 150m...... 58

3.6 Latitudinal transect along the northern boundary of zonal velcities from (a) HYCOM product prior to smoothing (b) HYCOM product post smoothing...... 59

3.7 An example of the hybrid bathymetry (blue) created from both GEBCO (black) and HYCOM (red) along the northern boudary. To create the hybrid bathymetry, HYCOM bathymetry was used for the outermost boundary bathymetry (100% HYCOM) and decreased by 5% per grid point until it was fully replaced by GEBCO bathymetry in the interior. For this example, the hybrid bathymetry comprises of 60% HYCOM bathymetry and 40% GEBCO bathymetry within the northern boundary zone. The box on the left shows the progression from pure HYCOM bathymetry to the Hybrid bathymetry to pure GEBCO bathymetry for all the boundaries...... 61

3.8 Sponge layer for model domain boundaries where (a) shows the decrease of viscosity across 60 grid points where the highest viscosity starts at the outermost (grid point = 0) while (b) indicates the width of the changing viscosity across the sponge layer. Units are in m2s-1...... 61

3.9 Monthly climatology of observed sea level derived from UHSLC (black line) and SNIO-ROMS model (red) from tide gauge locations at (1) Sri Lanka, Colombo (2) Cochin, India (3) Minicoy, India (4) Male, Maldives (5) Hanimaadhoo, Maldives (6) Gan, Maldives. All units are in [m]. . . . . 65

3.10 Seasonal sea surface temperature climatology difference between the

model and satellite from 2006-2016 during the (a) NEM and (b) SWM

with contour interval of 0.5°C. Positive (red) values indicate an

overestimation of SST while negative (blue) values denote an

underestimation of SST...... 68

viii 3.11 Q-Q plots of measured vs predicted values from climatology (2006-2016) for (a) zonal velocities during the (a) NEM and (b) SWM as well as the meridional velocities during the (c) NEM and (d) SWM...... 69

3.12 Comparison of Argo (black) and SNIO-ROMS (red) climatological monthly mean temperature profiles at 8 locations identified around Maldives and Sri Lanka (see Figure 3.3). Note that the y-axis are at different scales...... 70

3.13 Volume transport from model output at 80.5°E across 2.5°S-2.5°N, relative to a fixed depth of 80m for the NEM (black) and SWM (white) during 2006-2016...... 71

3.14 Seasonal mean circulation of velocities (arrows) and speed (colormap) for (a) NEM and (b) SWM from climatology (2006-2016) ...... 73

3.15 Map of depths of the 26°C isotherm (D26) relative to the mean during the NEM (left column) and SWM (right column). Shallower (deeper) values where the colormap is warmer (cooler) identify locations where the 26°C isotherm (D26) is uplifted (suppressed)...... 77

3.16 Top to bottom: NEM model output for December, January and February. Left panel shows the longitudinal temperature transect for Maldives along 72°E and the right panel shows the longitudinal temperature transect for Sri Lanka along 81°E. Surface circulation and temperature for the model domain is shown in the middle panel. Red lines indicate transect locations. 81

3.17 Top to bottom: SWM model output for June, July and August. Left panel shows the longitudinal temperature transect for Maldives along 74°E and the right panel shows the longitudinal temperature transect for Sri Lanka along 81°E. Surface circulation and temperature for the model domain is shown in the middle panel. Red lines indicate transect locations...... 85

3.18 Daily time series of zonal velocity and temperature along with longitudinal/latitudinal transects of key upwelling and downwelling events during the NEM at four selected locations in the model domain. Negative (positive) zonal velocities indicate westward flow (eastward) and are indicated in blue (red). Contour map denotes the region’s bathymetry...... 88

Daily time series of zonal velocity and temperature along with 3.19 longitudinal/latitudinal transects of key upwelling and downwelling events during the NEM at four selected locations in the model domain. Negative (positive) zonal velocities indicate westward flow (eastward) and are indicated in blue (red). Contour map denotes the region’s bathymetry. 91

3.20 Flow patterns during the SWM showing the presence of the Sri Lanka Done to the east of Sri Lanka...... 95 ix

4.1 Snapshot of chlorophyll concentrations around Sri Lanka and the Maldives on 12th December 2017 (Adapted from EOSDIS worldview). . . 101

4.2 Model domain and bathymetry with particle release locations for the SOURCE (red boxes) and ADVECTION (blue box) experiments. The blue boxes mark areas where the IME has been observed to occur...... 109

4.3 Monthly simulations for SOURCE – NEM experiment. Particle trajectories were plotted separately for clarity to show the evolution of the surface pathways from (a) 80.5°E and 81.5°E (b) 82.5°E (c) 83.5°E (d) 84.5°E and 85.5°E (e) 86.5°E and 87.5°E...... 113

4.4 (Top panel): Locations where particles were tracked near Sri Lanka (Box 2) and along the Maldives west coast (Box 1W). (Bottom panel): The percentage of particles that arrived in Box 1 and 2 from the different seeding locations during the NEM. Note that the y-axis has different scales for visualization purposes...... 114

4.5 Monthly simulations for SOURCE – SWM experiment. Particle trajectories were plotted separately for clarity to show the evolution of the surface pathways from (a) 73.5°E and 74.5°E (b) 71.5°E and 72.5°E (c) 69.5°E and 70.5°E (d) 68.5°E and 67.5°E (e) 66.5°E and 65.5°E. . . . 115

4.6 (Top panel): Locations where particles were tracked at Maldives (Box 1E), Sri Lanka (Box 2) and the Sri Lanka Dome (Box 3). (Bottom panel): The percentage of particles that arrived in Box 1 and 2 from the different seeding locations during the SWM. Note that Box 1E and 2 have similar y-axis scales but Box 3 has a different scale...... 116

4.7 (Top panel): Locations where particles were tracked at Maldives (Box 1W), Sri Lanka (Box 2) and the Sri Lanka Dome (Box 3). (Bottom panel): The percentage of particles that arrived in Box 1W and 2 over time from the different seeding locations during the NEM...... 118

4.8 (Top panel): Locations where particles were tracked at Maldives (Box 1),

Sri Lanka (Box 2) and the Sri Lanka Dome (Box 3). (Bottom panel): The

percentage of particles that arrived in Box 1,2 and 3 over time from the

different seeding locations during the SWM. Note that the y-axis (%) is

different for each of the boxes...... 119

4.9 DEPTH –NEM Experiment: Monthly mapped trajectories of particles released from 83.5°E (green), 84.5C (yellow) and 85.5°E (orange) at depths (a) 50m (b) 100m (c) 150m...... 121

x 4.10 The percentage of particles advected into Box 1W (Maldives) and Box 2 (Sri Lanka) and the depths and locations released from ...... 121

4.11 Mean vertical depth and standard deviation (error bars) of the particles during the NEM for particle simulations released at depth (a) 50m (b)100m (c)150m for every 7 days...... 123

4.12 DEPTH –SWM Experiment: Monthly mapped trajectories of particles released from 65.5°E (black), 71.5°E (green) and 74.5°E(blue) at depths (a) 50m (b) 100m (c) 150m...... 125

4.13 The percentage of particles advected into Box 1E (Maldives), Box 2 (Sri Lanka) and Box 3 (Sri Lanka Dome) and the depths and locations released from. Note that the colour scale is different for each box...... 125

4.14 Mean vertical depth and standard deviation (error bars) of particles during the SWM for particle simulations released at depth (a) 50m (b)100m (c)150m for every 7 days...... 126

4.15 Schematic illustration of current systems and nearshore inputs to the IME in the NIO. Currents and recirculation features occurring during the NEM (SWM) are depicted in blue (red) while year-round currents like the SECC are shown in black. The size of the coloured arrows around the Maldives, India and Sri Lanka show the directionality and relative contribution of particles to the wake zones (boxes) ...... 133

5.1 Bathymetry map of Maldives and Sri Lanka and locations where the flow structure of the island wakes were compared (red boxes)...... 147

5.2 (a) Southern Oscillation Index (SOI) and (b) Indian Ocean Dipole Mode Index (DMI). The shaded boxes identify the pure IOD years used for comparison, 2008 (positive IOD, red), 2014 (neutral IOD, green) and 2013 (negative IOD, blue) ...... 150

5.3 From top to bottom: Annual mean of surface current speeds (colourmap) and velocities (arrows), SST anomalies and chlorophyll concentrations during (a) Positive IOD year (b) Neutral/Weak IOD year and (c) Negative IOD year...... 155

5.4 Mean wind speed and direction during July for (a) Neutral/Weak (neu- IOD) year (b) Positive (p-IOD) year and (c) Negative (n-IOD) year. Wind velocities were obtained from the ECMWF-ERA Interim dataset and gridded to the model domain...... 156

5.5 Vorticity profile at Thiladhunmathi atoll during (a) Positive IOD – July 2008 (b) Neutral/Weak IOD – July 2014 (c) – July Negative IOD 2013. Top panel displays the surface vorticity while the bottom panel shows the 157

xi variation of the vorticity through depth at 50m intervals up till 150m. The red line is a reference point for the wake through depth......

5.6 Vorticity profile at Ari Atoll during (a) Positive IOD – July 2008 (b) Neutral/Weak IOD – July 2014 (c) – July Negative IOD 2013. Top panel displays the surface vorticity while the bottom panel shows the variation of the vorticity through depth at 50m intervals up till 150m. The red line is a reference point for the wake through depth...... 158

5.7 Vorticity profile at Huvadhoo atoll during (a) Positive IOD – July 2008 (b) Neutral/Weak IOD – July 2014 (c) – July Negative IOD 2013. Top panel displays the surface vorticity while the bottom panel shows the variation of the vorticity through depth at 50m intervals up till 150m. The red line is a reference point for the wake through depth...... 159

5.8 Vorticity profile at Sri Lanka during (a) Positive IOD – July 2008 (b) Neutral/Weak IOD – July 2014 (c) – July Negative IOD 2013. Top panel displays the surface vorticity while the bottom panel shows the variation of the vorticity through depth at 50m intervals up till 150m. The red line is a reference point for the wake through depth...... 160

5.9 Spatial distribution of Okubo-Weiss (OW) Parameter around the Maldives during November 2011 with positive OW (red) being shear- dominated and negative OW (green) being vorticity dominated and the (b) shear-dominated exceedance profiles where OW > 2 standard deviations of the regional mean OW (�) and (c) vorticity-dominated exceedance profiles where OW < -2�...... 162

5.10 (a) Spatial distribution of Okubo-Weiss (OW) Parameter around Sri Lanka during November 2011 and the (b) shear-dominated exceedance profiles where OW > 2 standard deviations of the regional mean OW (�) and (c) vorticity-dominated exceedance profiles where OW < -2�...... 162

5.11 Energy budget analysis around the Maldives during p-IOD, neu-IOD and n-IOD years where (a)-(c) shows the annual mean EKE and the spatial distribution of (d) – (f) KmKe (barotropic) (g)-(i) PeKe (baroclinic) and (j-l) FeKe (wind) ...... 165 5.12 (a) Time series of barotropic (KmKe), baroclinic (PeKe) and wind (FeKe) energy on the left axis and mean EKE (black) on the right axis for a positive IOD year. Note the different units for both axes. (b) Percentage distribution of KmKe (red), PeKe (green) and FeKe (blue) from the sum of KmKe, PeKe and FeKe. Note that the sum of the energy budget terms does not necessarily equate to the total mean EKE in (a)...... 166

5.13 (a) Time series of barotropic (KmKe), baroclinic (PeKe) and wind (FeKe) energy on the left axis and mean EKE (black) on the right axis for a positive IOD year. Note the different units for both axes. (b) Percentage 167 xii distribution of KmKe (red), PeKe (green) and FeKe (blue) from the sum of KmKe, PeKe and FeKe. Note that the sum of the energy budget terms does not necessarily equate to the total mean EKE......

(a) Time series of barotropic (KmKe), baroclinic (PeKe) and wind (FeKe) 5.14 energy on the left axis and mean EKE (black) on the right axis for a negative IOD year. Note the different units for both axes. (b) Percentage distribution of KmKe (red), PeKe (green) and FeKe (blue) from the sum of KmKe, PeKe and FeKe. Note that the sum of the energy budget terms does not necessarily equate to the total mean EKE in (a)...... 168

5.15 Energy budget analysis around the Maldives during p-IOD, neu-IOD and n-IOD years where (a)-(c) shows the annual mean EKE and the spatial distribution of (d) – (f) KmKe (barotropic) (g)-(i) PeKe (baroclinic) and (j-l) FeKe (wind)...... 171

5.16 (a) Time series of barotropic (KmKe), baroclinic (PeKe) and wind (FeKe) energy on the left axis and mean EKE (black) on the right axis for a negative IOD year. Note the different units for both axes. Also, KmKe and PeKe were scaled by 10^2 for plotting clarity. (b) Percentage distribution of KmKe (red), PeKe (green) and FeKe (blue) from the sum of KmKe, PeKe and FeKe. Note that the sum of the energy budget terms does not necessarily equate to the total mean EKE in (a)...... 172

5.17 (a) Time series of barotropic (KmKe), baroclinic (PeKe) and wind (FeKe) energy on the left axis and mean EKE (black) on the right axis for a negative IOD year. Note the different units for both axes. (b) Percentage distribution of KmKe (red), PeKe (green) and FeKe (blue) from the sum of KmKe, PeKe and FeKe. Note that the sum of the energy budget terms does not necessarily equate to the total mean EKE in (a)...... 173

5.18 (a) Time series of barotropic (KmKe), baroclinic (PeKe) and wind (FeKe) energy on the left axis and mean EKE (black) on the right axis for a negative IOD year. Note the different units for both axes. (b) Percentage distribution of KmKe (red), PeKe (green) and FeKe (blue) from the sum of KmKe, PeKe and FeKe. Note that the sum of the energy budget terms does not necessarily equate to the total mean EKE in (a)...... 174

xiii List of Tables

3.1 Vertical setup for the model ...... 57 3.2 Experiment name and period used from HYCOM for boundary 59 conditions...... 3.3 Observational data sets used for model validation ...... 63 3.4 Summary of model validation statistics for sea surface height obtained 66 from tide gauges (UHSLC)...... 3.5 Model skill statistics for SST and current surface velocities using 67 climatology from MODIS and OSCAR...... 3.6 Summary of key findings in Chapter 3 ...... 96 4.1 Summary of the performed numerical experiments, methods and 108 objectives ...... 4.2 Summary of key findings from each experimental run for both monsoon 127 periods......

xiv Acknowledgements

One of my favourite things to read in a thesis has always been the Acknowledgements page because it reminds me of the saying about how it takes a village to raise a child, or in this case, a young oceanographer. My village is special because of the all the people who have taken the time to mentor, teach me and shower me with unconditional love and support.

To my supervisor, Chari Pattiaratchi, thank you for supporting me through thick and thin. I still remember our first conversation we had where we talked about the Einstein’s Tea leaf paradox, the first of many interesting conversations about science and even life. More importantly, you treated me like a human being after all my accident(s) and you filled in a void that my family had long abandoned. You have taught me so much about being an oceanographer but you also gave me the best example about the kind of person I want to be.

To my co-supervisor Sarath Wijeratne, who always nags me about using “we” instead of “I” in my writing and which I probably still do. I don’t know how I could have done all this modelling without you. Even though I still get rather mortified when you tell others how I built 14 grids to get it right for the model, I’m grateful to have had this experience. Thanks to your patience, I went from someone with no modelling experience to someone who don’t blow things up...immediately. I’ve enjoyed learning from you but most of all, I’ve enjoyed working with you.

To Liah Coggins, my PhD would have been very different without your critical eye and constructive feedback. Thank you for taking the time to mentor me and having matcha- coffee days and always making me want to be the best version of myself.

To Nick D’Adamo, who has known me since I was an Honours student and paved the way for me to meet Chari and join the Indian Ocean community. Your enthusiasm and encouragement has always been a big motivation for my approach in working within this community.

To Anas Ghadouani, thank you for always being such a force of positivity and convincing me in the first week when I moved to Perth that I had not been scammed into moving across the country. You are a great example of how to stay passionate about research and how to impart that through teaching.

To Mirjam van der Mheen, Mauricio van der berg, Victoria Camilleri-Asch and Jonathon Mitchell who were critical in the last few weeks of my PhD in ensuring that I lived off more than coffee and sliced bread and keeping me sane. You guys are the best neighbours and I’m so grateful to have you constantly checking in and making sure that my cat and I were not dead amidst a pile of laundry…

To Lauren Peel, my sea-going roommate and all-round cheerleader in life. Thank you for always cheering me up with cool science facts, believing in me and never judging my ugly crying/sea - sick face. xv To my research group colleagues, both past and present, Jess Kolbulz, Moritz (Mo-Bro) Wandres, (Shark) Sammy Andrzejaczek, Philippa Wilson, Tanziha Majahabin, Chen Miaoju, Hadi Bahmanpour, Leigh MacPherson and Hannah Calich - Thank you for the laughter, the snacks and the camaraderie. But most of all, thank you for letting me be a part of this little office family.

To Ivica Janekovic, Simone Cosioli, Yasha Hetzel, Paul Thomson and Mun Woo, thank you for taking the time to always share your knowledge on how to improve my research and for all the encouragement (morning tea). Thanks to Ivica for constructive discussions about ROMS and always asking me if I know things when my facial expression says otherwise…haha. Special shoutout to Paul who gave me my sea legs and taught me all things chlorophyll.

To my friends – Emily Lester, Charlotte Birkmanis, Victorien Paumard, Antoine Dillinger, Camille Grimaldi, Choi Eunjoo, Chloe Power, Maria Kuznetsova, Maxim Khudyakov, Shawn and Raeanne Lee, Rukaiya Malik, Olivia Eisenbach, Madeline Green, Thea Waters, Amy Foulkes, who have just been phenomenal human beings since I moved to Australia 9 years ago. Special mention to my Southeast Asian (SEA) academic sister, Rabitah Daud who had the joy of suffering through ROMS with me and screaming at our computers.

To Agi Gedeon, Andrew Middleditch, Paul Lethaby and Christine Pequignet who were there at the beginning of my candidature and made me feel welcome to Perth and always checking in with lots of handy advice. Agi, I hope I finally get round to making laksa after I submit this thesis.

To my Australia family, the Post and Duffy family, who made the Blue Mountains a second home and looked after me when I was at my lowest.

To my friends in Singapore who have given me nothing but support and love over the past decade even after I moved to Australia to pursue my dreams and always make it feel like I have never left when I visit. To Fei Say and Yen Chia, who always seemed to have this uncanny ability to call or visit when I was feeling low, I can’t believe that you’ve stuck with me for 17 years since I told you I was going to be a marine scientist even though everyone else thought I was nuts. Thank you for fiercely believing in me and always keeping me grounded.

To my parents who have tried to support me in the best way they knew how. I will always love and miss you.

To my brother, Justin, thank you for all your love and support through the years, I hope you find your dream soon.

To my partner, Jérémie Giraud, you have been my rock in every storm and I am excited to start the next chapter of our life together. Thank you for quite simply, Everything.

xvi Funding

This research was supported by an Australian Government Research Training Program (RTP) Scholarship. I would also like to thank the following organizations for supporting my research.

Research Training Program & Australian Postgraduate Award AUD $25,849/annum Australian Government Department of Education and Training

UWA Top-Up Scholarship AUD $3,151/annum The University of Western Australia

Ad-Hoc Scholarship AUD $30,000 Professor Chari Pattiaratchi (UWA)

Chris Lawlor Scholarship AUD $7,500 Society for Underwater Technology

Early Career Scientist Grant AUD $2,000 International Association for the Physical Sciences of the

Oceans

Early Career Scientist Travel Grant CAD $1,100 International Council for Exploration of the Sea &

North Pacific Marine Science Organization

Postgraduate Conference Travel Grant AUD $900 UWA Postgraduate Students Association

Conference Travel Grant AUD $500 UWA Graduate Research School

xvii Authorship declaration

This thesis contains work that has been prepared for publication. Note that due to the completeness of each chapter a small amount of repetition, particularly in the description of the study site and methodology, was unavoidable.

Details of the work: Su, D.L.W., Wijeratne, E.M.S., Pattiaratchi, C.B. Numerical investigation of monsoonal influence on the Island Mass Effect around Maldives and Sri Lanka. To be submitted to Geoscientific Model Development Location in thesis: Chapter 3 Student contribution to work: > 80%

Details of the work: Su, D.L.W., Wijeratne, E.M.S., Pattiaratchi, C.B. Ocean connectivity pathways between Maldives and Sri Lanka. To be submitted to Biogeosciences. Location in thesis: Chapter 4 Student contribution to work: > 80%

Details of the work: Su, D.L.W., Wijeratne, E.M.S and Pattiaratchi, C.B., Influence of the Indian Ocean Dipole Mode on island wake development off Maldives and Sri Lanka. To be submitted to Journal of Geophysical Research (Oceans). Location in thesis: Chapter 5 Student contribution to work: > 80%

Student Signature

Date: 8th January

We certify that the student statements regarding their contribution to each of the works listed above are correct

Coordinating Supervisor Co-Supervisor

Date:+BOVBSZ E.M.S. Wijeratne

xviii

Chapter 1 Introduction

1.1 Overview

The formulation for the Island Mass Effect (IME) hypothesis first began with

Darwin’s Paradox in 1836 when Charles Darwin was on the (Cocos) Keeling

Islands in the eastern Indian Ocean. Darwin observed the nearshore biological productivity around the remote islands in an otherwise oligotrophic ocean and in his words, “Stumbling across the ecosystem of a coral reef in the middle of an ocean was like encountering a swarming oasis in the middle of a desert.”. 120 years after Darwin’s observations, Doty and Oguri (1956) attempted to the test the

IME hypothesis around the Hawaiian Islands, i.e. that oceanic primary productivity is enhanced upon closer proximity to island-reef ecosystems. This was the first comparative study for the IME and they found that the planktonic biomass around the island of Oahu increased by at least two orders of magnitude in the nearshore regions compared to offshore waters (Doty and Oguri,1956).

Since then, most of our current knowledge for the IME has been from case studies across every ocean basin, including but not limited to Madeira Island (Caldeira et al., 2002a) and the Canary Islands in the (Barton et al., 2004;

1 Chapter 1. Introduction

Basterretxea et al., 2002); the Aldabra and Cosmoledo atolls in the Indian Ocean

(Heywood et al., 1990); the Galapagos and Marquesas Islands in the Pacific Ocean

(Martinez and Maamaatuaiahutapu, 2004); and, even the sub-Antarctic Prince

Edward Island archipelago (Boden, 1988). Such was the prevalence of the IME across these geographically diverse locations that it raised the question about whether the IME was a ubiquitous feature of all island-atoll systems. This was later partly substantiated by Gove et al. (2016) who conducted a basin-wide investigation across the Pacific and found that the IME was present amongst 91% of the surveyed island-atoll-reef systems. The study also revealed that the variation in IME intensity could be explained by differences in the bathymetry and geomorphology of the island-atoll-reef systems since the resulting interaction with oceanic currents would drive vertical transport of sub-surface nutrient waters into the photic zone and enhance primary productivity (Gove et al., 2016). Due to the IME’s ecological significance in supporting higher trophic levels, an understanding of the physical mechanisms behind the IME is crucial in a changing ocean climate.

In the northern Indian Ocean, the bi-annual reversing monsoon winds and ocean currents (Figure 1.1) add an additional layer of complexity in studying the IME since the ocean currents are key to structuring the IME. The Indian Ocean is also influenced by two major climate modes, the El Niño Southern Oscillation (ENSO)

2

and the Indian Ocean Dipole (IOD) Mode, both of which affect monsoonal rainfall and wind strength in the Indian Ocean (Pui et al., 2012). Under a climate change scenario, the periodicity and intensity of these forcing events might change, which will then feedback into the local climate through wind variability and mesoscale dynamics (Lu and Ren, 2016a; Lu and Ren, 2016b). Since the IME is intrinsic to fisheries and ecosystem productivity, a detailed investigation into the IME’s generation mechanisms and response to large scale forcing events like the IOD would have direct societal impacts in the Indian Ocean that supports about a third of the world’s population (Lu et al., 2017).

Figure 1.1 10-year averaged wind speed and surface currents in the northern Indian Ocean obtained from ECMWF and HYCOM respectively. (a), (c) Northeast monsoon winds and currents; (b), (d) Southwest monsoon winds and currents. SMC – South Monsoon Current; NMC – North Monsoon Current. The magnitude of the current speeds is represented by the coloured arrows.

3 Chapter 1. Introduction

IME’s are mainly generated through the interaction between island topography

(both surface and sub-surface) and the prevailing ocean currents that generate complex three-dimensional circulation at different temporal and spatial scales to enhance the local primary productivity. This could be due to enhanced vertical mixing that may act as a ‘stirring’ effect and/or generation of upwelling of nutrient rich water to the photic zone through island wake effects. Thus, in the study of IMEs, an understanding of flow topography interactions is a fundamental requirement and is the subject of this thesis. The study region lies within the northern Indian Ocean encompassing the Maldives archipelago and Sri Lanka

(Figure 1.2).

Throughout the literature, the IME has been characterized by regions of high chlorophyll concentrations and cooler sea surface temperatures (SST) due to the upwelled deep water (Palacios, 2002; Caldeira et al., 2002b; Messié and Radenac,

2006; Hasegawa et al., 2009; Tchamabi et al. 2017). In the Indian Ocean, evidence of the IME has been supported by remote sensing studies accompanied by few in- situ measurements (Sasamal, 2006; de Vos et al., 2013) yet little is known about its variability on seasonal and inter-annual time scales or its generation mechanisms (Sasamal, 2006 ; de Vos et al., 2013). Following on from the work of

Gove et al. (2016) who highlighted the influence of topography on IME intensity, the Maldives and Sri Lanka also provide an opportunity to understand the

4

sensitivity of these processes due to their distinct bathymetric differences (Figure

1.2). Thus, the overarching theme of this thesis is to understand the temporal

(seasonal and inter-annual) and spatial variability of the physical mechanisms behind the IME around the Maldives and Sri Lanka. In the northern Indian

Ocean, the seasonal variability in the system is mainly due to the monsoon forcing. The inter-annual variability is introduced through the effects of IOD and

ENSO events.

Almost 200 years since Darwin’s first foray into the Indian Ocean, this thesis will attempt to reconcile the physics behind the biological paradox of these “ocean oases” in the northern Indian Ocean.

5 Chapter 1. Introduction

Figure 1.2 (a) Global map with study region identified in red (b) Bathymetry of study region in the northern Indian Ocean.

1.2 Objectives and Approach

Throughout the thesis, numerical modelling techniques will be used to address the following objectives:

1. Examine the influence of the monsoon currents on the variability of the IME around Sri Lanka and the Maldives (Chapter 3).

2. Define the role of monsoon currents in the connectivity between the Maldives and Sri Lanka (Chapter 4) 6

3. Determine the influence of the Indian Ocean Dipole and El Nino Southern Oscillation (ENSO) events on island wake circulation and upwelling processes (Chapter 5).

The seasonal in the northern Indian Ocean is governed by two monsoon seasons that result from the differential heating and cooling of the landmass to the north of the Indian subcontinent. The Northeast Monsoon

(NEM) occurs between December - February followed by the Southwest Monsoon

(SWM) that occurs between June to August along with two inter-monsoon periods from March to May and September to November (Figure 1.1). Between each monsoon period, there is a seasonal reversal in the open ocean currents that flow between the Arabian Sea and Bay of Bengal via Sri Lanka (Figure 1.1,

Shankar 2002). The Southwest Monsoon Current (SMC) flows from west to east during the SWM with a reversal of the currents, the westward flowing Northeast

Monsoon Current (NMC) occurring during the NEM (Figure 1.1). The reversal of the monsoon current system influences the flow topography interaction and subsequent IME development around the Maldives archipelago and Sri Lanka.

Remote sensing work by Sasamal (2006) revealed that the regions of higher chlorophyll concentrations shift from west to east between the monsoons, corresponding to the reversing current patterns. These seasonal changes are addressed in Objectives 1 and 2.

7 Chapter 1. Introduction

IOD events are characterized by anomalous wind forcing that result in differential

SST anomalies on either side of the Indian Ocean basin, depending on the phase of the climate mode. IOD events are known to have a significant influence on the

Indian Ocean equatorial zonal currents at inter-annual time scales. This can potentially influence island wake development and upwelling processes, two key

IME mechanisms, and is addressed in Objective 3.

1.3 Thesis Structure

The introductory and background remarks of this chapter are followed in Chapter

2 with an overview of the IME in terms of concept and the physical setting of the

Indian Ocean. The overall theme and structure of the thesis chapters is best described by Figure 1.3 which summarizes the timescales and physical components of the IME addressed in each chapter.

Chapter 3 describes the three-dimensional high resolution model setup using the

Regional Ocean Modelling System (ROMS) whose outputs would be used in

Chapters 4 and 5. Information about the model domain and topography, surface forcing and boundary conditions are described in detail and further evaluation of the model skill is carried out using available observational data. A comparison of

IME physical processes around the Maldives and Sri Lanka is also carried out for daily and seasonal timescales.

8

Chapter 4 examines the role of the monsoon currents as transport pathways for the IME between Maldives and Sri Lanka. The chapter aims to answer whether input for the IME is sourced locally or advected from further origins. This is addressed through integrating the model outputs from Chapter 3 with a

Lagrangian particle tracking model to quantify the contribution of local mixing processes versus the monsoon currents to IME input.

Chapter 5 focuses on several locations around the Maldives and Sri Lanka where the IME occurs to compare its three-dimensional flow structure across different phases of the IOD. The locations of the IME correspond to regions of high eddy activity and island wakes; thus, the structure of the IME is described in similar context. Here, I will show that the influence of the IOD on IME structure can be described in terms of its eddy energy budget and how it impacts the formation of prominent recirculation features like the Sri Lanka Dome.

Lastly, Chapter 6 provides a synthesis of the main findings and conclusions in each of the thesis chapters and offers a discussion of potential avenues for future work. A point to note would be that Chapters 3 to 5 have been prepared for submission to peer-reviewed journals, hence there would be a certain degree of repetition in the introduction and methods sections for the sake of completeness.

9 Chapter 1. Introduction

Due to the structure of this thesis, there are overlapping themes across the chapters. For instance, a common theme of Chapters 3 and 4 is the monsoonal variability of the IME physical processes with a focus on wind and current forcing.

Chapter 5 examines similar physical forcing and processes but on interannual time scales (Figure 1.3).

10

Figure 1.3 General overview of the proposed thesis chapters and schematic of the temporal scales and forcing addressed in each chapter.

11 Chapter 1. Introduction

This page has been intentionally left blank.

12

“I don’t know anything but I do know that everything is interesting

if you go into it deeply enough”

- Richard Feynman

13

Chapter 2 Background

2.1 General Theory

2.1.1 The Island Mass Effect

The Island Mass Effect (IME) refers to the enhanced primary productivity that occurs around oceanic islands in comparison to the surrounding waters (Doty and

Oguri, 1956) due to flow topography interaction. This sets up the IME as an ecosystem generator since the enhanced productivity can influence food-web dynamics by supporting higher trophic levels and migration patterns. Often, IME locations coincide with major fisheries and aggregations of marine megafauna

(Cowen and Castro, 1994; Anderson et al., 2011; Gove et al., 2016). Through remote sensing technologies, the presence of the IME around islands is often clearly démarcated by the chlorophyll-a pigments present in the phytoplankton blooms (Hasegawa et al., 2009).

Several mechanisms may create an IME such as strong tidal mixing around the island, which enhances vertical mixing locally (Neill and Elliott, 2004); uni or bi- directional flow past the island that generates a wake often extending downstream

(Chang et al., 2013); freshwater runoff or local wind shear stirring the surface

13 Chapter 2. Background layer sufficiently to cause local upwelling (Gove et al., 2016) (Figure 2.1).

However, for the scope of this thesis, the IME mechanism of interest would be focused on -topography interactions. This IME mechanism can be further subdivided into three processes as outlined by Hasegawa et al. (2009) and is as follows 1) passive advection from the deep via vertical mixing 2) the shedding of cyclonic eddies or stable island wakes that would upwell deep nitrate-rich water into the surface euphotic layer, stimulating primary productivity in the form of a phytoplankton bloom 3) the development of a weaker bloom in the immediate lee of the island, associated with nutrient injection and longer residence time of upwelled water (Hasegawa et al., 2009). Understanding these processes are crucial in a changing climate as they may offer insight on whether the efficiency of the system can continue under different forcing (e.g. interannual circulation variability due to the ENSO and IOD).

Globally, the IME has been observed in several case studies; Caldeira et al. (2002) reported the enrichment of planktonic biomass off the Madeira Island in the north-eastern part of the Atlantic Ocean. Intensification of oceanic production around islands has also been reported off Barbados (Cowen and Castro, 1994), the Sombrero and St. Croix Islands (Corredor et al. 1984), the Canary Islands

(Basterretxea et al., 2002; Barton et al., 2004) and the Aldabra and Cosmoledo atolls (Heywood et al., 1990; Barton and Simpson, 1990). In general, the IME can

14

be identified by three parameters, areas of cool SST, high chlorophyll-a concentrations and regions of high vorticity due to the flow interaction (Caldeira and Sangrà, 2012).

Figure 2.1 Concept diagram of several mechanisms that may generate the IME (from Gove et al. 2016). These include current-topography interactions that can drive vertical transport of water masses via upwelling, downstream mixing and eddies, and internal waves; island-associated inputs, such as submarine groundwater discharge and outflow from rivers.

2.1.2 Island Wakes

It is well established that the nature and magnitude of the biological response induced by island flows is strongly dependent on residence times for nutrients retained in the vicinity of the island (Coutis and Middleton, 2002). The flow separation around the island typically results in a wake downstream and 15 Chapter 2. Background depending on the wake’s stability may result in vorticity shedding (Pattiaratchi et al., 1987). There are two distinct type of island wakes – shallow water wakes and deep water wakes. The oceanographically important distinction between shallow and deep water wakes, is dependent on the location of the dominant boundary stress, i.e. whether it is associated with the nearshore bottom or the lateral side of the island (Tomczak, 1988). In the shallow water case, bottom drag is the primary source of vorticity generation (Wolanski et al., 1984; Furukawa and Wolanski, 1998 ;Alaee et al., 2004; Neill and Elliott, 2004). Furthermore, lateral vorticity generated by bottom stress can be tilted into vertical vorticity of significance in the wake flow (Dong and McWilliams, 2007). In the contrasting deep water case, the influence of bottom drag and vortex tilting are not important

(Dong et al., 2007).

Shallow wakes occur when considering islands in shallow shelves or estuaries where nearshore bottom friction dominates (Wolanski et al., 1984; (Pattiaratchi, et al., 1987; Alaee et al., 2004; Neill and Elliott, 2004). In a shallow water wake, the bottom-controlled upwelling has been likened to Einstein’s tea leaf paradox about how secondary circulation causes tea leaves to converge to the centre of the tea-cup (Blaise et al., 2007). In a perfect cyclostrophic balance with low water level at the centre, the radially outward centrifugal acceleration should balance the radially inward pressure gradient. In this bottom-up control mode, the bottom

16

friction reduces the centrifugal acceleration, thus allowing the bottom water to move radially inward and upwell at the centre (Wolanski et al., 1984).

Deep water wakes occur leeward around tall islands surrounded by a deep bathymetry where bottom influence can be neglected and topographic and wind forcing are the primary sources of vorticity generation (Dong et al., 2007). Here, bottom friction is unimportant and upwelling processes may be controlled from the top. Under the top down control scenario, Coutis and Middleton (2002) observed that upwelling and downwelling tended to occur in areas of surface flow divergence and convergence respectively. The consequent loss of water or surface divergence in the lee induces upwelling. One such example would be in the

Hawaiian Islands where the blocking effect of the Hawaiian Islands bathymetry on the oceanic flow is similar to those of the islands’ topography on the wind

(Kersalé et al., 2011; Jia et al., 2011). This results in eddy generation because of

Ekman pumping induced by wind shear in the island’s wake. Similar dynamics are observed around Gran Canaria (Jiménez et al., 2008) which is influenced by both the trade winds and the . The significance of wind forcing alone has also been observed in the Philippines where the absence of a background current through the archipelago isolates the role of the wind (Pullen et al., 2008).

During the boreal winter, monsoon surge winds trigger eddy formation and propagation in the lee of the Philippines (Pullen et al., 2008).

17 Chapter 2. Background

2.2 The Northern Indian Ocean

2.2.1 The Monsoons

The Indian Ocean (IO) is the third largest ocean basin in the world, yet its circulation is the least well understood of the open ocean basins due to being poorly sampled in both space and time. (Hood et al., 2015). In contrast to the

Atlantic and Pacific oceans, the IO’s northern boundary is landlocked by the

Asian continent. The differential heating of the Asian continent results in a strong land-sea contrast, which drives the strongest monsoon system on Earth (Schott et al., 2009). The monsoon winds, north of 10°S are unique since they reverse twice a year which have implications for the seasonal circulation variability.

During the boreal winter, which is also known as the Northeast Monsoon (NEM) period from December to February, the rapid cooling of the Asian continent creates a sharp north-south pressure gradient between the IO and the continental landmass (Tomczak and Godfrey, 1994). The north-easterly cold winds from the

Himalayas are blown across the Northern Indian Ocean and converges with the southeast trade winds at about 10°S to form the Intertropical Convergence Zone

(ITCZ) (Figure 2.2a) (Hastenrath and Greischar, 1991).

In contrast, during the boreal summer known as the Southwest Monsoon (SWM) that occurs between June to August, the Asian continent heats up, resulting in a

18

low-pressure system over the northern and central Indian subcontinent. This creates a south-north pressure gradient between the landmass and the South

Indian Ocean and when combined with the Coriolis, causes southwesterly winds to blow from the tropical Indian Ocean and Arabian Sea towards the Indian subcontinent landmass (Figure 2.2b) (Tomczak and Godfrey, 1994). This period dominates the annual wind cycle over the Indian Ocean (Tomczak and Godfrey,

1994).

Figure 2.2 Monsoon wind stress (vectors) and depths of 20°C isotherm (colour shaded) in the Indian Ocean for (a) NEM - January, (b) SWM - June, (c) SWM- August and (d) Intermonsoon - November. Wind stress climatology (1990-1996) is based on National Centres for Environmental Prediction (NCEP) (Kalnay et al., 1996) and monthly mean isotherm (1992-2001) from Simple Ocean Data Assimilation (SODA) (from Schott et al., 2009).

19 Chapter 2. Background

Between the two major monsoon periods, the NEM and SWM, monsoon transition periods occur during April to May for the NEM and October to November for the

SWM (Yamagata et al., 2013). During the monsoon transition periods, strong westerly winds are dominant in the equatorial region and generate equatorial eastward currents knowns as Wrytki jets that transport warm water towards the eastern IO (Wyrtki, 1973).

As a result of the reversing monsoon wind forcing, the oceanic surface circulation in the Northern IO reverses seasonally (Schott and McCreary, 2001; Shankar et al.,2002). This creates two distinct seasonal circulations, with upwelling favourable circulations during the SWM and downwelling circulations during the

NEM respectively (Schott and McCreary, 2001; Shankar et al., 2002). A schematic representation of the summer and winter monsoon circulations is presented in

Figure 2.3 and identifies the major current branches that were studied in Schott et al (2001; 2009). Throughout this study, the main currents of interest are the

Southwest Monsoon Current (SMC), the Northeast Monsoon Current (NMC), the

Southern Equatorial Current (SEC) and its counterpart, the Southern Equatorial

Counter Current (SECC).

The westward flowing SEC is driven by south-easterly winds and originates from the Pacific Ocean via the Indonesian Throughflow (ITF)(Schott and McCreary,

20

2001). The net ITF inflow to the IO was ~6 Sv and its interannual variability associates with the ENSO cycle (Gordon and Fine, 1996; Gordon et al.,2003;

Potemra, 2005; Wijeratne et al., 2018). In the eastern IO (east of ∼105°E), the

SEC has been observed between 7°S and 15°S while in the west IO, the SEC has been observed between 10°S and 20°S (Quadfasel et al., 1996). Upon reaching the east coast of Madagascar, the SEC bifurcates at 17°S into the Northern and

Southern extensions of the East (Figure 2.3). The Northern extension of the , also known as the NEMC, flows into the East African Coastal Current which converges with the southward flowing western , the Somali Current (SC). This confluence between 2°N

-4°N, feeds the eastward flowing Southern Equatorial Counter Current (SECC)

(Figure 2.3). A point to note is that the SC undergoes seasonal reversal and during the SWM, the SC flows poleward but flows equatorward during the NEM (Schott and McCreary, 2001). This affects the latitudinal range of the SECC between both monsoons.

The partitioning of the northern part of the IO basin by the Indian subcontinent also has an influence on the circulation, as it limits northward heat transport and only weakly ventilates the northern IO thermocline (Schott et al., 2009). The division of the northern IO by the continent creates two sub-basins, the Arabian

Sea and the Bay of Bengal which both have distinct water mass temperature and

21 Chapter 2. Background salinity characteristics. Excess evaporation makes the upper layers in the western part, Arabian Sea, saltier than the rest of the IO, while large runoff from river draining the Indian subcontinent makes the eastern part, the Bay of Bengal significantly fresher (Hood et al., 2015). The presence of the continental landmass also means that water exchange in the Northern IO is primarily facilitated by the monsoonal currents, the NMC and the SMC (Figure 2.3). The monsoon currents extend over the entire IO basin but several parts of the currents form at different times and it is only when they are in their mature phase that they can facilitate trans-basin exchange (Shankar et al., 2002).

During the NEM, the NMC flows westward across the IO basin transporting low salinity water (~33) from the Bay of Bengal into the eastern Arabian Sea, where it is entrained around the Lakshadweep Islands before continuing along the west coast of India as the West India Coastal Current (WICC) (Figure 2.3, Figure 2.9,

Wyrtki, 1973; Schott et al., 2001; Shankar et al., 2002). The NMC peaks during

January with current strength approaching 0.2 ms-1 south of Sri Lanka and 0.3 ms-1 in the southwestern Arabian Sea.

Conversely, during the SWM, the eastward flowing SMC transports high-salinity water (~36.5) from the Arabian Sea into the Bay of Bengal. The SMC in the

Arabian Sea is a continuation of the SC and the coastal current off (Figure

22

2.3; Schott et al., 2001; Shankar et al., 2002) It flows eastward and southeastward across the Arabian Sea and around the Lakshadweep Islands, before flowing eastwards south of Sri Lanka into the Bay of Bengal (Figure 2.3; Figure 2.8;

Schott et al., 2001; Shankar et al., 2002). During June, the SMC is separated from the Sri Lanka coast by a narrow westward current (Vinayachandran and

Yamagata, 1998). East of Sri Lanka, the SMC branches into two: one branch turns north-eastward to enter the Bay of Bengal during the entire season while the other continues eastward (Vinayachandran and Yamagata, 1998). The eastern branch of the SMC eventually terminates as the SWM progresses but the northeastern branch of the SMC forms the bulk of the current that flows into the

Bay of Bengal and is the link between the Bay of Bengal with the rest of the northern IO (Vinayachandran and Yamagata, 1998). A notable feature unique to the SWM is the development of a cyclonic recirculation known as Sri Lanka Dome

(SLD) located along 85°E, and is often accompanied by cooler SST

(<25°C)(Vinayachandran and Yamagata, 1998). Overall, the SMC peaks during

July with current strength approaching 0.4m ms-1 in parts of the bay but when depth-averaged over 50m, it has been found to last between May to September

(Shankar et al., 2002) (Figure 2.3).

Ship drift data further constrained the latitudinal and vertical range of the monsoon currents with the geostrophic flow of the NMC located north of 2°N-

23 Chapter 2. Background

3°N while the geostrophic flow of the SMC was restricted to a narrower range, north of about 4°N (Schott et al., 1994). The structure of the monsoonal currents were found to be mostly confined in the top 100m and early measurements found the NMC to carry a mean transport of about 12 Sv while the SMC was estimated to have a volume transport of 8 Sv between the range of 3°N-5°N (Schott et al.,

1994).

24

Figure 2.3 Schematic representation of the large-scale circulation of the Indian Ocean during the (a) winter monsoon (Northeast Monsoon) and the (b) summer monsoon (Southwest Monsoon) from Schott et al. (2009). Currents identified are the South Equatorial Current (SEC), South Equatorial Countercurrent (SECC), Northeast and Southeast Madagascar Current (NEMC and SEMC), East African Coastal Current (EACC), Somali Current (SC), Southern Gyre (SG) and Great Whirl (GW) and associated upwelling wedges (green shades), Southwest and Northeast Monsoon Currents (SMC and NMC), South Java Current (SJC), East Gyral Current (EGC), and Leeuwin Current (LC). Subsurface currents are shown in purple. Depth contours shown are for 1000 m and 3000 m (grey). Red vectors (Me) indicate directions of meridional Ekman transports. ITF indicates Indonesian Throughflow (from Schott et al.,2009).

25 Chapter 2. Background

A defining characteristic of the SWM is the Indian Summer Monsoon Rainfall

(ISMR) which brings about 80% of the Indian continent’s annual precipitation

(Jin and Wang, 2017). A monsoon year can be classified as either a strong or weak monsoon year based on the rainfall index across India (Zheng et al., 2016).

Strong monsoon years were identified as years where the June to August (JJA)

ISMR rainfall was at least 10% higher than the rainfall climatology computed between 1951 to 2014 (Zheng et al., 2016). In contrast, weak monsoon years were considered as years where the JJA rainfall was 10% less than the seasonal climatology (Figure 2.4; Zheng et al., 2016). Understanding the interannual variability of the ISMR is crucial since it is influenced by the IO’s climate modes

(Ashok et al., 2001) and may affect the development and intensity of the IME.

Furthermore, a stronger Indian monsoon has been hypothesized to lead to a cooler

IO (Jin and Wang, 2017).

26

Figure 2.4 Evolution of rainfall during ISMR from 1951 to 2014 (from Zheng et al. 2016). Strong (red) and weak (blue) ISMRs are identified by the departure of JJAS rainfall of each year from the JJAS rainfall climatology computed over the period 1951– 2014, whose departure values are larger than +10%, smaller than 10%, and within 10% and 10% of the seasonal climatology, respectively. The dashed line denotes a value of 110% (90%) of seasonal climatology. The years of strong and weak ISMRs are denoted by the numbers over the bars.

The seasonal circulation in the IO also affects the biogeochemical variability throughout the IO basin and in turn, the primary productivity (Wiggert et al.,

2006). During the NEM, the northeasterly winds drive convective mixing in the

Arabian Sea, creating winter blooms of phytoplankton in the Central and Arabian

Sea (Figure 2.5 (A) ; Wiggert et al., 2006, 2009). In contrast, the SWM has the highest primary productivity due to the upwelling favorable wind and current circulations as evidenced from remote sensing data (Figure 2.5 (C), Wiggert et al., 2006, 2009). In the western Arabian Sea, vertically-integrated chlorophyll- a concentrations can exceed 40 mg m−2 (Hood et al., 2017). During both transition

27 Chapter 2. Background monsoon periods, the primary productivity is mostly constrained to the coastline

(Figure 2.5).

Figure 2.5 Seasonal climatology of SeaWiFS chlorophyll for (A) NEM – Jan (B) Intermonsoon 1 – Apr (C) SWM – Aug (D) Intermonsoon 2 - Oct. The climatology was created using monthly Level 3 Standard Mapped Image data from September 1997 through January 2002, which were obtained from the Goddard DAAC (http:// daac.gsfc.nasa.gov/) (from Wiggert, et al., 2006)

2.2.2 Climate Modes

The IO is influenced by two prominent climate modes of interannual variability, the El Niño Southern Oscillation (ENSO) and the Indian Ocean Dipole (IOD)

(Cai, 2005). The duration of each mode throughout the year is summarized in

Figure 2.6. ENSO is the dominant mode of natural climate variability, and is 28

characterized by recurrent warming and cooling of the eastern equatorial Pacific

(El Niño) and changes in the zonal pressure gradient over the western equatorial

Pacific (the Southern Oscillation)(Schott et al., 2009). ENSO influences the IO via atmospheric teleconnections associated with the Southern Oscillation. During

El Niño years, the IO warms up but experiences cooling during La Niña years

(Chowdary and Gnanaseelan, 2007).

Figure 2.6 Seasonality of Indian Ocean Climate Modes from (Schott et al., 2009)

The IOD is a coupled ocean-atmospheric mode in the tropical Indian Ocean

(Sreenivas et al., 2012). During positive IOD years, due to increased upwelling near the Sumatra coast, cooler sea surface temperatures (SST) are observed in the south eastern IO, while warmer SST are observed in the western IO (Figure

2.7; Saji et al., 1999). This leads to droughts over the Indonesian region and heavy rains and floods over East Africa. In contrast, during negative IOD years, the south-eastern IO experiences enhanced downwelling and increased SST while cooler SST are observed in the western IO (Figure 2.7; Saji et al., 1999). The role of equatorial forcing has a great impact on the eddy generation mechanism in the

29 Chapter 2. Background northern IO, especially in the Bay of Bengal (Sreenivas et al., 2012). For instance, since an IOD event induces anomalous remote winds in the northern IO, this affects sea level variability, which may then affect the periodicity and intensity of planetary waves, e.g. Kelvin waves (Chowdary and Gnanaseelan, 2007; Aparna et al., 2012; Sreenivas et al., 2012). Sreenivas et al. (2012) showed that the observed annual cycle of coastal Kevin waves and their associated radiated Rossby waves play a prominent role in the evolution of mesoscale eddies in the Bay of

Bengal. Similarly, Palastanga et al. (2006) have also shown a link between mesoscale eddy activity around Madagascar and the IOD and, that inter-annual changes in large-scale remote forcing by the SEC associated with the IOD could potentially affect the number and intensity of eddies in the Channel and the transport of the East Madagascar Current.

The IOD can also affect the ISMR on its own and can weaken or strengthen the influence of ENSO on the ISMR (Ashok et al., 2001). Due to the existence of positive and negative events for both climate modes, the influence on ISMR depends on the phase and amplitude of the IOD and ENSO (Ashok et al., 2001).

This can affect potentially affect circulation variability due to the change in volume transport in the ocean from the freshwater input of the ISMR. This observation, along with the frequent occurrence of intense IOD events in the last decade has prompted investigations into whether the moving correlation between

30

the IOD and the ISMR changes from decade to decade and in particular, its role in the weakening of the monsoon-ENSO correlation (Ashok et al., 2001).

Under a climate change scenario where ocean SST anomalies were increased in model sensitivity experiments, extreme positive IOD events are expected to increase due to the weakening of the westerly equatorial winds and faster warming in the west, leading to more flooding in eastern Africa and droughts and forest fires in and Australia (Ashok et al., 2001). This would result in colder and more saline waters in the eastern equatorial IO and warmer, fresher waters in the western IO (Zheng et al., 2013). Such temperature anomalies would lead to a suppression of productivity in the Arabian Sea but enhanced productivity in the eastern equatorial IO (Wiggert et al., 2009). However, this could also have an unexpected benefit for some regions of the IO, such as the west coast of India where upwelling of the subsurface anoxic waters can be detrimental for coastal resources (Vallivattathillam et al., 2017),

31 Chapter 2. Background

Figure 2.7 Schematic of the Indian Ocean Dipole Mode and the resulting changes in convection pathways and equatorial thermocline depth at (a) neutral phase (b) positive phase (c) negative phase (Bureau of Meteorology, Australia).

2.2.3 Study Region

The island of Sri Lanka and the Maldives Archipelago occupy a unique location due to their position at the crossroads of water exchange between the Arabian

32

Sea and the Bay of Bengal. This section will expand on the monsoon circulation earlier discussed in Section 2.1 with respect to Sri Lanka and the Maldives and the schematic in Figure 2.9 summarizes the circulation patterns.

As outlined earlier in Section 2.1, the westward flowing NMC transports low salinity water from the Bay of Bengal during the NEM (Figure 2.8; Schott and

McCreary, 2001). The East Indian Coastal Current (EICC) is the western boundary current of the Bay of Bengal and flows equatorward past Sri Lanka to merge with the NMC, flowing from east to west (Figure 2.8; Schott et al. 1994).

During December, the NMC is fed by the EICC and has a velocity of ~0.4 ms-1 southwest of Sri Lanka (Shankar et al., 2002). The currents then flow beyond

70°E before turning to flow around the clockwise Lakshadweep eddy and northward along the western Indian coastline before flowing into the poleward

WICC (Schott et al. 1994). At its mature phase during January to February, the

NMC reaches speeds greater than 0.4 ms-1 south of Sri Lanka (Shankar et al.,

2002).

Prior to the start of the SWM, traces of the SMC can be observed as an eastward flow in southern Bay of Bengal between 4°N to 8°N (Schott and McCreary 2001;

Shankar et al., 2002). As the SWM progresses, the development of the anti- clockwise Lakshadweep eddy along the southwest coast of India modifies the SMC

33 Chapter 2. Background current flow in the Arabian Sea. When the SMC flows along the eastern boundary of the Arabian Sea, it is augmented by the equatorward flowing WICC (Shankar et al., 2002). In the southern Arabian Sea, there is an eastward flow at 5°N between 65°E and 75°E; this flow then merges with the SMC south of Sri Lanka

(Figure 2.8; Schott et al. 1994; Schott & McCreary 2001; Shankar et al., 2002).

(Figure 2.8; Schott & McCreary 2001; de Vos et al. 2013). The SMC flows along the southern coast of Sri Lanka from west to east between the equator and Sri

Lanka (Figure 2.8; Schott & McCreary 2001; de Vos et al. 2013). After passing the coast of Sri Lanka, the currents form an anti-clockwise eddy known as the Sri

Lanka Dome (SLD) located at 83°E and 7°N (Vinayachandran and Yamagata

1998, de Vos et al. 2013). The western arm of SLD drives a southward current along the eastern coast of Sri Lanka whilst the remainder flows northward along the eastern Indian coast as the EICC (Figure 2.9; Schott et al. 1994;

Vinayachandran and Yamagata 1998). A point to note that as the SWM sets in, the northward flowing EICC decays in the north and by August, the current flow is weak north of 10°N (Durand et al., 2009).

34

Figure 2.8 Schematic of the major recirculation features in the proposed study area during the NEM (left panel) and SWM (right panel). SLD represents the recirculation feature known as the Sri Lanka Dome (from de Vos et al., 2013).

Several studies have investigated the influence of the monsoonal circulation on biological productivity in the region around Sri Lanka and Maldives

(Vinayachandran et al; 2004; de Vos et al.,2013 ; Strutton et al., 2015). A few features of importance in this region include the cold pool off the southern tip of

India (Rao et al., 2006), the SLD and significant phytoplankton blooms around the Maldives and Sri Lanka (Vinayachandran et al., 2004; Sasamal, 2006; Yapa,

2012). These regions of high chlorophyll concentrations coincide with areas of cool

SST and high salinity (Figure 2.9), fulfilling the criteria used to characterize the presence of the IME. The advection of these chlorophyll wakes also correspond to the directionality of the major monsoon currents and can be used as a tracer to identify IME locations and island wake effects in this region (Figure 2.9;

Vinayachandran, et al., 2004; Sasamal, 2007; Strutton et al., 2015).

35 Chapter 2. Background

Figure 2. 9 The seasonal climatology of remotely sensed SST and currents (left column of panels), salinity (center column) and chlorophyll (right column) for a region of the IO encompassing the RAMA (Research Moored Array for African–Asian–Australian Monsoon Analysis and Prediction) mooring at the equator, 80.5°E (magenta dot) (from Strutton et al., 2015)

2.2.4 Sri Lanka

Sri Lanka is located at 5°-9°N, 79°-81°E and its southwest coast has one of the narrowest (<20 km) and steepest continental shelves recorded globally

(Wijeyananda, 1997). This means that typical shelf processes are unable to explain the upwelling and subsequent productivity along the south coast of Sri

Lanka. Remote sensing studies have identified high chlorophyll-a concentrations

36

along the south point of India as well as the south coast of Sri Lanka (Figure

2.10; Wiggert et al., 2006; Yapa, 2009, 2012; de Vos et al., 2013). In addition, chlorophyll-a is also seen along the east coast of Sri Lanka at the peak of the

SWM, indicative of the presence of the SLD. However, the two mechanisms behind the south coast upwelling region and the SLD are quite different.

Numerical model simulations done by de Vos et al (2013) found that upwelling along the south coast of Sri Lanka was a product of flow convergence and that the upwelling centre was dependent on the wind driven flow along the island coastline. This was used to explain the relatively high chlorophyll concentrations along the south coast and why there was a year-round presence of a population of blue whales throughout the year, despite it not being the SWM period. Instead, the SWM wind forcing enhances the existing background mechanism of secondary circulation along the coastline, to increase primary productivity (de Vos et al.

2013). On the other hand, the SLD develops as a response to the local wind field

(Vinayachandran and Yamagata, 1998) and is analogous to a headland eddy due to the interaction between the SMC and Sri Lanka (de Vos et al. 2013). The

Ekman pumping within the dome is responsible for upwelling cooler nutrient rich water to the surface. The SLD peaks in July but decays in September upon the arrival of a Rossby wave that was due to a reflection of the Wyrtki jet along the eastern IO boundary (Vinayachandran and Yamagata, 1998). However, for most of the modelling studies conducted in this area, the models that were used only

37 Chapter 2. Background examined the surface circulation (de Vos et al. 2013) or the coarse resolution was unable to resolve the fine scale three-dimensional vertical mixing processes or did not account for salinity variability (Vinayachandran and Yamagata, 1998).

Figure 2. 10 Monthly composite images of chlorophyll-a derived from SeaWiFS during 1998-2002 showing the evolution of the chlorophyll bloom around the southern tip of India and Sri Lanka during the summer monsoon.(Hood et al., 2015)

2.2.5 The Maldives archipelago

The Maldives archipelago is located to the south of India 73°E and to the west of

Sri Lanka and forms a barrier to the NMC (SMC) during the NEM (SWM)

(Sasamal, 2007). The channels through the Maldives have varying widths that generate turbulent flow upon interaction with the monsoonal and equatorial currents, creating wake eddies that develop offshore from the atolls (Sasamal,

2007). Current speeds through the outer channels are typically greater than 1.5 ms-1 while current speeds within the deeper channels are in the range of ~0.5 –

0.8 ms-1 (CUSP, 2011). Sasamal (2006) has shown using MODIS derived

38

chlorophyll-a data that the IME on the western coast responds to the NMC and the decay of the IME phenomena was observed along with the reduction in the westward flow from the NMC (Figure 2.11). However, although the westward flow can be seen through the altimeter data, it does not always correspond to a significant chlorophyll plume e.g. March 2004 (Sasamal, 2007). This leads to the question as to why the presence of these eddies do not always coincide with the occurrence of a chlorophyll plume. This could be related to the depth and extent of which these wake eddies produce upwelling. In addition, there has been a lack of characterization for this counterpart feature along the eastern coast during the

SWM due to high cloud cover present in the MODIS imagery (Sasamal, 2007).

The ecological implications of this “monsoonal swapping” of the IME has already been observed in the Maldives where manta rays, Mobula alfredi, were observed to aggregate along the downstream sides of the atolls and would switch sides bi- annually as the monsoon currents changed (Anderson et al., 2011). This reiterates the importance of the IME in supporting higher trophic levels in the open ocean.

39 Chapter 2. Background

Figure 2. 11 MODIS derived chlorophyll plume (left panel) around the Maldives and SST (right panel) during the NEM (from Sasamal, 2006)

2.3 Numerical Models

A review of the IO dynamics was presented in Section 2.2 and a recurring theme throughout much of the cited literature was the paucity of in-situ measurements and the reliance on remote sensing and modelling techniques to understand these ocean processes. Despite recent progress and interest over the last 50 years, there is still an insufficient amount of observational data available for this region. Thus, numerical ocean models are still relied on to provide insight into seasonal and inter-annual dynamics and can be used to resolve sub-mesoscale features like island wakes. Furthermore, these models can be used to test hypotheses about the contribution from different forcing, such as wind and topographical sensitivity.

40

2.3.1 The Regional Ocean Modelling System

The model of choice throughout the thesis is the Regional Ocean Modelling

System (ROMS) and has been previously used to study the surface circulation around Sri Lanka (de Vos et al., 2013) as well as island wake circulation in the southern California Bight (Dong and McWilliams, 2007), the Hawaiian Islands

(Kersalé et al., 2011), the Marquesas Archipelago (Raapoto et al., 2018),

Fernando de Noronha Island (Brazil)(Tchamabi et al., 2017), and coastal upwelling in New Caledonia (Marchesiello et al., 2010).

ROMS is a split-explicit, free surface, terrain-following coordinate model that solves barotropic and baroclinic momentum equations separately (Shchepetkin and McWilliams, 2003, 2005). By separating the solutions to the barotropic and baroclinic modes of the momentum equations, the split-explicit approach reduces the number of time-stepping operations required, thus improving computational efficiency (Shchepetkin and McWilliams 2005). The associated gain in computational efficiency allows for simulations to be performed at a higher resolution, making it ideal for studies focused on mesoscale and sub-mesoscale features.

ROMS solves the incompressible primitive equations under the Boussinesq and hydrostatic approximations. The hydrostatic approximation assumes that the

41 Chapter 2. Background buoyancy force is balanced by the vertical pressure gradient. The primitive equations of motion (Eq. 2.1, 2.2) under the Boussinesq and hydrostatic approximations are discretized in an orthogonal, curvilinear coordinate system in the horizontal and a stretched, terrain-following coordinate system in the vertical direction (Song and Haidvogel, 1994; Shchepetkin and McWilliams, 2003, 2005).

ROMS is also coupled with the non-linear equation of state and advective/diffusion schemes for potential temperature and salinity. For further detail on the below equations, please refer to Cushman-Roisin and Beckers (2011).

• Horizontal momentum equations

�� �� + �⃗ . �� − �� = − + � + � (Eq 2. 1) �� ��

�� �� + �⃗ . �� − �� = − + � + � (Eq 2. 2) �� ��

• Advective-diffusive equations for potential temperature, T (°C) and salinity,

S (psu)

�� (Eq 2. 3) + �⃗ . �� = � + � ��

�� + �⃗ . �� = � + � (Eq 2. 4) ��

42

• Hydrostatic and mass balance

�� �� = − (Eq 2. 5) �� �

• Continuity equation for an incompressible fluid

�� �� �� + + = 0 (Eq 2. 6) �� �� ��

• Equation of state for seawater, where P is total pressure (N m-2)

� = � (�, �, �) (Eq 2. 7)

For the above equations:

• �, �, � are the components of the velocity,�⃗ (ms-1) in �, �, � respectively.

• �, �, �: horizontal and vertical coordinates

• �: Coriolis parameter (s-1)

• �: dynamic pressure (m2 s-2)

• g: Acceleration due to gravity (m s-2)

• �, �, �, � : forcing terms

• �, �: lateral momentum dissipation terms

• �, �: diffusivity for temperature, T and salinity, S

-3 -3 • �: reference seawater density (1027 kg m ), �: seawater density (kg m )

43 Chapter 2. Background

One of the major challenges in modelling is resolving the vertical resolution in the surface ocean mixed layer versus having a deeper model depth (Griffies et al.,

2000). In ROMS, this has been addressed using vertical stretching functions that increase the resolution in the surface layers (Song and Haidvogel, 1994;

Shchepetkin and McWilliams, 2003). This is important since an accurate representation of the surface mixed layer proccesses is relevant to our study due to its focus on deep island wakes.

2.4 Summary

The chapter introduced the general theory and concept behind the IME hypothesis with a focus on how island wake circulation can develop the IME. The chapter also provided a detailed summary of the monsoon dynamics present in the IO as well as an introduction to its two main climate modes, ENSO and the

IOD. In addition, the chapter also reviewed previous work done in the study region and ongoing research challenges associated with the region. Lastly, an overview of the Regional Ocean Modelling System was presented, and its advantages and relevance for the project was also addressed.

44

This page has been intentionally left blank.

45

“What the four seasons of the year mean to the European, the one season of the monsoon means to the Indian.

It is preceded by desolation; it brings with it hopes of spring; it has the fullness of summer and the fulfilment of autumn all in one”

- Khushwant Singh

46

Chapter 3 Monsoonal variability of the Island Mass Effect around Sri Lanka and The Maldives

3.1 Summary

Regional circulation in the Northern Indian Ocean (NIO) is unique since it develops in response to the bi-annual reversing monsoonal winds. The monsoon currents mirror this reversal, which have implications for the development of the

Island Mass Effect (IME) in the NIO. The IME is characterized by areas of high chlorophyll concentrations around islands and was identified by remote sensing to be located around the Maldives and Sri Lanka in the NIO. The IME around the Maldives was observed to swap between the monsoons to downstream of the incoming monsoonal current whilst a recirculation feature known as the Sri Lanka

Dome developed off the east coast of Sri Lanka during the Southwest Monsoon.

To understand the physical mechanisms underlying this monsoonal variability of the IME, a numerical model based on the Regional Ocean Modelling System

Framework was implemented and validated. The model was able to realistically simulate the known regional circulation and was used to investigate the structure of the IME in terms of its temperature and velocity. Results revealed that

47 Chapter 3. Monsoonal variability of the Island Mass Effect around Sri Lanka and the Maldives downwelling processes were prevalent along the Maldives for both monsoon periods but was applicable only to latitudes above 4°N since that was the extent of the monsoon current influence. For the Maldives, atolls located south of 4°N, were influenced by the equatorial currents. Around Sri Lanka, upwelling processes were responsible for the IME during the SWM but with strong downwelling during the NEM. In addition, there were also regional differences in intraseasonal variability for these processes. Overall, the strength of the IME processes was closely tied to the monsoon current intensity and was found to reach its peak when the monsoon currents were at the maximum.

3.2 Introduction

During the Northeast Monsoon (NEM), satellite ocean colour imagery revealed surface chlorophyll concentrations downstream of the Maldives west coast that were relatively higher compared to the surrounding waters (Figure 3.1). A ‘tail’ of higher surface chlorophyll concentrations was also observed along the coastline of Sri Lanka (Figure 3.1). In contrast, a similar signature of high chlorophyll concentrations was observed separately for the east coast of the Maldives (Figure

3.1f) and Sri Lanka during the Southwest Monsoon (SWM). The surface chlorophyll concentrations were also observed to be advected into a recirculation feature along the east coast of Sri Lanka (Figure 3.1e). These areas of high

48

chlorophyll concentrations are identified as areas where the Island Mass Effect

(IME) occurs, which is defined by Doty and Oguri (1956) as an enhancement of primary productivity in the surrounding waters of islands. This increase in primary productivity can be attributed to several non-exclusive mechanisms such as tidal mixing, internal waves and lee eddies formed by flow disturbance or

Ekman pumping (Gove et al., 2016). All of them involve mixing processes in the water column to bring deeper, nutrient-rich water into the photic zone, thus stimulating productivity. As such, the IME is also characterized by regions of cool

SST and high current speeds due to these mixing processes (Caldeira et al., 2002).

The IME is ecologically significant for its role in supporting higher trophic levels as well as influencing migratory patterns of marine megafauna and localization of fisheries around islands (Anderson et al., 2011; Palacios, 2002). In addition, the presence of the IME has been observed in various case studies worldwide such as around the Hawaiian Islands, Barbados (Cowen and Castro 1994), Cosmoledo atolls (Heywood et al., 1990), Madeira Island (Caldeira et al., 2002), the Great

Barrier Reef, Australia (Hamner and Hauri, 1981) and the Galapagos Archipelago

(Palacios, 2002).

49 Chapter 3. Monsoonal variability of the Island Mass Effect around Sri Lanka and the Maldives

Figure 3.1 Daily snapshots of surface chlorophyll-a concentrations around Sri Lanka and the Maldives during the Northeast Monsoon (a) – (d) and the Southwest Monsoon (e) – (h). (d) and (h) displays multiple von Karman vortex streets swirling downstream from the Huvadhoo Atoll (Figure 3.2), Maldives during the Northeast Monsoon and the Southwest Monsoon respectively. Images adapted from EOSDIS Worldview.

The seasonality of the NIO is defined by two major monsoons, namely the NEM and SWM. The key difference between the monsoons lie in the dynamics of the bi-annual monsoonal wind forcing which influences the circulation of the Indian

Ocean basin (Shankar et al., 2002). During the NEM, the winds blow from the north east towards the south west direction while the winds during the SWM

50

come from the opposite direction. This reversal is mirrored in the circulation of the Northeast Monsoonal Current (NMC) that flows westwards from the Bay of

Bengal during the NEM and the Southwest Monsoon Current (SMC) flows eastward during the SWM. Consequently, due to this difference in directionality and strength of wind forcing, the type of circulation features, and nature of the

IME, developed are unique to each monsoon period (Figure 3.1).

To date, there have been several studies by Anderson et al., (2011), de Vos et al.,

(2013) and Sasamal (2007) that have identified the IME in the Northern Indian

Ocean (NIO) via remote sensing data and numerical modelling. However, the paucity of in-situ measurements in this region as well as the effect of cloud cover during the SWM meant that there have been limitations in resolving the mechanisms underlying IME development. In addition, coarser global/regional models currently are unable to resolve the fine-scale features implicit to the IME.

In the present study, we focus on the development of the IME created from the flow disturbance around the island of Sri Lanka and the Maldives archipelago created by the reversing monsoon currents. Specifically, we refer to island wake processes that have been induced by the oceanic currents. These islands occupy a unique location at the crossroads of water exchange between the higher salinity

Arabian Sea and the lower salinity water from Bay of Bengal and are

51 Chapter 3. Monsoonal variability of the Island Mass Effect around Sri Lanka and the Maldives bathymetrically distinct from each other (Figure 3.2). The orography between the islands is also vastly different with the highest point of elevation in Sri Lanka at

2524 m above sea level while the highest area of elevation in the Maldives is less than 10m (Google Earth, 2018). This would offer an interesting point of comparison for wind wakes in the lee of Sri Lanka versus current induced wakes around the Maldives.

The aim of this chapter is to examine the influence of monsoon currents on the variability of the IME around Sri Lanka and the Maldives archipelago. This was undertaken using a high resolution three-dimensional numerical model adapted for the region to characterize the structure of the IME across seasonal time scales.

The chapter is organized as follows: Section 3.3 describes the model configuration in detail while Section 3.4 evaluates the robustness of the model against observational data; Section 3.5 examines the variability of the IME at both seasonal and intra-seasonal scales and Section 3.6 is the discussion followed by the concluding remarks.

52

Figure 3.2 Bathymetry map of (a) Maldives and (b) Sri Lanka.

3.3 Methodology 3.3.1 Model setup

The numerical model was built based on the Rutgers version of the Regional

Ocean Modelling System (ROMS) (http://www.myroms.org) with significant adaptations for the Maldives and Sri Lanka. It has been previously used to study the surface circulation around Sri Lanka (Vos et al., 2013) as well as island wake circulation in the southern California Bight (Caldeira et al., 2005 ; Dong and

McWilliams, 2007), the Hawaiian Islands (Kersalé et al., 2011), the Marquesas

53 Chapter 3. Monsoonal variability of the Island Mass Effect around Sri Lanka and the Maldives

Archipelago (Raapoto et al.,, 2018), Fernando de Noronha Island (Brazil)

(Tchamabi et al., 2017), and coastal upwelling in New Caledonia (Marchesiello et al., 2010).

ROMS is a split-explicit free surface, terrain-following vertical coordinate oceanic model and resolves the incompressible primitive equations (Cushman-Roisin and

Beckers, 2011) using the Boussinesq approximation and hydrostatic vertical momentum balance (Shchepetkin and McWilliams, 2003, 2005). These equations are discretised on an orthogonal, curvilinear, Arakawa C-grid and a stretched, terrain-following coordinate system in the vertical direction (Song and Haidvogel,

1994; Shchepetkin and McWilliams, 2003, 2005).

The model grid domain extends from 7°S - 15°N, 65°E - 88°E (Figure 3.3).

Bathymetry was obtained from the General Bathymetric Chart of the Oceans

(GEBCO) 30 arc-second gridded product (https://www.gebco.net/) and coastline data was extracted from GSHGG coastlines (ngdc.noaa.gov). Potential horizontal gradient errors were minimized by smoothing the bathymetry till the recommended slope factors for the Beckman and Haidvogel number, rx0 <0.2

(Beckmann and Haidvogel, 1993) and the Haney number, rx1 (also known as the hydrostatic inconsistency number) 5

54

numerical stability for the grid. The bathymetry was smoothed in the following steps (1) a Shapiro filter was applied twice over the original bathymetry (2) High rx0 areas >0.2 were identified (Figure 3.4) (3) Iterative pointwise smoothing were applied at these high rx0 areas. This was found to be a good balance in maintaining realism and enabling grid stability. rx1 is sensitive to the vertical parametrization in the model which is dependent on the number of vertical layers, the thermocline depth, Tcline and the surface and bottom control parameters θS and θb respectively (Table 3.1). In addition, an accurate representation of surface mixed layer processes is sensitive to the vertical resolution at the surface and therefore to the choice of stretching functions, i.e. the vertical terrain-following stretching function, VStretch and vertical terrain-following transformation function,

VTransform (Table 3.1). The model domain has a depth range of 10m-5000m with 30 vertical levels (Figure 3.5) and a summary of the parameters used in the setup is provided in Table 3.1. along with the vertical resolution across the bathymetry

(Figure 3.5). The thermocline depth was estimated from the ARGO climatology profile (Table 3.3) averaged for the entire model domain. The resulting horizontal grid resolution was approximately 3 km per cell in the 860 x 860 grid cell structure and the model domain and bathymetry is shown in Figure 3.3.

55 Chapter 3. Monsoonal variability of the Island Mass Effect around Sri Lanka and the Maldives

Figure 3.3 Model domain showing bathymetry and locations of model validation points. The observational data used for model validation are denoted in blue circles for selected tide gauge sites (Table 3.2). The locations of ARGO profile data used to compare model results are represented in yellow shaded boxes [A1-A8]. The RAMA mooring, M1 is denoted by the pink triangle and a summary of the time periods and variables used for validation is summarized in Table 3.2.

56

Figure 3.4 Gridded map for grid domain values of (a) rx0 and (b) rx1. The recommended values are rx0 <0.25 and 5

Table 3. 1 Vertical setup for the model

Parameter Value

Number of vertical layers, N 30

S-coordinate surface control parameter, θS 6

S-coordinate bottom control parameter, θb 0.3

S-coordinate surface/bottom layer width, Tcline 150

Vertical terrain-following stretching function, VStretch 2

Vertical terrain-following transformation function, VTransform 4

57 Chapter 3. Monsoonal variability of the Island Mass Effect around Sri Lanka and the Maldives

Figure 3.5 Vertical grid in the SNIO-ROMS where the top panel is the zoomed inset for the top 200m and the lower panel shows the vertical resolution over the bathymetry to a depth of 5000m. The red line indicates where the thermocline, Tc is set to 150m.

The initial and open boundary data (salinity, temperature, sea surface height and velocity fields) for the model were obtained from the 1/12° global US Navy Hybrid

Coordinate Ocean Model (HYCOM) (Chassignet et al., 2007) reanalysis product

(Table 3.2). Prior to creating the boundary files, the reanalysis product was gradually smoothed to 0 between depths of 700-2500m to avoid the potential development of bottom artificial currents (Figure 3.6).

58

Figure 3.6 Latitudinal transect along the northern boundary of zonal velcities from (a) HYCOM product prior to smoothing (b) HYCOM product post smoothing.

Table 3.2 Experiment name and time period used from HYCOM for boundary conditions.

HYCOM: https://www.hycom.org/data/glbu0pt08

Period Experiment

Jan 2005 – Jan 2012 GLBu0.08 – 19.1

Jan 2013- Aug 2013 GLBu0.08 – 90.9

Aug 2013 – Apr 2014 GLBu0.08 – 91.0

Apr 2014 – Apr 2016 GLBu0.08 – 91.1

Apr 2016 – Dec 2016 GLBu0.08 – 91.2

Due to the difference in bathymetry from the z-coordinate HYCOM reanalysis product and model s-coordinate grid, a hybrid bathymetry using both the GEBCO bathymetry and HYCOM bathymetry was created to improve interpolation at

59 Chapter 3. Monsoonal variability of the Island Mass Effect around Sri Lanka and the Maldives the boundaries (Figure 3.7). The revised reanalysis product and hybrid bathymetry were then gridded to the model grid boundaries for the final boundary files. Tidal constituents were obtained from the TPXO7 global tide model with a

1/4° resolution (Egbert and Erofeeva, 2002) to generate the tidal forcing that was applied to the four open boundaries using the Flather condition (Flather, 1976) and Chapman implicit boundary condition for the elevation. For the baroclinic mode (temperature, salinity and baroclinic momentum), a combination of

Orlanski-type radiation boundary conditions were applied with nudging

(Marchesiello et al.,, 2001). Vertical mixing processes were parameterised with the non-local K-profile boundary layer scheme (Large et al.,1994) implemented for both surface and bottom boundary layers.

In addition, a sponge layer was applied across 60 grid points (approximately 180 km) where the viscosity was increased linearly from the interior to a maximum viscosity of 300m2s-1 at the exterior (Figure 3.8). Explicit lateral viscosity was null everywhere in the model domain, except along the sponge layer near the open boundaries. The model was nudged towards daily HYCOM data along a linearly tapered nudging band along the open boundaries that had the same dimensions as the sponge layer (60 grid points).

60

Figure 3.7 An example of the hybrid bathymetry (blue) created from both GEBCO (black) and HYCOM (red) along the northern boudary. To create the hybrid bathymetry, HYCOM bathymetry was used for the outermost boundary bathymetry (100% HYCOM) and decreased by 5% per grid point until it was fully replaced by GEBCO bathymetry in the interior. For this example, the hybrid bathymetry comprises of 60% HYCOM bathymetry and 40% GEBCO bathymetry within the northern boundary zone. The box on the left shows the progression from pure HYCOM bathymetry to the Hybrid bathymetry to pure GEBCO bathymetry for all the boundaries.

Figure 3.8 Sponge layer for model domain boundaries where (a) shows the decrease of viscosity across 60 grid points where the highest viscosity starts at the outermost (grid point = 0) while (b) indicates the width of the changing viscosity across the sponge layer. Units are in m2s-1.

61 Chapter 3. Monsoonal variability of the Island Mass Effect around Sri Lanka and the Maldives

At the surface, the model was forced with three-hourly atmospheric (10m wind), heat (sensible and latent heat flux, shortwave radiation) and freshwater

(evaporation, precipitation) fluxes at a 1/8° resolution obtained from the

European Centre for Medium Range Weather Forecast (ECMWF) ERA-Interim reanalysis dataset (Dee et al., 2011). To prevent significant drift of sea surface temperature (SST) and salinity (SSS), the simulated surface data (SST and SSS) were relaxed to daily surface fields from HYCOM and a heat correction of 37°C was also applied.

One of the challenges of modelling this region is due to the paucity of in-situ data.

This results in the atmospheric forces being poorly constrained due to the limited availability of the observational satellite data, particularly during the SWM period with its increased cloud cover. This can lead to further and larger biases in the model forced simulations. Biases induced by the forcing can be further reduced by an assimilation of the ocean observational data (Zhang et al., 2007).

However, as we use a reanalysis product for the initial state and boundary conditions, this too has its limitations since the reanalysis product assimilates all available data products for this region and the conundrum in this is its dependence on data availability and quality.

62

3.3.2 Validation Data Sets

Numerical results from the model were compared with several observational data sets for the following variables: sea surface height, temperature, salinity and surface current velocities. Spin-up time took about 6 months to reach static equilibrium and is dependent on initial conditions used, in this case hydrography from HYCOM. As such, model results from 2006 onwards were used. A summary of the time period, data source and variable used for model comparison are presented in Table 3.3 along with the HYCOM reanalysis experiment numbers.

Table 3.3 Observational data sets used for model validation

Variable Period Source

Sea surface Sri Lanka: UHSLC: height May 2006 - https://uhslc.soest.hawaii.edu/ Oct 2016 Maldives: Jan 2006 – Dec 2016 India: Oct 2011 – Dec 2016

Temperature, Jan 2006 – Dec ARGO: salinity 2016 http://www.godac.jamstec.go.jp/argogpv/e/

Current velocities Jan 2006 – Dec OSCAR: 2014 https://www.pmel.noaa.gov/tao/drupal/disdel/

Sea surface Jan 2006 – Dec MODIS: temperature 2016 https://modis.gsfc.nasa.gov/data/dataprod/

63 Chapter 3. Monsoonal variability of the Island Mass Effect around Sri Lanka and the Maldives

3.4 Model Evaluation

To have confidence in the ability of the model to adequately represent the regional circulation, validation with observational data is necessary. The following statistical metrics were used for validation – Pearson correlation coefficient

(Coefficient of determination, R2), the Root-Mean Square Error (RMSE), the model bias, the mean absolute error (MAE) and Willmott model skill (Willmott et al., 1982).

The Willmott model skill was calculated as:

(� − �) ����� = 1 − (Eq 3. 1) [(� − �) + (� − �)]

Where �, = Observed values, �,= Modelled values, n = No. of measurements

64

Figure 3.9 Monthly climatology of observed sea level derived from UHSLC (black line) and SNIO-ROMS model (red) from tide gauge locations at (1) Sri Lanka, Colombo (2) Cochin, India (3) Minicoy, India (4) Male, Maldives (5) Hanimaadhoo, Maldives (6) Gan, Maldives. All units are in [m].

When comparing the model performance for the sea surface height, the model generally underestimated the sea level at the specified locations but performed better around Sri Lanka and India compared to the Maldives (Figure 3.9, Table

3.4). Possible reasons for this error could be due to the complex bathymetry around the Maldives that may have been overly smoothed to avoid potential horizontal pressure gradient errors and that the minimum depth of the model was set to 10m which may be deeper than where the tide gauges were situated at. In addition, some of the locations extracted from the model could not be exactly matched to the locations of the tide gauges since those fell within the land mask area of the model.

65 Chapter 3. Monsoonal variability of the Island Mass Effect around Sri Lanka and the Maldives

The model performed well in simulating the sea surface temperature during both the NEM and SWM (Skill >0.6, Table 3.5). It tended to slightly overestimate

SST in the region of the SECC (2°S - 2°N) but underestimate SST around the islands (Figure 3.10, Table 3.5). The overall mean difference between the model and observations was less than 0.25°C for both monsoons (Figure 3.10).

Table 3.4 Summary of model validation statistics for sea surface height obtained from tide gauges (UHSLC).

The climatology of the NEM and SWM for SST and velocities were computed from 11 years of model output and used for the validation. Comparisons between the climatology of the MODIS SST and the model SST indicate that although the bias is low for both monsoons (Table 3.5), the general distribution of SST difference are not consistent (Figure 3.10). The model tends to underestimate SST around the islands, particularly the shelf area between India and Sri Lanka by a

66

mean difference of ~1.8°C (Figure 3.10). This discrepancy could be due to the bottom reflectance from the shallow depth in this shelf region (<10m) which has been known to affect the estimation of MODIS measurements (Jiang et al., 2017).

Conversely, the overestimation of temperatures occurs closer to the SECC region

(Figure 3.10). During the NEM, the model has a lower positive bias compared to the SWM climatology and the largest difference in SST is localized around the

Lakshadweep Islands. However, during the SWM, the largest SST difference is more pronounced between the shelf region of Sri Lanka and the south point of

India. Overall, the model exhibits a strong skill (Skill>0.6, Table 3.5) to simulate the surface SST.

Table 3.5 Model skill statistics for SST and current surface velocities using climatology from MODIS and OSCAR.

Monsoon RMSE BIAS MAE R2 SKILL

SST 0.40 0.18 0.25 0.82 0.79 NEM U 0.19 0.03 0.13 0.60 0.70 (DEC/JAN/FEB) V 0.09 -0.02 0.07 0.44 0.41

SST 0.43 0.24 0.32 0.65 0.70 SWM (JUN/JULY/AUG U 0.25 -0.01 0.18 0.88 0.67 V 0.13 0.02 0.08 0.39 0.34

67 Chapter 3. Monsoonal variability of the Island Mass Effect around Sri Lanka and the Maldives

Figure 3.10 Seasonal sea surface temperature climatology difference between the model and satellite from 2006-2016 during the (a) NEM and (b) SWM with contour interval of 0.5°C. Positive (red) values indicate an overestimation of SST while negative (blue) values denote an underestimation of SST.

In terms of velocities, the model performed slightly better for the zonal velocities

(Skill >0.65, Table 3.4) compared to the meridional velocities (Skill <0.5, Table

3.5) where the mean difference for the zonal velocities and meridional velocities were 0.03ms-1 and 0.06ms-1 for the NEM and SWM respectively. A plausible explanation for the overall lower model skill could be due to the coarse resolution of the OSCAR satellite product at 1/3°. Based on the Q-Q plots for the observations with the model in Figure 3.11, the model tends to over predict velocities during the NEM while it underestimates velocities during the SWM

(Figure 3.11).

68

Figure 3.11 Q-Q plots of measured vs predicted values from climatology (2006-2016) for (a) zonal velocities during the (a) NEM and (b) SWM as well as the meridional velocities during the (c) NEM and (d) SWM.

Model validation with the ARGO vertical profiles were conducted at 8 locations around the Maldives and Sri Lanka (Figure 3.3). The mean monthly climatology of ARGO temperature during 2006-2014 was compared with the mean monthly climatology from the model at those areas and plotted with depth (Figure 3.12).

Overall, the model had good agreement with the Argo data and the simulated profiles were able to represent the changes in the thermocline depth for the different regions in the model domain. However, the model tended to

69 Chapter 3. Monsoonal variability of the Island Mass Effect around Sri Lanka and the Maldives underestimate temperatures compared to the observational data and this difference is apparent at depths less than 600m.

Figure 3.12 Comparison of Argo (black) and SNIO-ROMS (red) climatological monthly mean temperature profiles at 8 locations identified around Maldives and Sri Lanka (see Figure 3.3). Note that the y-axis are at different scales.

70

Volume transport at the RAMA mooring (M1, Figure 3.3) was calculated from the model at 80.5°E, 2.5°S-2.5°N and depth integrated over 80m for both the NEM and SWM during 2005-2016. The negative direction of the volume transport indicates that the main directionality of the currents heading westwards and the transport values were in good agreement with Mcphaden et al., (2015). The mean transport values were in the range of 9-11 SV (Figure 3.13) while Mcphaden et al., (2015) had transports of 5-15SV. The westward mean flow was attributed to being from the mean residual of the monsoon currents above the equator

(Mcphaden et al.,, 2015). The SWM had higher volume transport with a mean value of 10 SV compared with the NEM mean value ~7 SV.

Figure 3.13 Volume transport from model output at 80.5°E across 2.5°S-2.5°N, relative to a fixed depth of 80m for the NEM (black) and SWM (white) during 2006-2016.

71 Chapter 3. Monsoonal variability of the Island Mass Effect around Sri Lanka and the Maldives

3.5 Results

3.5.1 Seasonal mean circulation

The climatology for the mean surface current speed for the entire domain during the SWM was ~0.28 ms-1 while the mean speed during the NEM was ~0.22ms-1

(Figure 3.14). During the NEM, current speeds (>0.3ms-1) along the east coast of

India indicate the presence of the East Indian Coastal Current (EICC) as it flows southwards along the coastline of Sri Lanka. However, along the southern coast of Sri Lanka, current speeds increase (>0.5ms-1) before part of the current continues to form part of the West Indian Coastal Current (WICC) along the west coast of India (Figure 3.14). Several island wakes develop along the western coastline of the Maldives with the largest, in terms of spatial extent and current speed (>0.5 ms-1), located at 5°N and 1°N (Figure 3.14). All the wakes flow westwards and the largest wake at ~1°N extends up to 780km. The SECC was also identified between ~4°S-6°S throughout the model domain and has a mean current speed of ~ 0.4ms-1.

During the SWM, the SMC (speed>0.5 ms-1) enters the model domain from ~4°N-

8°N and flows eastwards against the northern tip of the Maldives archipelago and past the southern coast of Sri Lanka. Here, the SMC continues to form part of a recirculation feature identified as the Sri Lanka Dome (SLD). In contrast to the

72

NEM, the EICC and WICC reverses direction along the India coastline with weaker current speeds (speed < 0.35ms-1). A second major inflow current from the eastern boundary (speed > 0.4ms-1) between 0°N-3°S appears as an extension of the SEC (Figure 3.14).

Figure 3.14 Seasonal mean circulation of velocities (arrows) and speed (colormap) for (a) NEM and (b) SWM from climatology (2006-2016).

To identify locations of upwelling and downwelling, we extracted and mapped the depth of the 26°C isotherm (henceforth, referred to as D26) from the climatology

(Figure 3.15). D26 is more often used in the Northern Indian Ocean (McCreary et al.,, 1993; Rao et al.,, 2006 ; Mcphaden et al.,, 2015 ; Ali et al.,, 2015) but D20 is used more in the Southern Indian Ocean. Here, we define upwelling

(downwelling) processes to be when the local D26 is shallower (deeper) than the domain-averaged mean depth for D26. There are several distinct differences between the NEM and SWM, the first being that downwelling is more prevalent

73 Chapter 3. Monsoonal variability of the Island Mass Effect around Sri Lanka and the Maldives during the NEM and upwelling during the SWM. The mean depth for D26 during the NEM is ~112m while during the SWM, the mean depth for D26 is ~100 m.

During the NEM, downwelling occurs on both sides of the Maldives island chain but upwelling occurs within the archipelago and downstream from Kolhumadulu

Atoll (73.5°E, 0.5°N). Upwelling occurs along the south coast of Sri Lanka and

India (Figure 3.15). D26 reaches its shallowest at approximately 83°E, 13°N

(D26<50m), indicating strong upwelling present.

At the start of the NEM during December, the mean depth of D26 was ~113m and downwelling occurs on the western coastline of the Maldives and is deepest along 3°N and 0.5°N. Around the eastern coastline of Sri Lanka, D26 shallows to depths of less than 50m from the surface between ~7°N to 13°N. Downwelling intensifies during January with D26 being suppressed to depths greater than 160m and the overall mean depth increasing to ~115m. The downwelling signal is strongest along the northern tip of the Maldives at ~7°N and extends further westward across ~4°N. In contrast, upwelling strengthened along the east coast of India and downwelling occurred at ~86°E, 12°N. Upwelling reached its maximum along the south coast of Sri Lanka during January with D26 being shoaled to less than 80m from the surface. In February, downwelling intensity subsides overall (mean depth of D26 at ~112m) but is more prevalent across the domain with downwelling occurring both around Maldives and the south coast of

74

Sri Lanka Sri Lanka. However, upwelling also develops along the western coastline of Maldives at ~5°N and downstream from Kolhumadulu Atoll.

Downwelling during the SWM tends to occur to the south of 5°N for the Maldives but occurs at the northern tip of the Maldives. Upwelling intensifies along the south coast of Sri Lanka where the mean depth of D26 has been shoaled by approximately 20m to the surface. In addition, strong upwelling between 82°E -

86°E identify the general location of the Sri Lanka Dome (Figure 3.15).

The SWM is characterized by increased wind stress with monsoonal currents flowing eastward and known to be favourable for upwelling. However, upwelling is continuous throughout the domain and there are key differences around the

Maldives and Sri Lanka. The mean depth of D26 across the domain is ~104m in

June and shallows to 102m in July and shallowest in August at ~97m. Upwelling is at a maximum both spatially and in intensity at the northern tip of the

Maldives at ~6°N in June but subsides in July. In August, downwelling develops at this location. Below 5°N, downwelling is prevalent along the eastern coastline from June to July but the reverse occurs during August. At ~1°N, D26 is gradually uplifted from ~142m in June to a maximum of ~49m in August. At Sri Lanka, upwelling begins at the south coast in June and reaches its maximum in August.

In the region where the Sri Lanka Dome has been known to develop, upwelling

75 Chapter 3. Monsoonal variability of the Island Mass Effect around Sri Lanka and the Maldives begins at ~82°E,8°N and broadens during July, reaching its maximum at ~27m depth from the surface. The WICC along the western coast of India also intensifies upwelling throughout the SWM.

76