1 2 A1. Calibration The amplitude of the back-scattered signal is calibrated to volume backscattering 3 cross-section σb in units m−1 (a measure of the plume's ability to scatter sound 4 waves, which is dependent on the suspended-particle concentration and 5 turbulence-induced temperature fluctuations within the plume) using the following 6 sonar equation: 7 8 9 VSS = 20 ∗ log10(abs(datain )) − SL − D − G + TL − 20 ∗ log10(RS) −10 ∗ log10(dV). (1) Within the equation, VSS = 10log10 σb is the volume scattering strength, which is 10 the logarithm of the volume backscattering cross-section; abs(datain ) is the 11 envelope of the raw back-scattered signals;SL is the source level of the transmitter; 12 D is the directivity of the transmitter; G is the combined gain at both the transmitter 13 and receiver ends;TL is the transmission loss caused by the absorption of sound by 14 seawater and spherical expansion;RS is the receiver sensitivity, which is used to 15 convert the units of the raw signal (datain ) from machine units to μPa; and dV is the 16 volume element insonified by sound waves. Among those parameters mentioned 17 above, SL and RS are measured through laboratory experiments; TL is calculated 18 using the Francois-Garrison model (Francois et al. 1982) with constant parameters: 19 depth = 2500 m, salinity = 35 psu, temperatue = 2 °C, and PH = 7. 20 21 A2. Doppler Velocity Measurement 22 We measure the component Vr (in the direction of the acoustic line-of-sight) of 23 plume velocity from the Doppler frequency shift fD (Hz) of acoustic backscatter 24 using the covariance method described in Jackson et al. 2003. The following equation 25 gives the relation between Vr and fD , Vr = 26 27 1 N cfD 2fc T p w fD = 2π∆t angle [∑n=1 ∫t=0 E(t)E ∙∗ (t + ∆t)dt ′ ] (2) (3) 28 In equation (2), c is the sound speed and fc ~ 400 kHz is the sonar operation 29 frequency in the Doppler mode. In equation (3), the operator angle[ ] calculates the 30 phase angle in radians of a complex number. E(t) is a demodulated complex signal 31 corresponding to a given azimuthal beam and a given ping, whose amplitude and 32 phase reflect the amplitude of the acoustic backscatter and its phase shift relative to 33 the transmitted pulses. The integral in equation (3) estimates the autocorrelation 34 function of E(t) at the time lag ∆t. Note that a rectangular window with length 35 Tw = 1 ms is used to truncate the received signal. A summation over NP = 40 pings 36 at each elevation angle reduces the uncertainty in the measurement caused by 37 turbulence and background noise. In addition, the standard deviation 𝑉𝑠𝑡𝑑 of 𝑉𝑟 is 38 calculated over the 40 pings, which is used as a measure of the statistical uncertainty 39 in the Doppler measurements (Jackson et al. 2003). 40 A3. Entrainment Coefficient Calculation 41 Tao et al. 2013 estimate the entrainment coefficient α of hydrothermal plumes 42 rising in stably stratified static environments using numerical simulations. Their 43 estimates are reported to be consistent with experimentally determined values but 44 somehow not explicitly given in their paper. We herein recalculate the entrainment 45 coefficient using the simulation results and method presented in Tao et al. 2103. 46 47 According to Morton et al. 1956, the maximum rising height of a buoyant plume can be calculated as 1 Zm = 48 B 4 Ce (N03 ) , (4) 49 50 where B0 (m4 ⁄s 3 ) is the initial buoyancy flux of the plume and N (1/s) is the 51 buoyancy frequency, and Ce = 1.148α−1/2 52 (5) 53 54 is a universal constant. By performing linear regression on the simulation results of 55 Zm with different combinations of B0 and N, Tao et al. 2013 obtain Ce = 4.05. 56 Substituting into equation (5) gives α = 0.083. 57 A4. Spectrum Calculation 58 The time series samples of vertical volume flux, centerline vertical velocity, and 59 expansion rate are not evenly spaced. The sampling is semi-regular with the sampling 60 interval oscillating around 3 hours with small amplitude. In addition, the time series 61 has gaps concentrating mostly in the sector before Oct 9th (see Figure 6 (a), (b) in the 62 main text). As result, a modified periodogram (Lomb-Scargle periodogram, Scargle 63 1982) is used for spectral analysis of the time series. The Lomb-Scargle periodogram 64 has the following mathematical form 65 66 PX (ω) = N 1 [∑j Xj cosω(tj −τ)] { 2 ∑N 2 j cos ω(tj −τ) 2 + [∑N j Xj sinω(tj −τ)] 2 ∑N j sin ω(tj −τ) 2 }, where τ is defined by 67 N tan(2ωτ) = (∑N j sin2ωt j )⁄(∑j cos2ωt j ). 68 Within the above equation, ω = 2πf and f is the frequency of interest; X is the time 69 series data; N is the number of samples within the time series. The results shown in 70 Figure 7 of the main text are normalized to the variance of the time series (PX (w)⁄σ2X ) 71 and are thus dimensionless. 72 73 The Lomb-Scargle periodogram has well defined statistics for unevenly spaced 74 samples. Assuming Xj are independent samples of normally distributed noise, the 75 significance level of a spectral peak Z = PX (w) (the chance of observing a spectral 76 peak greater than Z ) is (Scargle 1982). Z 77 N QZ = 1 − (∫ χ2k (z)dz) , k = 2, 0 78 where χ2k is the probability density function of the chi-square distribution with k 79 degrees of freedom. In order to smooth the spectrum, periodograms calculated at 80 different levels along the plume axis (S = 5 to 15 m at 0.5-m intervals) are averaged to 81 give the result shown in Figure 7. Since a total of 21 periodogram are used for 82 averaging, assuming the noise at different levels is independent, the significance level 83 of the smooth spectrum becomes 84 QZ = 1 − (∫0 N MZ 2 χk (z)dz) , k = 2M, M = 21. 85 The dashed horizontal line in Figure 8 denotes the value of Z with the significance 86 level of 0.05(5%). 87 88 89 A5 Isolation of the Inertial Oscillations We isolate the inertial oscillations from the mean plume centerline vertical 90 velocity component 〈Wc 〉 by fitting the following sinusoidal function to the time 91 series of 〈Wc 〉, Wci = C + Acos(2πfc t) + Bsin(2πfc t) 92 1 t+T 〈Wc (t)〉dt is the mean value of the time series, fc = 1.5 cycle/day 93 where C = 94 is the inertial peak in the spectrum (Figure 7 in the main text), A and B are optimal 95 coefficients corresponding to the minimum of the mean square fitting error ∫ T t R= 96 97 Wci (t))2 dt. The amplitude of the inertial oscillations is Amp = √A2 + B2 , 98 99 1 t+T ∫ (Wc (t) − T t and the phase is B Pha = arctan (A). 100 101 A moving window of length T = 4 days is used to segment the time series to yield 102 the variation of the inertial oscillation amplitude as a function of time shown in Figure 103 9 in the main text. 104 Note that we use the least-square-fit method describe above instead of the 105 common band-pass filters (Thomson et al. [1990]) to extract the inertial oscillations 106 because the 〈Wc 〉 time series is unevenly sampled and has gaps. The least-square-fit 107 method is immediately applicable to unevenly sampled time series; whereas one has 108 to interpolate the 〈Wc 〉 time series onto an evenly spaced time axis before band-pass 109 filtering, and the interpolation process will introduce bias into the result. 110 111 112 A6 Quantifying the tidal loading effects According to Morton et al. [1956], the centerline vertical velocity Wc of a 113 buoyant plume injecting into a uniform, static background fluid from a point source 114 (B0 ≠ 0, M0 =0, Q0 =0, where M0 = 1/2b2 W02 and Q0 = b2 W0 are the initial 115 momentum and volume fluxes) can be calculated as 1/3 B Wc = q ( z0 ) 116 , (6) λ2 B0 = 1+λ2 b2 W0 gαt ∆T. 117 (7) 118 Within the equations, B0 (m4 /s3 ) is the initial specific buoyancy flux of the plume; 119 z is the height above the source; q is an empirically determined constant, which is 120 assumed ~1; b, W0 and ∆T are the orifice radius, initial vertical flow rate and 121 temperature anomaly of the plume; αt is the thermal expansion coefficient; λ = 1.2 122 is the empirically determined ratio between the e-folding radii of the cross-sectional 123 distributions of temperature and vertical flow rate; g = 9.8 m/s2 is the gravitational 124 acceleration. 125 126 A plume from a finite source ( B0 ≠ 0, M0 ≠0, Q0 ≠ 0) is asymptotically 127 identical to the plume from a point source (B0 , 0, 0) at vertical distance Z0 below the 128 vent orifice. Therefore we rewrite equation (6) as 129 1/3 B Wc = (z+Z0 ) (8) 0 130 before applying it to the hydrothermal plume observed in this study. Using the 131 formulations developed by Morton 1959 with minor modification, Z0 is calculated as 132 1/|γ| Z0 = 3.253 − 3.126|γ|3/2 ∫−1 (t 5 + 1)−1/2 t 3 dt, 133 (9) 134 γ5 = 1 − Г, 135 Γ = 22 5α−1 (1 + λ )B0 V0−5 Q20 , 9 2 1/2 136 where α = 0.08 is the entrainment coefficient and V0 = M0 . Substituting the 137 following parameters into equation (7) to (9) gives the mean values of Wc at 5, 10, 138 and 15 m above the orifice. Note that the parameters are chosen to match those 139 measured at Grotto. Increasing W0 and ∆T by 40% and 6% (see Section 4.2 of 140 the main text) and substituting into equations (7) to (9) along with the other 141 parameters gives the perturbed values of Wc at each height. Table A1 summarizes 142 the results. The ratios of deviations to means are also given in Table 2 in the main 143 text. 144 Z = 5, 10 and 15 m; 145 ∆T = 350 °C; 146 W0 = 0.3 m/s; 147 b = 0.1 m 148 αt = 10−4 / °C 149 Table A1: Z (m) Mean (m/s) Perturbed (m/s) Deviation/Mean 5 0.0414 0.0462 11.6% 10 0.0355 0.0399 12.5% 15 0.0320 0.0361 12.9% 150 151 152 153 Figure A1 154 155 156 Figure A1. Histogram of the mean volume flux <Q> time series. The red dotted lines denotes the cut-offs of the central 80% quantile. 157 Figure A2 158 159 Figure A2. Power Spectrum of a year-long seafloor pressure time series (April 1st, 2011 – Apri 1st, 160 161 162 163 2012) measured by a mooring (RCM-NE, NEPTUNE Canada) deployed approximately 3 km to the north of Grotto within the axial valley. The dashed lines denote the 98% confidence interval of the spectrum (solid line). Note that the spectrum has two spikes at semi-diurnal and diurnal frequencies, and no spike within the inertial band. 164 References: 165 1. R.E. Francois and G.R. Garrison, Sound absorption based on ocean measurements: 166 Part I: Pure water and magnesium sulfate contributions, Journal of the Acoustical 167 Society of America, 72, 896-907, 1982. 168 2. Jackson, D. R., C. D. Jones, P. A. Rona, and K. G. Bemis, A Method for Doppler 169 acoustic measurement of black-smoker flow fields, Geochemistry Geophysics 170 Geosystems, 4, 1095-1107, doi:10.1029/2003GC000509, 2003. 171 3. Scargle, J. D., Studies in astronomical time series analysis. II. Statistical aspects of 172 spectral analysis of unevenly spaced data, The Astrophysical Journal, 263, 173 835-853, 1982. 174 4. Morton, B. R., Forced Plumes, Journal of Fluid Mechanics, 5, 151-163, 1959 175 5. Morton, B. R., G. Taylor, and J. Turner, Turbulent Gravitational Convection from 176 Maintained and Instantaneous Sources, Proceedings of the Royal Society of 177 London, 234, 1-23, 1956.