You are on page 1of 9

Ferney Sandoval Sanchez - CID: 01605474

CIVE97064 - Air-Sea Interaction Dynamics 2019-2020


Report - Coursework
1 INTRODUCTION

This report provides a summary of the air-sea dynamic interaction for the two settings of wind-forced waves generated in the
lab practice. This report acts a resource in order to understand the momentum flux from the overlying wind to the water column
below. To achieve the above, the data recovery from the wind-wave channel in the lab was processed to parametrize the systems
and identify the relevance of the wind speed and fetch.

2 ATMOSPHERIC BOUNDARY LAYER (ABL)

The ABL is the lowest part of the atmosphere and its behavior is directly influenced by its contact with a planetary surface.
The planetary surface acts as a non-slip condition and the velocities start in zero. This boundary generates rapid fluctuations,
turbulence and efficient vertical mixing. For the study cases, the planetary surface is the water surface. An important fact related
to this interface is that this surface is mobile and due to that the roughness elements are not constant.

2.1 Wind speed profile

Figure 1. Wind speed profile setting 12 and setting 16

From Figure 1, is appreciate that the velocity distribution of setting 16 is greater than the setting 12, this supposed that more
energy is develop and carry. Also, from the speed profile in the logarithmic scale is possible to observed that from an usual
profile the part that is modelled is the logarithmic layer and if the profile is extrapolate is possible to identify the roughness
length in the viscous sub-layer. On the other hand, related to the accuracy of the data there are slightly errors that create a
noise shape and this is directly associated with an human error due to the wrong location of the measured device (must be
place straight perpendicular to the wind direction).

2.2 Quantities Estimation

Once, the wind speed measurements are made is possible to determine the fluxes and quantities associated to the ABL. To do
that, 2 different approaches were considered: 1) Using the function “fit” within the Curve fitting toolbox of MATLAB; 2) Bulk
Aerodynamic Method (BAM). Below a description of the equations involved for the BAM. Also, it is important to specify that
in order to use this method is required to assume constant stress layer and zero velocity in roughness height.

𝜅[𝑢(𝑧2 ) − 𝑢(𝑧1 )] 𝑧1 𝑢∗ 𝑧 𝑢∗ 2
𝜏 = 𝜌𝑜 𝑢∗ 2 ; 𝑢∗ = 𝑧 ; 𝑧𝑜 = ; 𝑢(𝑧) = ln ( ) ; 𝐶𝐷10 =
ln( 2⁄𝑧1 ) 𝑢(𝑧 )𝜅 𝜅 𝑧𝑜 𝑢10 2
𝑒𝑥𝑝 ( 1 ⁄𝑢∗ )

Table 1. Quantities and parameters for each wind speed. (Setting 16 and Setting 12)

𝒖∗ [𝒎/𝒔] 𝒛𝒐 [𝒎] 𝒖𝟏𝟎 [𝒎/𝒔] 𝑪𝑫𝟏𝟎 𝝉 [𝑷𝒂]


Approach
Set 16 Set 12 Set 16 Set 12 Set 16 Set 12 Set 16 Set 12 Set 16 Set 12
Fit-Log 0.310 0.288 3.2e-04 4.7e-04 7.84 6.99 0.0016 0.0017 0.1193 0.1026
BAM 0.311 0.289 2.9e-04 5.7e-04 7.94 6.89 0.0015 0.0018 0.1204 0.1037

Rev. 0 14-03-2020 Page 1 of 9


Ferney Sandoval Sanchez - CID: 01605474
CIVE97064 - Air-Sea Interaction Dynamics 2019-2020
Report - Coursework

Figure 2. Fitting wind speed from fit Matlab and BAM to flab measurements

From the results presented in Table 1 and Figure 2, it is appreciated that the fitting function results fit more than the BAM
results to the lab measurements. This might be because the BAM only considered the relation between two heights while the
fitting MATLAB function condense the whole domain of the input data into a regression matching the measurements. Although,
for the study case de 11 measurements where interrelated to create 10 different BAM correlation and then finding the mean
value remain some errors. Perhaps the BAM gives a good approximation, still getting some errors relate to an empirical
approximation. Furthermore, is possible to appreciate that for both settings the BAM approach the solution differently. It means
that in setting 12 approach the measure shape oversizing the velocity values meanwhile in setting 16 underestimate the velocity.
Finally, the relative error getting between the 2 approaches oscillates between 0.3%-20% this is due to the assumption that
turbulence, viscosity of water, atmospheric stability and dissipation are ignored.

On the other hand, out of the good fitting obtained to the lab measurements, the overall results are consistent to the wind speed
profile because it is clearly that the turbulence in the setting 16 is greater than in the setting 12 due to bigger 𝒖∗ , value that is
an empirical parameter to describe turbulence. Similarly, is evident than the gradient rate to let profile fully develop is bigger
for the setting 16 than setting 12, perhaps it starts from lower height it reaches a higher velocity. Also, it is possible to appreciate
that the Drag coefficient is greater for the setting 16, that explain the reason the profile develops slowly at the begin and then
far away of the viscous sub layer goes rapidly. Additionally, is evident that there is a momentum flux.

To conclude, the regime that characterize both settings is the aerodynamically rough regime this statement is supported to the
parametrizing with u10 and CD10 and the Yelland et al., [1996] graphic in Figure 3. This figure evidence the momentum transfer
and is consistent and reliable to Charnock’s law. The domain of rough regime is defined after the change of slope in the data.
Roughly assumption could be considered that this rough regime is defined after u10 exceeds 6 m/s and CD10 is over 0.001.

Figure 3. Drag coefficient averaged

2.3 Experiment in Gulf of Mexico

Analogous if the experiment is tend to be perform to a platform in the Gulf of Mexico it is highly recommended to measured
some other parameters such as wind direction, virtual temperature, air temperature, sea surface temperature and relative
humidity, barometric pressure, radiation. Hence, there are several reasons to include these new parameters in the analysis:

Rev. 0 14-03-2020 Page 2 of 9


Ferney Sandoval Sanchez - CID: 01605474
CIVE97064 - Air-Sea Interaction Dynamics 2019-2020
Report - Coursework
1. General technics in the field such as automated weather stations measuring meteorological and observational satellites
sampling are not accurate because weather varies in surface layers and the data is not available for all altitudes. While
direct observations are more accurate and possible in the lab. This meteorological parameters help describing TKE.
2. Several assessments related to changes in air quality due to oil and gas production are develop base on these parameters.
3. This are meteorological inputs for dispersion modeling that can be useful assessments to quantify and diagnosticate human
comfort. Coastal communities in the Gulf of Mexico can be affected by the thermodynamic variations of the sea breeze.

3 FETCH-LIMITED WIND-GENERATED WAVES

3.1 Wave series

Wind generated waves due to the stress generated in the interface. Below the water surface elevation for each gauge and setting
is presented.

Figure 4. Wave surface elevation series for each gauge (Left) Setting 12 (Right) Setting16

From Figure 4, is appreciate the clear dependence of the water surface disturbance to the fetch and the velocity. This explains
why there is a proportional relation between these three parameters. As higher wind velocity excites the water large amplitudes
and develop waves are generated, as well, if the is more distance travel by the blowing wind. Also, as Figure 4 shows farther
gauge spends more time to generate a significant disturbance. However, is clearly that is easier to reach the gauge in the setting
16 due to higher velocity in the wind channel.

3.2 Stationary analysis

Figure 5. Stationary analysis setting 16 (left) normal scale (right) log scale

Rev. 0 14-03-2020 Page 3 of 9


Ferney Sandoval Sanchez - CID: 01605474
CIVE97064 - Air-Sea Interaction Dynamics 2019-2020
Report - Coursework
Figure 5 shows the stationary analysis develop in order to identify the data span at which the waves stop being time dependent.
This analysis identifies when the peak of the amplitude start oscillation on time but remains constant in space. Basically, means
when the wave reaches its steady state. Base on Figure 5 the most suitable way to identify this point on time is by looking
through the logarithmic scale representation and due to the fetch focus just on the gauge 4 that is the last to reach the stationary
state. To conclude, the time at which the data begin being stationary was defined at 100 s for all of the settings and gauges.
Also, as the wind source was not turning in the recording lapse, there is no drop in the amplitude and till the end of the records
is considered stationary.

3.3 Spectrums

Once a stationary state is defined is possible to determine the three main types of spectrums: discrete variance spectrum,
variance density spectrum (Evar(f)) and energy density spectrum (E(f)). Instead of representing the sea-state with the amplitude
spectrum these spectrums are adopted because they describe the amplitude quantities in an statistical form making the data
more manageable. For example, the sum of the variances of the wave components is equal to the variance of the sum of the
wave components. Additionally, because in linear wave theory the sum of potential and kinetic energy of the wave is
proportional to the variance. this variance spectrums allows to link the amplitude to the energy of the waves.

Nevertheless, there are some difference between these spectrums. First, the discrete variance spectrum is based on discrete
frequencies, whereas nature does not select such discrete frequencies and all frequencies are present at sea. Then, to solve the
above statement, the variance density spectrum is constructed to distribute the variance over a constant frequency interval Δf.
As this distribution remain creating slightly jumps or discontinuities is recommended that Δf--> 0. Finally, as was presented
before the advantage of this spectrums is the possibility to link with energy of wave. Due to that, the last spectrum is developed
in order to describe the 1-D energy density spectrum in terms of the 1-D variance density spectrum. In Figure 6 the spectrums
are presented and normalized by the peak frequency the biggest spectrum is always associated to the last gauge.

Table 2. Spectrum Resume

SPECTRUM UNITS EQUATION


Discrete variance spectrum [m2 ] 1 2
𝑎
2 𝑖
Variance density spectrum (Evar(f)) [m2 Hz-1] 1 2
𝑎 /Δ𝑓
2 𝑖
Energy density spectrum (E(f)) [kg s-2 Hz-1] 1
𝜌𝑔 𝑎𝑖 2 /Δ𝑓
2

Figure 6. (Top-Left) Discrete variance spectrum (Top-right) Variance density spectrum (Low) Energy density spectrum for all gauge in
setting 16

Rev. 0 14-03-2020 Page 4 of 9


Ferney Sandoval Sanchez - CID: 01605474
CIVE97064 - Air-Sea Interaction Dynamics 2019-2020
Report - Coursework
3.4 Quantities Estimation

Once the spectrum are built is possible to identify the short-term statistical characteristics such as peak wave frequency derived
from the frequency associated to the peak energy in the spectrum, significant wave height derived and mean wave period
derived from the spectral moments of the variance density spectrum. Then, analyze the behavior of this quantities as a function
of fetch. The below equations were required.
10 10
𝑚
𝑚𝑜 = ∫ 𝐸(𝑓)𝑑𝑓 ; 𝑚2 = ∫ 𝑓 2 𝐸(𝑓)𝑑𝑓 ; 𝐻𝑠 = 4√𝑚𝑜 ; 𝑇̅𝑜 = √ 𝑜⁄𝑚2
1 1

Figure 7. Parameters (Left) Setting 16 (Right) Setting 12

Table 3. Parameters resume setting 16

Gauge 1 Gauge 2 Gauge 3 Gauge 4


Variables
X = 3.53 m X = 6.79 m X = 9.92 m X = 14.19 m
m0 2.505e-06 6.396e-06 9.815e-06 1.1768e-05
m2 6.127e-05 9.795e-05 8.408e-05 8.542e-05
Hs [m] 0.0063 0.0101 0.0125 0.0137
To [s] 0.20 0.26 0.34 0.37
fp [s-1] 4.80 3.76 2.69 2.48

Table 4. Parameters resume setting 12

Gauge 1 Gauge 2 Gauge 3 Gauge 4


Variables
X = 3.53 m X = 6.79 m X = 9.92 m X = 14.19 m
m0 4.263e-07 1.947e-06 2.784e-06 3.348e-06
m2 1.673e-05 4.866e-05 4.0796e-05 3.936e-05
Hs [m] 0.0026 0.0056 0.0067 0.0073
To [s] 0.15 0.20 0.26 0.29
fp [s-1] 6.37 4.78 3.56 3.20

The results obtained and presented in Figure 7, Table 3,Table 4 are reliable to the expected behavior. First, the magnitude of
the results of the setting 16 are larger and is due to the higher velocity associated to this setting. Second, the significant height
and the mean wave period increased proportionally to the fetch whereas the frequency decrease. This behavior is because of
the fetch that make the spectrum narrower, steeper and sharper. Also, results are accurate because the significant wave height
is smaller than the maximum height for the time series of each setting. Additionally, it’s possible to notice that if the average
wave period is calculated the value oscillated from Tp to 0.9Tp that is accurate base on lectures.

Rev. 0 14-03-2020 Page 5 of 9


Ferney Sandoval Sanchez - CID: 01605474
CIVE97064 - Air-Sea Interaction Dynamics 2019-2020
Report - Coursework
3.5 Fetch-Limited

Wind-Wave development follows the duration‐ and fetch‐limited growth laws in a steady wind forcing case. For this analysis
the duration dependence is neglected because of the stationary condition. However, remain a fetch limited condition. For
instance, at short fetches, the waves grow rapidly and create a young sea states, but gradually the growth slows down until it
eventually stops. Meanwhile, at long-fetches is possible to reach fully developed sea states with higher energy.

Also, by doing this experimental approach is clear that JONSWAP was right considering a peak-enhancement function in order
to correlate the fetch-limited condition with the spectrum shape. Since the fetch let the spectrum evolve and change. Higher
fetch creates narrower and sharper spectrums and concentrate the energy in lower frequencies.

Figure 8. Fetch-limited dependence of spectrum (left) Setting 16 (Right) Setting 12

3.6 Jacobian transformation

Since frequencies ω and wave numbers k are related through the dispersion equation for linear wave theory, the wave-number
spectrum can be obtained from the frequency spectrum. For the study case specifically, a deep-water assumption was defined
in order to simplify the dispersion equation without depth dependence because it is unknown.

𝐸(𝑓) → 𝐸(𝜔) → 𝐸(𝑘)

Then, is required to identify the Jacobian for spectrum transformation. From lectures is known that the Jacobian from frequency
to wave frequency is 2π. Hence, the only unknown is the Jacobian to move from wave frequency to wave-number.

𝑑𝜔
𝐸(𝑓) = 𝐸(𝜔) → 𝐸(𝑓) = 𝐸(𝜔) 2𝜋
𝑑𝑓

𝑑𝜔
𝐸(𝑘) = 𝐸(𝜔) → 𝐸(𝑘) = 𝐸(𝜔) 𝐶𝑔
𝑑𝑘

From the analysis is observed that the Jacobian from wave frequency to wave-number happens to be equal to the group velocity
of waves propagation. Now, as deep-water assumption had been done the group velocity can be written as wave propagation
velocity half’s (0.5c) and dispersion equation can be simplified to ω2=gk.

𝜔𝑔 𝑔
𝐽 = 𝐶𝑔 = 0.5 𝑐 = 0.5 𝜔⁄𝑘 = ⁄𝜔 2 = ⁄𝜔

The final transformation is presented below


𝑔⁄
𝐸(𝑘) = 𝐸(𝜔) 𝜔

Figure 9. wave-number spectrum (left) non-normalized (right) normalized

Rev. 0 14-03-2020 Page 6 of 9


Ferney Sandoval Sanchez - CID: 01605474
CIVE97064 - Air-Sea Interaction Dynamics 2019-2020
Report - Coursework
4 ORIGINAL DATA ANALYSIS

4.1 Wind data

In order to determine the atmospheric boundary layer quantities using the bulk aerodynamical method (BAM) in chapter 2, two
assumptions were made:

1. Constant stress layer.


2. Parametrization just depending on 2 heights

This means that the roughness length is going to be constant independent of the reference height selected to apply the method.
In this section a sensible analysis of this assumption is going to be develop based on that 55 different combinations can be
made with the input data (11 points for each setting).
11 11!
𝐶(11,2) = ( ) = = 55 𝑛𝐶𝑟
2 (2! (11 − 2)!)

Figure 10. roughness length behavior depending on h1 & h2 BAM Method

From Figure 10 is observed that the assumptions made had an incident error in the calculation of the roughness length because
the roughness length is fluctuating depending in the pair of heights is chosen. So it is possible to over or under estimated the
quantities required. Also, is clearly that there is a human effect in the lab recording that explains the behavior in which at some
pair of heights the roughness length is far away from main sets plotted. Due to that, is required to create filters an neglect the
data results that are not reliable to the main behavior.

4.2 Wave data

Related to the wave data two analysis where propose, the first relate to the accuracy of the spectrums and the second related to
the energy of the waves.

4.2.1 Noisy Spectra

From the data recorded is required to apply a smoothing method to the resultant spectrum to make it less noise. The noise
generated in the spectrums is due to the frequency of recording data and the turbine channel noise. In order to deal with this
fact, a methodology was applied splitting the steady signal into segments and developing Fourier transformation to each
segment to get the spectrums and afterwards ensemble the average. This methodology is applicable base on the stationary
condition. Otherwise, the average couldn’t be performed.

Rev. 0 14-03-2020 Page 7 of 9


Ferney Sandoval Sanchez - CID: 01605474
CIVE97064 - Air-Sea Interaction Dynamics 2019-2020
Report - Coursework

Figure 11. Noisy spectra behavior

From Figure 11 is observed that the spectrum gets less noisy as much segments divide the stationary signal. Nevertheless, still
statistical accurate. Moreover, let the spectrum be more visually understandable with one respective peak energy and wave
frequency.

4.2.2 Total energy of spectrum and fetch-limited condition

Defined the energy density spectrum is possible to determine the total energy within the spectrum. Base on that energy is
possible to observed that there is a proportional relation between the fetch. This reaffirms that the fetch plays an important role
to fully developed a sea state and the fetch-limited condition cannot be neglected an any type during an analysis.
10
𝐸𝑡𝑜𝑡 = 𝜌𝑔 ∫ 𝐸𝑣𝑎𝑟 (𝑓)𝑑𝑓
1

Figure 12. Total energy of spectrums vs fetch

From Figure 12, is observed that energy associated to setting 16 is greater and reliable with previous quantities estimated in
chapter 2 and 3. Further, the velocity plays an important role in defining the increasing rate because it is proportional to a2. As
show in the figure the growth gradient at setting 12 is smaller than for setting 16.

Rev. 0 14-03-2020 Page 8 of 9


Ferney Sandoval Sanchez - CID: 01605474
CIVE97064 - Air-Sea Interaction Dynamics 2019-2020
Report - Coursework
4.2.3 Probabilistic & Statistics

Once the sea-state is defined it’s possible to predict possible massive waves that are important for coastal and offshore design.
For this case, a random case is going to be analyzed in order to evaluate which velocity and fetch is the most likely to let that
wave occur. For example, The probability of a wave larger than 0.55 cm and number of wave heights (𝐻). It is considered that
the storm has a duration of 30 minutes and the water is assumed deep.

−𝐻 2
( ) 30 𝑚𝑖𝑛
𝑃(𝐻 > 0.55 𝑐𝑚) = 𝑒 8𝜎𝑛 2 ; #𝑤𝑎𝑣𝑒𝑠 = 𝑝(𝐻 > 0.55 𝑐𝑚)
𝑇𝑜

Table 5. Results of the statistical analysis

Gauge 1 Gauge 2 Gauge 3 Gauge 4


P(H>0.55) # waves P(H>0.55) # waves P(H>0.55) # waves P(H>0.55) # waves
Setting 16 0.22 1749 0.55 3466 0.68 3185 0.72 3130
Setting 12 1.4E-4 1 0.14 1147 0.26 1575 0.32 1773

Figure 13. Statistical results for each setting and fetch (Left) Probability (Right) # of waves

From Figure 13, is observed that the probability remain following the fetch-limit condition and wind velocity dependence than
previous analysis and quantities found in the document. However, the number of waves over 0.55 cm in a 30 min simulation
breaks this pattern for the setting 16. This behavior can be attribute to the broad banded spectrum in the gauge 2 for this setting.
For this spectrum probably the energy is more concentrate in larger frequencies and that is the reason why the slope of the
probability distribution against the fetch decrease. (See Figure 8).

Rev. 0 14-03-2020 Page 9 of 9

You might also like