Fmars 11 1332883
Fmars 11 1332883
Fmars 11 1332883
REVIEWED BY
Harvey (2017) forecast
Charles Reid Nichols,
Marine Information Resources Corporation, Sue Chen 1*, James A. Cummings 2, Jayaram Veeramony 3
United States
Vijay Tallapragada, and Justin S. Tsu 1
NCEP Environmental Modeling Center (EMC),
United States
1
Naval Research Laboratory, Marine Meteorology Division, Monterey, CA, United States, 2 Science
Applications International Corporation, Monterey, CA, United States, 3 Naval Research Laboratory,
*CORRESPONDENCE Ocean Sciences Division, Stennis Space Center, MS, United States
Sue Chen
sue.chen@nrlmry.navy.mil
KEYWORDS
1 Introduction
Tropical Cyclones (TC) are one of the most severe natural disasters that cause
devastating damage on societies worldwide from the combination of winds, tornados,
flood, and storm surge in both coastal and inland regions. The advancements of TC
dynamic models in recent decade have made steady inroads to improve deterministic and
probability forecasts of TC track, intensity, winds, and waves the surface wind and pressure field will generally result in higher
(Heming et al., 2019; Cangialosi et al., 2020). In particular, skill in the wave forecasts. However, direct improvement of wave
increased capabilities to observe the ocean wave energy spectra forecasts using data assimilation is also possible, but the sparsity of
routinely by operational buoy observing system or by targeted real-time wave data is a core issue for wave assimilation systems,
sampling of TC provide an opportunity to investigate the wave especially under tropical cyclone (TC) conditions. Buoy networks
spectra data assimilation impact on TC prediction. Inclusion of provide highly accurate estimates of wave spectrum statistics, but
wave observations to improve operational forecasting of wave these data are sparse due to the cost of deployment and
conditions has high impact in the coastal zones where a large maintenance of the buoys. In the U.S., the National Buoy Data
population lives. Center (NDBC) deploys fixed buoys near the coast that collect a
The fundamental concept underpinning spectral wave variety of observations useful in the monitoring of Hurricane wave
modeling is the energy balance equation. This states that the conditions. These measurements include accelerometers or
evolution of the wave spectrum is the sum of three source terms inclinometers on board the buoys that measure the heave,
describing the input of energy from the wind, the nonlinear transfer acceleration, and the vertical displacement of the buoy hull
of energy within the wave variance spectrum, and the dissipation of (NDBC, 2009; Hall et al., 2022), as well as winds and ocean
energy from wave breaking or shallow water processes (Komen temperatures. The in situ wave observations complement the
et al., 1994). Wave models compute an explicit representation of all satellite remote sensing of surface waves from altimeters, which
three source terms and the evolution of the wave spectrum, without has been available since the 1980s and assimilated by many
a priori assumptions about the spectral shape. Even though various operational centers worldwide. The altimeter wave estimates are
data assimilation techniques have been used to improve model valuable but they carry considerable uncertainty and the spatial and
predictions (see Bannister, 2017 for a review) in an operational temporal sampling characteristics of the altimeters limit their
environment, these techniques have been applied largely to the effectiveness (Wittmann and Cummings, 2004; Tran et al., 2010;
atmospheric and ocean circulation models. Wave model runs with Fan et al., 2012; Seemanth et al., 2021).
assimilated data have been shown to significantly reduce errors in More fundamentally, wave forecasting is a forcing problem
modeled wave parameters. Lionello et al. (1992) used an optimal while numerical weather prediction is an initial value problem. A
interpolation technique (OI) to assimilate significant wave heights considerable part of the ocean wave field is formed directly by
derived from satellite altimeter data to correct the total energy in the surface winds and near-surface pressure fluctuations in the
wave spectrum. A similar technique was used by Voorrips et al. atmosphere (i.e., wind sea components). As a result, errors in the
(1997) to assimilate significant wave height data from wave buoys. forecasting of wind-sea wave conditions are reduced when the skill
Smit et al. (2021) took the OI technique further by assimilating bulk of the weather models is improved with data assimilation. The
wave parameters including significant wave height and wave period assimilation of wave observations correct wave model errors, but if
from a large distributed buoy network, 129 free-drifting Sofar the wave field is corrected without also correcting the atmospheric
Spotters buoys, to analyze the potential impact on the total forcing field to match, the wave field will rapidly return to the state
energy. They reported the forecast improvement from wave dictated by the forcing. In this regard, a coupled ocean, atmosphere,
assimilation extends to 4 days. In all these cases, the aim was to wave modeling system that includes data assimilation in all of the
improve the wave energy in the modeled domain. Veeramony et al. coupled system components would be expected to maintain forecast
(2010) and Orzech et al. (2014) showed that for smaller domains skill for much longer than an uncoupled system.
focused on the nearshore region, a 4D variational assimilation Building upon the previous aforementioned studies, this study
(4DVar) system assimilating the wave spectrum to correct the investigates a wave model assimilation methodology that assimilate
wave boundary conditions provided significant improvements in both routinely available observations of wave energy spectra from
reproducing the directional wave spectra in the domain. Compared 13 NDBC buoys (Table 1; Kuik et al., 1988; Riley et al., 2019) and
to the 3DVAR data assimilation system, the added computation satellite altimeter derived altimeter significant wave height
cost of using 4DVAR may not be feasible for operational measurements in a fully coupled air-ocean-wave tropical cyclone
implementation because of wall time constrain. Conversely, data assimilation and forecast modeling system. Because there are
recent studies using a 3DVAR significant wave height many NDBC buoys in the Gulf of Mexico (GOM) region, the
assimilation show improvement of wave forecast up to 24 h Hurricane Harvey (2017) case is ideally used to study the impact of
(Saulter et al., 2020; Byrne et al., 2023). Using the most the directional wave spectra assimilation.
computational expansive ensemble-based assimilation of Hurricane Harvey entered the Gulf of Mexico on 23 August,
significant wave height method can further extend the wave rapidly intensified and made landfall near Port Aransas, Texas
forecast improvement out to 2.5 days (Houghton et al., 2023). around 10 p.m. on 25 August (03 UTC 26 August) as a devastating
The success of data assimilation in atmospheric and Saffir-Simpson category (CAT) 4 TC (Saffir 1973; Simpson, 1974).
oceanographic forecast systems is made possible by a considerable Hurricane Harvey’s rapid intensification near landfall was due to a
amount of routine observational data from a variety of sources. pre-existing very warm (~ 31°C), and well-mixed ocean
Data assimilation in atmospheric forecasts indirectly improve ocean temperature along the Texas Bight (Potter et al., 2019). They
and wave models through improved estimates of surface wind attributed the absence of cold ocean water in the shallow Texas
stresses, surface pressure, and air-sea fluxes (Sanabia et al., 2013; Bight shelf enabled Hurricane Harvey to draw from this warm
Chen et al., 2017). For wave models, specifically, improvements in ocean energy source and rapidly intensified from CAT3 to CAT4
Global model forecasts from the NOAA Global Forecasting the sea surface and infers SWH from the shape and position of the
System (GFS), NRL Global Ocean Forecasting System (GOFS 3.1; leading edge of the waveform data. The altimeter SWH retrieval
Metzger et al., 2010), and global WW3 provide the initial and lateral algorithms have difficulty deriving SWH in very low wave height
boundary conditions for COAMPS simulations. Additional to the environments and also near shore where radar reflections from land
GFS initial condition, a TC bogus is used to replace the GFS contaminate the signal. SWH is an integral wave parameter and can
circulation in TC areas based on the JTWC warning message be derived from 2D directional wave spectra obtained
(Komaromi et al., 2021). Procedures to insert a TC bogus vortex from observations.
into COAMPS-TC consist of several steps. Step 1 removes the GFS
TC vortex winds. This is accomplished by removing the vorticity 2.2.2 Spectral wave data
and divergence component of the GFS winds. Step 2 performs direct Spectral wave data are obtained from fixed buoys. For fixed
solver iteration to balance the pressure, non-divergent winds, and buoys the wave measurements are inferred from the accelerometers
thermodynamics variables using a non-linear balancing equation or inclinometers on board the buoys that measure the heave
and variational adjustment method (Barker, 1980). Step 3 computes acceleration or the vertical displacement of the buoy hull during
a Rankin vortex (Holland et al., 2010) that uses the radius of the wave acquisition time. A Fast Fourier Transform (FFT) is
maximum wind radius to build a sloping eyewall wind structure at applied to the data by the processor on board the buoy to
the TC location from the JTWC warning message. The bogus vortex transform the data from the temporal domain into the frequency
depth and structure of the secondary circulation are set empirically. domain. From this transformation wave energies with their
Step 4 blends large-scale winds with the Rankin vortex. The final associated frequencies are derived. A limitation of the fixed buoy
step 5 inserts the TC bogus vortex in nest 3 using a Barnes analysis network is that the buoys are primarily positioned in shelf and
scheme (Barnes, 1964). coastal areas and do not provide open-ocean directional wave
Eight 5-day coupled COAMPS simulations for Hurricane observations. However, they provide valuable in situ observations
Harvey are carried out from 12 UTC 24 to 00 UTC 30 August, to monitor and improve model initial ocean and wave conditions
2017 cycling every 12-h using the Navy Coupled Ocean Data ahead and underneath the hurricanes. For the study, 13 NDBC
Assimilation (NCODA) for the ocean (Cummings and Smedstad, buoys that observed sea surface temperature and wave spectrum are
2013) and wave component. NCODA assimilates ocean and wave assimilated into NCODA (Figure 1).
observations from both satellite and in situ data sources.
Previous studies have shown assimilation of ocean observation 2.2.3 Ocean and wave model data
improved air-ocean coupled COAMPS-TC intensity forecast assimilation method
(Sanabia et al., 2013; Chen et al., 2017). The current study uses a NCODA uses a 3DVAR scheme to initialize the ocean model
new wave assimilation method in NCODA. Three fully coupled air- and a 2DVAR scheme to initialize the wave model. The
ocean-wave COAMPS experiments are used to quantify wave implementation is based on a “weakly” coupled data assimilation
assimilation impact on Harvey’s forecast. They include approach whereby the coupled model is used to provide
experiments without the ocean and wave assimilation (NDA), background information for separate ocean, atmosphere, and
with the ocean assimilation only (ODA), and with the wave wave model analyses. The analysis increments generated from
assimilation only (WDA). Experiment ODA assimilates surface these separate analyses are then added back into the next cycle of
and subsurface ocean temperature and salinity from buoys, floats, coupled model background fields. The mismatch in time scales
drifters, and satellite. The ODA also assimilates the Altimeter sea among the atmosphere, ocean, and wave models is a major difficulty
surface height observations. Experiment WDA only assimilates the as each of the models must use the same time window in the weakly
2D directional wave spectrum derived from buoys and altimeter coupled assimilation. We have taken the approach of using an
significant wave height. appropriately short time window of 12 hours for all of the
model components.
NCODA is an oceanographic implementation of the Navy
2.2 Assimilation method and wave data Variational Atmospheric Data Assimilation System (NAVDAS), a
3DVAR technique developed for Navy numerical weather
Two sources of wave field observations are assimilated in the prediction systems (Daley and Barker, 2001). The 3DVAR
COAMPS WW3 model and are described here: analysis variables are temperature, salinity, geopotential (dynamic
height), and u, v vector velocity components. All ocean variables are
2.2.1 Significant wave height analyzed simultaneously in three dimensions. The horizontal
Significant wave height (SWH) is defined as the average wave correlations are multivariate in geopotential and velocity, thereby
height (from trough to crest) of the highest third of the waves in a permitting adjustments to the mass fields to be correlated with
given sampling period. Satellite altimeters are the primary means of adjustments to the flow fields. The velocity adjustments (or
providing near-global measurements of SWH. The measurements increments) are in geostrophic balance with the geopotential
are made at nadir along the satellite tracks with resolutions of 5 to 7 increments, which, in turn, are in hydrostatic agreement with the
km. The standard algorithm for retrieving SWH from satellite temperature and salinity increments. The NCODA 3DVAR
altimeters fits a modeled shape to the radar echo waveforms from problem is formulated in observation space as:
xa = xb + Pb H T (HPb H T + R)−1 ½y − H(xb ) Fourier coefficients (r1, r2), the mean wave direction (alpha1) and
the principal wave direction (alpha2) based on the Fourier series
where xa is the analysis vector, xb is the background vector, Pb is expansion originally developed by Longuet-Higgins et al. (1963).
the background error covariance matrix, H is the forward operator, This required a modification to the H forward operator in the
R is the observation error covariance matrix, and y is the 3DVAR. For other data types, the forward operator is spatial
observation vector. HP b HT is approximated directly by the interpolation performed in two or three dimensions by fitting a
background error covariance between observation locations, and surface to a 4x4 or a 4x4x4 grid point target and evaluating the
PbHT directly by the error covariance between observation and grid surface at the observation location.
locations. The quantity [y-H(xb)] is referred to as the innovation
Benoit (1992), (1994) describes the methods commonly used to
vector, [y-H(xa)] is the residual vector, and xa-xb is the increment reconstitute the directional wave spectra from the reported values.
(or correction) vector.
Of these, there are two main methods that are widely used: The
The oceanographic error covariances in the 3DVAR are maximum likelyhood method (MLM), which was developed and
separated into a background error variance and a correlation. The
presented in detail by Oltman-Shay and Guza (1984), and the
correlation is further separated into a horizontal and a vertical maximum entropy method (MEM) developed and presented in
component. Horizontal correlation length scales vary with location
detail by Lygre and Krogstad (1986). Earle et al. (1999) show that
and vertical correlation length scales vary with location and depth. while MEM provides better resolution, it produces artificially
Flow-dependence is introduced by modifying the horizontal
narrow directional spreading. For the purposes of this study, it
correlations with a tensor obtained from a model forecast was determined that the accuracy in directional spreading provided
background field that is used as a proxy for the circulation field.
by MLM is more important than the resolution since the model
Example flow fields include sea surface height, sea surface resolution used is 7.5°, which is reasonable for this application, but
temperature, and potential vorticity. The flow-dependent tensor
not high enough to necessitate using MEL. Once the 2D spectral
spreads innovations along rather than across the flow field density is obtained from the values reported by the buoys, the
gradients, which is a desirable outcome in the analysis. The
observed spectra are compared to the model spectra using nearest
discontinuous and non-homogeneous influence of coastlines in neighbor. The differences in wave energy as a function of frequency
the analysis uses a second tensor based on distance to land. Near
and direction are computed which form the (y – H(xb)) innovations
the coast the tensor rotates and stretches the horizontal correlations in the 3DVAR.
along the coast while minimizing or removing correlations into the
land. Error correlations are expected to be anisotropic near
coastlines due to horizontal advection from coastal currents that
flow parallel to the coast.
3 Results
Assimilation of wave observations into the wave model
component of the coupled system required several modifications
3.1 Buoy observations
to the 3DVAR. First, the wave model prognostic variable is wave
Buoys 42002, 42019, 42020, and 42035 have a minimum
energy as a function of location, direction, and frequency. As such,
distance to Harvey’s track in GOM about 74, 57, 37, and 37 nm
all analysis updates or corrections to the wave model must be done
respectively during Harvey’s traverse toward Texas coast. Buoy
to the model forecast directional wave spectra at each model grid
42002 and 42019 are located to the right of Harvey’s track and buoy
point. Second, the error covariances were modified to include not
42020 is located to the left of the Harvey’s track (Figure 1).
only correlations between observations and grid points but also
Interestingly because Harvey’s track moved back out to sea after
correlations within wave spectra in terms of frequency and
the first landfall on 26 August, this movement put buoy 42035 to the
direction. Assimilation of integral measures of wave energy, such
right of Harvey’s track at this time before the first landfall, then
as altimeter SWH, is achieved in two steps. First, an analyzed SWH
switched to the left of the track after Harvey’s final landfall on 30
increment field is added to the WW3 SWH forecast (Hf) to produce
August. Buoys 42002, 42019, and 42020 all have high values of
a corrected SWH analysis field (Ha). The analyzed wave model
spectral wave density > 50 m2 Hz-1 around 25-26 August when
spectrum (Fa) as a function of frequency (f) and direction (Q) is
these buoys were closest to the Harvey track. During this period, all
then obtained from the ratio of analyzed and forecast SWH fields to
three buoys exhibit a dominant wind-sea wave with range of wind
produce an updated forecast spectrum (Ff) using a simple scaling
speed (ms-1) of 10-27 ms-1 (1 ms-1 = 1.943844492 kt) mean wave
strategy,
period of 10-14s, and significant wave height of 3-7 m (top panel of
a = (H a =H f )2 Figures 2A–C). A research mooring south of Port Aransas at 20 m
water depth recorded a 10 m significant wave height on 26 August,
F a (f , Q) = aF f (f , Q)
2017 (Romero-Arteaga et al., 2022).
Assimilation of the NDBC buoy spectral wave data is more The significant wave height from buoy 42002 relaxes back down
straightforward in that the observations measure the directional to around 5 m after the passage of Harvey between 27-28 August
wave spectrum. However, the 2D wave data provided by the NDBC and has the signature of swell with mean wave period< 5 s. During
are in the form of the non-directional spectral wave density as a this time, buoys 42019 and 42020 mean wave period are ~ 6-8 s,
function of the frequency along with the normalized directional indicating a mixed wind and swell sea. All three buoys’ significant
wave height and mean wave period increase again after Harvey’s spectra density maxima on the two front quadrants of Harvey
track moved back out to the sea on 28 August. Compared to these before Harvey’s final landfall. The slow-down of Harvey’s track that
three buoys, buoy 42035 experiences a lower wind speed range of 8- turned back out to sea after the first landfall and subsequent
16 m s-1, mean wave period of 6-10 s, and 2-3 m significant wave landfalls reduced the wind intensity, hence the weaker 2D
height during 26-29 August (top panel of Figure 2D). The wind directional wave spectral density over two closest distance buoys
speed increases to 19 m s-1 but with receding significant wave height 42019 and 42035. The furthest buoy from Harvey’s track is 42002,
to 2 m and mean wave period ~ 7 s when Harvey’s track moved which has the largest 2D directional wave spectral density. For the
back out to sea. Buoy 42035 is located on the shallow Texas– earlier three time periods, waves from all four buoys show their
Louisiana continental shelf and because the wind direction (middle maximum 2D directional wave spectral density rotate from
panel of Figure 2D) during this time is northerly, prevents the high Harvey’s front left quadrant to rear-right quadrant. Additionally,
wave surge at this location. Buoy 42035 also has the lowest spectral buoy 42035 (Figure 4D) has the smallest 2D directional wave
wave density compared to the other three buoys (bottom panel spectral density compared to the other three buoys (Figures 4A–C).
of Figure 2D).
Figure 3 shows two 2D wave density spectra using MLM from
buoys 42002 (3a) and 42020 (3b) at 05 UTC (mid night Texaslocal 3.2 Wave data assimilation impact on
time) 25 August when Harvey reached a CAT 4 intensity. The COAMPS TC track and intensity forecast
maximum 2D wave spectra density are 0.36 and 0.68 m2 Hz-1,
respectively and located on the front-right quadrant (0° points to Three fully coupled air-ocean-wave COAMPS experiments are
north). The time series of mean 2D wave density spectra for all experiments without the ocean and wave assimilation (NDA), with
frequencies from these four buoys show four separate periods of 2D the ocean assimilation only (ODA), and with wave assimilation only
wave spectra density (m2 Hz-1 degree-1) on 25-26 August, 27-28 (WDA). Comparison of COAMPS forecast track from each
August, 28-29 August, and 30 August–1 September (Figure 4). All experiment shows all three COAMPS experiments have a smaller
four buoys show a fully developed sea from all directions after 30 track error over the ocean than after landfall (Figure 5). The best
August and maximum 2D directional wave spectra density track is from the International Best Track Archive for Climate
magnitudes at this time. They exhibit 1-3 2D directional wave Stewardship (IBTrACS) version 4 (Knapp et al., 2018) which shows
A C
B D
FIGURE 2
NDBC buoy observations from buoy (A) 42002, (B) 42019, (C) 42020, and (D) 42035. Top panel is the time series of buoy observed wind speed (m
s-1, blue line), mean wave period (s-1, orange line), and significant wave height (m, black line). Bottom panel is the time series of spectra wave density
(m2 Hz-1).
FIGURE 3
Buoy 42002 and 42020 2D directional wave spectral density (m2 Hz-1 degree-1) in polar coordinate on 05 UTC 25 August, 2017. At this time
Hurricane Harvey reached a CAT 4 intensity. The 2D directional wave spectral density is derived from the buoy 1D spectral wave density using the
MLM method.
Harvey degraded to tropical storm after the last landfall and moved ~ 3-4 nm compared to the no ocean and wave DA experiments
rapidly northeastward after 31 August. In contrast, the last 00 UTC (Table 2). The maximum reduction of absolute mean track error
28 August COAMPS runs for all three experiments have a from WDA experiment is about ~ 9 nm (1 nm = 1.852 km), i.e., 50
northward track and move much slower than the observation. nm compared to the NDA experiment’s 59 nm, at 66 h
The time series of mean absolute track error shows WDA (red forecast time.
line in Figure 6) is smaller for most of the forecast lead times. The 5- Figure 7A is a spaghetti plot of forecast maximum wind speed
day mean absolute track error for experiments NDA, ODA, and (kt) that show all COAMPS experiments have a small low intensity
WDA are 72.9, 73.2, and 69.1 nm, respectively. Assimilating wave bias compared to IBTrACS. The homogenous time series of mean
data clearly improves COAMPS Harvey track forecast by a mean of intensity bias of WDA experiment is better than the NDA and ODA
A C
B D
FIGURE 4
Time series of mean spectral wave density (m2 Hz-1 degree-1) from buoy (A) 42002, (B) 42019), (C) 42020, and (D) 42035.
TABLE 2 COAMPS 0-5 day forecast mean track and intensity errors for
each experiments.
The forecast period is from 12UTC 24 August - 00 UTC 30 August, 2017. The sample size for
each experiment is 184, i.e. 8 forecast cycles times 23 of 6 hourly data.
A B
FIGURE 7
(A) A speghettii plot of COAMPS forecast maximum wind speed (kt) and (B) COAMPS forecast maximum wind speed bias compared to IBTtraACS.
between WDA minus NDA and WDA minus ODA experiments wave height and mean wave period from all three experiments are
show the ODA experiment is as good as or better than the WDA larger than 0.5. But the mean wave direction correlation for all 3
assimilation. They both reduced the forecast error for all three wave experiments is lower than 0.5 with correlation of 0.34, 0.39, and 0.35
variables evaluated here compared to the NDA experiment. for the NDA, ODA, and WDA, respectively, A plausible cause of
However, the mean correlation of WDA is higher than ODA for lower wave direction correlation of WDA than ODA may be due to
significant wave height and absolute wave period but not for the MLM tendency to over broaden the 2D wave direction (Donelan
wave direction (Figure 12). The mean correlation of significant et al., 2015).
A B
C D
FIGURE 8
Maximum differences of COAMPS WDA minus NDA forecast validate at the Harvey’s first landfall time on 03 UTC 26 August, 2017. (A) 39 h forecast
from the 12 UTC 24 August cycle, (B) 27 h forecast from the 00 UTC 25 August cycle, (C) 15 h forecast from the 12 UTC 25 August cycle, and (D) 3
h forecast from the 00 UTC 26 August cycle. Color shadings are the wind speed (WSPD) differences. Blue contours represent significant wave height
(SIGH) differences with contour interval of 0.1 m. Solid blue lines are positive differences and dashed blue lines are negative differences. Black and
magenta arrowers depict the maximum differences of COAMPS WDA minus NDA forecast wind speed and direction, and current speed (CURR) and
direction, respectively. The maximum differences of wind & current vectors are shown on the title of each panel. Black-square line depicts Havey’s
best track from IBTrACS and gray- dot line is the WDA forecast track.
FIGURE 9
Time series of forecast mean (A) significant wave height (m) bias (dash lines) and RMSE (solid lines), and (B) mean absolute error differences between
the WDA minus NDA (red line) and WDA minus ODA (cyne line) experiments. The blue, cyane, and red lines in (A) are from the NDA, ODA, and WDA
experiments, respectively. The station and forecast cycles mean is computed at buoys 40002, 42019, 42020, and 42035 locations.
FIGURE 10
Time series of forecast mean (A) wave directional (°) bias (dash lines) and RMSE (solid lines), and (B) mean absolute error differences between the
WDA minus NDA (red line) and WDA minus ODA (cyne line) experiments. The blue, cyane, and red lines in (A) are from the NDA, ODA, and WDA
experiments, respectively. The station and forecast cycles mean is computed at buoys 40002, 42019, 42020, and 42035 locations.
FIGURE 11
Time series of forecast mean (A) absolute wave period (s) (dash lines) and RMSE (solid lines), and (B) mean absolute error differences between the
WDA minus NDA (red line) and WDA minus ODA (cyne line) experiments. The blue, cyane, and red lines in (A) are from the NDA, ODA, and WDA
experiments, respectively. The station and forecast cycles mean is computed at buoys 40002, 42019, 42020, and 42035 locations.
A B
FIGURE 12
The mean correlation of forecast and buoy observed time series for (A) significant wave height (m), (B) mean wave direction (°), and (C) mean
absolute wave period (s). The forecast cycle is from 12 UTC 25 August to 00 UTC 28 August, 2017.
References
Allard, R. A., Campbell, T. J., Smith, T. S., Jensen, T. G., Chen, S., Cummings, J. A., vol. II. (Springer Berlin Heidelberg, Berlin, Heidelberg), 303–343. Available at: https://
et al. (2010). Validation test report for the coupled ocean atmospheric mesoscale link.springer.com/chapter/10.1007/978-3-642-35088-7_13.
prediction system version 5.0 (Washington DC, U.S.A.: Naval Research Laboratory), Daley, R., and Barker, E. (2001). NAVDAS: Formulation and diagnostics. Monthly
172. NRL/MR/7322-10-9283. Weather Rev. 129 (4), 869–883. doi: 10.1175/1520-0493(2001)129%3C0869:NFAD%
Allard, R. A., Smith, T., Jensen, T. G., Chu, P. Y., Rogers, E., Campbell, T., et al. (2012). 3E2.0.CO;2"https://doi.org/10.1175/1520-0493(2001)129<0869:NFAD>2.0.CO;2
Validation test report for the Coupled Ocean/Atmosphere Mesoscale Prediction System Donelan, M. A., Babanin, A., Sanina, E., and Chalikov, D. (2015). A comparison of
(COAMPS) version 5.0: ocean/wave component validation. (Washington DC, U.S.A.: methods for estimating directional spectra of surface waves. J. Geophys. Res. Oceans
Naval Research Laboratory), 91, NRL Memorandum Report NRL/MR/7320–12-9423. 120, 5040–5053. doi: 10.1002/2015JC010808
Bannister, R. N. (2017). A review of operational methods of variational and ensemble Doyle, J. D., Jin, Y., Hodur, R., Chen, S., Jin, H., Moskaitis, J., et al. (2012). “Real-time
variational data assimilation. Q. J. R. Meteorol. Soc. 143 (703), 607–633. doi: 10.1002/ tropical cyclone prediction using COAMPS-TC,” in Advances in geosciences, vol. 28 .
qj.2982. ISSN: 1477870X. Eds. C.-C. Wu and J. Gan (World Scientific Publishing Company, Singapore), 15–28.
Barker, E. H. (1980). Solving for temperature using unnaturally latticed hydrostatic doi: 10.1142/9789814405683_0002
equations. Mon. Wea. Rev. 108, 1260–1268. doi: 10.1175/1520-0493(1980)108,1260: Earle, M. D., Steele, K. E., and Wang, D. W. C. (1999). Use of advanced directional
SFTUUL.2.0.CO;2 wave spectra analysis methods. Ocean Eng. 26, 1421–1434. doi: 10.1016/S0029-8018
Barnes, S. L. (1964). A technique for maximizing details in numerical weather map (99)00010-4
analysis. J. Appl. Meteorology Climatology 3, pp.396–pp.409. doi: 10.1175/1520-0450 Fan, Y., Lin, S. J., Held, I. M., Yu, Z., and Tolman, H. L. (2012). Global ocean surface
(1964)003<0396:ATFMDI>2.0.CO;2 wave simulation using a coupled atmosphere–wave model. J. Climate 25, 6233–6252.
Benoit,M. (1994).Extensive comparisonofdirectionalwave analysismethodsfrom gauge array doi: 10.1175/JCLI-D-11-00621.1
data. In Proc. 2nd Int. Symp. on Ocean Wave Measurement and Analysis (ASCE), 740–754. Gopalakishnan, S., Upadhayay, S., Jung, Y., Marks, F., Tallapragada, V., Mehra, A.,
Benoit, M. (1992). Practical comparative performance survey of methods used for et al. (2021). 2020 HFIP R&D activities summary: Recent results and operational
estimating directional wave spectra from heave-pitch-roll data. Coast. Eng. 1992, 62–75. implementation. doi: 10.25923/718e-6232
doi: 10.1061/9780872629332.005 Hall, C., Jensen, R. E., and Wang, D. W. (2022). Performance evaluation of the newly
Byrne, D., Horsburgh, K., and Williams, J. (2023). Variational data assimilation of operational NDBC 2.1-m hull. J. Atmospheric Oceanic Technol. 39, 861–880.
sea surface height into a regional storm surge model: Benefits and limitations. J. Geo. doi: 10.1175/JTECH-D-21-0172.1
Res. Ocean 16 (1), 1–14. doi: 10.1080/1755876X.2021.1884405 Heming, J. T., Prates, F., Bender, M. A., Bowyer, R., Cangialosi, J., Caroff, P., et al.
Campbell, T., Allard R., R., Preller, R., Smedstad, L., Wallcraft, A., Chen, S., et al. (2019). Review of recent progress in tropical cyclone track forecasting and expression of
(2010). Integrated modeling of the battlespace environment. Computing Sci. Eng. 12, uncertainties. Trop. Cyclone Res. Rev. 8 (4), 181–218. doi: 10.1016/j.tcrr.2020.01.001
36–45. doi: 10.1109/MCSE.2010.78 Holbach, H. M., Bousquet, O., Bucci, L., Chang, P., Cione, J., Ditchek, S., et al. (2023).
Cangialosi, J. P., Blake, E., DeMaria, M., Penny, A., Latto, A., Rappaport, E., et al. Recent advancements in aircraft and in situ observations of tropical cyclones. Trop.
(2020). Recent progress in tropical cyclone intensity forecasting at the National Cyclone Res. Rev. 12 (2), 81–99. doi: 10.1016/j.tcrr.2023.06.001
Hurricane Center. Weather Forecasting 35, 1913–1922. doi: 10.1175/WAF-D-20- Holland, G. J., Belanger, J. I., and Fritz, A. (2010). A revised model for radial profiles of
0059.1 hurricane winds. Monthly weather Rev. 138, 4393–4401. doi: 10.1175/2010MWR3317.1
Centurioni, L., Braasch, L., Di Lauro, E., Contestabile, P., De Leo, F., Casotti, R., et al. Houghton, I. A., Penny, S. G., Hegermiller, C., Cesaretti, M., Teicheira, C., and Smit,
(2017). A new strategic wave measurement station off Naples port main breakwater. P. B. (2023). Ensemble-based data assimilation of significant wave height from Sofar
Coast. Eng. Proc. 1, 36. doi: 10.9753/icce.v35.waves.36 Spotters and satellite altimeters with a global operational wave model. Ocean Model.
Charnock, H. (1955). Wind stress on a water surface. Quart. J. R. Meteor. Soc 81, 183, 102200. doi: 10.1016/j.ocemod.2023.102200
639–640. doi: 10.1002/qj.49708135027 Janssen, P. A. (1991). Quasi-linear theory of wind-wave generation applied to wave
Chen, S., Cummings, J., Doyle, J. D., Hodur, R. M., Holt, T., Liou, C. S., et al. (2003). forecasting. J. Phys. oceanography 21, 1631–1642. doi: 10.1175/1520-0485(1991)021%
COAMPS version 3 model description. (Washington DC, U.S.A.: Naval Research 3C1631:QLTOWW%3E2.0.CO;2
Laboratory), 143, NRL Report: NRL/PU/7500-04-448. Joyce, B. R., Gonzalez-Lopez, J., van der Westhuysen, A. J., Yang, D., Pringle, W. J.,
Chen, S., Cummings, J. A., Schmidt, J. M., Sanabia, E. R., and Jayne, S. R. (2017). Westerink, J. J., et al. (2019). US IOOS coastal and ocean modeling testbed: Hurricane-
Targeted ocean sampling guidance for tropical cyclones. J. Geo. Res. Ocean 122 (5), induced winds, waves, and surge for deep ocean, reef-fringed islands in the Caribbean.
3505–3518. doi: 10.1002/2017JC012727 J. Geophysical Research: Oceans 124, 2876–2907. doi: 10.1029/2018JC014687
Chen, S., Gaberšek, S., Doyle, J. D., Cook, J., Chu, P., Allard, P. R. A., et al. (2015). Knapp, K. R., Diamond, H. J., Kossin, J. P., Kruk, M. C., and Schreck, C. J. (2018).
Validation test report for the Coupled Ocean/Atmosphere Mesoscale Prediction System International best track archive for climate stewardship (IBTrACS) project, version 4.
(COAMPS) version 5.0: Air/Wave component validation (Washington DC, U.S.A.: [indicate subset used] (NOAA National Centers for Environmental Information).
Naval Research Laboratory), 68. NRL/MR/7322-10-9526. doi: 10.25921/82ty-9e16
Cummings, J. A., and Smedstad, O. M. (2013). “Variational data assimilation for the Komaromi, W. A., Reinecke, P. A., Doyle, J. D., and Moskaitis, J. R. (2021). The naval
global ocean,” in Data assimilation for atmospheric, oceanic and hydrologic applications, research laboratory’s coupled ocean–atmosphere mesoscale prediction system-tropical
cyclone ensemble (COAMPS-TC ensemble). Weather Forecasting 36 (2), 499–517. Saffir, H. S. (1973). Hurricane wind and storm surge Vol. 65 (Virginia, U.S.A: The
doi: 10.1175/WAF-D-20-0038.1 Military Engineer), 4–5.
Komen, G. J., Cavaleri, L., Donelan, M., Hasselmann, K., Hasselmann, S., Janssen, P., Sanabia, E. R., Barrett, B., Black, P. G., Chen, S., and Cummings, J. A. (2013). Real-
et al. (1994). Dynamics and modelling of ocean waves (Cambridge, UK: Cambridge Time upper-ocean temperature observations from aircraft during operational
University Press), 554. doi: 10.1017/CBO9780511628955 hurricane reconnaissance missions: AXBT demonstration project year one results.
Kuik, A. J., van Vledder, G., and Holthuijsen, L. H. (1988). A method for the routine Weather Forecasting 28, 1404–1422. doi: 10.1175/WAF-D-12-00107.1
analysis of pitch-and-roll buoy wave data. J. Phys. Oceanogr. 18, 1020–1034. Sanabia, E. R., and Jayne, S. R. (2020). Ocean observations under two major
doi: 10.1175/1520-0485(1988)018,1020:AMFTRA.2.0.CO;2 hurricanes: Evolution of the response across the storm wakes. AGU Adv. 1,
Lionello, P., Gunther, H., and Janssen, P. A. (1992). Assimilation of altimeter data in e2019AV000161. doi: 10.1029/2019AV000161
a global third-generation wave model. J. Geophys. Res. Ocean 97 (C9), 14453–14474. Saulter, A. N., Bunney, C., King, R. R., and Waters, J. (2020). An application of
doi: 10.1029/92jc01055 NEMOVAR for regional wave model data assimilation. Front. Mar. Sci. 7. doi: 10.3389/
Longuet-Higgins, M. S., Cartwright, D. E., and Smith, N. D. (1963). “Observations of the fmars.2020.579834
directional spectrum of sea waves using the motions of a floating buoy,” in Ocean wave Schuler, D. L., Lee, J. S., Kasilingam, D., and Pottier, E. (2004). Measurement of ocean
spectra (New Jersey, U.S.A.: Prentice-Hall), 111–136. Available at: http://www.jstor.org/ surface slopes and wave spectra using polarimetric SAR image data. Remote Sens.
stable/2414167. Environ. 91, 198–211. doi: 10.1016/j.rse.2004.03.008
Lygre, A., and Krogstad, H. E. (1986). Maximum entropy estimation of the Seemanth, M., Remya, P. G., Bhowmick, S. A., Sharma, R., Nair, T. B., Kumar, R.,
directional distribution in ocean wave spectra. J. Phys. Oceanography 16, 2052–2060. et al. (2021). Implementation of altimeter data assimilation on a regional wave
doi: 10.1175/1520-0485(1986)016<2052:MEEOTD>2.0.CO;2 forecasting system and its impact on wave and swell surge forecast in the Indian
Marchok, T. (2021). Important factors in the tracking of tropical cyclones in Ocean. Ocean Eng. 237, 109585. doi: 10.1016/j.oceaneng.2021.109585
operational models. J. Appl. Meteorology Climatology 60, 1265–1284. doi: 10.1175/ Simpson, R. H. (1974). The hurricane disaster—Potential scale. Weatherwise 27,
JAMC-D-20-0175.1 169–186. doi: 10.1080/00431672.1974.9931702
Martin, P. J., Edwards, K. L., Veeramony, J., Blain, C. A., and Campbell, T. J. (2020). Smit, P. B., Houghton, I. A., Jordanova, K., Portwood, T., Shapiro, E., Clark, D., et al.
VTR for implementation of wave-ocean coupling in COAMPS in both the nearshore and (2021). Assimilation of significant wave height from distributed ocean wave sensors.
offshore ocean. (Washington DC, U.S.A.: Naval Research Laboratory), 67. NRL Ocean Model. 159, 101738. doi: 10.1016/j.ocemod.2020.101738
Memorandum Report NRL/MR/7322–20-10176. Smith, T. A., Chen, S., Campbell, T., Martin, P., Rogers, W. E., Gaberšek, S., et al.
Martin, P. J., Rogers, E., Allard, R. A., Dykes, J. D., and Hogan, P. J. (2013). Tests of (2013). Ocean-wave coupled modeling in COAMPS-TC: A study of hurricane ivan,
parameterized langmuir-circulation mixing in the ocean’s surface mixed layer. (2004). Ocean Modelling 69, 181–194. doi: 10.1016/j.ocemod.2013.06.003
(Washington DC, U.S.A.: Naval Research Laboratory), 47. NRL Memorandum
Theurich, G., DeLuca, C., Campbell, T., Liu, F., Saint, K., Vertenstein, M., et al. (2016). The
Report NRL/MR/7320–13-9444, Naval Research Laboratory, SSC, MS 39529.
earth system prediction suite: Toward a coordinated US modeling capability. Bull. Am.
Metzger, E. J. H. E., Hurlburt, X., Xu, J. F., Shriver, A. L., Gordon, J., Sprintall, R. D., Meteorological Soc. 97, 1229–1247. doi: 10.1175/BAMS-D-14-00164.1
et al. (2010). Simulated and observed circulation in the Indonesian Seas: 1/12 global
HYCOM and the INSTANT observations. Dyn. Atmos. Ocean. 50, 275-300. doi: Thomson, J., Moulton, M., de Klerk, A., Talbert, J., Guerra, M., Kastner, S., et al.
10.1016/j.dynatmoce.2010.04.002 (2019). “A new version of the SWIFT platform for waves, currents, and turbulence in
the ocean surface layer,” in 2019 IEEE/OES twelfth current, waves and turbulence
National Hurricane Center. (2017). Costliest U.S. tropical cyclone tables update. measurement (CWTM) (New Jersey, U.S.A.: IEEE), 1–7. doi: 10.1109/
Available at: https://www.nhc.noaa.gov/news/UpdatedCostliest.pdf. CWTM43797.2019.8955299
NDBC. (2009). Handbook of automated data quality control. Available at: https://www. Tran, N., Vandemark, D., Labroue, S., Feng, H., Chapron, B., Tolman, H. L., et al.
ndbc.noaa.gov/publications/NDBCHandbookofAutomatedDataQualityControl2009.pdf. (2010). Sea state biasin altimeter sea level estimates determined by combining wave
Nichols, C. R., and Wright, L. D. (2020). The evolution and outcomes of a model and satellite data. J. Geophys. Res. 115, C03020. doi: 10.1029/2009JC005534
collaborative testbed for predicting coastal threats. J. Mar. Sci. Eng. 8, 612. Valle-Levinson, A., Olabarrieta, M., and Heilman, L. (2020). Compound flooding in
doi: 10.3390/jmse8080612 Houston-Galveston Bay during Hurricane Harvey. Sci. Total Environ. 747, 141272.
Nowotarski, C. J., Spotts, J., Edwards, R., Overpeck, S., and Woodall, G. R. (2021). doi: 10.1016/j.scitotenv.2020.141272
Tornadoes in hurricane harvey. Weather Forecasting 36, 1589–1609. doi: 10.1175/ Veeramony, J., Walker, D., and Hsu, L. (2010). A variational data assimilation system
WAF-D-20-0196.1 for nearshore applications of SWAN. Ocean Model. 35, 206–214. doi: 10.1016/
Oltman-Shay, J., and Guza, R. T. (1984). A data-adaptive ocean wave directional- j.ocemod.2010.07.008
spectrum estimator for pitch and roll type measurements. J. Phys. Oceanography 14, Voorrips, A. C., Makin, V. K., and Hasselmann, S. (1997). Assimilation of wave
1800–1810. doi: 10.1175/1520-0485(1984)014<1800:ADAOWD>2.0.CO;2 spectra from pitch-and-roll buoys in a north sea wave model. J. Geophys. Res. C: Oceans
Orzech, M., Veeramony, J., and Flampouris, S. (2014). Optimizing spectral wave 102 (C3), 5829–5849. doi: 10.1029/96JC03242. ISSN: 01480227
estimates with adjoint-based sensitivity maps. Ocean Dynamics 64, 487–505. Walsh, E. J., Fairall, C. W., and PopStefanija, I. (2021). In the eye of the storm. J. Phys.
doi: 10.1007/s10236-014-0700-2 Oceanography 51, 1835–1842. doi: 10.1175/JPO-D-20-0219.1
Potter, H., DiMarco, S. F., and Knap, A. H. (2019). Tropical cyclone heat potential Wittmann, P. A., and Cummings, J. A. (2004). “November. Assimilation of altimeter
and the rapid intensification of Hurricane Harvey in the Texas Bight. J. Geophysical wave measurements into WAVEWATCH III,” in 8th international workshop on wave
Research: Oceans 124, 2440–2451. doi: 10.1029/2018JC014776 hindcasting and forecasting(Virginia, U.S.A.: U.S. Defense Technical Information
Riley, R., Hall, C., Stewart, R., DiNapoli, S., and Wang, D. W. (2019). “NDBC OWL wave Center), 14–19. Available at: https://apps.dtic.mil/sti/pdfs/ADA436548.pdf.
system development,” in 2019 IEEE/OES twelfth current, waves and turbulence measurement Wurman, J., and Kosiba, K. (2018). The role of small-scale vortices in enhancing
(CWTM) (Virginia, U.S.A.: IEEE), 1–6. doi: 10.1109/CWTM43797.2019.8955248 surface winds and damage in Hurricane Harvey, (2017). Monthly Weather Rev. 146,
713–722. doi: 10.1175/MWR-D-17-0327.1
Romero-Arteaga, A., Ruiz de Alegrı́a-Arzaburu, A., and Esquivel-Trava, B. (2022).
Spatial variability of surface waves and nearshore currents induced by hurricane harvey WW3DG, T.W.I.D.G (2019). User manual and system documentation of Wavewatch III
along the southern texas coast. J. Mar. Sci. Eng. 10, 1722. doi: 10.3390/jmse10111722 version 6.07. Available at: https://polar.ncep.noaa.gov/mmab/papers/tn276/MMAB_276.pdf.