2.2.3 Lagrangian experiments
To identify transport routes and confluence zones in the study region,
we conducted several numerical experiments using a particle advection
model developed by the Ocean-Atmosphere Interaction working group at the
Institute of Atmospheric Sciences and Climate Change of UNAM. This model
has already been used to successfully model connectivity between coral
reefs in the southern GoM (Sanvicente-Añorve et al., 2018, 2019). The
Lagrangian model is implemented in MATLAB and operates offline with
archived ocean velocity data from HYCOM, integrating with a second-order
Runge-Kutta method. To simulate horizontal diffusion, we applied a
random walk scheme (Fabbroni, 2009; Lynch et al., 2015; Majda & Kramer,
1999; Okubo & Levin, 2001), determining a diffusion coefficient of 13,
by doing sensitivity tests with values in the range of 0 to 14
m2s-1 (Döös et al., 2011; Döös &
Jönsson, 2013; LaCasce & Ohlmann, 2003; Lara-Hernández et al., 2019).
To calibrate the model, we used 21 GDP surface drifters released between
2017 and 2018 and optimized the diffusion coefficient to minimize the
daily distance between the drifters and the corresponding numerical
particles’ positions after a 45-day advection. After 45 days, we
calculated the daily distance between the drifter and the synthetic
drifter positions. The optimal value of the diffusion coefficient
resulted from finding the minimum distance between the numerical
particles and the drifters in more than 50% of the particles. After
calibrating the model, we conducted Lagrangian experiments by releasing
1,866,600 numerical particles (100 particles for each initial trajectory
position) during a climatological year, every third day, starting
January 1st. The release polygon covered the Central
West Atlantic, between -45oW and
-29oW longitude and 0o and
8oN latitude, and consisted of a 17 x 9 node grid with
a spatial resolution of 1o. The Lagrangian model was
integrated forward in time during a climatological year comprised of 365
days, starting when the particles were released. To complete a year of
Lagrangian simulations, the field of climatological velocities was
repeated consecutively. We used the HYCOM climatological daily data and
the spatial patterns obtained from the SOMs analysis to drive the
particle model with currents. The SOMs patterns were chronologically
ordered according to the Best Match Units (BMUs), thus identifying the
corresponding pattern for each day of the year. With this approach, we
will demonstrate that the patterns obtained from SOMs are comparable, in
Lagrangian terms, to the HYCOM data climatology. Besides forcing the
Lagrangian experiments with currents from the HyCOM climatology and the
SOMs patterns, we also forced them with these currents plus 1% or 2%
of wind from the CFSR climatological daily data. To analyze the
relationship between particle advection and the cLCS, we conducted a
case study on the east coast of the YP. For this area, we delimited an
irregular polygon to identify particles approaching the coastal region
and calculated the yearly and monthly accumulated sum of particles.