The Gulf of Tehuantepec is one of the most productive regions in the northeastern tropical Pacific. However, in-situ studies of chlorophyll-a (Chl-a) during the rainy season remain scarce. Under stratified conditions, surface Chl-a may not accurately represent its vertical distribution, because substantial concentrations can occur at depth. This study examined the horizontal and vertical distribution of Chl-a and its relationships with mixed-layer depth (MLD), thermocline depth (TD), and thermocline strength (TS) in the Gulf of Tehuantepec and adjacent offshore waters during June 2010. Chl-a concentrations and CTD profiles were obtained at continental shelf and offshore stations during an oceanographic survey. Surface Chl-a showed a marked shelf-offshore gradient, with higher concentrations over the continental shelf and oligotrophic conditions offshore. In contrast, Chl-a integrated over the sampled depth range did not differ significantly between shelf and offshore stations. However, their vertical profiles differed significantly, showing that similar integrated Chl-a values can result from different distributions with depth. This finding indicates that surface Chl-a alone provided an incomplete representation of water-column Chl-a under rainy-season conditions. DCMs occurred mainly between 10 and 30 m. All DCMs over the continental shelf occurred within the mixed layer, whereas offshore DCMs showed a more heterogeneous vertical distribution relative to the mixed layer and thermocline. Depth-integrated Chl-a was positively correlated with MLD and TD, whereas DCM depth was not significantly correlated with MLD, TD, or TS. Overall, surface and subsurface Chl-a patterns differed between shelf and offshore waters and were associated with contrasting hydrographic conditions. These findings highlight the importance of vertical observations for characterizing Chl-a distribution during the rainy season.
Marine shallow gas emissions threaten marine engineering and ecosystems, with hydrodynamics playing a key role. However, due to complex environmental interactions, accurate prediction remains challenging. This study presented a proof-of-concept investigation based on a short-term (11-day) in-situ dataset from Zhoushan coastal waters. Two machine learning algorithms-Random Forest (RF), and Long Short-Term Memory neural network (LSTM)-were used to predict the emission process under hydrodynamics. In-situ data-including rate of bottom water temperature (rT), significant wave height (SWH), significant wave period (SWP), seafloor pressure (P), acoustic amplitude (AMP), residual flow velocity (V_res_std), vertical velocity fluctuation (w_std), direction angle (θ), and dissolved methane concentration (DMC) were used. Model performance was evaluated using R2 and error metrics including mean absolute error (MAE), mean squared error (MSE), root mean squared error (RMSE), and mean absolute percentage error (MAPE), combined with radar charts and Taylor diagrams. Results showed that the LSTM model incorporating SWP and V_res_std achieved the most balanced performance (R2 = 0.847, σ_ratio = 0.837), while the autoregressive LSTM yielded the lowest RMSE (10.565 nmol/L) but exhibited notable bias in variability reproduction (σ_ratio = 0.765). The RF model with three lagged methane features demonstrated competitive correlation (R = 0.932) and captured general temporal trends reliably, yet consistently underestimated peak concentrations (σ_ratio≈0.72). Feature importance analysis identified recent methane concentration (DMC_lag_1) as the dominant predictor (importance>0.95). Excluding temporal autocorrelation, SWP and θ emerged as the most influential hydrodynamic drivers, followed by w_std and V_res_std, while SWH, AMP, P, and rT contributed minimally. SHAP analysis further revealed that higher direction values promoted methane predictions, SWP exhibited bidirectional effects, lower V_res_std promoted predictions, higher AMP inhibited predictions, and positive temperature change rate promoted predictions while negative values inhibited them. Sliding-window cross-correlation analysis quantified that enhanced wave activity and vertical mixing (SWH, SWP, w_std) exerted negative driving effects through dilution (mean lags: 55, 36, and 51 min). In contrast, V_res_std exhibited a positive influence, likely through enhanced pore-water advection, with a mean lag of 46 min. Given the limited observational period, this work is intended as a methodological demonstration rather than a generalizable or high-precision forecasting tool. The findings highlight the potential and current limitations of applying machine learning to short-term shallow gas emission prediction.
The surface turbulent transport processes in the central Baltic Sea during summer are investigated using Lagrangian trajectory simulations forced by an operational ocean model and Finite Size Lyapunov Exponent (FSLE) and related diagnostics of horizontal stirring. The spatial distribution of Lagrangian Coherent Structures (LCS) and the local prevalence of attracting versus repelling LCS are presented. Coastal areas are characterised by strong stirring in contrast to the weakly stirred offshore regions. The along-trajectory absolute value of relative vorticity and total strain attest to the important role of eddies, fronts and jets in driving horizontal stirring in the ocean model. Strong asymmetry in LCS distribution between the eastern and western coast of Gotland is found and its consequences on the transport of algae blooms and pollutants are discussed.
Trophic function of any ecosystem is primarily governed by predator-prey interactions that influence the dynamics and the fate of communities. This study combines daily observations of fishery catch and gut-content analysis collected over three years and two months with an Ecopath model to characterise the trophic structure and functioning of Caranzalem Bay, a tropical estuarine bay in Goa, India. The empirical observations revealed substantial temporal variability in species composition and biomass, particularly among lower and mid-trophic groups. A balanced 44-functional-group Ecopath model was used to quantify ecosystem structure, energy flow, and trophic interactions, with relatively low internal cycling (Finn's Cycling Index, 4.37%). Primary producers provided the larger contribution to energy transfer, while detrital pathways increased overall transfer efficiency. Small-sized predators including elasmobranchs and catfishes consumed benthic prey, while large predators of the same groups preferred pelagic prey. Transient forage fish and megabenthic species showed large temporal fluctuations, with size-dependent prey preferences among predators. Notably, cannibalism, ontogenetic dietary shift, and intraguild predation added trophic pathways and complexity within the food web. In this shallow bay habitat, benthic, demersal, and pelagic distinctions were diffused owing to trophic linkages across the strata, wherein shrimps and brachyuran crabs showed strong system-wide trophic influence. With access to adjacent habitats and the open-ended nature of the bay, the motile mid- and upper-trophic-level megafauna can commute, resulting in a pulsatile nature of catch and bycatch, with a high degree of stochasticity in predator-prey encounters.
Tides profoundly influence nutrient supply and phytoplankton biomass in semi-enclosed bays by regulating water stratification and mixing; however, how tides regulate these coupled nutrient-production dynamics remain poorly understood due to the lack of high-frequency observations. Here, we conducted continuous measurements of dual water isotopes, hydrological parameters, nutrients, chlorophyll a (Chl-a), and total suspended matter (TSM) at 2-h intervals over a 24-h period during spring and neap tides in summer and winter in Zhanjiang Bay (ZJB), a typical tide-dominated semi-enclosed bay in Southern China. During the summer neap tide, weaker tidal forcing allowed stronger freshwater input (18.7%) and intensified stratification, resulting in the lower Chl-a level despite a higher terrestrial nutrient contribution, which was confined largely within the thin surface layer. During the spring tide, strong currents weakened the stratification (freshwater contribution ∼15.5%), sustaining nutrient supply to the euphotic zone without causing sediment resuspension (TSM remained low). Consequently, Chl-a was high. By contrast, in the winter spring tide, intense mixing caused severe sediment resuspension and high TSM, which, together with tidal dilution, most likely reduced light availability and thus lowered Chl-a. However, winter neap tides featured weak hydrodynamics, low TSM, and favorable light conditions, promoting phytoplankton growth (higher Chl-a level). Thus, whether tidal processes enhance or suppress phytoplankton growth depends on the balance they establish between nutrient supply and light availability. In summer, moderate mixing with weak stratification (spring tide) promotes blooms, while overly strong stratification (neap tide) suppresses them; in winter, intense mixing (spring tide) likely induces light attenuation and low Chl-a, whereas weak mixing (neap tide) favors higher Chl-a. These findings provide isotope-based insights for the ecological management of tide-dominated bays.
