9:45 AM - 10:00 AM
[ACG40-04] High frequency time series analysis of extinction coefficient in turbid estuaries with macro tidal ranges by using mooring obserbation
Keywords:extinction coefficient, Ariake Bay, estuary
・Introduction
Variable factors in the ocean extinction coefficient include seawater, dissolved organic matter in the water column, particulate organic matter, and phytoplankton. In estuaries, rivers provide abundant nutrients and the suspended sediments of terrestrial origin supplied, and floating mud due to cohesive response caused by mixing with seawater resulting in turbidity. On the other hand, the water quality including suspended sediments changes due to the tidal asymmetry and lateral advection, as in an environment exposed to high light condition during ebb. Therefore, although turbidity is high in the estuary, turbidity is considered to be affected by tides, and the extinction coefficient is expected to fluctuate in synchrony with fluctuates in turbidity. Since light extinction plays a significant role in the production balance of benthic-pelagic systems, elucidating fluctuates in the light intensity is an important insight for understanding the primary ecosystems in estuaries. However, few studies have focused on light extinction in tidal flats and estuaries, and in-situ studies on long-term light extinction are very rare. In this study, we conducted the in-situ mooring system survey in the Midori River estuary facing the Ariake Bay from October to December 2021, and estimated the extinction coefficient at 10-minute intervals at the research site. Then, the validity of the estimated extinction coefficient is verified, and a prediction model of the extinction coefficient is constructed from the measured water quality data using a random forest model. Finally, the relative importance of explanatory variables in the predictive model is compared, and variable factors cause variation in the extinction coefficient are discussed.
・Materials and methods
Photosynthetically active radiation (PAR), chlorophyll-a (Chl-a) concentration, water depth, water temperature, and turbidity just above the sea bottom were measured at 10-minute intervals from October to December 2021 using a mooring system at two locations in the Midori River estuary (Nakasu and Kamenzu). The atmospheric PAR was measured on the rooftop of Prefectural University of Kumamoto as the surface PAR for calculating the extinction coefficient. Maintenance of the mooring system was performed 10 times during the research period, and the same parameters of the mooring system were measured at the same time using a multi-parameter water quality meter for calibration of the mooring system. A prediction model was constructed using random forest regression with the estimated extinction coefficient as the objective variable and four water quality parameters as explanatory variables: turbidity, Chl-a concentration, water temperature, and depth difference at 10-minute intervals to determine whether the tide was rising or ebb. The IncNodePurity indicates the relative importance of the explanatory variables, was calculated from the predictive model.
・Results and Discussion
Fig 1 shows the calibration results of the extinction coefficient. Since the in-situ extinction coefficient were estimated using two points the atmospheric PAR and the PAR above the sea bottom, the estimated extinction coefficient was necessary to confirm their reliability. Regression analysis was conducted between the extinct coefficient estimated by the mooring system and those obtained from the vertical distribution using the multi-parameter water quality meter, and a high correlation (r2 = 0.89 for the Nakasu and r2 = 0.67 for the Kamesu) was found, indicating that the extinction coefficient estimated by the mooring system is reliable. Fig 2 shows the time series fluctuations of the extinction coefficient without nighttime data. The results support previous studies, as the trend of variation over short time intervals should be more influential than the seasonal trend. The ratio of the minimum to maximum values of the extinction coefficient during a day was more than hundredfold on some days, indicating that long term and high frequent monitoring is essential to accurately determine the fluctuation of the extinction coefficient in the estuary. Fig 3 shows the results of the predicted extinction coefficient by random forest regression. The coefficients of determination exceeded 0.5 at both sites, indicating the predictive model was able to construct a model with high prediction accuracy. IncNodePurity was higher at both sites in the order of turbidity, water temperature, Chl-a concentration, and depth difference, indicating turbidity was the most important factor.
Variable factors in the ocean extinction coefficient include seawater, dissolved organic matter in the water column, particulate organic matter, and phytoplankton. In estuaries, rivers provide abundant nutrients and the suspended sediments of terrestrial origin supplied, and floating mud due to cohesive response caused by mixing with seawater resulting in turbidity. On the other hand, the water quality including suspended sediments changes due to the tidal asymmetry and lateral advection, as in an environment exposed to high light condition during ebb. Therefore, although turbidity is high in the estuary, turbidity is considered to be affected by tides, and the extinction coefficient is expected to fluctuate in synchrony with fluctuates in turbidity. Since light extinction plays a significant role in the production balance of benthic-pelagic systems, elucidating fluctuates in the light intensity is an important insight for understanding the primary ecosystems in estuaries. However, few studies have focused on light extinction in tidal flats and estuaries, and in-situ studies on long-term light extinction are very rare. In this study, we conducted the in-situ mooring system survey in the Midori River estuary facing the Ariake Bay from October to December 2021, and estimated the extinction coefficient at 10-minute intervals at the research site. Then, the validity of the estimated extinction coefficient is verified, and a prediction model of the extinction coefficient is constructed from the measured water quality data using a random forest model. Finally, the relative importance of explanatory variables in the predictive model is compared, and variable factors cause variation in the extinction coefficient are discussed.
・Materials and methods
Photosynthetically active radiation (PAR), chlorophyll-a (Chl-a) concentration, water depth, water temperature, and turbidity just above the sea bottom were measured at 10-minute intervals from October to December 2021 using a mooring system at two locations in the Midori River estuary (Nakasu and Kamenzu). The atmospheric PAR was measured on the rooftop of Prefectural University of Kumamoto as the surface PAR for calculating the extinction coefficient. Maintenance of the mooring system was performed 10 times during the research period, and the same parameters of the mooring system were measured at the same time using a multi-parameter water quality meter for calibration of the mooring system. A prediction model was constructed using random forest regression with the estimated extinction coefficient as the objective variable and four water quality parameters as explanatory variables: turbidity, Chl-a concentration, water temperature, and depth difference at 10-minute intervals to determine whether the tide was rising or ebb. The IncNodePurity indicates the relative importance of the explanatory variables, was calculated from the predictive model.
・Results and Discussion
Fig 1 shows the calibration results of the extinction coefficient. Since the in-situ extinction coefficient were estimated using two points the atmospheric PAR and the PAR above the sea bottom, the estimated extinction coefficient was necessary to confirm their reliability. Regression analysis was conducted between the extinct coefficient estimated by the mooring system and those obtained from the vertical distribution using the multi-parameter water quality meter, and a high correlation (r2 = 0.89 for the Nakasu and r2 = 0.67 for the Kamesu) was found, indicating that the extinction coefficient estimated by the mooring system is reliable. Fig 2 shows the time series fluctuations of the extinction coefficient without nighttime data. The results support previous studies, as the trend of variation over short time intervals should be more influential than the seasonal trend. The ratio of the minimum to maximum values of the extinction coefficient during a day was more than hundredfold on some days, indicating that long term and high frequent monitoring is essential to accurately determine the fluctuation of the extinction coefficient in the estuary. Fig 3 shows the results of the predicted extinction coefficient by random forest regression. The coefficients of determination exceeded 0.5 at both sites, indicating the predictive model was able to construct a model with high prediction accuracy. IncNodePurity was higher at both sites in the order of turbidity, water temperature, Chl-a concentration, and depth difference, indicating turbidity was the most important factor.