5:15 PM - 7:15 PM
[BCG06-P06] Hydroclimatic changes during the Early Jurassic oceanic anoxia: analysis of terrestrial palynomorph using machine learning
Keywords:Toarcian, Toyora Group, carbon isotope, machine learning, object detection
The Toarcian oceanic anoxic event (T-OAE; ~183 Ma) involved a major disturbance in the global carbon cycle, depleting oxygen in oceans and resulting in a mass extinction among marine organisms. An increase in storm deposits during the T-OAE is generally interpreted as the result of perturbations in the global hydrological and carbon cycles. In order to understand their mechanisms, it is also necessary to understand the nature of these storm deposits and estimate their possible origin.
In this study, we conducted a geological survey of the Lower Jurassic Nishinakayama Formation at Sakuraguchidani in Toyota-cho, southwestern Japan (western Panthalassa) and analyzed the sedimentology, organic carbon stable isotopic composition (δ13Corg), and palynomorph abundance throughout the T-OAE. Additionally, machine learning was used in order to reproduce manual microscope observation. In particular, an object detection program was developed to detect palynomorphs in microscope images. The speed and consistency of object detection using machine learning make it an effective method for parallel analyses of T-OAE sections around the world.
Stable carbon isotope analysis identified a negative ~2‰ shift in δ13Corg over a 20 cm interval at the onset of the T-OAE, which is an important chemostratigraphic marker globally. Within this interval, sandstone layers interpreted as storm deposits are present approximately every 8 cm, reflecting increased terrestrial flux throughout the T-OAE. δ13Corg measurements in these sandstone horizons are similar to those in neighboring mudstone layers, suggesting similar responses of terrestrial and marine organic matter to paleoenvironmental changes throughout the T-OAE. Assuming a constant depositional rate throughout this interval, calibrated by U-Pb ages, the isotope shift is estimated to have occurred within ~1,300 years. Similarly, sandstone layers were deposited at an estimated interval of every 500 years.
Palynomorph quantification using machine learning revealed an enhancement of terrestrial organic matter influx to the ocean during the T-OAE main phase in western Panthalassa relative to corresponding intervals at Tethyan T-OAE sections. A relatively high diversity in palynomorph particle size was observed at the same interval, suggesting their deposition resulting from increased storm activity.
In this study, we conducted a geological survey of the Lower Jurassic Nishinakayama Formation at Sakuraguchidani in Toyota-cho, southwestern Japan (western Panthalassa) and analyzed the sedimentology, organic carbon stable isotopic composition (δ13Corg), and palynomorph abundance throughout the T-OAE. Additionally, machine learning was used in order to reproduce manual microscope observation. In particular, an object detection program was developed to detect palynomorphs in microscope images. The speed and consistency of object detection using machine learning make it an effective method for parallel analyses of T-OAE sections around the world.
Stable carbon isotope analysis identified a negative ~2‰ shift in δ13Corg over a 20 cm interval at the onset of the T-OAE, which is an important chemostratigraphic marker globally. Within this interval, sandstone layers interpreted as storm deposits are present approximately every 8 cm, reflecting increased terrestrial flux throughout the T-OAE. δ13Corg measurements in these sandstone horizons are similar to those in neighboring mudstone layers, suggesting similar responses of terrestrial and marine organic matter to paleoenvironmental changes throughout the T-OAE. Assuming a constant depositional rate throughout this interval, calibrated by U-Pb ages, the isotope shift is estimated to have occurred within ~1,300 years. Similarly, sandstone layers were deposited at an estimated interval of every 500 years.
Palynomorph quantification using machine learning revealed an enhancement of terrestrial organic matter influx to the ocean during the T-OAE main phase in western Panthalassa relative to corresponding intervals at Tethyan T-OAE sections. A relatively high diversity in palynomorph particle size was observed at the same interval, suggesting their deposition resulting from increased storm activity.