Inverse modeling of turbidity currents using artificial neural network: verification for field application

Abstract

Abstract. Although in situ measurements observed on modern frequently occurring turbidity currents have been performed, the flow characteristics of turbidity currents that occur only once every hundreds of years and deposit turbidites over a large area have not yet been elucidated. In this study, we propose a method for estimating the paleo-hydraulic conditions of turbidity currents from ancient turbidites by using machine learning. In this method, we hypothesize that turbidity currents result from suspended sediment clouds that flow down a steep slope in a submarine canyon and into a gently sloping basin plain. Using inverse modeling, we reconstruct seven model input parameters including the initial flow depth, the sediment concentration and the basin slope. Repeated numerical simulation using one-dimensional shallow water equations under various input parameters generates a dataset of the characteristic features of turbidites. This artificial dataset is then used for supervised training of a deep learning neural network (NN) to produce an inverse model capable of estimating paleo-hydraulic conditions from data of the ancient turbidites. Only 3,500 datasets are needed to train this inverse model. The performance of the inverse model is tested using independently generated datasets. Consequently, the NN successfully reconstructs the flow conditions of the test datasets. In addition, the proposed inverse model is quite robust to random errors in the input data. Judging from the results of subsampling tests, inversion of turbidity currents can be conducted if an individual turbidite can be correlated over 10 km at approximately 1 km intervals. These results suggest that the proposed method can sufficiently analyze field-scale turbidity currents.

Inverse Analysis of Experimental Scale Turbidity Currents Using Deep Learning Neural Networks

“…In the case of turbidity currents, however, it is impossible to obtain sufficient data sets of in‐situ measurements of flow characteristics for developing a DNN inverse model. Instead of using in‐situ measurements of turbidity currents in nature, Naruse and Nakao (2020) generated numerical data sets of turbidites using a forward model. The generated data sets were input into a DNN model to explore the functional relationship between turbidites and initial flow conditions.…”

“…After this network training process, the DNN model can estimate flow conditions from new turbidite data. Naruse and Nakao (2020) performed inverse analysis using a trained DNN model on field scale numerical test data sets generated by a forward model. Their results showed that the DNN model can reconstruct flow properties from numerical test data sets and was robust against noise in input data.…”

“…CAI AND NARUSE 10.1029/2021JF006276 2 of 32 Here, we implemented a forward model and a DNN inverse model. The forward model was implemented with the same governing Equations as Naruse and Nakao (2020), but the numerical scheme and closure Equations were modified to accommodate experimental scale simulations and improve the accuracy of the calculation. The DNN model was trained with the experimental scale numerical data sets.…”

Inverse Analysis of Experimental Scale Turbidity Currents Using Deep Learning Neural Networks

“…Since previous methods to estimate flow conditions for turbidites were either overly simplified (Baas et al, 2000), incapable of reproducing graded beds (Falcini et al, 2009), accurate but computationally intractable for natural scale turbidity currents (Lesshafft et al, 2011), or low in accuracy (Parkinson et al, 2017), a method that is both accurate and not computationally intractable should be developed. To resolve the aforementioned issues, Naruse and Nakao (2020) proposed a new method for inverse analysis of turbidite deposits using deep learning neural networks (DNN). A DNN model is a machine-learning computing system that works as a universal function approximator (Liang & Srikant, 2016), meaning that an unknown function governing the relationship between observations within a domain is explored and approximated.…”

“…In the case of turbidity currents, however, it is impossible to obtain sufficient data sets of in-situ measurements of flow characteristics for developing a DNN inverse model. Instead of using in-situ measurements of turbidity currents in nature, Naruse and Nakao (2020) generated numerical data sets of turbidites using a forward model. The generated data sets were input into a DNN model to explore the functional relationship between turbidites and initial flow conditions.…”