SBIR Phase I: Near Shore Bathymetry and Coastline Modeling in the Arctic

Polarctic Llc, Stafford VA

Investigators

Abstract

The broader impact of this SBIR Phase I project will be leveraging the power of Artificial Intelligence and Machine Learning (AI/ML) to better understand how to construct models for complicated systems with incomplete information, such as the coastline and nearshore of the Arctic. Natural beach processes are not stagnant shorelines but rather dynamic systems that shift, migrate, fill and erode. Accurate seafloor depths and nearshore charts are important for forecasting weather, tsunami, and storm surge events that can impact local communities, infrastructure, and shipping industries. The necessity of safe navigation for civilian ships not rated for ice operations and dynamic geographical needs is driving the need for innovative modeling. Traditional mapping techniques are incapable of updating constantly evolving bathymetric navigation models quickly and accurately to suit the needs of business and communities. AI/ML technology has much to offer and investigate in terms of different architectures, applications and possible modifications. Some of the largest roadblocks to implementing current AI/ML modeling are that they require data, cannot directly include current scientific understandings of systems, and are considered a black box for understanding the way the system is solved. The proposed work will teach AI/ML models to learn from more than just data but also scientific theories, and then to write out solutions that can be interpreted by humans This SBIR project will develop a method to integrate data and models in a dynamic learning environment. Systems using AI/ML have the capability to identify critical patterns in nonlinear, complex systems when good training data is available. However, this process does not create new insights on system dynamics. Beach physics and coastal morphology are actively researched today and have led to high quality traditional coastal modeling codebases, both phenomenologically generated and physics-based, but these are not easily updated. This project will design an architecture using analytical functions and data to develop solutions as a codebase, which can then be reviewed and compiled as a traditional coastal model, allowing for novel AI/ML methods without requiring the user to learn new readable formats. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.