Predicting the Future of Sediment: A Parametric Regression Approach to Harbor Shoaling

Accurate prediction of shoaling rates and dredge volumes is critical for optimizing the USACE’s navigation channel maintenance strategies and controlling dredging costs. Existing methodologies—such as traditional linear regression and advanced tools like the Corps Shoaling Analysis Tool (CSAT)—are constrained by limited survey data sets used as inputs, by their tendency to underestimate complex sedimentation processes, and by their inability to integrate long- and short-term shoaling trends. Emerging machine learning approaches, while not limited by the same process-based assumptions, remain constrained primarily by the availability and frequency of survey data sets in most harbors. This study introduces a segmented parametric linear regression methodology that blends long-term (η, 5–10 years) and short-term (κ, 1–3 years) shoaling rates using tunable weighting factors, enabling site-specific calibration of predictive models. The sediment volumes and shoaling rates, and the distinction between short- and long-term trends are determined from the existing survey data for each section of navigation channel. The approach was validated across three harbors—Naval Station Mayport, Palm Beach Harbor, and Tampa Harbor—encompassing diverse geomorphic, hydrodynamic, and data availability conditions. Results demonstrate that the proposed methodology improves predictive accuracy to over 80% in most cases and exceeds 90% for some forecasting windows, outperforming USACE’s current shoal forecasting tool CSAT by a substantial margin in percent error and relative root mean square error. Incorporation of synthetic data, derived through linear interpolation, further enhanced accuracy and extended reliable forecast horizons for data-limited systems. The demonstrated ability to reduce forecast error has significant implications for minimizing instances of dredging contracts being unawardable due to a large discrepancy with the internal government estimate (bid busts), avoiding costly contract modifications (requests for equitable adjustments) due to improper estimation of the level of work to be performed, and optimizing dredging budgets at district, divisional, and national levels.