Degree

Doctor of Philosophy (PhD)

Department

Oceanography & Coastal Sciences

Document Type

Dissertation

Abstract

Freshwater diversions alter the timing, magnitude, and pathways of water delivery to receiving estuaries, yet the downstream hydrodynamic and ecological consequences of both operational releases and non-operational leakage are not fully understood. This dissertation examines how upstream basin-scale conditions drive freshwater diversion operations at the Bonnet Carr\'e Spillway (BCS) and alter water transport and water quality vulnerability in the Lake Pontchartrain Estuary (LPE), Louisiana, a shallow, semi-enclosed system subject to recurrent Cyanobacterial Harmful Algal Blooms (cyanoHABs).

A hybrid Convolutional Neural Network--Long Short-Term Memory (CNN--LSTM) forecast model, \texttt{GaugePredict}, was developed to predict daily downstream discharge and water level from basin-wide upstream observations across lead times of one to thirty days. Applied to the Mississippi River Basin, the model reproduced the timing and magnitude of discharge and water level at the BCS with high skill through ten days and moderate skill at fifteen days, with declining performance at longer horizons. One-day lead forecasts were sufficiently accurate to fill short gaps in gauge records caused by sensor outages.

The influence of diversion flows on estuarine transport in LPE was evaluated using exposure time ($E_t$), the cumulative duration a water parcel remains within a defined region including instances of re-entry. Under open diversion conditions, high tributary discharge reduced $E_t$ by more than half relative to low discharge, while closed diversion scenarios showed little sensitivity to tributary variability. Spatially, low discharge produced extensive zones of prolonged retention along the northern and western shores, whereas high discharge promoted flow separation and short-circuiting of diversion water through tidal inlets, indicating that tributary forcing modulates diversion transport pathways.

Spatial $E_t$ estimates from simulations of an operational year (2020) and a non-operational leakage year (2021) were compared with NASA CyAN $\mathrm{CI}_{\mathrm{cyano}}$ composites to evaluate the relationship between retention and cyanoHAB development. Bloom timing and spatial extent were broadly consistent with regions of elevated exposure time in both years. The operational year exhibited rapid early-summer bloom expansion coinciding with diversion activity, while the non-operational year showed a delayed onset and more persistent late-summer bloom period, consistent with bloom expression driven by background retention and reinforced by episodic forcing including tributary inflow and storm-driven circulation changes.

Date

3-22-2026

Committee Chair

Matthew Hiatt

LSU Acknowledgement

1

LSU Accessibility Acknowledgment

1

Download

Available for download on Monday, March 22, 2027

Share

COinS