Nitrogen is an essential nutrient for primary production in water bodies, however, nitrogen levels higher than their natural levels in coastal waters can lead to a decline in marine water quality. Recent scientific studies suggest that increases in nutrient loads can reduce corals' heat tolerance levels, making them susceptible to bleaching stress. Increasing nutrient (nitrate- NO3) levels can lead to coral reef decline because nutrient enrichment increases coral reef macroalgae productivity and expands the macroalgal densities on coral reefs. Macroalgae production tends to shade the corals, which results in a reduction of water exchange in the corals, leading to chemical disturbances. For instance, the Gulf of Eilat was affected by the strong upwelling of nutrient-rich waters, which resulted in algal blooms that covered the reef in thick filamentous algae and eventually led to the death of the coral. Therefore, accurate prediction and monitoring of nitrogen levels in water bodies is crucial in combating and preventing the detrimental effects of nitrogen pollution on marine ecosystems. Our case study was focused on developing a nitrate prediction model using Sentinel-2 satellite data combined with field data from the Belize Barrier Reef System (BBRS). A machine learning model called Deep Neural Networks (DNN) was implemented to account for the complexity and non-linear relationships between NO3 and other water quality parameters. After the model was trained and assessed for prediction capability, it was validated using an independent test dataset. DNN algorithm achieved a correlation coefficient (R2) of 0.68, mean squared error (MSE) of 0.14, and normalized root mean squared error (NRMSE) of 9.2% for the test set. This study found that DNN shows promise in observing general seasonal patterns and anomalies for NO3 concentrations in the BBRS.
Objective 1: Creating an estimation model for total suspended sediment (TSS) concentration by combining satellite date and in situ measurements.
With a working TSS model, time-series plots are created to understand the spatio-temporal distribution and dynamics of TSS in the BBRS over the past few years.
The model is then used to create maps of suspended sediment distribution in the BBRS.
Objective 2: Creating an estimation model for nitrate levels using satellite data and in situ measurements.
With a working nitrate model, time-series plots are created to understand patterns of nitrate levels in the BBRS over the past few years. This is useful in identifying nutrient-sensitive zones within the BBRS.