Investigation of sea level anomalies along the Indian coastline using tide gauge data: a physics-guided machine learning approach
Abstract
The Indian coastline falls within low-elevation coastal zones that are densely populated.
Recent observations indicate, with high confidence, that the rate of sea level rise in the In-
dian Ocean has been accelerating over the past two decades beyond previous expectations.
However, limited research has focused on the physical factors that influence India’s coastal
sea level rise. Coastal dynamics are distinct from open ocean dynamics due to the influ-
ence of land interactions, shallow depths, and complex interactions between waves, tides
and currents. This study aims to address this gap by developing a model using a physics-
informed machine learning approach to investigate the dominant physical drivers of coastal
sea level variability along the Indian coast. Daily climatology was removed from the sea
level and all other physical variables to obtain anomalies and focus on which anomalies in
the factors drive the sea level anomaly. The model prediction showed a positive correlation
with the original signal, and the results were interpreted using the SHAP (Shapley Additive
exPlanations) algorithm. This analysis identified upper ocean heat and salt content as the
most dominant physical factors determining coastal sea level anomalies along the Indian
coastline. Thermal expansion in the Indian Ocean, driven by global warming, has been
well established in previous studies, clearly manifesting along the Indian coast. In addi-
tion, changes in salinity largely influenced by freshwater influx also play a crucial role.
While the melting of polar ice caps is a well-known contributor to global sea level rise,
our analysis highlights that the retreat of Himalayan glaciers is particularly significant for
the Indian Ocean region. As these glaciers melt, they release large amounts of freshwater
into river systems, which ultimately discharge into the Indian Ocean, altering local salinity
patterns and further contributing to coastal sea level rise. Furthermore, prediction intervals
were estimated for one year of test data using quantile regression techniques, where more
than 70% of the observations fall within these predicted intervals.