|dc.description.abstract||Climate change is affecting the shipping industry in Paraná River in Argentina. Lower levels
of precipitation cause the water levels to decrease, which limits the amount of cargo a ship can
load. Accurate predictions of water level in shallow rivers are therefore crucial in reducing the
risk of under- or overestimating the cargo during trades.
This thesis aims to create a predictive model for the maximum allowable draft for vessels to
pass through Paraná River, using historical levels of precipitation and precipitation forecasts.
The maximum allowable draft levels are obtained using reported draft from official AIS data.
In the process, we have studied the seasonal pattern of precipitation and tested different types
of moving averages to calculate the precipitation variable. Further on, we investigated
different statistical methods, including generalized additive models and linear and multiple
regression models. Using the generalized additive model, the precipitation variable explained
~40% of the variance in maximum allowable draft. A time-based cross-validation method was
utilized to predict future maximum draft levels, based upon precipitation forecasts. Finally,
the modelling accuracy using AIS data is compared to the modelling accuracy of using
monthly reported water levels in Paraná River.
The result suggests that the high-frequency AIS reported draft can perform better than monthly
reported water levels, and therefore provide a better estimation basis for pricing of cargoes.
The model provides estimated water level margins 1-12 weeks ahead of time and may be used
by shipping companies to improve cargo estimation and reduce risk.||en_US