Sea Surface Temperature Prediction using Proper Orthogonal Decomposition
As a part of the work submitted to 2020 COMAP's Mathematical Contest in Modeling (MCM) and won the Outstanding Winner award (top 19 out of 13753 teams).
Abstract
Gradual rise of the ocean temperature results in a significant shift in the distribution of herring and mackerel. This situation may cause catastrophe for the small fishing companies in Scotland, for they may not be able to catch sufficient fish in their current location. Our team was asked to identify the most likely distribution of herring and mackerel in the next 50 years, and give practical suggestions to Scotland fishermen based on our study results.
We start by analyzing the sea surface temperature data for the past 39 years and use Proper Orthogonal Decomposition to decompose the original dataset into two parts: the spatial term POD modes and the temporal term time coefficients. After retaining the 3 most significant POD modes, we assume that the POD modes are stable over time and achieve our sea surface temperature prediction through the prediction of their corresponding time coefficients. We proceed to estimate the locations of herring and mackerel over the next 50 years based on their habitat temperature. Our conclusion is that in the Atlantic Ocean north-west to Scotland, herring and mackerel will move northward away from the British Isles over the next 50 years. And in the North Sea, herring will gradually move toward east direction.