Estimation of Chlorophyll-a From Oceanographic Properties - An Indirect Approach.

Remote sensing has been widely used to determine marine chlorophyll by the property that Chl-a reflects electromagnetic radiations for certain wavelengths. Similarly, the same property can be used to determine other oceanographic properties as well such as nitrates, phosphates, iron concentration, etc. Now, it is not always possible to create Chl-a sensitive wavelengths because of hardware limitations and therefore there always exists a need to estimate Chl-a based on other oceanographic properties. For this purpose, supervised machine learning-based regression techniques can be utilized which can be used to train the model to predict marine chlorophyll based on other oceanographic properties. This also shows the dependencies of Chl-a with these oceanographic features. The experiment have been conducted on data obtained from Marine Copernicus Hindcast program where the data have been converted into time series, used different preprocessing techniques and applied regression algorithms. The experiment has obtained an R2 score of up to 0.904. The model can be used to remotely monitor Chl-a concentration in ocean based on other oceanographic properties like nitrates, phosphates, iron concentration, etc.