Processing of Remote Sensing Ocean Parameter Using Downscaling and Machine Learning Techniques

Ocean Data monitoring and prediction is widely studied using various techniques. The remote sensing data is available for parameters like Sea Surface Temperature (SST), Chlorophyll, and Salinity etc. The goal is to utilize the data and train the system, so that it will be useful in predicting the future data. The paper aims at processing the remote sensing information primarily focused on SST and Chlorophyll parameters using the Downscaling technique and Random Forest methodology. The mean value of the Spatial Distribution is calculated using the multivariate regression model, whose input is the course and the fine resolution data. The temperature data is gotten from INCOIS (Indian National Centre for Ocean Information Services) Site using AVHRR Sensor (Advanced Very High Resolution Radiometer). OCx algorithm is used to process the Chlorophyll data and the images are pre-processed using Gnomonic Projection. We process the pixel data using the prediction model and the outcome is measured in terms of Accuracy and AUC (Area under the Curve) of ROC Curve. The prediction model is compared with Nearest Neighbor (kNN) and Logistic Regression (LR), via the standard parameters Precision, Recall and Accuracy wherein the accuracy of our model stands at 0.943 which is significantly better than the other two (kNN and LR).