Wide-ranging effects on the environment, society, and stability of the entire world, sea ice study and how it interacts with climate change is of highest importance. Sea ice in the Arctic and Antarctic is melting rapidly, which shows the urgent requirement to address the underlying their negative effects on environment and society. Using information from the NSIDC DAAC collection, inspection done using a Deep Learning-CNN. This dataset contains a number of measures include sea ice concentration, extent, sea surface temperature, and others, provided in daily, weekly, and monthly formats. Deep Learning-CNN employs features as input alongside a on its own layer of hidden nodes to estimate parameters like temperature, extent, and concentration of sea ice. Deep Learning-CNN algorithm produces Ice temperature, ice extent, and ice concentration that is less noisy with better that from image analysis. Each stage in a ConvNets is composed of various filter banks with fundamental activation algorithms such as RELU Utilizing multiple stages, a ConvNet has the capacity to acquire hierarchical features across various levels. The Deep Learning analysis extends to sea ice concentration, extent, and temperatures, compared with estimates derived from the data employing the ASI algorithm. By using CNN, Mean Error Rate is minimum − 0.003215, − 0.01629 respectively, but RMSE is reached to 0.12618 and 0.17713 for the training and testing datasets Here, RMSE is minimum, when compare to other algorithms like ASI or MLP for both training and testing dataset. The findings illustrate the influence of altering modifying the levels in accordance with the supplied patch size.