AdaBoost-based satellite detection of summer daytime sea fog and low clouds in ice floe fields of the Arctic

Abstract

Under the trend of global warming, the Arctic region has experienced increasing warming and accelerating melting of sea ice, resulting in the opening of Arctic shipping routes with significant navigational potential. However, intense water vapor releases and a cold underlying surface can lead to frequent occurrences of sea fog and low clouds over the Arctic ice surfaces. Conducting detection of sea fog and low clouds is therefore highly important for ensuring the safety of Arctic shipping routes. In this study, a satellite detection algorithm for summer daytime sea fog and low clouds in ice floe fields of the Arctic has been proposed using the time series remote sensing data from the Moderate-Resolution Imaging Spectroradiometer (MODIS) and the AdaBoost method. Methodologically, three spectral indices, the normalized difference sea-water index (NDSWI), the sea-ice recognition index (SIRI), and the mid-high-cloud recognition index (MHCRI), are first constructed by analyzing the variations in reflection spectra for sea fog and low clouds, sea ice, sea water, and mid-high clouds. Additionally, three texture features, namely, the homogeneity, contrast, and entropy, of the brightness temperature at 11.030 μm (BT_{11.030 μm}) are calculated using the gray-level co-occurrence matrix (GLCM). Subsequently, a strong learner classification model of the AdaBoost ensemble learning algorithm was built by adopting the samples of the spectral indices and texture features above and the weak learner of the decision. Finally, the residual mid-high clouds are removed through the threshold at BT_{11.030 μm}. Verification indicated that the probability of detection (POD), false alarm rate (FAR), and critical success index (CSI) values were 84.44%, 9.45%, and 77.52%, respectively. This research supports the accurate detection of sea fog and low clouds in the Arctic, thereby ensuring safe navigation of Arctic shipping routes.