Application of robust regression on sea surface temperature data in the indian ocean

Sea Surface Temperature (SST) is the temperature of the water near an ocean's surface. It plays a critical role in the interaction of the Earth's surface and atmosphere. However, not all time series data on SST are complete to affect climate change prediction. To address such issues, the median imputation is used to deal with the missing data. In data analysis, outlier is unavoidable, and robust statistical methods are required. The presence of outlier which is common in the dataset leads to an error in the result. To remedy this problem, the robust regression has been proposed. The missing values of SST in the Indian Ocean got imputed by using the median imputation approach. We then construct the robust regression model using the S-estimator and LTS-estimator.The R-squared (R ) values of the S-estimator and LTS-estimator were 2 0.5670 and 0.6033, respectively. When the data was contaminated with 1%, 2%, 3%, 4% and 5% outliers, the R2 values of the LTS-estimator were 0.5899, 0.5811, 0.5740, 0.5767 and 0.5699. The study's findings revealed that the LTS-estimator is a better model compared to the S-estimator in terms of robustness, as is evidenced by the highest R2 value.