Reduced Major Axis Regression

In geosciences, satellite remote sensing data are usually required to validate against ground-based instrumental measurements. However, both datasets depend on instrumental calibration, which may introduce errors in the datasets. Therefore, the use of appropriate regression analysis is required to model the relationship between dependent and independent variables, which can consider errors in both variables. However, in general, the OLS regression method is most commonly used in geosciences to model the relationship between two variables, and it is assumed that the independent variable is measured without error, although this rarely occurs. On the other hand, the RAM method models the relationship between two variables considering errors in both the dependent as well as independent variables. Therefore, RMA regression should be considered more suitable for satellite remote sensing applications. To demonstrate the performance of RMA against OLS, remotely sensed satellite-based aerosol and rainfall data were validated against ground-based measurements. The results showed that the RMA regression is better able to model the relationship between the variables compared to OLS regression in terms of slope, small MBE, RMSD, and MSE. This study recommends that selecting a proper regression method is essential for modeling relationships in geoscience data and that RMA regression is more suitable than OLS.