Evaluating machine learning methods and satellite images to estimate combined climatic indices

The reflections recorded on satellite images have been affected by various environmental factors. In these images, some of these factors are combined with other environmental factors that cannot be distinguished. Therefore, it seems wise to model these environmental phenomena in the form of hybrid indicators. In this regard, satellite imagery and machine learning methods can play a unique role in modeling and data mining of climatic phenomena as a result of their significant advantages, including their availability and analysis. Therefore, addressing the improvement and expansion of machine learning methods and modeling algorithms using remote sensing data is inevitable. In this study, 7 well-known machine learning algorithms are applied with different initial data to show that satellite images are able to estimate the combined indices more accurately. A new index (HT) is also defined by combining the quantities of relative humidity and temperature. Then, machine learning algorithms are trained for each of these three quantities. For each of the temperature and relative humidity quantities, four optimal bands were selected using the PCA method, then a combination of these optimal bands was determined for the HT index. Two criteria are used to validate the results: Root Mean Square Error (RMSE) statistic and comparing the map of the interpolation method with the result of this study. RMSE values show that satellite imagery could have a high ability to model and estimate composite indices. Classification-KNN-Coarse and Ensemble-Bagged Trees with accuracy of 79.8626 % and 84.9281% are identified as the best machine learning methods for temperature and relative humidity, while the best accuracy to estimate the HT index is 92.8792% for Matern 5/2 GPR. Therefore, it can be said that by changing the methods of database preparation, the modeling results can be changed effectively in order to train models.