Temporal analysis of multi-spectral instrument level and surface reflectance data sets for seasonal variation in land cover dynamics by using Google Earth Engine

Abstract

By rapid growth in programming tools, accessibility to end consumer computing power, and the availability of free satellite data, the data science and remote sensing fields have begun to converge in recent years. Before this major processing time is wasted in collection of data. Google Earth Engine easily overcomes above problem; it contains data from different satellites and has power of processing and computation also. Well known data provider satellites are present in the library of GEE and users can easily process and track real time data from these satellites over GEE. “Sentinel”, a mission of the European Space Agency and “Landsat”, an American Earth observation satellite have been used in a variety of remote sensing applications. GEE makes these data sets available to the general public. These datasets are utilised for computing and analysis purposes. The objective of this study is to find change in study area by using above discussed two satellite data, over each season of year on different category of classification (Random Forest, CART, GTB and SVM). This work focuses on improving the classification accuracy of different classification algorithm by reviewing training samples and analyzing post-classification with image differencing in the algebraic technique. Because Landsat data have a medium spatial resolution, therefore point-wise computation was used. Lastly, we also detect which data sets are working better on an appropriate machine learning algorithm, so after final calculation we estimate accuracy of each algorithm by using confusion matrix and kappa.

Keywords

• GEE
• remote sensing
• classification
• Landsat
• Sentinel
• satellite data