Potential Evaluation of SMA and NSMA Methods Implementation to Extract Sub-pixel Impervious Surfaces from Landsat 8 Imagery (OLI), (Case Study: Sari)

Authors

1 PHD student in Tehran university

2 Master of remote sensing and GIS

3 Master of Science (MSc) in Remote Sensing and GIS of Tehran University

4 Assistant Professor of Remote Sensing and GIS University of Tehran

5 Ph.D. Student in Remote Sensing and GIS University of Tehran

Abstract

Impermeable surfaces are one of the integral part of modern urban environments, and in fact new cities are merged with intrusive levels. The increasing in impervious surfaces caused dramatic changes in local and global scale environmental interactions, especially in urban environments, so this phenomenon has increased the importance of impervious surfaces monitoring in recent decades. But also according to the extent and changes of thesesurfaces, monitoring asfield studies is not achievable.However, remote sensing technology provides potential opportunities of identifying and monitoring of urban environment and consequently these surfaces. This article aims to evaluate the ability of implementing of Spectral Mixture Analysis (SMA) and Normalized Spectral Mixture Analysis (NSMA) methods on Landsat 8 satellite images toimpervious surfacesSub-pixel extraction in northern regions of Iran. To achieve this goal, we used OLI images acquired on 27/03/2014 as model inputs and WorldView-2 image acquired on 03/27/2014 as a reference map to evaluate the accuracy results of the model. First, we applied Principal Component Analysis (PCA) to extract Endmember and then SMA and NSMA methods to obtain land cover fraction from any Endmember. Alsowe compared the results of these two methods implementation on the original images and panchromatic combined images. The accuracy of estimated impervious surfaces, evaluated using a 3*3 pixels window random sampling method. The results indicated that the NSMA model combined by panchromatic image band has the best performance to estimate impervious surfaces with an 8.2% Root Mean Square Error (RMSE) and 0.93 coefficient of determination (R2).

Keywords

Main Subjects

Files

Share

How to cite

Statistics