Integration of Sentinel-1 and Sentinel-2 for Classification of Small Urban Areas in Rural Landscape aided by Google Earth Engine

Gordana Kaplan,Majid Aghlmand

doi:10.3390/iecg2020-08542

Abstract

Rapid economic development and population growth lead to fast urban expansion in both urban and rural landscapes. Accurate and updated mapping of urban expansions is curtail in urban and territorial planning for sustainable and strategic urban development. Using Earth Observation (EO) technologies, classification of urban areas in a rural landscape is more challenging than big cities. In this regard, in this paper, we aim at assessing the integration of Sentinel-1 and Sentinel-2 satellite data for classifying small urban areas in rural landscape in Google Earth Engine (GEE). Images of close dates from Sentinel-1 and Sentinel-2 were selected, preprocessed, and integrated to develop a machine learning classification through a Support Vector Classification (SVM) classifier. We have also added vegetation indices to the investigated dataset. As a study area, two rural areas in the Republic of North Macedonia has been selected. The results showed that the integration of Sentinel-1 and Sentinel-2 performed better than Sentinel-2 alone, with accuracy higher than 90%. For future studies, we recommend testing the dataset to different study areas and adding different EO data for obtaining even higher accuracy.

Highlights

Urban expansion has been prompted by the rapid economic development and the significant population growth in the last few decades
This paper investigated the potential of Sentinel-1 and Sentinel-2 for extracting urban areas in rural landscapes within Google Earth Engine (GEE)
The area is mainly used for agriculture, and there are large areas of greenhouses that can be misclassified with urban areas

Summary

Introduction

Urban expansion has been prompted by the rapid economic development and the significant population growth in the last few decades. The population in the urban areas has risen drastically causing environmental concerns Even though this situation is more obvious in the big cities, urban growth is affecting the rural landscapes, causing changes in the land cover [1]. The relatively coarse spatial resolution often cannot meet specific project requirements of urban landuse/landcover classification, especially in a complex urban-rural interface. For this task, for accurate classification, researchers use high-resolution imagery (< 5m). Even though researchers have agreed that extracting urban areas in the rural landscape can be challenging using medium-resolution satellite imagery, taking into consideration the latest developments in the remote sensing field, in this study we use Sentinel imagery integrated into. The Strumica-Radovis valley located in the structural basin of the Strumica River in the Republic of North Macedonia has been selected

Study area

Materials and methods

Findings

Conclusions