Open AccessFEATURE SELECTION OF VARIOUS LAND COVER INDICES FOR MONITORING SURFACE HEAT ISLAND IN TEHRAN CITY USING LANDSAT 8 IMAGERY
Author(s) -
Nikrouz Mostofi,
Mahdi Hasanlou
Publication year - 2017
Publication title -
journal of environmental engineering and landscape management
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.514
H-Index - 28
eISSN - 1822-4199
pISSN - 1648-6897
DOI - 10.3846/16486897.2016.1223084
Subject(s) - urban heat island , land cover , support vector machine , urbanization , normalized difference vegetation index , feature selection , vegetation (pathology) , consistency (knowledge bases) , vegetation cover , selection (genetic algorithm) , geography , kernel (algebra) , linear regression , physical geography , feature (linguistics) , regression , remote sensing , environmental science , land use , computer science , climate change , meteorology , statistics , mathematics , geology , machine learning , artificial intelligence , ecology , medicine , oceanography , pathology , biology , linguistics , philosophy , combinatorics
Recently, scientists have been taking a great interest in Global warming issue, since the global surface temperature has been significantly increased all through last century. The surface heat island (SHI) refers to an urban area that has higher surface temperatures than its surrounding rural areas due to urbanization. In this paper, Tehran city is used as case study area. This paper tries to employ a quantitative approach to explore the relationship between land surface temperature and the most widespread land cover indices, and select proper (urban and vegetation) indices by incorporating supervised feature selection procedures using Landsat 8 imageries. In this regards, genetic algorithm is incorporated to choose best indices by employing kernel base one, support vector regression and linear regression methods. The proposed method revealed that there is a high degree of consistency between affected information and SHI dataset (RMSE = 0.9324, NRMSE = 0.2695 and R2 = 0.9315).