Nikrouz Mostofi; Hossein Aghamohammadi Zanjiirabad; Alireza Vafaeinezhad; Mahdi Ramezani; Amir Houman Hemmasi
Abstract
Introduction Surface temperature is considered to be a substantial factor in urban climatology. Italso affects internal air temperature of buildings, energy exchange, and consequently the comfort of city life. An Urban heat island (UHI) is an urban area with a significantly higher air temperature ...
Read More
Introduction Surface temperature is considered to be a substantial factor in urban climatology. Italso affects internal air temperature of buildings, energy exchange, and consequently the comfort of city life. An Urban heat island (UHI) is an urban area with a significantly higher air temperature than its surrounding rural areas due to urbanization. Annual average air temperature of an urban area with a populationof almost one million can be one to three degreeshigher than its surrounding rural areas. This phenomenon can affect societies by increasing costs of air conditioning, air pollution, heat-related illnesses, greenhouse gas emissions and decreasing water quality. Today, more than fifty percent of the world’s population live in cities, and thus, urbanization has become a key factor in global warming. Tehran, the capital of Iran and one of the world’smegacities, is selected as the case study area of the present research. A megacity is usually defined as a residential area with a total population of more than ten million. We encountered significant surface heat island (SHI) effect in this area due to rapid urbanization progress and the fact that twenty percent of population in Iran are currently living in Tehran.SHI has been usually monitored and measured by in situ observations acquired from thermometer networks. Recently, observing and monitoring SHIs using thermal remote sensing technology and satellite datahave become possible. Satellite thermal imageries, especially those witha higher resolution, have the advantage of providing a repeatable dense grid of temperature data over an urban area, and even distinctive temperature data for individual buildings.Previous studies of land surface temperatures (LST) and thermal remote sensing of urban and rural areas have been primarily conducted using AVHRR or MODIS imageries. Materials and Methods Recently, most researchers use high resolution satellite imagery to monitor thermal anomalies in urban areas. The present study takes advantage of themost recentsatellite in the Landsat series (Landsat 8) to monitor SHI, and retrieve brightness temperatures and land use/cover types.Landsat 8 carries two kind of sensors: The Operational Land Imager (OLI) sensor has all former Landsat bands in addition of three new bands: a deep blue band for aerosol/coastal investigations (band 1), a shortwave infrared band for cirrus detection (band 9), and a Quality Assessment (AQ) band. The Thermal Infrared Sensor (TIRS) provides two high spatial resolution thirty-meter thermal bands (band 10 and 11). These sensors use corrected signal-to-noise ratio (SNR) radiometric performance quantized over a 12-bit dynamic range. Improved SNR performance results in a better determination of land cover type. Furthermore, Landsat 8 imageries incorporate two valuable thermal imagery bands with 10.9 µm and 12.0 µm wavelength. These two thermal bands improve estimation of SHI by incorporating split-window algorithms, and increase the probability of detectingSHI and urban climatemodification. Therefore, it is necessary to design and use new procedures to simultaneously (a) handle the two new high resolution thermal bands of Landsat 8 imageries and (b) incorporate satellite in situ measurement into precise estimation of SHI.Lately, quantitative algorithms written for urban thermal environment and their dependent factors have been studied. These include the relationship between UHI and land cover types, along with its corresponding regression model. The relation between various vegetation indices and the surface temperature was also modelled in similar works. The present paper employ a quantitative approach to detect the relationship between SHI and common land cover indices. It also seeks to select properland coverindices from indices like Normalized Difference Vegetation Index (NDVI), Enhanced Vegetation Index (EVI), Soil Adjusted Vegetation Index (SAVI), Normalized Difference Water Index (NDWI), Normalized Difference Bareness Index (NDBaI), Normalized Difference Build-up Index (NDBI), Modified Normalized Difference Water Index (MNDWI), Bare soil Index (BI), Urban Index (UI), Index based Built up Index (IBI) and Enhanced Built up and Bareness Index (EBBI). Tasseled cap transformation (TCT) which is a method used for Landsat 8 imageries, compacts spectral data into a few bands related to thecharacteristics of physical scene with minimal information loss. The three TCT components, Brightness, Greenness and Wetness, are computed and incorporated to predict SHI effect.The main objectives of this research include developing a non-linear and kernel base analysis model for urban thermal environment area using support vector regression (SVR) method, and also comparing the proposed method with linear regression model (LRM) using a linear combination of incorporated land cover indices (features). The primary aim of this paper is to establish a framework for an optimal SHI using proper land cover indices form Landsat 8 imageries. In this regard, three scenarios were developed: a) incorporating LRM with full feature set without any feature selection; b) incorporating SVR with full feature set without any feature selection; and c) incorporating genetically selected suitable features in SVR method (GA-SVR). Findings of the present study can improve the performance of SHI estimation methods in urban areas using Landsat 8 imageries with (a) an optimal land cover indices/feature space and (b) customized genetically selected SVR parameters. Result and Discussion The present study selects Tehran city as its case study area. It employs a quantitative approach to explore the relationship between land surface temperature and the most common land cover indices. It also seeks to select proper (urban and vegetation) indices by incorporating supervised feature selection procedures and Landsat 8 imageries. In this regards, a genetic algorithm is applied to choose the best indices by employing kernel, 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).
Ramin Mokhtari Dehkordi; Reza Shahhoseini
Abstract
Extended Abstract Introduction Nowadays, studying urban expansion is very importantin developing countries. Rapid growth of cities has devastating environmental impacts,and irreparable economic and social consequences. Moreover, studying urban expansion is of great importance for managers and planners ...
Read More
Extended Abstract Introduction Nowadays, studying urban expansion is very importantin developing countries. Rapid growth of cities has devastating environmental impacts,and irreparable economic and social consequences. Moreover, studying urban expansion is of great importance for managers and planners of a society. Land surface temperature (LST) is one of the important parameters in urban-regional planning.Urban heat, which is usually referred to as urban heat island, can affect human health, theecosystem, surrounding air, air pollution, urban planning, and energy management. The phenomenon of urban heat island (UHI) is closely related toland-use changes in urban areas, especially when natural surfaces turn intoimpermeable urban surfaces, and increases heat flux and reduces latent heat. Materials & Methods In this study, a collection of Landsat-5 multi-temporal satellite images received in 1986, 1989, 1993, 1998, 2001, 2008, and Landsat 8 multi-temporal satellite images received in 2013, 2015 and 2017, was used along with night images of the MODIS sensor recieved in 2001, 2008, 2013, 2015, 2017 (on the same day Landsat-5 and Landsat-8 satellite images were received). In order to classify land cover and calculate land surface temperature usingLandsat 5, Landsat 8 and MODIS sensorsatellite images, initial pre-processing (radiometric and geometric corrections)was performed.In order to classifyland cover in the study area, training areas were selected using Google Earth andthen, land cover classification was carried outusing Neural Network Algorithm. Since, classifying urban areas wasthe priority ofthe present study, Normalized Difference Built-up Index (NDBI) was also used.Ultimately, pixelidentified by classification algorithm and NDBI index was allocated tourban areas. A simple relationship suggested by the United States Geological Survey (USGS) was used to estimate land surface temperature from Landsat-5 imageries.Split-window algorithm was also used to estimate land surface temperature from Landsat-8 and MODIS imageries. Since, Landsat-8 and MODIS imageries were collectedwith only afew hours (or less than that)time difference, and their thermal bands’spectral rangeswere close to each other, Landsat-8 thermal bands’emissivity coefficient with a higher spatial resolution (30 m) was used to calculate land surface temperature from MODIS images. Results & Discussion Classifying land cover in Shahr-e Kordusing Landsat-5 and Landsat-8 imageries received in 1986, 1989, 1993, 1998, 2001, 2008, 2013, 2015, and 2017 indicated that in this31-year time period,residential areas were approximately duplicatedand reached from 1004 hectares to 2112 hectares. Analysis of land surface temperature maps using Landsat 5, and Landsat 8 imageries indicated that urban areas and areas with dense vegetation had lower surface temperatures compared to areas with thin vegetation cover. Therefore, land surface temperature of urban areas is lower than the surrounding areas. However, land surface temperature obtained from MODIS imageries indicated that land surface temperature of urban areas is higher at nights. Therefore, urban heat islands in this city occur at nights. Results indicated that with increasingexpansion of urban areas, urban heat islands also intensifyat nights. Conclusion Although, Shahr-ekordis a less developed urban area as compared to other Iranian metropolises,expansion of its constructed areas can stillhave negative effects on the environment and climate of the region. The present study investigates urban growth, and itsinfluence on land surface temperature and occurrence of urban heat island. Thermal maps produced in the present study indicated that daytime air temperature of this city was relatively lower than other regions. But this is not the case at nights: compared to other areas,residential areas have a higher temperature at nights. This indicates the existence of a heat island in the city, and possibly have adverse and devastating effects on humidity, reduces precipitation, changes local winds and the climate. Results also indicate that urban expansion have directlyaffected urban heat islands. Thus, urban heat islandshave intensified and expanded during this time period. Therefore, it is concluded that there is a direct relationship between land surface temperature and land use type.