Application of split-window algorithm to study urban heat island in Yazd county

Introduction
LST (LST) is one of the important parameters that affect the physical, chemical, and biological processes of the earth as well as environmental science and urban planning. Human activities such as land use changes and the development of urban areas led to an increase in the LST and the appearance of thermal islands. The main source of climate data such as temperature are synoptic stations, however, it is impossible and time-consuming to use traditional methods to estimate the LST for all types of earth conditions, on the other hand, synoptic stations only measure temperature information for specific points, and the obtained values are only related to that specific point; while according to the land cover and other conditions, the temperature in different parts of a region is different compared to the temperature recorded for a specific point and can be several degrees celsius lower or higher, therefore, it is necessary to use scientific methods that provide the possibility of calculating the temperature of any point on the earth's surface. At present, remote sensing images due to features such as wide and continuous coverage, low cost, timeliness, and the ability to obtain information in reflective and thermal ranges, are suitable tools for extracting LST and land use maps. Spatial analysis is one of the important subjects in the temporal and spatial evaluation of land surface data, which can be used to examine the spatial and temporal changes of spatial data in a region. Given that the data that are examined in environmental studies are not independent of each other in most cases and their dependence is due to the location of the observations in the studied space, which are called spatial data; Due to the existence of a spatial correlation between the data, the usual statistical methods are not a suitable method for examining these data, and spatial statistics can be used as a suitable option for analyzing these data. The aim of this research is to extract LST and land use map of Yazd county using a remote sensing technique. In this study, the spatial autocorrelation of LST in Yazd city and the identification of hot thermal clusters have been investigated using the global Moran statistic and the Getis-Ord GI statistic.
 
Materials and Methods
In this research, Landsat 8 satellite’s multi-spectral and thermal images have been used to extract the land use and LST in the study area, After performing the necessary corrections in the preprocessing stage, the land use map of the study area was prepared in 5 classes (built-up, vegetation cover, water bodies, bare land, and rock) using the support vector machine method and the overall accuracy and kappa coefficient were used to evaluate the classification result. In the next step, LST was extracted by the split window method. The relationship between LST and Soil adjusted vegetation index (SaVI) was investigated using regression analysis. In order to identify the spatial pattern of the LST, the global Moran index was used and hot spots were identified by Getis-Ord GI statistics.
 
Results and Discussion
Our findings show that the kappa coefficient and overall accuracy were equal to 0.96% and 98.99%, respectively, bare lands are the most, and water bodies have the least area, equal to 76.16 and 0.09%, respectively. The average LST was 50.83°C. The result showed that the type of land use had an effect on LST, the water bodies had the lowest, and barren lands had the highest mean LST, equal to 36.91 and 52.13 °C, respectively. Vegetation is one of the factors that regulate the LST, areas without vegetation have a maximum LST and areas with high density vegetation have minimum LST .Based on the results, the vegetation quality of the study area was poor and its average temperature was 45.61°C. The mean of SAVI index was equal to 0.09 and correlation analysis showed a negative correlation between SAVI index and LST (r = -051). The analysis of spatial correlation with global Moron indexes showed that the LST of Yazd has a spatial structure, in other words, LST is distributed in a cluster form, Based on the results of the Getis-Ord GI statistic, the area of hot and cold spots was equal to 66.86% and 27.4%, respectively. In general, parks, cultivated lands, tree and forest cover and water areas, formed the cold spot areas of yazd city, and the hot spot areas of yazd city were located in the industrial areas and surrounding urban lands, hospitals, passenger terminals, gas stations, places near busy roads and bare and uncovered lands.
 
Conclusion
The results showed a strong relationship between land use and LST. Based on the results, the LST data of Yazd has a spatial structure pattern, barren lands and industrial areas formed hot thermal islands, and vegetation and water bodies formed cold thermal islands in the study area; the wide area of barren lands, the lack and poor vegetation cover due to the lack of rainfall and drought are factors affecting the LST and the creation of hot thermal islands in the study area. The result showed a negative relationship between LST and SAVI, the vegetation of the study area is weak and its temperature is high. Considering the role of vegetation in adjusting LST, it is recommended to take necessary management measures in order to improve the quality of vegetation and reduce bare land in the study area, and also prevent the conversion of natural land uses into built-up land. The results of this research can be used by managers and planners for better urban management. The results of this research confirm the capability of remote sensing in environmental studies, it is suggested to identify thermal islands in other seasons and at night and compare the results with the results of this research.