Introduction

Variations in land use/land cover (LULC) and change have significant effects on the local, regional, and global environment. Urbanization is a major environmental concern. Land-use changes and land cover conversion contribute to unsustainable development (Yang et al. 2009), increasing surface temperatures (Liu et al. 2006), air pollution (Wang et al. 2020), water pollution, greenhouse gas emissions, and intensification of urban heat islands (UHIs) (Shao et al. 2006; Chan and Yao 2008).

These changes profoundly affect biodiversity, local and regional climates (Luck and Wu 2002), and landscapes (Jain et al. 2020). They affect not only the composition of surface cover but also the flows of matter and energy (Wang et al. 2016), surface roughness (Shi et al. 2006), evaporation rates (Sun and Zhou 2010), and soil quality (Deng et al. 2009). One of the main reasons that soil is covered with impermeable surfaces is urbanization. In fact, water management in these areas has been affected by urbanization and is induced by soil sealing, which expands the connectivity of the flows (Keesstra et al. 2018). Urbanization and industrialization cause serious variation in the capacity of the soil by soil sealing (Cerdà et al. 2018; Visser et al. 2019).

Unprecedented variations in urban land use have significantly modified energy exchanges between Earth surfaces and the atmosphere, leading to increased urban temperatures in contrast to the temperatures of the surrounding areas (Buyantuyev and Wu 2010). Increasing air temperature in cities (compared to the countryside) is known as the UHI, a term coined by Luke Howard in 1833 in London, England (Sobrino et al. 2004).

The natural growth of Iran’s metropolises and increasing rural-to-urban migration have degraded cities’ environments and increased pollution. In some parts of Iranian cities, temperatures have increased more than in other districts, with temperatures that remain closer to those of the surrounding areas. The areas with increasing temperatures have dense concentrations of more geometrical buildings, residential and industrial developments, high energy consumption rates, and greater motor vehicle densities. Because of the negative effects of UHIs on residents, it is important for urban planning (Choudhury et al. 2019). UHIs not only affect the living environment (Konopacki and Akbari 2002) but also increase energy consumption and threaten human health (Changnon et al. 1996).

One critical aspect of the thermal environment and the UHI is the surface temperature. Jiang and Tian (2010) investigated the effect of urban coverage on land surface temperature (LST) and showed that changing the surface coverage is the most crucial element behind LST increases. Li et al. (2011), examining the effect of urban landscapes on the heat island of Shanghai, China, indicated that there is a significant positive relationship between cover types and temperature. Su et al. (2012) found a significant relationship between cover type and heat island formation. Nassar et al. (2016) showed that UHIs generate their biggest impacts during daytime hours in summer and at night during winter. Zhao et al. (2019) determined the spatial distribution of LST trends in the central Himalayas from MODIS satellite products and found that land surface temperatures decreased with increasing elevation. Furthermore, nighttime temperatures are more sensitive to land surface changes than daytime temperatures.

Portela et al. (2020) analyzed LST variations in urban development and its effects on regional climate and surface energy flows in southeastern Brazil. Landsat images were used to extract land surface thermal indices. They found that the highest surface temperatures were in urban and industrial areas with buildings, impermeable pavement, and less vegetation. Shafizadeh-Moghadam et al. (2020) modeled spatial variations of urban land surface temperature for Tehran, Iran, using land use, solar radiation, elevation, distance to main roads, and normalized difference vegetation index (NDVI). Results from four machine learning (ML) models—random forest (RF), generalized additive model (GAM), boosted regression tree (BRT), and support vector machine (SVM)—revealed that NDVI, land use, and elevation accounted for 86% of the spatial variation in surface temperature. Unexpectedly, land surface temperatures near main roads were lower due to vegetation along Tehran’s streets and highways. The performance analyses of the models showed that there was no significant difference between them.

This study examines the influence of the various land covers on LSTs in Ahvaz in southern Iran. Land-use variation, LST variations, UHI index, and urban thermal field variance index (UTFVI) were investigated to identify and describe places with heat island. The results of this study provide a better understanding of the spatial effects of environmental and human conditions on urban LST variations, which can facilitate environmental management and urban planning to reduce urban LST variations.

Material and methods

Study area

Ahvaz, the capital of Khuzestan Province, has a population of 1.4 million. It is located between 31° 14′ 12″ N and 31° 23′ 17″ N and 48° 33′ 48″ E and 48° 47′ 12″ E in southwestern Iran (Fig. 1). It is located at an average of 12 m above sea level (Borojerdnia et al. 2020). Ahvaz is the seventh metropolis in Iran and one of the most significant economic and industrial centers in Iran. This city is located in the hottest areas of Iran and has low vegetation. Area of 21020.21 hectares and about 400 hectares belongs to the Karun River. Karun with 950 km long is known as the biggest and longest river of Iran. The minimum temperature recorded in the city is 15.6°C, and the maximum is 39.5°C (Asadifard and Masoudi 2018). The average precipitation is 213 mm, most of which falls from September to May (Borojerdnia et al. 2020). Khuzestan Province borders Iraq, Kuwait, and Saudi Arabia, and, like its neighbors, experiences dust storms. Geologically, the study area is quartz, and the surface deposits are silt and sand (Zarasvandi et al. 2011). The soil textures are loamy, sandy loam, and loamy clay. Soil pH ranges from 7.8 to 9.1, indicating that the soil in the region is alkaline (Ghanavati et al. 2019).

Fig. 1
figure 1

Khuzestan Province is located in southwestern Iran (top left) and Ahvaz is near the center of the province (top right)

Full size image

Research methods

Several steps were required to conduct this research (Fig. 2).

Fig. 2
figure 2

Flowchart of different steps of research in the study area

Full size image

Data

Thermal data were acquired from Landsat 4, 5, 7, and 8 satellite images as the average of all available images per year. Annual averages from 1995 to 2018 were used to prepare an LST map for the city of Ahvaz. The LSTs were estimated for 1995, 2000, 2010, 2015, and 2018. Data from 1995 to 2016 were used to create a series of annual land-use maps.

Estimating UHI index, UTFVI, and creating a composite

The LST maps were classified into five temperature classes using means and standard deviations to create a UHI index (Xu et al. 2013) (Table 1).

Table 1 Urban heat island (UHI) (Xu et al. 2013)
Full size table

The UTFVI is used to quantitatively describe the effect of the UHI quantitatively and is calculated using Eq. 1 (Liu and Zhang 2011; Zhang et al. 2006):

$$ \mathrm{UTFVI}=\frac{\Delta \mathrm{T}}{T_i}=\frac{T_i-{T}_M}{T_i} $$
(1)

where Ti is the land use surface temperature ( LST of a certain point), and TM is the average temperature of the entire study area. To prepare an ecological evaluation map of Ahvaz’s heat islands, each of the six classes created for the UTFVI classification map was assigned excellent, good, normal, bad, worse, and worst according to the values (Table 2).

Table 2 UTFVI categories (Zhang et al. 2006)
Full size table

A composite map of UTFVI and the UHI index was generated using the combined function to examine the degree of consistency among the classes for the two indices.

Preparation of land use map

Land use maps were prepared from Landsat satellite data from 1995, 2000, and 2016 using the support vector machine (SVM) method. SVM is an efficient method for accurately distinguishing land use. Previous studies (Shafri and Ramle 2009, Otukei and Blaschke 2010; Taati et al. 2015) have shown that ML algorithms perform well in this task. SVM is a supervised non-parametric statistical method introduced by Vapnik (1999). Its main features are that it can use fewer training samples and achieve higher accuracy than other classification methods (Mantero et al. 2005; Mountrakis et al. 2011). Samples from all the bands form the boundaries of the classes using an optimization algorithm. An optimal linear decision boundary was calculated to separate the classes. The samples or training points closest to the decision boundaries are the support vectors. The algorithm finds a hyperplane that is compatible with the training data and can separate the data sets from each other. The optimal separating hyperplane is an area that can minimize incorrectly classified pixels using training data (Mountrakis et al. 2011). In this study, the number of training points for each year was held constant and was divided into two sets: 70% were used for training and classification, and 30% were used for testing the resulting model. All training samples were polygons with a diameter of 22 m.

Investigation of different land uses

According to the land uses types of the study area, temperature variations in 20 years of each land use were determined. In the next step, the class of very hot in each of the eight districts of Ahvaz was extracted. The sum of the number of pixels related to the hottest class of the heat island for each district of Ahvaz was calculated separately. Thus, districts with the highest number of heat islands were identified.

Results and discussion

Preparation of land surface temperature map of the study area

Thermal data were acquired from Landsat 4, 5, 7, and 8 archives to investigate trends in the land surface temperature (LST) variations of Ahvaz from 1995 to 2018. The thermal statistics were derived from LST maps (Table 3). The results show that in 23 years, the average, minimum and maximum LST are 21.50°C and 47.56°C, respectively. From 2013 to 2018, the number of summer images exceeded 20 and winter images exceeded 18. It is noteworthy that June, July, August, and September months for summer and December, January, February, and March months for winter were used. During this period, the temperature range varied from 38.30 to 41.16°C. Temperatures were also mapped for 1995, 2000, 2016, 2017, and 2018 (Fig. 3). The minimum temperature during these years was 1995 and the maximum occurred in 2018. In 1995, LST ranged from 21.50 to 44.92°C, while in 2018, the range was from 23.85 to 47.56°C. These results indicate that LST increased over the 23 year study period.

Table 3 Statistical data of surface temperature from 1995 to 2018
Full size table
Fig. 3
figure 3

The land surface temperature in the study area in 1995, 2000, 2016, 2017, and 2018

Full size image

Index map of UHI

The UHI index map was prepared with five temperature categories (very cold, cold, temperate, hot, and very hot) (Fig. 4). In the six selected years, the rank (most to least) of these five classes, based on percentages of the study region in each category, was temperate, hot, very cold, cold, and very hot.

Fig. 4
figure 4

UHI index classification map of Ahvaz at different time intervals

Full size image

A comparison of the UHI indexes in 1995 and 2018 revealed that the total areas of pixels classified as very cold, temperate, and hot increased, while the areas of very hot and cold decreased. The area of very cold zones decreased from 5.63% in 1995 to 5.16% in 2018, a 0.47% reduction. The area of spaces categorized as very hot increased from 3.79% in 1995 to 5.93% in 2018, a 2.14% increase. Investigation of the spatial extents and time variations of heat island intensity in Ahvaz revealed that these trended upward.

UTFVI map

The UTFVI map was prepared with six classes: best, better, normal, bad, worse, and worst (Fig. 5). The highest percentages of the study area from 1995 to 2018 were classified as best and worst. Better, normal, and bad classes comprised similar amounts, although their rank orders varied. For example, in 1995, after areas of best (47.83%) and worst (41.52%) heat, the others were normal (2.76%), bad (2.70%), good (2.65%), and worse (2.54%). In 2018, however, after best (50.21%) and worst (38.60%) were better (2.88%), normal (2.86%), bad (2.73%), and worse (2.73%) (Table 4). Thus, most of the study area was either the best (UTFVI <0) or worst (UTFVI> 0.020). Areas with optimal thermal conditions (UTFVI <0) were either water, green space, or agricultural land in the city. Previous studies have shown that these land uses are typically in the best zone (UTFVI <0) (Portela et al. 2020). Land uses that relate to extreme heat, such as bare land and residential covers, are coincident with zones of worst heating (UTFVI> 0.020). In addition, where vegetation is abundant, these areas are located in the excellent category UTFVI because they have the best temperature conditions. In contrast, areas without vegetation and high-temperature conditions, such as bare lands and residential areas, are in the worst category of UTFVI (Guha et al. 2018; Portela et al. 2020).

Fig. 5
figure 5

UTFVI index classification map of the study area

Full size image
Table 4 Classification of ecological evolution index UTFVI and their percentage of area in Ahvaz during 1995–2018
Full size table

Results of composite map of UTFVI and UHI index

Creating a composite map of UTFVI and the UHI index shows that the areas with the worst UTFVI classification of worst coincide with areas classified as either very hot, hot, or temperate UHI. None of the worst areas was classified as cold or very cold (Fig. 6).

Fig. 6
figure 6

Composite map of UTFVI and UHI index

Full size image

Preparation of land use map

A land-use map was prepared for 1995, 2000, and 2016 with lands classified as agricultural, residential, bare lands (less than 7% vegetated), green spaces, or water (Fig. 7). The SVM accuracies for 1995, 2000, and 2016 were 84%, 86%, and 93%, respectively. The percentage coverage of each land-use class and the number of training points are tabulated (Table 5). The rank of land use coverage in the study area in 1995 and 2000 was bare land, residential, green space, agricultural, and water. In 2016, the order was residential, agricultural, bare land, and water. The trend from 1995 to 2016 was an increase in the residential area and agricultural land, and a decrease in bare land and green space. The water content remained relatively constant. The residential area increased by the largest percentage over the 16 years (Fig. 8). Growing pressure on land and unsuitable use is cussed to destroy land resources, so a sustainable approach to land use and land management is needed (Keesstra et al. 2018).

Fig. 7
figure 7

Land use classes of the study area during selected years

Full size image
Table 5 Percentage of landuse and training data
Full size table
Fig. 8
figure 8

Landuse pattern by years

Full size image

Land surface temperature and different land-use types

Next, the relationship between land surface temperature (LST) and each land use was examined (note that LST in 2016, 2017, and 2018 was compared to the land uses of 2016). The results reveal that the highest average annual LST is correlated to bare lands and residential areas. The lowest LSTs were found in the locations of water bodies and green spaces. These relationships are consistent with the literature (Amiri et al. 2009; Oliveira et al. 2011; Maimaitiyiming et al. 2014; Huang et al. 2019). Increasing impermeable levels and decreasing vegetation in urban environments have increased the heat capacity and thermal conductivity rate (Oke 1995; Weng 2001; Bouyer et al. 2009; Song and Wang 2015). Urban areas have different energy balances compared to non-urban areas. The differences in the nature of these regions lead to an energy imbalance in the urban environment, causing severe environmental problems, such as dangerous thermal conditions for the residents. Increased population density and the expansion of residential land use exacerbate urban heating. Schatz and Kucharik (2014) indicated that one of the major reasons for the spatial change in temperature patterns in urban areas is the built-up environment. The nature of different land uses dictated their surface temperatures, so spatial variations of LST were strongly correlated (Table 6). From 1995 to 2018, the average LST increased for each land-use category. In 1995, the average temperature of bare lands was 36.85°C, and in 2018, it was 41.15°C, a 4.3°C increase. LST increases from 1995 to 2018 were 3.66°C for residential areas, 4.94°C for green space, and 4.09°C for agricultural land. For water, however, it was 8.32°C. Thus, the rank order of average LST change was water, green space, bare land, agricultural land, and residential land (Fig. 9). The largest percentage of very hot land was in 1995 in district 8 of Ahvaz, the location of residential and industrial land use. In 2016, 2017, and 2018, the largest percentage of land classified as very hot was in district 2; most of this district was bare land with less than 7% vegetation. The results of other studies (Oke et al. 1981; Bokaie et al. 2016) showed a close relationship between urban morphology and heat island intensity. The formation of various UHI patterns is influenced by LULC types and urban geometry. Moreover, there is an LULC parameter affecting UHI, including bare land, which can be important (Shojaei et al. 2017; Li et al. 2018; Morais et al. 2018). The UHI can be a mixed pattern of temperature variation for bare land, urban areas, water, and vegetation lands. Thus, the UHI phenomenon can be clearly explained by the different LULC. The UHI in the study area is also located in the bare lands around the city, which are often uninhabited. Research has reported that an effective reduction strategy to decrease UHI effects is the increase in vegetation cover area (Rosenfeld et al. 1995; Ashie et al. 1999; Tong et al. 2005; Yang and Wang 2015; Wang et al. 2016). Wang et al. (2016) in the Phoenix metropolitan area have found that the cooling effect of developed vegetation has been stronger than heating effect of urbanization. Vegetation cover causes an increase in soil carbon, while the loss of carbon in the soil has a positive response to rising temperatures. Therefore, soil management can help reduce climate change by increasing vegetation (Cox et al. 2000; Lal 2004; Keesstra et al. 2016).

Table 6 Land use information in different years
Full size table
Fig. 9
figure 9

Land surface temperature pattern in land use

Full size image

Accuracy of the results

Five locations classified as very hot were examined (Fig. 10). These areas showed that the very hot conditions coincided with either industrial areas or barren lands in Ahvaz. Previous studies have demonstrated that bare soil and impermeable surfaces (asphalt, concrete, and buildings) absorb more solar radiation (Gartland 2012; Sodoudi et al. 2018; Shafizadeh-Moghadam et al. 2020; Alahmad et al. 2020). Lower heat holding capacity, less reflection, and greater thermal conductivity cause these surfaces to reradiate greater amounts of energy to the lower atmosphere, leading to greater increases in local ambient temperatures than in the surrounding areas.

Fig. 10
figure 10

Map of very hot islands in Ahvaz and zoomed examples by Google Earth images. 1—bare land, 2—Ahvaz piping plant, 3—Iran carbon factory, 4—reactor manufacturing, and 5—Khuzestan steel company

Full size image

Limitation

However, some limitations of this study should be reported. First, some of the satellite imagery was distorted and we could not report their results. This problem is limited to presenting a complete analysis of the results. In these cases, future researchers should improve their methods for using distorted satellite imagery. Finally, the SVM algorithm is used to separate different land uses, while other algorithms can be tested and the best algorithm is evaluated.

Conclusion

LST is one of the main thermal parameters used to study urban environments. By studying the spatial distribution of LST in Ahvaz, Iran, areas with the highest temperatures were identified as heat islands related to specific land uses. One of the critical factors for understanding the changing geography of LST is understanding the ratios of land use in a study region and how they change over time. The SVM method was used to classify the thermal patterns and to study variations in the patterns over specific time intervals. The results of the classification show that the best classification accuracy occurred with the 2016 data. By considering the land use variation in Ahvaz at four different times, the greatest changes in proportions occurred with bare lands, residential areas, and green spaces. These variations show the conversion of bare land to residential land use, bare land, and green space to agricultural land use. According to the results of LST, the maximum temperatures recorded for Ahvaz in 2017 occurred in areas of bare land (42.93°C) and residential land (40.06°C), depicting the heating of urbanizing areas and denuded landscapes. Residential land uses, due to human decisions and materials (asphalt, concrete, and buildings), and bare lands, due to aridity and high thermal absorption, are usually urban spaces that experience the highest temperatures. The urbanization of Ahvaz eliminated a large amount of green space and barren land was replaced by industrial developments, buildings, and other infrastructure. The results of this study can inform urban planners and environmental managers in the region with tangible evidence of LST, land use, and coincident changes in both. This knowledge can be applied to forecasting and methods of adapting growth in ways that change the trend of intensification of the UHI.