Introduction

In recent decades, many cities around the world have experienced unprecedented population growth. More than half of the world population live in urban areas, which is estimated to reach 75% of the population by 2050 (Flies et al., 2017). According to the first census conducted in Iran in November 1956, less than 31.7% of the population lived in cities, which increased to 61% and 74% in 1996 and 2016, respectively (Mahmoudian & Ghassemi-Ardehayi, 2014; SCI 2017). Rapid urbanization is associated with population growth, changes in mortality, migration from rural to urban areas, increasing external and internal pressure factors, merging villages into neighbouring cities and changes in urban boundaries (Assari & Mahesh, 2011). By increasing urban population, cities may face several challenges such as uncontrolled spatial expansion, declining urban infrastructure and facilities and environmental pollution. In addition, population growth and human activities in urban areas can result in a significant expansion of built-up areas and disruption of natural ecosystems. These processes cause land use and land cover change (LULCC) from natural and semi-natural ecosystems such as forest, agriculture and range to build-up and artificial ecosystems. Therefore, declining urban quality has been recently observed in most urban areas of the world (Hua et al., 2017; Huang et al., 2009; Xu et al., 2019). In developing countries, this problem can lead to other problems such as poverty, social problems and mismanagement of resources, which exacerbate the challenges posed by urbanization. Additionally, the negative effects on the environment and population growth in these areas may deplete natural resources (Ameen & Mourshed, 2017; Campbell-Lendrum & Corvalán, 2007). In this regard, it is essential to assess the changes in urban ecosystems since it can provide a scientific basis for governments to protect the environment and achieve sustainable and comprehensive development at ecological, economic and social levels (Akbari & Rezaey, 2018; Hang et al., 2020; Xu et al., 2019). Recent advances in remote sensing (RS) and geographic information systems (GIS) have enabled ecologists to rapidly measure ecological and regional spatio-temporal variations. These two technologies are among the most important and effective tools for monitoring ecological features and variations (Hu & Xu, 2018; Jing et al., 2020; Tuvdendorj et al., 2019; Wen et al., 2019).

Assessments made based on ecosystem quality indicators seem very challenging due to the complexity of ecosystems and the intricate process of assessing ecological quality. This challenge is even more complex for urban environments since not only the interactions of ecosystem components but also the relationship between human activities and the types of pollution should be considered. In fact, urban environments are receptive to all kinds of human and ecosystem interactions/conflicts, and the natural parts of the cities are unfortunately falls victim to these conflicts. Since natural ecosystems can guarantee the life and quality of the cities, it is vital to monitor the quality components of urban areas for having healthy and sustainable cities. The ecosystem quality of the cities has been assessed by several indicators (Guo et al., 2017; Hu & Xu, 2018; Yue et al., 2019) including Normalized Difference Vegetation Index (NDVI) or Leaf Area Index, Land Surface Temperature (LST), Land Surface Moisture (LSM), Normalized Difference Built-up Index (NDBI), Index-based Built-up Index (IBI) and Normalized Difference Impervious Surface Index (NDISI) (Ariken et al., 2020; Guo et al., 2017; Hu & Xu, 2018; Wen et al., 2019; Yue et al., 2019). However, most of these indices are not comprehensive enough and cannot assess all the important and effective aspects of urban quality. Hence, researchers have recently started to pay attention to composite indices. Su et al., (2019) prepared a four-layer index-based framework to evaluate the quality of 14 Chinese cities along the Silk Road Economic Belt in terms of structure, function, process and development and found that the quality of most cities was gradually improving. In another study, Das et al., (2020) examined the ecosystem quality of Kolkata Metropolitan City in India using the modified Vigor-Organization-Resilience (VOR) index and reported that the quality of the city has been deteriorating, which is spatially heterogeneous. In addition, Casey Keat-Chuan (2020) used Avifauna-Based Biophysical Index (ABI) to investigate the ecological landscape of Ipoh City in Malaysia and maintained that the ABI can be generalized to other tropical cities in the world. Furthermore, Chen et al. (2020) developed an index to map the ecological space quality of Pearl River Delta Metropolitan Region in China and found that the overall ESQ was good while the trend was slightly downward in different years. Meng et al. (2021) built a comprehensive index to evaluate ecological developments in some pilot cities in China and found significant progress in their ecological quality.

The Remote Sensing-Based Ecological Index (RSEI) is one of the most practical tools recently developed based upon the integration of RS and GIS techniques. RSEI has been used to assess the quality of various types of ecosystems such as wetlands (Jing et al., 2020), islands (Wen et al., 2019), cities (Hu & Xu, 2018; Yue et al., 2019; Zhai et al., 2019), farms (Tuvdendorj et al., 2019) and forests (Xu et al., 2019). This index, first introduced by Hu and Xu (2018) to evaluate the ecological quality of Fuzhou City in China, has been modified and used to detect the ecological quality and spatial heterogeneity of urban landscapes. Xu et al. (2018) used RSEI to spatially assess the ecological impacts of population growth and urban development in Xiong’an New Area, north China. The results showed that the expansion of impervious areas by urban development had a significant effect on the ecological quality of the region. In another study by Zhai et al. (2019), the ecological quality of Xinjiekou City (China) was assessed by RSEI in three intervals of 1990, 2002 and 2013, and the findings indicated the declining ecological quality of the city. Several other studies have used this index to examine the quality of the cities (e.g. Bai et al., 2019; Liu et al., 2020; Sun et al., 2020; Xiong et al., 2021; Yang et al., 2019; Zhu et al., 2020).

The review of previous studies indicated that most studies were conducted in China (e.g. Hu & Xu, 2018; Jing et al., 2020; Shan et al., 2019; Sun et al., 2020; Wen et al., 2019; Xu et al., 2018; Zhai et al., 2019) and in cities with humid climates (e.g. Wen et al., 2019; Xu et al., 2019; Zhai et al., 2019; Zhu et al., 2020). In other words, few studies have been conducted in other countries (except China) and countries with arid and semi-arid ecosystems. In addition, previous studies mostly focused on only spatial variation or temporal changes in a 2-year period (e.g. He et al., 2017; Hu and Xu 2018; Niu & Li, 2020; Singh et al., 2017; Sun et al., 2020; Xu et al., 2018, 2019; Zhu et al., 2019). In fact, determining the ecological status in longer periods of time can provide a better picture of the ecological status of a region.

Isfahan, as the third most populous city in Iran and an important tourism hub, joined the UNESCO Creative Cities Network in 2015. This city, which is known as one of the top four cities hosting the highest number of immigrants in Iran, has undergone a lot of changes and intensive development in recent years. The different geographical location and ecological conditions of this city, in comparison to the cities in the previous studies, make this study somehow different. Additionally, the international importance of Isfahan, as a cultural heritage, increases the importance of studying its ecological quality.

Therefore, the present study aims to examine the spatial and temporal changes of Isfahan from the perspective of ecological conditions and determine the factors which positively or negatively affect the ecological quality of the city.

Method

Study area

Isfahan is located between the latitudes of 32°29′-32° 50′ N and the longitudes of 51°29′-51° 52′ E, at a distance of 448 km from the south of Tehran, capital of Iran (Fig. 1). The main part of the only permanent river of the Iranian Central Plateau (Zayandeh-Rud River) passes through the city of Isfahan. The lowest and highest altitudes of the city are 1550 m (near Zayandeh-Rud River) and 2232 m (in Mount Soffeh), respectively. The mean annual rainfall and temperature of the city are 121.1 mm and 16.2 °C, respectively (Assari et al., 2017; Hosseiniebalam & Ghaffarpasand, 2015; Shirani-bidabadi et al., 2019). Over the past 40 years, Isfahan has been the second largest destination for immigration after Tehran. This city is the destination for approximately 7% of all migrants coming from Khouzestan, Chaharmahal-Bakhtiari and Lorestan provinces (Mahmoudian & Ghassemi-Ardahaee, 2014). With a population of approximately 2.0 million, Isfahan is the third most crowded city, following Tehran and Mashhad, in Iran. Approximately 30 and 70% of its population live in the south and north of Zayandeh-Rud, respectively. The population of Isfahan has grown more than 7 times over the past half century. Government policies for promoting investment, economic prosperity and industrialization have caused a huge wave of migration to this city (Ghahraei et al., 2019). Human activities such as constructions and land-use changes have peaked in recent years due to the unprecedented population growth in this city, which has been exacerbated by improper management of natural resources.

Fig. 1
figure 1

Location of the study area

Full size image

Data collection and pre-processing

In order to investigate the ecological quality changes in the study area over a period of 15 years, the Landsat Thematic Mapper (TM) and Landsat Operational Land Imager (OLI) satellite images from 2004, 2009, 2014 and 2019 were downloaded for free from the USGS (United States Geological Survey) website (USGS, 2019). All images almost were from the summer months (Table 1). After pre-processing of the collected images, the RSEI was synthesized in QGIS software.

Table 1 Landsat images used for analysis
Full size table

RSEI calculation

The RSEI equation is composed of four indices, namely, NDVI (Greenness), LSM (Humidity), NDBSI (Dryness) and LST (Heat) based on the framework of pressure-state-response (PSR) which is one of the most widely used ecological quality evaluation frameworks. The PSR framework is deeply rooted in notion of causality and the concept of selected indices in three classes of anthropogenic pressure (P), environmental state (S) and climatic responses (R) (Guo et al. 2017a; Hu and Xu, 2018b; Xu et al., 2019; Yue et al., 2019).

$${RSEI}=1-{PCA }({NDVI},{ LSM},{ LST},{ NDBSI})$$
(1)

where PCA refers to the principal component analysis.

Normalized Difference Vegetation Index

The NDVI, as one of the most important and commonly used indices to calculate the vegetation cover (Eq. 2), is based on the difference in the spectral reflectance of the red and near-infrared bands, which is a function of vegetation cover (Binh et al., 2005; Gandhi et al., 2015; Javadzarin et al., 2018; Yin et al., 2012).

$${NDVI}= \left({\uprho }_{{NIR}}-{\uprho }_{{Red}}\right)/\left({\uprho }_{{NIR}}+{\uprho }_{{Red}}\right)$$
(2)

where \({\uprho }_{\mathrm{Red}}\) and \({\uprho }_{\mathrm{NIR}}\) are the reflections from red and near-infrared bands, respectively. The numerical range of the formula is from − 1 to + 1. In the areas with dense vegetation, the index tends to + 1 while it drops to − 1 in areas with low vegetation density (Huang et al., 2009).

Land surface moisture

The LSM index is commonly used to determine the ecological quality of a region due to the direct relationship between the components of Tasseled cap transformation (TCT), including brightness, humidity and greenness with the characteristics of land surface (Zawadzki et al., 2016). This index indicates soil and vegetation moisture, which shows the ecological characteristics of an ecosystem (Guo et al., 2017). The SWIR is the most sensitive region to the surface moisture (Yue et al., 2019). The LSM index is calculated as follows:

$$\begin{aligned}{{LSM}}_{{TM}}= & \ {0.03151\uprho }_{{Blue}}+{0.2021\uprho }_{{Green}}+{0.3102\uprho }_{{Red}}\\&+{0.1594\uprho }_{{NIR}}-{0.6806\uprho }_{{SWIR}1}-{0.6109\uprho }_{{SWIR}2}\end{aligned}$$
(3)
$$\begin{aligned}{{LSM}}_{{OLI}}=& \ {0.1511\uprho }_{{Blue}}+{0.1973\uprho }_{{Green}}+{0.3283\uprho }_{{Red}}\\&+{0.3407\uprho }_{{NIR}}-{0.7117\uprho }_{{SWIR}1}-{0.4559\uprho }_{{SWIR}2}\end{aligned}$$
(4)

where LSMTM and LSMOLI are the land surface moisture in TM and OLI satellite sensors, respectively. In addition, \({\uprho }_{\mathrm{Blue}}\), \({\uprho }_{\mathrm{Green}}\), \({\uprho }_{\mathrm{Red}}\), \({\uprho }_{\mathrm{NIR}}\),\({\uprho }_{\mathrm{SWIR}1},\) and \({\uprho }_{\mathrm{SWIR}2}\) represent the spectral regions of blue, green, red, near-infrared and short-wave infrared band 1 and short-wave infrared band 2, respectively.

Land surface temperature

Thermal infrared imaging (TIR) provides an effective way to gather useful information about LST on a regional and global scale as most of the energy detected by the sensor is emitted directly from the ground (Atitar & Sobrino, 2009).

The remote radiometer measures radiant surface temperature as well as the energy reflected from red and near-infrared bands of the electromagnetic spectrum, which can be used to quantify changes in land cover (Afrasiabi Gorgani et al., 2013; Solanky et al., 2018; Vlassova et al., 2014).

This index can be used to define the thermal islands in cities. Many studies showed that cities are usually warmer than their surrounding rural areas (Rizwan et al., 2008; Stewart, 2011; Tomlinson et al., 2011). The LST index was calculated using Eq. 5 (Asgarian et al., 2015; Chatterjee et al., 2017):

$${LST}= \frac{{{T}}_{{sensor}}}{1+\left(\uplambda \times {{T}}_{{sensor}}/\uprho \right){ln\varepsilon }}$$
(5)

where λ indicates the wavelength of the emitted waves (11.435 μm for Landsat 5–7 and 10.9 μm for the band 10 of Landsat 8), ρ is the constant value (1.438 × 10−2 mK) and ε represents the emission rate from the land surface which is calculated as follows:

$$\varepsilon=\left\{\begin{array}{lc}0.995& \; \; {NDVI}\;\leq\;0\\0.970&0\;<\;{NDVI}\;\leq\;0.157\\1.0094\;+0.047\;{In}\;{NDVI}&0.157\;<\;{NDVI}\;\leq\;0.727\;\varepsilon=\\0.986 & \qquad \; {NDVI}\;>\;0.727\end{array}\right.$$
(6)

where \({\mathrm{T}}_{\mathrm{sensor}}\) is the at-satellite brightness temperature calculated by Eqs. 7 and 8:

$${{T}}_{{sensor}}= \frac{{{K}}_{2}}{{ln}\left({{K}}_{1}/{{L}}_{\uplambda }+1\right)}$$
(7)
$${{L}}_{\uplambda }={Gain}\times {DN}+{Bias}$$
(8)

where \({\mathrm{L}}_{\uplambda }\) is the at-sensor spectral radiance, Gain shows the band-specific multiplicative rescaling factor and Bias indicates the band-specific additive rescaling factor. Additionally, DN indicates the digital numbers in each pixel, and K1 and K2 are the calibration coefficients for the thermal bands of TM/ETM+/OLI.

For TM, ETM+ and OLI: K2 = 1260.56 K, 1282.71 K and 1321.08 K. K1 = 607.76, 666.09 and 774.89 mW cm−2 (Yue et al., 2019).

Normalized Differential Build-Up and Bare Soil Index

NDBSI includes the index-based built-up index (IBI) as well as barren land areas with no vegetation cover (deforested or abandoned areas). The NDBSI is a combination of IBI and SI (soil index). Build-up area and bare soil reflect more SWIR than NIR, which are calculated as follows (Essa et al., 2012):

$${NDBSI}= \left({IBI}+{SI}\right)/2$$
(9)
$$\begin{aligned}{BI}= & \left[\left({\uprho }_{{SWIR}1}+{\uprho }_{{red}}\right)-\left({\uprho }_{{nir}}+{\uprho }_{{blue}}\right)\right]\\&/\left[\left({\uprho }_{{SWIR}1}+{\uprho }_{{red}}\right)+\left({\uprho }_{{nir}}+{\uprho }_{{blue}}\right)\right]\end{aligned}$$
(10)
$$\mathrm{IBI}=\frac{\left\{{2\uprho }_{\mathrm{SWIR}1}/\left({\uprho }_{\mathrm{SWIR}1}+{\uprho }_{\mathrm{nir}}\right)-\left[{\uprho }_{\mathrm{nir}}/\left({\uprho }_{\mathrm{nir}}+{\uprho }_{{red}}\right)+ {\uprho }_{{green}}/\left({\uprho }_{{green}}+{\uprho }_{{SWIR}1}\right)\right]\right\}}{\left\{\left\{{2\uprho }_{{SWIR}1}/\left({\uprho }_{{SWIR}1}+{\uprho }_{{nir}}\right)+\left[{\uprho }_{{nir}}/\left({\uprho }_{{nir}}+{\uprho }_{{red}}\right)+ {\uprho }_{{green}}/\left({\uprho }_{{green}}+{\uprho }_{{SWIR}1}\right)\right]\right\}\right\}}$$
(11)

After calculating the indices, principal component analysis (PCA) was used in QGIS to determine the relative importance of the variables and generate RSEI maps. PCA is an important method for compressing multidimensional data which can help eliminate the effect of co-linearity between variables (Mishra et al., 2017; Seddon et al., 2016).

The function leads to an accurate and rapid assessment of the ecological quality of a region. Since the range of values varies in different indices, all indices should be normalized before applying PCA so that the range of numbers is between 0 and 1 (Jolliffe & Cadima, 2016). In the RSEI map (Fig. 3), the closer pixel value to 1 leads to better ecological status, while the closer value to 0 leads to worse ecological status. Furthermore, as shown in Fig. 4, the values of the RSEI are classified into 5 classes with intervals of 0.2 and named as poor, fair, moderate, good and excellent to show the ecological quality of the region in more details.

Spatial heterogeneity analysis

Spatial autocorrelation

The values of pixels in the RSEI map are either close to each other or very different when the difference between neighbouring pixels is not easy to detect (especially for those with negative correlations). The spatial autocorrelation analysis compares a given map with a map in which values are randomly distributed. The Moran correlation coefficient shows how much each place is correlated with its surrounding places. The purpose is to create a map with values between 1 and − 1. Accordingly, when similar values in the neighbourhood are greater than the random value, the correlation is positive; otherwise, it is negative. The Moran index indicates the degree of homogeneity in different years. The spatial correlation in this study was calculated by Geoda software. This method is divided into two parts: global spatial autocorrelation and local spatial autocorrelation (Amaral & Anselin, 2014; Anselin et al., 2006). The global spatial autocorrelation is calculated as follows (Chen, 2013):

$${I}_{{g}}=\frac{N{\sum }_{{i}}{\sum }_{{j}}{w}_{{ij}}\left({x}_{{i}}-\mu \right)\left({x}_{{j}}-\mu \right)}{\left({\sum }_{{i}}\sum_{{j}}{w}_{{ij}}\right)\sum_{{i}}{\left({x}_{{i}}-\mu \right)}^{2}}$$
(12)

where wij, xi, xj, μ and N indicate the normalized weights, RSEI value in the ith pixel, RSEI value in the jth pixel, mean RSEI value of the study area and the total number of pixels, respectively. The Moran’s index is approximately + 1 for places with complete correlation while it is approximately − 1 for places which are completely non-correlated (Anselin et al., 2006).

Since Ig cannot determine hot and cold spots, local spatial autocorrelation was used to calculate these values to display the Local Indicator of Spatial Analysis (LISA) (Eq. 13).

$${I}_{{l}}=\frac{{x}_{{i}}-\mu }{{\sum }_{{i}}{\left({x}_{{i}}-\mu \right)}^{2}}\sum_{{j}}{w}_{{ij}}\left({x}_{{j}}-\mu \right)$$
(13)

Next, a map was obtained to divide the RSEI map into five categories, namely, high-high (hot spots), low-low (cold spots), low–high, high-low and not significant.

Results and discussion

Ecological status

This study is divided into three main stages as shown in Fig. 2. In this study, the RSEI was used to detect spatiotemporal changes in the ecological quality of Isfahan City in 2004, 2009, 2014 and 2019 by scoring the PC1 of four indices (NDVI, NDBSI, LSM and LST). As shown in Table 2, the eigenvalue of PC1 in each year is more than 70% (ranging from 79.30% to 83.53%). Therefore, it can be concluded that PC1 shows the most variability among all four indices compared to other PCs. Therefore, PC1 is used to represent four index variables.

Fig. 2
figure 2

Flowchart of study stages

Full size image
Table 2 Contributions of four indices to the first principal component (PC1)
Full size table

Further, there are two opposite groups of indices based on their contribution to ecological quality of the city. The results show that NDVI and LSM are in one group while LST and NDBSI are in another group to make RSEI. In addition, as can be seen in Table 3, the mean RSEI is always less than 0.40, which indicates the not-so-appropriate condition of the city. Additionally, the different values of the index in different years indicate a non-constant and fluctuating trend (Fig. 3). Based on the results, the average of RSEIs in 2004, 2009, 2014 and 2019 was 0.34, 0.37, 0.26 and 0.30, respectively. The index increased by 8.82% from 0.34 to 0.37 between 2004 and 2009. Furthermore, NDVI and LSM, which have a positive effect on ecological quality, increased by 8.57% and 2.27%, respectively. NDBSI and LST, which impose a negative effect on the ecological quality, decreased by − 4.61% and − 7.69%, respectively. The ecological quality significantly declined by − 29.72% from 0.37 to 0.26 between 2009 and 2014. At the same time, NDVI and LSM decreased by − 34.21% and − 20.00%, respectively, while NDBSI and LST increased by 17.74% and 3.33%, respectively. The RSEI increased by 20% from 0.26 to 0.30 while NDVI and LSM increased by 0.56% and 11.11%, respectively, during 2014–2019. Further, NDBSI and LST decreased by − 5.47 and − 8.06, respectively. Based on the results, the highest and lowest RSEIs were observed in 2009 and 2014, respectively (Fig. 3).

Table 3 Means of RSEI and four indices for different study years
Full size table
Fig. 3
figure 3

Satellite images (false colour composite) for 2004, 2009, 2014 and 2019 (left) and RSEI map for 2004, 2009, 2014 and 2019 (right)

Full size image

As mentioned before, this index was divided into 5 classes at 0.2 intervals (Fig. 4) for better analysis and representativeness of RSEI. The classes are named from low to high quality: poor, fair, moderate, good and excellent, respectively, which is similar to previous studies (Hu & Xu, 2018; Jing et al., 2020; Shan et al., 2019; Sun et al., 2020; Wen et al., 2019; Zhai et al., 2019; D. Zhu et al., 2020).

Fig. 4
figure 4

Classified RSEI maps of Isfahan in the last 15 years

Full size image

Table 4 shows the numerical values corresponding to the RSEI classes. The overall ecological quality of Isfahan is in the fair class (level 2), with an area percentage of about 82%, 76%, 59% and 83% in 2004, 2009, 2014 and 2019, respectively. In addition, it is obvious that areas with poor ecological quality account for only a small part of the study area in the studied years, except for 2014 when the area of this class reached 30% of the study area. Despite the significant decrease in 2019, this amount was still more than 2004 and 2009. The fair class declined over from 2009 to 2014 as compared to 2004, but increased again in 2019. The moderate class was upward until 2009, which experienced a decrease in 2014, and then an increase. The changing trend of the good class was upward in all intervals except for 2014. The same trend was observed for the excellent class.

Table 4 Classification of the study area in different years based on the RSEI value
Full size table

As shown in Fig. 4, the poor, fair, moderate, good and excellent areas shown in red, orange, yellow, light green and green, respectively, indicate differences in the ecological quality levels of different areas of the city. Based on the classification results (Tables 4 and 5), approximately 97% of the total study area belonged to the moderate and fair classes in 2004. The fair class in this year accounted for the largest percentage of the entire study area (89.19%). However, the area of the fair class decreased at the expense of the moderate class in 2009. On the other hand, the area of the poor class increased from 0.01% in 2009 to 30.61% in 2014. Further, the sum of fair and moderate classes decreased from 97 to 67% between 2009 and 2014. This change was partially compensated and increased to 94% in 2019. In addition, the majority of the areas in the southern, south-eastern and eastern parts are classified as poor in 2014 and 2019 (Fig. 4).

Table 5 Different ecological quality classes (ha)
Full size table

Comparison of RSEI with LULCC map

LULCC maps were prepared for each year (Fig. 5) to better interpret the dynamic ecological quality status of the city and the impact of land use/land cover changes. Since land use and land cover as well as their changes could have a significant impact on terrestrial ecosystems and their quality (Kafy et al., 2020; Mahato & Pal, 2018; Xiao & Weng, 2007), it can be identified as a good method for analysing the results from RSEI in the present study. Based on the ecosystem of Isfahan, five land use types (bare land, vegetation, build-up, rock, and water) were selected to classify in four study years. Table 6 shows the area percentage of each land use.

Fig. 5
figure 5

Land use/land cover change by the year

Full size image
Table 6 Area percentage of land use type in the four study years
Full size table

As shown in Fig. 5 and Table 6, most changes in LULCC map are related to water. The water level of this river decreased by 100% in 2014 due to the cut-off of the Zayandeh-rud River and returned again in 2019. In addition, the vegetation change percentage, as a positive index to increase ecological quality, is about + 4%, − 5% and + 1% in 2004–2009, 2009–2014 and 2014–2019 intervals, respectively. In the same period, this value was approximately − 8%, + 23% and − 10% for the bare land class, − 3%, − 1% and 0% for rock class and + 2%, + 2% and + 1% for the build-up class. Additionally, all these three classes are inversely related to ecological quality.

As previously mentioned, immigration, uncontrolled expansion of built-up areas and the arid and semi-arid climate of the city have made this city as one of the most vulnerable urban areas in Iran (Assari & Mahesh, 2011; Isfahan Municipality, 2018). In addition, the manipulation of natural resources by human as well as several natural challenges disrupted the ecological balance of the city. Remote sensing-based indices were developed to present a comprehensive view of the ecological status of the city due to the importance of the ecological quality of the ecosystems and difficulty in producing the time series ecological images. Since the RSEI index is obtained using different aspects and characteristics of an ecosystem, it can be considered as a reliable index to illustrate the ecological condition of a region or ecosystem. The findings of this study confirmed the high efficiency of this index in determining the ecological quality, which is in line with the results of previous studies using RSEI (Guo et al., 2017; Hu & Xu, 2018; Jing et al., 2020; Wen et al., 2019; Yue et al., 2019). The RSEI values were always at a low level (less than 0.4), which indicated the fair ecological quality of the city (level 2). Furthermore, there was a fluctuating trend in ecological condition in the four study years. Additionally, RSEI increased from 0.34 to 0.37 between 2004 and 2009. Further, this index had a downward trend in 2009–2014, which increased from 0.26 to 0.30 in 2019.

It is necessary to further analyse the studied indices to detect the main causes of sinusoidal changes. The NDVI index representing the vegetation cover was 0.35, 0.38, 0.25 and 0.39 in 2004, 2009, 2014 and 2019, respectively. Since Isfahan is located in a semi-arid region, the vegetation cover of the city includes agricultural lands and urban parks. As shown in LULCC map (Fig. 5) and Table 6, the percentage of vegetation in these years has a direct relationship with NDVI. In other words, the NDVI decreased when the percentage of vegetation cover decreased in 2014 by about 5%. In addition, the average LSM indicated that soil and vegetation moisture, which was 0.88 in 2004, reached 0.90 in 2009. The value of this index decreased again in 2014 and reached 0.72, which increased to 0.8 by increasing again in 2019. Thus, this trend could be due to the fluctuations in the water level of Zayandeh-Rud River. The mean NDBSI in the region showed that human pressures in the region were 0.65 and 0.62 in 2004 and 2009, respectively. The index increased to 0.73 in 2014, but decreased to 0.69 in 2019.

Average rainfall and temperature of the study months (summer months) in the region showed that the temperature in Isfahan was 28.4, 25.5, 28.8 and 28.5 C in 2004, 2009, 2014 and 2019. Additionally, the rainfall was recorded for four study years as 0, 3.4, 0 and 1.8 mm. Therefore, rainfall can be an important factor showing RSEI changes during the studied period (SCI, 2020). These not-so-severe climate change-induced effects in the region were exacerbated by human impacts, which reduced water supplies in recent years. The depletion of water supplies to the mentioned problems led to the manipulation of water flow upstream of Zayandeh-Rud River (closing the upstream dam or transferring water to other neighbouring areas) and reduced volume of water entering the city in recent years (Isfahan Municipality 2016). To evaluate the relationship between the inflow of water into Isfahan and its ecological quality, this study analysed the inflow rate of Zayandeh-Rud River (at Mousian Station) during the studied period (Regional Water Company of Isfahan 2019) (Table 7).

Table 7 The flow rate of water entering Isfahan
Full size table

As shown in Table 7, the RSEI had a lower rate during the years the river was dry and the water flow was cut off or was temporarily flowing. The results showed that RSEI had a non-uniform and wavy trend in the studied years. Although there was a slight improvement (3%) in 2009 compared to 2004, it dropped sharply to 0.26 in 2014. The trend increased to 0.30 in 2019 but was still lower than the RSEI amounts in 2004 and 2009.

The total volume of water entering Isfahan at intervals of 5 years showed a non-steady trend. The total volume of the incoming water had a decreasing trend, dropped from 25,977.33 m3/s in the 5-year period before 2004 to about 11,413.80 m3/s in the 5-year period before 2014 and increased to 19,834.72 m3/s in 2014–2019, which is still lower than 2004. The temporal comparison of the fluctuations in the index and water flow entering the city showed a close relationship between the two factors. Furthermore, when less water entered the city, the vegetation cover and wetness dropped and the quality index of the city decreased accordingly. When the river water flow was cut off for a part of the year or the whole year or a smaller volume entered Isfahan, the agricultural water rights decreased, which caused the agricultural lands to remain barren or led to changes in their usage.

Liu et al. (2020) used RSEI to assess the ecological change of Pingtan County (a coastal city in China) and reported a constantly declining downtrend in RSEIs. They claimed that the declining RSEI trend was mainly related to a reduction in the vegetation cover and farmlands. Sun et al. (2020) studied the ecological quality of Tableland Region in China and highlighted temperature, potential evapotranspiration and precipitation as the most important natural factors which have a key role in the ecological quality of the region. Zhu et al. (2020) introduced land use land cover changes caused by urban expansion as the main driver of the deteriorating ecological condition of Zhengzhou City in China.

Bai et al. (2019) found land use land cover changes acted as a disturbing factor and negatively affected the ecological quality of the Mount Wutai World Cultural Landscape Heritage Protected Area in China, which in contrast to the findings of the present study. In this study, changes in water inflow and climatic factors were introduced as the important factors affecting the ecological quality of Isfahan, which could be related to the differences in the geographical conditions of the studied areas. Isfahan is located in a semi-arid region where the flow of water as a scarce commodity could have a vital impact. Figure 6 shows the area of the regions experiencing ecological changes in 2004 to 2019. As can be seen, when the volume of water entering Isfahan reached its minimum, the city experienced the least improvement and the most degradation. In 2014–2019 period, the sudden increase in the volume of water entering the city led to the most improvement and the least damage.

Fig. 6
figure 6

RSEI changes in Isfahan during the four study years (ha)

Full size image

Additionally, the results showed that when a large percentage of the classified area changed to the poor ecological class (level 1) in 2014, the eastern and southern parts suffered more ecological damage than other parts. It is noteworthy that these areas mostly contain rock and build-up areas. Kafy et al. (2020) and Setturu et al. (2013) stated that land uses such as rock, build-up areas and barren lands are more sensitive to an increase in temperature and drought compared to other land uses, which could be related to the lack of vegetation. Furthermore, there were numerous agricultural lands in eastern and southern regions which became barren due to insufficient water in 2014. Therefore, human tensions and negative natural elements could decrease ecological quality, and the parts with denser buildings, low density of vegetation or those highly affected by human activities will be more damaged than other areas, which is consistent with the results of the previous studies (Ariken et al., 2020; Liu et al., 2020; Wen et al., 2019). Hence, human manipulation and natural phenomena were recognized as the most important factors on the variation of ecological quality in Isfahan City, which is in line with the findings of other researchers (Hang et al., 2020; Hu & Xu, 2018; Sun et al., 2020; Wen et al., 2019; Xu et al., 2019).

Since the changes in RSEI in each period were very close to the real conditions of the region, RSEI could be considered as a suitable and useful index for determining the ecological quality of built-up areas, which is in line with the findings of past researches (Ariken et al., 2020; Guo et al., 2017; Shan et al., 2019; Xu et al., 2019; Yue et al., 2019; Zhai et al., 2019).

LISA clustering map was used to better understand the spatial ecological quality of RSEI characteristics (Fig. 7). LISA map can show fragile and spatially significant ecological environments by distribution of HH, LL, HL and LH parts. High values surrounded by high neighbouring are shown as high-high (HH), high values surrounded by low neighbouring are displayed as high-low (HL), low values surrounded by low neighbouring are demonstrated as low-low (LL) and low values surrounded by high neighbouring are shown as low–high (LH). As can be seen in Fig. 7, HH is located in the central parts of the city and the borders of the river, which are covered by vegetation. Additionally, LL is more frequent in regions with lack of vegetation cover or areas further destroyed by humans such as eastern and southern parts.

Fig. 7
figure 7

LISA map of Isfahan in 2004, 2009, 2014 and 2019

Full size image

Based on the LISA results, the high-high areas were more expanded in 2014 than 2004 and 2009, especially towards the western parts, with few scattered spots in the centre, east, and north of the city. In 2019, the high-high areas were dispersed mostly in the central parts and increased in some areas near the Zayandeh-Rud. During this year, the low-low areas increased in many areas and scattered throughout the region.

The results of the Moran scatter plots showed a steady decline in the correlation from 2004 to 2019 since it changed from about 0.91 in 2004 and 2009 to 0.88 in 2014 and 0.87 in 2019 (Fig. 8). The spatial correlation of Moran points was more clustered than random points. The correlation in study area decreased over time. However, the spatial correlation was still high in 2019. A decrease in the Moran index indicated a decrease in homogeneity in different years. When homogeneity decreased, the points with high ecological quality separated from each other and became fragmented. Urbanization, physical expansion of the built-up areas and manipulation of the natural flow of the river and its consequences led to the dispersal of the areas with high ecological quality and reduced homogeneity in the region, which could be the reason for why the high-high spots were more scattered in 2019 than 2004.

Fig. 8
figure 8

Moran scatter plots of Isfahan in 2004, 2009, 2014 and 2019

Full size image

Conclusion

In the present study, the ecological quality of Isfahan was evaluated using the RSEI. This study is the first application of this index for a metropolitan city in a developing country located in an arid and semi-arid region. The results showed that increasing urban activities and changes caused by human interventions as well as climate changes disturbed the ecological balance and reduced the ecological quality of the city. In addition, the results of the Moran index indicated that although an improvement in terms of ecological quality was observed after 2014, the hot spots were still more scattered than past years. In other words, the correlation of areas with high ecological quality was always declining (2004–2019), which could be attributed to human interventions. Increasing human activities, poor management of resources and climatic changes in dry, fragile ecosystems (e.g. in Isfahan City) led to unfavourable changes in the region. The findings imply that many of the degraded areas could be rehabilitated if resource management practices are modified. Therefore, it is suggested that more attention be paid to managing the urban ecosystems and their recourses in an ecological resilience way. Last but not least, it is recommended that agricultural and vegetated lands around and inside the region be preserved as much as possible and that the cost of creating green space be considered in the construction process of urban areas.