Introduction

Although a small percentage of the world’s land surface is covered with urban areas, these areas are the primary source of most of the world’s environmental issues, including carbon emissions, energy and resource overuse, and biodiversity and ecosystem degradation (Grimm et al. 2015; Mcdonald et al. 2008; Seto et al. 2012). Indeed, the poorly planned and controlled expansion of urban structures in combination with the shrinking of ecologically valuable landscapes is making urban development unsustainable (Peng et al. 2017; Feist et al. 2017). As a result, finding a way to ensure the structural stability and functional security of natural ecosystems for sustainable urban development has become a global challenge (Cumming and Allen 2017). The structural and functional changes caused by uncontrolled urban expansion can undermine the ecosystem services (ESs) provided by the ecological infrastructure and threaten the cities’ environmental security and sustainable development. Therefore, to preserve the ecological security of cities, urban planners and administrators are always looking for urban planning measures and methods that can help them regulate and control urban structure in support of the stability of ecosystem functions (Zhaoxue and Linyu 2010: 1394).

One major way to preserve sustainability in urbanizing landscapes is to create a landscape ecological security pattern in order to effectively protect ecological security by managing the interactions between landscape patterns and ecological processes (Wu et al. 2013; Yu et al. 2006). The ecological security approach has emerged as an effective instrument for the fundamental assessment of ecosystem structures and functions that are threatened by urban expansion and economic development (Solovjova 1999; Kullenberg 2002; Ma et al. 2004; Eckersley 2005) and thus a powerful tool for protecting the ecological security of regional ecosystems (Su et al. 2013; Berkes and Folke 1998; Haeuber and Ringold 1998; Devuyst et al. 2001; Ehrlich 2002).

The concept of landscape ecological security pattern was introduced by Yu (1996), who believed that protecting ecological security could be the most useful approach for preserving ecologically valuable processes and landscape patterns (Yu 1995, 1996). According to Yu, ecological security patterns can preserve the structural integrity, functions and processes of an ecosystem (Zhang et al. 2015). Ecological security patterns can serve as a potent tool for protecting regional ecological security by preserving the important parts of the landscape and important ecological processes (Yu 1996; Kattel et al. 2013; Peng et al. 2018a). From the perspective of landscape ecology and the interaction between landscape patterns and ecological processes, creating and optimizing ecological security patterns can help prevent poor landscape protection, manage the conflict between environmental protection and economic development, and preserve the ESs that are needed in a healthy environment (Yu 1996; Kattel et al. 2013). Research into ecological security patterns can be conducted at the city (Peng et al. 2019), province (Peng et al. 2018b), region (Zhang et al. 2017), country (MacMillan et al. 2007), or even global scales.

Indicators of ecological security play a key role in the assessment of this quality. It should also be noted that different ecosystems need their own unique indicators in different temporal and spatial dimensions (Zhao et al. 2006). One of the indicators of ecological security is the quality of ESs (Chen et al. 2018) and landscape patterns (Liu and Chang 2015). In recent years, many countries have started to pay more attention to research on ESs and landscape patterns with an ecological security approach (Gao et al. 2012; Zhang et al. 2016; Liu et al. 2018; Chen et al. 2018; Ma et al. 2019) and using the findings as a basis for decision-making about preserving socioeconomic balance as well as ecosystems (Chen et al. 2018).

Ecological security of the earth has been a major source of concern for governments as well as academia since the late 1970s (Bächler et al. 1995; Matthew et al. 2002). Over the years, this issue has received increasing attention (Peng et al. 2019; Wang et al. 2019; Liu et al. 2018) and has been the subject of many studies across the world (Zhang et al. 2017; Yang et al. 2018; Li et al. 2019a, b; Yang and Cai 2020; Li et al. 2020). Also, various tools have been developed for measuring this security. But given the complexity of this subject, there is still ample room and ongoing efforts for advancement in this field and the aforementioned tools. Of the numerous studies conducted in this field, many have assessed the current ecological security (Zhang et al. 2017; Li et al. 2020), some have only focused on ecological landscape issues (Liu et al. 2018; Ma et al. 2019), and some have attempted to make a connection between ecological security and ecological footprint (Yang and Cai 2020; Hao et al. 2017). Also, many of these studies have evaluated the ecological security of a landscape in the past as well as the present (Li et al. 2019a; Wang et al. 2019; Wu et al. 2019), some have also discussed the issues concerning ESs and ecological security (Qin et al. 2019; Pasgaard et al. 2017), and others have focused on the past ecological security of a region (Lu et al. 2020; Chen et al. 2020).

This study presents a spatial–temporal analysis of land-use/cover changes and their ecological effects in a region using InVEST 3.7.0 and FRAGSTATS 4.2.1 models. These analyses are meant to provide, through the assessment of ecological security with a comprehensive approach to landscape ecology and ESs, the scientific information that might benefit the management and monitoring of the Lavasanat watershed in Tehran, Iran. Today, one of the major challenges in this watershed is the trend of land-use change, which can be attributed to the encroachment of the Tehran metropolitan area into nearby natural and agricultural lands especially since 2000. These land-use changes have resulted in the expansion of previously built areas in the form of fast urbanization, and also the emergence of new residential areas in the Lavasanat region, which highlight the need to address the issue to prevent further damage.

In particular, the objectives of the present study are as follows:

  1. 1.

    Spatial–temporal analysis of LULC changes from 2000 to 2020

  2. 2.

    Analysis of changes in landscape metrics and ESs (water yield) in response to LULC changes

  3. 3.

    Analysis of changes in ecological security due to LULC changes from 2000 to 2020 in terms of landscape metrics and ESs (water yield).

Materials and methods

Study area

This research has been performed in the Lavasanat catchment, which is located in Shemiranat County of Tehran province (Fig. 1). The area of this catchment is 52933 hectares, and its inhabitants are 25,000 people. The city of Lavasan is the largest residential area in this catchment. Stretched between 51° 24′ and 51° 50′ east longitudes and 35° 46′–36° 03′ north latitudes, Lavasanat watershed is bound on the north by the Noor county, on the west by the Karaj county, on the east by the Damavand county, and on the south by the city of Tehran (Rahmani Fazli et al. 2017). Lavasanat watershed consists of three sub-watersheds called Kond, Afjeh, and Lavarak, and its main rivers and tributaries discharge directly into the Latyan Dam reservoir (Talari 2016).

Fig. 1
figure 1

Location and land-use classes of the Lavasanat

Full size image

Methodology

In this study, first, a spatial–temporal analysis was conducted on LULC changes in the study area, and then, the InVEST 3.7.0 and FRAGSTATS 4.2.1 models were used to assess these land-use changes, and finally, an analysis was performed to determine changes in the region’s ecological security in terms of landscape metrics and ESs (water yield) (Fig. 2).

Fig. 2
figure 2

Schematic diagram for mapping and modeling ecological security

Full size image

Spatial–temporal analysis of land-use/cover changes

LULC images for the 20-year period of interest (2000, 2010, and 2020) were prepared using two sets of data. Initially, the authors intended to only use Landsat satellite images (Table 1). However, since many parts of the study area contained patches of land-uses in low-density villages and also many land-uses were too close to be distinguished in Landsat satellite images, for such areas, manually made land-use maps were digitized using high-resolution spatial images of the area. Ultimately, the results of the two methods were merged to create a coherent LULC map for the period of interest.

Table 1 Description of satellite images
Full size table

To fix some errors in the images, they were subjected to a pre-processing, which involved either geometric or atmospheric correction depending on the image classification. The radiometric correction was performed on multi-time images, i.e., images taken at different seasons or years or by different sensors (Du et al. 2002). Before being used in digital analysis, satellite data were examined to ensure standard quality in terms of geometric or radiometric errors such as stripe noises, misaligned scan lines, and duplicate pixels, and atmospheric errors such as cloudy spots.

The geometric correction was conducted by the use of cadastral maps, visual interpretation, Google Earth, and field trips. Since satellite images were in raster format and the pixel rows and columns in these images were not aligned with terrestrial coordinates, for georeferencing, the digital data of an acceptable number of points with appropriate distribution had to be extracted from the images. Therefore, the data of 20 ground control points were extracted from the 1:25,000 digital maps and used in the geometric correction. These points were selected so that they would cover the entire area and would be detectable on the ground and clearly visible on the image. Then, the regression method in the software ENVI5.3 was used to transfer the DN1 values of pixels from the original image to the created image and perform the geometric correction. In this process, while selecting the RMS2 values (root mean square error), the authors tried to use those with RMS < 1.

In view of the land-uses present in the region, the research objectives, and the spatial and radiometric resolution of the images, the lands were classified into five lands-uses, including residential areas (urban, rural, and industrial areas and roads), bare lands, rangelands, water bodies, and agricultural lands, using the supervised classification method called maximum likelihood in ENVI5.3. After this process, the accuracy of the classification was evaluated by two measures: overall accuracy (OA) and kappa coefficient (KC).

Spatial–temporal changes assessment

Landscape patterns metrics

Considering the research objectives and the studies previously conducted in this field, and also the overlap between some metrics, this study used six metrics at the class and landscape level (Table 2). Landscape patterns were calculated using FRAGSTATS 4.2.1(McGarigal et al. 2012).

Table 2 Landscape patterns metrics
Full size table

Assessment of ecosystem services

LULC changes due to the expansion of residential and other built areas as well as agriculture are among the most important human factors that affect ecosystem services (Li et al. 2007). One of the most valuable ecosystem services is water yield (Brisbane 2007), which happens to be more vulnerable than other ecosystem services to severe land-use changes (Tao et al. 2016). Typically defined for dry ecosystems (Brauman et al. 2007), this ecosystem service depends on watershed characteristics such as topography, vegetation cover, land-use, climate and other parameters (De Groot et al. 2010). Therefore, measuring changes in the water yield ecosystem service in response to land-use change is an effective way to assess the environmental costs and benefits of land-use planning. Thus, modeling the ecosystem services (water yield) of a watershed can help governments to monitor and predict the effect of economic development policies and the consequent land-use changes and adopt proper plans and policies accordingly (Lang et al. 2017).

Modeling and simulation

InVEST modeling

The water yield (WY) in the Lavasanat watershed was determined using the WY model in the software InVEST3.7.0. This model predicts and demonstrates how changes in land-use patterns affect WY and its functions in different sectors. Over the years, this tool has been used for analysis and evaluation of WY in a large number of studies and management programs across the world (Rahimi et al. 2020; Hu et al. 2020; Yang et al. 2019; Hamel and Guswa 2015; Boithias et al. 2014). The WY model of InVEST is based on the Budyko curve (Budyko 1974), which is an empirical function relating the ratio of actual evapotranspiration to the ratio of potential evapotranspiration to precipitation.

In this model, first, the mean annual water yield \(Y\left( x \right)\) of each pixel of the landscape is calculated as follows (Sharp et al. 2019):

$$Y\left( x \right) = \left( {1 - \frac{{{\text{AET}}\left( x \right)}}{P\left( x \right)}} \right) \cdot P\left( x \right)$$
(1)

where \({\text{AET}}\left( x \right)\) is the mean annual actual evapotranspiration of pixel x and \(P\left( x \right)\) is the mean annual precipitation of the same pixel.

In the InVEST model, the input land-uses are divided into two classes, vegetated and non-vegetated, and the \(\frac{{{\text{AET}}\left( x \right)}}{P\left( x \right)}\) ratio of each class is calculated differently. For vegetated land-uses, this ratio is calculated based on the Budyko curve of Fu (1981) and Zhang et al. (2004):

$$\frac{{{\text{AET}}\left( x \right)}}{P\left( x \right)} = 1 + \frac{{{\text{PET}}\left( x \right)}}{P\left( x \right)} - \left[ {1 + \left( {\frac{{{\text{PET}}\left( x \right)}}{P\left( x \right)}} \right)^{\omega } } \right]^{{{\raise0.7ex\hbox{$1$} \!\mathord{\left/ {\vphantom {1 \omega }}\right.\kern-\nulldelimiterspace} \!\lower0.7ex\hbox{$\omega $}}}} { }$$
(2)

where \({\text{PET}}\left( x \right)\) is the potential evapotranspiration of pixel x, and \(\omega \left( x \right)\) is a non-physical parameter that depends on the natural climatic-soil characteristics of the pixel. Potential evapotranspiration, PET (x), is defined as follows:

$${\text{PET}}\left( x \right) = K_{{\text{C}}} \left( {l_{x} } \right) \cdot {\text{ET}}_{^\circ } \left( x \right)$$
(3)

In this equation, \({\text{ET}}_{^\circ } \left( x \right)\) is a parameter reflecting the climatic conditions of the area based on the evapotranspiration of a reference plant in that area, and \(K_{{\text{C}}} \left( {l_{x} } \right)\) is mainly determined by the vegetation characteristics of the land-use in the pixel (Allen et al. 1998).

\(\omega \left( x \right)\) is an empirical parameter that can be expressed as a linear equation of \(\frac{{{\text{AWC}}*N}}{P}\), where N is the number of events per year, and AWC is the water content available for plant use (mm). This linear equation is as follows (Donohue et al. 2012):

$$\omega \left( x \right) = Z\frac{{{\text{AWC}}\left( x \right)}}{P\left( x \right)} + 1.25$$
(4)

Here, \({\text{AWC}}\left( x \right)\) is the available water content (mm) in pixel x, which depends on soil texture and effective rooting depth. \({\text{AWC}}\left( x \right)\) determines the amount of water that the soil retains for plant use. This parameter is obtained by multiplying the plant available water capacity (PAWC) by the root-restricting layer depth or the plant rooting depth, whichever is lower (Yang et al. 2019):

$${\text{AWC}}\left( x \right) = {\text{Min}}\left( {{\text{Rest}} \cdot {\text{layer}} \cdot {\text{depth}},\;{\text{root}} \cdot {\text{depth}}} \right) \cdot {\text{PAWC}}$$
(5)

The root-restricting layer depth is the depth beyond which the roots cannot penetrate into the soil because of its physical and chemical properties. The plant rooting depth is usually considered to be the depth at which 95% of the root biomass occurs. PAWC is the difference between the soil's water-holding capacity and the plant wilting point (Yang et al. 2019).

The hydrogeological parameter Z is an empirical constant between 1 and 30, which reflects the local precipitation pattern, precipitation intensity, seasonal climate change, and topographic features of the watershed. There are three methods to obtain Z (Sharp et al. 2014): (1) using the formula N × 0.2, where N is the number of rainy days in a year (Donohue et al. 2012) [the lower the number of rainy days in an area, the lower the Z value (Redhead et al. 2016)]; (2) using the global ω estimate (Liang et al. 1994; Xu et al. 2013); (3) estimation after calibration with actual runoff data (Canqiang et al. 2012).

For non-vegetated land-uses (such as urban areas, wetlands, etc.), the actual evapotranspiration \({\text{AET}}\left( x \right)\) is obtained directly from the reference evapotranspiration \({\text{ET}}_{^\circ } \left( x \right)\) using the following equation:

$${\text{AET}}\left( x \right) = {\text{Min}}\left( { K_{{\text{c}}} \left( {l_{{\text{c}}} } \right) \cdot {\text{ET}}_{^\circ } \left( x \right),P\left( x \right)} \right)$$
(6)

In this equation, \({\text{ET}}_{^\circ } \left( x \right)\) is the reference evapotranspiration and \(K_{{\text{c}}} \left( {l_{{\text{c}}} } \right)\) is the land-use-specific evaporation factor.

Results and discussion

Assessment of land-use/cover changes

After preparing the maps of LULC classes, the changes occurred during the study periods were examined (Fig. 3). In the validity assessment of the obtained classification, the overall accuracy and kappa coefficient of the classification for the years 2000, 2010, and 2020 were calculated to 95.72, 96.26, and 95.32 percent and 0.948, 0.943, and 0.936, respectively. These results indicated the land-use classifications are acceptably accurate and match the reality on the ground (Table 3). The classification done with the maximum likelihood algorithm showed that there are five classes of land-use in the area: residential areas (urban, rural, and industrial areas and roads), bare lands, rangelands, water bodies, and agricultural lands (Fig. 3).

Fig. 3
figure 3

Land-use Lavasanat watershed for years 2000, 2010, and 2020

Full size image
Table 3 Area statistics of LU/LC classification during the study period
Full size table

The prepared land-use maps and data collected from satellite images clearly demonstrate the expansion of residential areas because of population growth and rising demand for housing and the consequent urban expansion. On the contrary, agricultural lands have shrunk over the study period. The growth of residential areas and the shrinking of agricultural lands between 2010 and 2020 are quite evident. Given the current trend of land-use change, one can expect a sharp decrease in the area of agricultural lands in the coming years. Rangelands of the watershed have shrunk between 2000 and 2010, but have grown between 2010 and 2020, a trend that can be attributed to the conversion of agricultural lands into rangelands in the second period. The area of bare lands in the watershed has decreased in both periods (Table 3).

Landscape pattern analysis at different levels

Analysis of landscape pattern at class level

  1. (1)

    Changes in land-use classes in terms of CA: The total area of residential land uses in the Lavasanat watershed in 2000, 2010, and 2020 has been 432.4, 689.82, and 1040.86, respectively. This increase clearly reflects the expansion of built areas in the region. The total area of bare lands in these years has decreased from 19,117.28 to 19,079.35 and ultimately to 19,015.51. In these years, the total area of agricultural lands has also decreased from 3224.36 to 3204.7 and then to 2231.91. The area of rangelands has decreased from 29,702.61 in 2000 to 29,600.46 in 2010 but has then increased to 30,286.13 in 2020 because of land-use changes (Fig. 4a—CA).

  2. (2)

    Changes in land-use classes in terms of PLAND: The percentage of landscape (PLAND) covered by residential areas has increased from 0.81% in 2000 to 1.3% in 2010 and then to 1.96% in 2020. For bare lands, this metric has decreased from 36.11 to 36.04% and ultimately to 35.92%, showing the slight shrinkage of bare lands in the region. For agricultural lands, PLAND has been 6.28% in 2000, 6.5% in 2010, and 4.21 in 2020, which demonstrate an accelerating reduction. Between 2000 and 2010, the percentage of the landscape covered by rangelands has decreased from 56.11 to 55.92%, but between 2010 and 2020, this metric has bounced back up to 57.21% (Fig. 4b—PLAND).

  3. (3)

    Changes in land-use classes in terms of the NP: As the results show, the NP of residential areas in 2000, 2010, and 2020 has been, respectively, 1626, 2943, and 3774, which demonstrate an increasing trend and is indicative of the growth and expansion of built areas in the landscape. The NP of agricultural lands in the same years has been, respectively, 1248, 1648, and 1842. The NP of bare lands has also had a rising trend, increasing from 113 in 2000 to 197 in 2010 and to 403 in 2020. The NP of rangelands has decreased from 10,084 in 2000 to 9861 in 2010, but shows an increasing trend between 2010 and 2020 because of land-use changes (Fig. 4c—NP).

  4. (4)

    Changes in land-use classes in terms of IJI: According to the results obtained for this index, between 2000 and 2010, the IJI of residential areas has increased from 54.72 to 63.54, which means an increasing level of contact with patches of different classes. But from 2010 to 2020, the IJI of residential areas has dropped to 55.39, probably because residential areas have become more continuous and consolidated. IJI of rangelands in the years 2000, 2010, and 2020 has been, respectively, 55.39, 55.05, and 66.11, which show a slight increase in the second 10-year period. For bare lands, the metric (IJI) has been rising consistently, increased from 2.31 in 2000 to 6.3 in 2010 and then to 14.32 in 2020. For agricultural lands, IJI has increased from 33.63 in 2000 to 46.43 in 2010 but then has decreased to 42.79 in 2020 (Fig. 4d—IJI).

  5. (5)

    Changes in land-use classes in terms of LSI: The LSI of residential areas in 2000, 2010, and 2020 has been, respectively, 54.72, 74.82, and 99.9, reflecting the increasingly complex and irregular shape of these land-uses. For agricultural lands, LSI has increased from 43.68 to 52.26 and then to 54.69, which signifies the fragmentation of the patches of agricultural land and consequently the increasing geometrical irregularity and complexity of these patches. The same can also be said for rangelands, for which LSI has increased from 65.3 in 2000 to 65.65 in 2010 and to 73.08 in 2020. There has also been a slight increase in the LSI of bare lands in the years 2000 (60.12), 2010 (60.66), and 2020 (62) (Fig. 4e—LSI).

  6. (6)

    Changes in land-use classes in terms of LPI: For residential areas, LPI has shown a consistent trend, increasing from 0.19% in 2000 to 0.56% in 2010 and to 0.91% in 2020. For the bare lands class, this metric has decreased from 8.12% in 2000 to 4.32% in 2010 and ultimately to 4.13% in 2020. This trend can be attributed to the shrinking total area of this class and the increasing fragmentation of the patches of bare land in the watershed. Between 2000 and 2020, the LPI of agricultural lands has also decreased consistently (from 0.82 to 0.59 and then 0.36), which can be attributed to the expansion of residential areas and the complete disappearance of the patches of farmlands. For rangelands, LPI in the years 2000, 2010, and 2020 has been 43.93, 43.37, and 41.1, respectively (Fig. 4e—LPI).

Fig. 4
figure 4

Changing trend of landscape pattern at class level

Full size image

Analysis of landscape pattern at landscape level

The total NP in the entire watershed in the years 2000, 2010, and 2020 has been, respectively, 13,073, 14,651, and 17,016. This increasing trend is indicative of increased fragmentation of the landscape because of the land-use changes occurred in the watershed (Fig. 5a—NP; Table 4). PD in the landscape during the same years has been 24.69, 27.67, and 32.14, respectively, which also reflects the increase in the number of landscape patches (Fig. 5b—PD; Table 4). These trends indicate the rising trajectory of landscape fragmentation in the watershed. LSI of the Lavasanat watershed for the years 2000, 2010, and 2020 was calculated to 51.34, 54.33, and 59.72, respectively, which means the landscape is becoming increasingly complex and irregular in terms of the geometry of its land-uses (Fig. 5c—LSI; Table 4). This increased complexity has made the landscape patches unstable. In other words, the level the degradation has increased with the increasing geometrical complexity of land-uses. Thus, the increasingly irregular shape of the landscape is indicative of man-made degradation and disturbances in the Lavasanat watershed.

Fig. 5
figure 5

Changing trend of landscape pattern at landscape level

Full size image
Table 4 The overall landscape index changes in the research area
Full size table

SPLIT of the watershed for the years 2000, 2010, and 2020 was calculated to 4.82, 5.03, and 5.63, respectively. The increase in this metric is also a sign of fragmentation in the landscape (Fig. 5d—SPLIT; Table 4). ED represents the ratio of the perimeter of each patch to its area. The ED of the watershed in the years 2000, 2010, and 2020 was calculated to be 86.86, 92.06, and 101.42, respectively. This trend is indicative of the increasing degradation and deformation of different patches in the landscape (Fig. 5e—ED; Table 4). The increase in the IJI of the landscape from 38.86 in 2000 to 45.05 in 2010 and then to 50.54 in 2020 is reflective of the increasing blending of patches of land-uses over these years (Fig. 5f—IJI; Table 4).

Water yield ecosystem service

To build and execute a WY model, it was necessary (Hu et al. 2020) to collect and prepare a series of inputs including the maps of precipitation, potential evapotranspiration, root-restricting layer depth, plant available water content, land-uses, vegetation, and watershed boundary (Tables 5 and 6), and also the biophysical table in the CVS format (Table 7).

Table 5 The used data characteristics for water yield
Full size table
Table 6 Data statistics and range of variation for InVEST model inputs
Full size table
Table 7 Biophysical table used for the baseline InVEST water yield model
Full size table

After preparing these inputs, the model of InVEST 3.7.0 was used to simulate the temporal and spatial changes of WY for the years 2000, 2010, and 2020 based on climate and land-use data (Fig. 6). These simulations showed that the WY of the entire Lavasanat watershed in 2000 has been 2,641,734.816 m3. For this year, the mean WY, potential evapotranspiration, and actual evapotranspiration of each individual pixel were 4.990, 1884.446, and 406.748 mm, respectively. For the year 2010, the WY of the entire watershed was estimated to be 3,318,950.915 m3, and the mean WY, potential evapotranspiration, and actual evapotranspiration of each pixel were calculated to be 6.270, 1806.965, and 370.340 mm, respectively. The WY of the watershed for the year 2020 was 7,737,201.215 m3. For this year, the mean WY, potential evapotranspiration, and actual evapotranspiration of each pixel were calculated to be 14.616, 1535.086, and 530.245 mm, respectively. These model calculations showed that the WY in the Lavasanat watershed is increasing. This increase can be attributed to the observed land-use changes and particularly the rampant expansion of built areas (residential land-uses).

Fig. 6
figure 6

The used data for InVEST model

Full size image

Of the watershed’s total WY, 1,677,926.367m3 for the year 2000, 2,287,145.055m3 for the year 2010, and 4,908,786.651m3 for the year 2020 have been related to residential areas. The total area of residential areas in the entire watershed has increased from 4,820,578.505m2 in 2000 to 6,885,513.787m2 in 2010 and then to 10,407,948.705m2 in 2020 (Fig. 7).

Fig. 7
figure 7

Spatial distribution of water yield/fractp (estimated actual evapotranspiration fraction per pixel)/aet (estimated actual evapotranspiration per pixel) in the Lavasanat watershed for years 2000, 2010, and 2020

Full size image

Relationship between land-use/cover changes and ecological security

Human activities damage ecosystems by destroying their structures and affecting ecological processes (Salvati and Carlucci 2014). Urbanization is the most important driver of economic development, but undeniably, it is also a major source of environmental problems such as rapid land-use changes (He et al. 2014; Li et al. 2011; Kong et al. 2017; Aljoufie et al. 2013). Ecological and environmental problems caused by rapid land-use changes are in most cases fairly severe. Furthermore, rapid land-use changes may cause fundamental changes in the quality of environmental ecosystems in the short term (Juanita et al. 2019; Wu et al. 2019).

In the absence of advance planning and science-based strategizing, built areas may expand into neighboring environments that play a key role in preserving the ecological security of the entire region. In this study, spatial–temporal analysis of land-use changes in the studied watershed showed that from 2000 to 2020, the watershed’s residential areas have been expanding, its agricultural lands have been shrinking, and there has been not much change in other land-uses (Fig. 3, Table 3). These results are consistent with the findings of Wu et al. (2020), which showed that the expansion of residential and urban areas directly affects ecological security.

Relationship between landscape patterns and ecological security

Identification and monitoring of land-use/cover changes could be an intricate process (Sun and Zhou 2016: 121), especially because such changes may affect ecological processes and functions (Su et al. 2012: 297). One major way to understand and track land-cover and land-use changes is to analyze changes in landscape patterns (Fan and Ding 2016: 156). Landscape patterns can be quantified in terms of landscape metrics (Li and Wu 2004: 224). Using the combination of land-use maps and landscape metrics, one can identify areas with critical changes in ecosystem services at the landscape level. Such knowledge will offer valuable opportunities to improve future land management plans and practices (Tolessa et al. 2017: 49). Landscape structure (composition, configuration, and connectivity) plays an essential role in conserving biodiversity and providing ecosystem services (Zhang and Gao 2016: 42).

Destructive human activities such as land conversion and fragmentation have profoundly affected the landscape of the Lavasanat watershed, leaving a severe adverse environmental impact on this region. This has reduced the area of natural lands in the region and increased their dispersion. The findings show that rapid urbanization and population growth in the Lavasanat watershed have altered its ecological structure, resulting in the conversion of agricultural lands into other land-uses. Most of the agricultural lands that have undergone such conversion have turned into residential areas. Further, the results show a decrease in the number and total area and a change in the shape and size of patches of agricultural land in the Lavasanat watershed, causing these patches to turn into small-sized irregularly shaped fragments. The observed changes in the landscape structure indicate that the destruction of the landscape can mostly be attributed to the emergence of new small-sized patches. These results are consistent with the findings of Tong et al. (2017) and Aljoufie et al. (2013) who attributed the changes in land-use and landscape patterns to rapid urbanization and population growth.

The results indicate that poorly controlled real estate development has greatly affected the ecological structure of the Lavasanat watershed. For example, from 2000 to 2020, large agricultural lands, which are of greater ecological value, have been consistently fragmented and replaced with smaller patches with lower ecological value, thus leading to lower ecological security, which is consistent with the results of similar studies (Alberti and Marzluff 2004; Su et al. 2010: 61; Asfaw and Worku 2019: 145). Essentially, the expansion of residential areas has acted as an interrupting factor, breaking the initial continuity of the watershed’s vast natural expanses and undermining its balance and functional dynamism, thereby causing or exacerbating environmental problems such as air pollution, soil erosion, and increased runoff.

Relationship between ecosystem services and ecological security

The concept of ecosystem services and its application in urban environments can offer useful opportunities for gaining a better understanding of the interactions between humans and nature (Tobias 2013). Ecosystem service indicators play a key role in the assessments of ecological security. Therefore, the needs of humans and ecosystems should be analyzed in this framework to determine the deficiencies of these services (Chen et al. 2018). Water supply is a primary component of ecosystem services (Brisbane 2007) that highly depends on natural, economic, and human activities (Sun et al. 2016; Yang et al. 2016; Jie et al. 2015; Liquete et al. 2011). Human activities affect the yield, volume, and availability of water resources by affecting climate, land-use, and water quality (Smith 1997; Chen et al. 2016; Liu et al. 2017).

The results of the modeling of WY in the Lavasanat watershed for the last two decades showed an increase in WY because of land-use changes caused by human activities. These results showed that the WY related to residential areas has increased by 1.36-fold from 2000 to 2010 and by 2.14-fold from 2010 to 2020, which sum up to a 2.92-fold increase over these 20 years. Correspondingly, the total area of residential areas in the watershed has increased by 1.42-fold from 2000 to 2010 and by 1.51-fold from 2010 to 2020, which sum up to a 2.15-fold increase for these 20 years. These results are consistent with the findings of several other studies that have demonstrated the great impact of human activities and the expansion of residential areas on the volume and availability of water resources (Smith 1997; Chen et al. 2016; Yang et al. 2016; Sun et al. 2016; Liu et al. 2017).

Analyses show that the total WY of the watershed has experienced a 1.25-fold increase from 2000 to 2010 and a 2.33-fold increase from 2010 to 2020, resulting in a 2.92-fold increase in water yield over the entire period between 2000 and 2020. Our results indicated that climatic data can have a significant impact on WY. For example, an increase in precipitation of the year 2020 was found to cause a sudden increase in the WY of this year. This result is consistent with the results of many other studies that have shown the great effect of precipitation on the outputs of the WY model (Rahimi et al. 2020; Boithias et al. 2014; Terrado et al. 2014). Finally, it was also observed that the watershed’s WY is greatly influenced by climatic factors, among which precipitation is the most effective. Nevertheless, the effect of the type and density of vegetation cover on WY should not be ignored. Thus, considering the importance of securing water supply for the people living in the region and the ongoing problems in this regard, the model can be used to obtain a relative estimate of WY and assess the role of vegetation in changing the WY so as to recharge underground aquifers.

Conclusion

While the environment consists of countless interconnected systems and components, in modern times, LULC changes have emerged as one of the most important determinants of change in ecological systems (Mondal and Southworth 2010: 1716). In this study, the LULC maps obtained from satellite images (2000, 2010, and 2020) and physical data were used to simulate urban expansion and LULC changes in the Lavasanat watershed. Then, landscape metrics (at the class and landscape level) and ecosystem services (water yield) were used to assess the ecological security of this area.

One of the most important processes commonly observed in urbanizing landscapes is fragmentation, which represents the disturbances caused by human activities in the landscape structure and function (Ahern and Andre 2003: 65–93). Hence, landscape fragmentation also has an impact on the region’s ecosystem services and functions (Peng et al. 2016: 1083). The results of this study showed a rapid growth in the region’s residential areas, which has caused the landscape to disintegrate and fragment, thereby undermining the region’s ESs and consequently ecological security over the past 20 years. For example, the increasing number of patches of built areas in the region and their integration with each other has increased the region’s water yield, or in other words, led to a significant positive relationship between LPI and water yield in residential areas, which is consistent with the results of Hu et al. (2020). While the results showed a degree of shrinking and degradation in rangelands and bare lands of Lavasanat watershed, these changes are not as critical as changes in agricultural lands, although this could be because most of these lands are located in areas with high slopes and elevations, which are out of human reach.

The results also showed that given the short distance between urban and rural areas and the presence of several low-density and scattered settlements in the region that face no physical or geographical barriers for expansion, the existing patches of built areas are likely to continue growing and join together. Because of this trend, in the future, the region will have fewer patches, less fragmentation, simpler and more regular geometries, and more cohesiveness and continuity at the landscape level. At the class level, the results show that with the continuation of the current trend, there will be a decrease in the number of all kinds of patches except bare lands. While in the built areas, this will be due to the joining of small patches emerged in previous years, in other classes, this will be mostly due to the conversion of small patches to residential areas.

Moreover, the results of the InVEST scenario-generator model and the water yield model showed that the expansion of built (residential) land-uses in the studied region has led to significantly increased water yield and, consequently, increased runoff. It was also found that the sudden increase in water yield in 2020 can partly be attributed to increased precipitation due to climate change in this year. This demonstrates the great importance of precipitation as an input of the InVEST model, which has also been proven in many other studies (Yang et al. 2019; Hamel and Guswa 2015; Marquès et al. 2013). Also, the maps produced by the model indicate that the expansion of residential areas and the shrinking of agricultural areas and rangelands (because of conversion into residential areas) have disrupted the water balance of the Lavasanat watershed, causing increased runoff in this area. These changes will certainly have significant environmental consequences, such as increased soil erosion, reduced regional groundwater recharge, and easier contaminant displacement.

In conclusion, it should be stated that further expansion of infrastructure and human activities in the Lavasanat region without attention to its capacities will certainly pose many problems for the ecological security of the Lavasanat watershed in the future. Therefore, to avoid these problems, it is essential to formulate management plans in line with the principles of sustainable development, make an assessment of landscape potentials, and make sure that the region’s resources are utilized properly so as to prevent unintended change in the region’s structure and preserve its spatial cohesion. Also, it is necessary to adopt a new development plan for the Lavasanat watershed with a focus on the protection and preservation of its green areas, because otherwise the continuation of the current trends is expected to lead to further degradation of natural lands and the increased expansion of built land-uses and specifically residential areas.