Introduction

Qanat (phonetically in the Persian language Qanāt or Kāriz) is defined as an underground construction system, which linearly collects and delivers groundwater from highland mountainous zones to flatland residential or agricultural areas (Naghibi et al. 2015, 2018). A typical Qanat has many different parts, including the gallery or underground tunnel for collecting and conveying water, the mother well; the shaft wells; the outlet; and the surface channels to distribute water (Yazdi and Khaneiki 2013; Kharrat et al. 2019).

Qanat system, as a great human heritage, contributes to the sustainable management of groundwater (Salih 2006). Historically, the Qanat system has provided water for human settlements for centuries, particularly in arid and semiarid regions in more than 34 countries (Fattahi 2015). Nowadays, Qanats have a wide distribution in many places around the world, such as the Middle East (Iran, Oman, and Afghanistan), North Africa (Tunisia, Algeria, and Morocco), and Mediterranean Europe (Spain and Italy) (Al-Ghafri et al. 2003).

In the initial literature review, Beaumont (1971) noted that the Qanat system was an important feature of the Persian (Iranian) landscape during the Achaemenid Empire (550–330 B.C.). Now, the oldest (~ 900 years ago) and longest (~ 80 km) features of the active Qanat system seem to belong to Yazd city in central Iran (hydrological basin of Siahkuh), which has been classified as a national and world heritage monument (Yazdi and Khaneiki 2013; Kharrat et al. 2019). In this regard, Papoli-Yazdi and Khaniki (2000) have noted central Iran as a source of Qanat (Kariz) civilization in the world. Furthermore, Qanat water systems have supplied the majority of Iranian settlements for several centuries (Abbasnejad et al. 2016).

Qanat systems have been constructed in areas with low-yielding aquifers but in proximity to mountains and hills with a high potential for groundwater resources (Nasiri and Mafakheri 2015). The constructional technology of the Qanat system depends on the environment and culture of the Iranian region, and in the year 2020, approximately 40,000 lines of Qanat water lines with 4.7 billion cubic meters of mean annual discharge have been recognized in the data center of the Iranian Ministry of Agriculture (IMA 2020), entitled national Qanat geo-information system. A review of the time series of the Qanat data between 2008 and 2018 indicates that the mean annual discharge of the Qanat water system has declined nearly by 50%. This declining value mostly is due to climate change and the overloaded consumption of groundwater in Iran (Zamanirad et al. 2018; Khaneiki 2019).

Major parts of Iran are faced with very limited water availability, and more than 90% of the country is under arid or semiarid conditions (Behling et al. 2022). Besides, the climate change and water scarcity facts in Iran are very intensive among the Middle Eastern countries with an increase of 2.6 °C in mean temperatures and a 35% decline in precipitation in the next decades (Mansouri Daneshvar et al. 2019a). Together with the springs and wells, the Qanat system is categorized as an important groundwater resource type in Iran (Pourghasemi and Beheshtirad 2015). Different environmental variables, such as topography, hydrology, and lithology characteristics, can influence groundwater resource types. Thus, many researchers have attempted to model and map the geographical groundwater features (Oh et al. 2011; Ozdemir 2011; Cortez and Embrechts 2013; Pourtaghi and Pourghasemi 2014; Rahmati et al. 2015; Pourghasemi and Beheshtirad 2015; Naghibi et al. 2016).

Particularly, researchers have addressed Qanat potential map using several environmental factors, implementing empirical models in geographical information systems (GIS) in recent years (e.g., Naghibi et al. 2015; Golkarian et al. 2018). However, there is no observed attempt Goto assess the efficiency of each variable in the construction of the current situation of Qanat water lines and locations in Iran. Particularly, there have been no studies to represent the certain relationships between climatic or geological variables and the Qanat water system, separated in the different parts of Iran. Hence, the novelty of the research is to implement new development of a GIS-based statistical method to understand the causal evidence of the Qanat water system in the major hydrological basins of Iran.

Given the aforementioned arguments, the main aim of this paper is to efficiency assessment of the environmental variables in the construction of the Qanat water system in Iran. For this purpose, the paper develops a set of environmental variables in GIS as the high-resolution gridded and independent input layers in addition to the dependent distribution of the constructional lines of the Qanat geo-information system. Using the statistical method of the area under the curve (AUC) of receiver operating characteristic (ROC), the efficiency of 20 environmental variables is assessed in the constriction of the Qanat water system in Iran. The AUC index is a standard measure, which can be used for assessing the accuracy and discrimination of distributive models (Lobo et al. 2008; Jiménez-Valverde 2012) by analyzing the efficiency of the variables (Gonçalves et al. 2014).

Geographical setting of the study area

Based on the underground water data of Iran (WRM 2020), the total surface of Iran has been divided into 30 major hydrological basins, which have the hydrological homogenous to conduct the total surface and underground water into the plain or aquifer outlets (Fig. 1a). The general names, codes, statistics, and physical data of these basins are shown in Table 1. The largest basins are included four cases of Kavir, Lut, NamakTehran, and Jazmurian basins with a total area of 595,000 km2 (approximately 37% of the total Iran area), which are categorized as the continental endogen basins with outlets inside the Iran plateau. Furthermore, the most value of mean annual groundwater discharge belongs to the NamakTehran basin (~ 10,000 Million m3) due to its largest share of population density in Iran (~ 20% of the total Iran population). In the present study, divisions of these 30 major hydrological basins are considered spatially as study areas (cases) to correspond to the distribution of dependent and independent variables in GIS.

Fig. 1
figure 1

Geographical position of the study areas including a. 30 major hydrological basins of Iran and b. spatial distribution of Qanat water lines

Table 1 General names, codes, statistics, and physical data of 30 major hydrological basins in Iran (after WRM 2020)

The paper develops a set of environmental variables in GIS as the high-resolution gridded and independent input layers in addition to the dependent distribution of the constructional lines of the Qanat geo-information system. In this regard, the spatial distribution of Qanat water lines (counts of Qanat lines) and their mean annual discharge (based on WRM 2020) are assumed as dependent variables of the research shown in Fig. 1b. The mentioned Qanat data are divided based on 30 major hydrological basins in Table 2, which revealed the largest share of constructional lines (> 4000 lines) and annual discharge of the Qanat water system (> 300 Million m3/y) belongs to NamakTehran, Kavir, and Lut basins in central Iran. However, the largest share of Qanat discharge from total groundwater discharge (> 25%) belongs to Siahkuh, Baluchestan, and Mashkil basins in central and southeastern Iran.

Table 2 Distribution of Qanat water lines and their mean annual discharge based on 30 major hydrological basins in Iran (after WRM 2020)

Methodology

Data preparation

In the present research, some national and international databases were considered to prepare environmental variables. In the first step, the spatial and statistical data for the Qanat water system, i.e., the constructional distribution of Qanat water lines, were obtained from the water resources management company of Iran (https://data.wrm.ir), as the dependent variable of the research (WRM 2020). In this regard, the counts of Qanat lines in each basin were assumed as the dependent variable. The Qanat line refers to the length of the underground tunnel, which is reflected as a line in the shape files in GIS with an integer value. The source data layer has been created between 2018 and 2020 in the vector with a spatial scale of 1:250,000.

In the next step, about 20 independent environmental variables were gained from several GIS-based databanks to define their efficiency assessment in the construction of the Qanat water system in Iran. In this regard, environmental variables were categorized as [1] basin total area, [2] basin total plains, [3] total groundwater discharge, [4] mean elevation height, [5] mean annual temperature, [6] mean annual precipitation, [7] drought hazard index, [8] climatic aridity index, [9] geological landforms, [10] soil units, [11] geophysical gravity index, [12] fault density, [13] land covers, [14] settlement density, [15] population rate, [16] urban points, [17] drainage density, [18] basin runoff coefficient, [19] peak flood discharge, and [20] annual groundwater consumption. The method for preparing and mapping mentioned data at the country level is adopted in GIS software as observed in research by Rabbani et al. (2021).

The mean values of annual temperature and precipitation, reanalyzed from the historical data, were obtained from a long-term gridded global database of world climate (https://www.worldclim.org/data/index.html). Besides, the climatic zone of the study area was reproduced in GIS based on the historical bioclimatic data of the aforementioned database (see Fig. 2a–c). This set of global climate layers for the period of 1970–2020 in the raster format has been downloaded with a spatial resolution of about 1 km2 at 30 s of a longitude/latitude degree. Topography-based data, e.g., elevation heights, basin locations, and plain territories were obtained from the global digital elevation model (http://www.gdem.aster.ersdac.or.jp/search.jsp) (see Fig. 2d). This layout with a spatial resolution of ~ 90 m (converted to 100 m) has been created in 2011 as geo-referenced tagged image file format (GeoTIFF).

Fig. 2
figure 2

Input environmental variables to efficiency assessment of the construction of the Qanat water system including a. mean annual temperature, b. mean annual precipitation, c. climatic zones, and d. mean elevation heights

The soil data were considered from a global soil information dataset, namely soil grids (https://soilgrids.org) (see Fig. 3a). International soil information center released the first version of soil grids at 1 km spatial resolution in 2010 (Hengl et al. 2014). The geological landform data in the GIS-based raster format (see Fig. 3b) and the geophysical airborne gravity data were taken from the national geosciences database of Iran (http://www.ngdir.ir) These gridded data layers have been produced in 2011 with the spatial scale 1:250,000 (see Fig. 3c). Land covers and land-use types were derived from the global land-use and land-cover change database (http://geosimulation.cn/download/GlobalSimulation/Global_MODIS_2010) (see Fig. 3d). This data layer retrieved from the land-cover types of MODIS products in 2010 at an original resolution of 1 km (Li et al. 2017). Furthermore, some GIS-based vector and raster data for the physical environment, such as drainage network, settlement, and urban points, fault lines, and groundwater consumption (see Fig. 4), were gathered based on 2018 ESRI vector shape files with the spatial scale 1:250,000 from the NGDIR (2020). The last reports of population data (2020), adapted in the study basins, were used from the statistical center of Iran (http://www.amar.org.ir).

Fig. 3
figure 3

Input environmental variables to efficiency assessment of the construction of the Qanat water system including a. geology (major landforms), b. pedology (major soil units), c. geophysical gravity, and d. land-cover map

Fig. 4
figure 4

Input environmental variables to efficiency assessment of the construction of the Qanat water system including a. drainage density, b. settlement density, c. fault density, and d. annual groundwater consumption

GIS-based spatial analysis

In this paper, the research analysis is based on the GIS approach and statistical method of the AUC index. The GIS software application in a grid environment can be used as the basic analysis tool for spatial data management in each environmental study (Malczewski 2006). The spatial analysis using GIS can also help in managing big data quantification and the geo-environmental visualization (Bagherzadeh and Mansouri Daneshvar 2013). The integration of spatial and temporal research in the GIS can provide an improved basis for addressing multi-criteria decision-making approaches in the hydrological studies and Qanat system researches (e.g., Ozdemir 2011; Oh et al. 2011; Lee et al. 2012; Naghibi et al. 2015, 2018; Golkarian et al. 2018).

All independent variables with different classes were mapped through GIS (as an elaborated method by Ebrahimi et al. 2019) and were converted to gridded (raster) factor layouts with a pixel resolution of 1 × 1 km. Then, these factor layouts were transferred as attribute tables based on each hydrological basin in the zonal statistical extension of GIS. Some attribute tables can represent the classes of an independent variable (such as three classes of alluvial sediments, evaporate deposits, and carbonate rocks in the geological map or three classes of low, moderate, and high in the fault density map), and some other attribute tables can represent raw integers (such as mean annual precipitation amounts or mean elevation heights).

Data statistical analysis

Some independent indices were used in the present study based on hydrological modeling and analysis. For instance, the runoff coefficients of different basins were determined by overlapping the environmental variables of precipitation, elevation, and land physical characteristics (e.g., Del Giudice et al. 2014; Mousavi et al. 2019). Besides, determination of the peak flood discharge in each basin was determined based on the rational formula as expressed below equation (Parak and Pegram 2006; Heidari et al. 2021):

$$Qp = Cr \times Pi \times A \times 3.6$$
(1)

where Qp is the flood peak discharge (m3/s), Cr is the runoff coefficient (unitless), Pi is the maximum precipitation intensity (mm/h), and A is the watershed area (km2).

Ultimately, the efficiency of 20 environmental variables was assessed in the constriction of the Qanat water system in Iran, using the national databases and statistical method of the area under the curve (AUC) of receiver operating characteristic (ROC). In this regard, the overall performance of a model is drawn by plotting the independent variable (test variable) or specificity on the x-axis against the dependent variable (state variable) or sensitivity on the y-axis in a ROC curve (Liuzzo et al. 2019; Pirnia et al. 2019). Thereafter, a confusion matrix is needed to produce the ROC plot, as a contingency table for describing the performance of the model (Yang and Berdine 2017). In the confusion matrix (Table 3), four conditions can be defined below: [1] true positive (TP) or perfect sensitivity, [2] false negative (FN), [3] false positive (FP), and [4] true negative (TN) or specificity. In the present study, the dependent actual state variable is defined as the Qanat water lines, and the independent test variables are defined as 20 environmental variables.

Table 3 Confusion matrix of the study based on the major hydrological basins

For plotting ROC curves under the ‘Analyze’ tab in the SPSS, the positive (= 1) actual state of Qanat water lines is categorized as values > 200 number of lines. Contrarily, the negative (= 0) actual state of Qanat water lines is categorized as values < 200 number of lines. Besides, the positive and negative classes or integers for test variables are categorized as valid = 1 (or high ~ 1) and null = 0 (or low ~ 0) values of the environmental variables, respectively. For this purpose, the classes were transferred as separated columns in SPSS with binary values of valid (= 1) or null (= 0) for the basin cases. Besides, the integers were converted as contiguous digits with standardized and dimensionless measures ranging from low (~ 0) to high (~ 1).

According to the ROC plots, the area under the curve (AUC) is the most commonly used index, defining the probability of a test with the average specificity toward sensitivity values of the ROC plots (Khatami et al. 2023). According to Table 4, the acceptable, excellent, and outstanding discriminating values for the test variable should indicate the AUC > 0.6, whereas poor and random chance gives the AUC < 0.6 (Lasko et al. 2005; Khatami et al. 2023). In this research, the AUC indices are used to examine the research idea, regarding the efficiency assessment of the environmental variables in the construction of the Qanat water system in Iran. Ultimately, the Pearson correlation coefficients between the dependent and independent variables were produced to indicate the constant relationships between the Qanat water system and environmental parameters. The correlation test method is in line with the methodology of the worldwide research performed to distinguish the relationships between environmental effects and anthropogenic systems (e.g., Sravari 2019; Mansouri Daneshvar et al. 2019b).

Table 4 Determination of discriminations for AUC values (after Khatami et al. 2023)

Results

Estimation of the climatic and geological data

The feasibility of the construction of Qanat systems depends on several climatic, geographical, and environmental characteristics (Nasiri and Mafakheri 2015). On this basis, input environmental variables are categorized to efficiency assessment of the construction of the Qanat water system based on the study basins. The first category of variables is the climatic data, including climatic variables of annual temperature, precipitation, drought, and aridity indices, which are given in Table 5 using the zonal statistical extension of GIS for each basin. Moreover, the geological data are considered for efficiency assessment, including geological landforms, soil units, gravity index, and fault density (Table 6). The tables revealed that the lowest precipitation records are observed in the three basins of Siahkuh, Hirmand, and Lut (< 100 mm), and the highest temperature records are included Baluchestan and Sadij basins (> 24 °C). The highest drought hazards over 40 years (from 1975 to 2015) revealed a trend of more frequent occurrence in the basins of Aras and Urmia over the northwestern regions of Iran (adapted by Mansouri Daneshvar et al. 2013a). Meanwhile, the highest aridity index of climatic zones of Iran including slight to extreme heat stresses, is observed in 21 major hydrological basins over the central and eastern regions of Iran (adapted by Mansouri Daneshvar et al. 2013b).

Table 5 Inventory of environmental variables of 30 major hydrological basins in Iran, including mean annual temperature, mean annual precipitation, drought hazards, and climatic zones
Table 6 Inventory of environmental variables of 30 major hydrological basins in Iran, including geological landforms, soil units, gravity index, and fault density

On the other hand, the basins can be classified based on their geological characteristics of three major landforms of alluvial sediments, evaporate deposits, and carbonate rocks. The distribution of the landforms revealed that alluvial sediments or the composition of alluvial deposits with other landforms covers the large numbers of basins (24 basins) dominantly in central and eastern parts of Iran. These basins also include the soil units of aridisols and salt flats, which are the indicators of the arid and semiarid regions. The distribution of gravity index and fault density values indicated that all basins with thick alluvial sedimentations or other deposits, low fault density and tectonic, and low density of crust rocks (Mansouri Daneshvar 2015) could be categorized as low classes of gravity index (negative values), such as playa fields in Lut, Kavir, and Siahkuh basins. Contrarily, the composition of metamorphic and volcanic landforms with high fault density and tectonic activity could produce high classes of gravity index (positive values), such as metamorphic zones in the Baluchestan and Qaraqum basins.

Estimation of the hydrological and anthropogenic data

The runoff coefficient and peak flood discharge are important parameters to estimate flood discharge in basin management (Zeinali et al. 2019). According to the rational and empirical models and equations (e.g., Eq. 1), the values of runoff coefficient and peak flood discharge in the basins were typically determined by overlapping land covers, elevation characteristics, and climatic precipitation data (Mousavi et al. 2019; Heidari et al. 2021). In this regard, the basins with high values of runoff coefficient were identified as Mond and Sadij basins in southern Iran, which are covered dominantly by pastureland and rangeland. However, high values of peak flood discharge (> 4500 m3/s) have been recorded for NamakTehran and Kavir basins in central Iran, which are covered by bareland and saltland dominantly. Based on the drainage density data, the high values can belong to the Haraz, Qarasu, and Atrak basins in northern Iran, which are dominantly covered by woodland and dense pastureland, without any spatial relation with the locations of the Qanat water system (Table 7).

Table 7 Inventory of environmental variables of 30 major hydrological basins in Iran, including land covers, runoff coefficient, average of peak flood discharge, and drainage density

Meanwhile, groundwater plays a dominant role in the sustainable development of human societies, especially in arid and semiarid countries (Pourghasemi and Beheshtirad 2015). Each Qanat system collects groundwater for end users of human activities in the urban and rural settlements and agricultural consumptions. One of the sets of environmental variables can be titled the anthropogenic characteristics, population rate, settlement density, urban points, and groundwater consumption (Table 8). On this basis, the study basins can be classified based on the different statuses. In this regard, the familiar basins with a high rate of population, settlements, urban points, and water consumption are entitled as NamakTehran and Kavir basins.

Table 8 Inventory of environmental variables of 30 major hydrological basins in Iran, including annual groundwater consumption for agriculture uses, settlement numbers, population rates, and main cities

Discussion

Relations between the Qanat system and basin physical parameters

The ROC plots for the determination of AUC indices were produced between the independent environmental variables and dependent distribution of Qanat lines to reveal the efficient characteristics, of constructing the Qanat water system, which is shown as a comprehensive matrix in Table 9. The ROC plots between the Qanat system and basin physical parameters, e.g., basin total area, basin total plains, basin total groundwater discharge, and basin mean elevation heights, were estimated with the AUC indices from 0.634 to 0.830 (i.e., AUC > 0.6), revealing acceptable and excellent discrimination. It means that overall basin physical parameters have an effective role regarding the higher construction of the Qanat system in Iran. In other words, the construction of Qanat systems is observed where the basins have large values of the area, plain extension, groundwater discharge, and elevation height. The mean elevation height was recognized as the best parameter influencing the Qanat system with an exact AUC equal to 0.830 (Fig. 5a). Based on the physical parameters, previous scholars have divided the constructional pattern of Qanat systems, followed by the geomorphic and topographic shapes of the areas (Ahmadi et al. 2010).

Table 9 Efficiency assessment of the environmental variables in the construction of the Qanat water system based on the AUC values
Fig. 5
figure 5

ROC plots to sensitivity analysis between Qanat water lines and some selected variables a. mean elevation heights, b. geological landforms, c. t land covers, and d. groundwater consumption for agricultural uses

Relations between the Qanat system and climatic variables

The relationships between climatic variations and groundwater discharge in the Qanat systems have been confirmed in the literature (e.g., Beaumont 1971; Khaneiki 2019), while the constant effects of climatic variables in the construction of Qanat lines have not been studied previously. In the present study, based on the ROC plots for analysis of efficient relations between the state variable of Qanat water lines and test variables of most climatic factors, e.g., mean annual temperature, mean annual precipitation, and annual drought hazards, the AUC indices were calculated from 0.281 to 0.557 (i.e., AUC < 0.6), revealing poor conditions or no discrimination.

Although most of the Qanat systems are situated in areas with annual precipitation below 300 mm (Beaumont 1971), our result concludes that the elemental factors of climate, such as temperature and precipitation, have poor discrimination and not enough controlling role in the creation and construction of Qanat system in Iran.

Scholars noted that the Qanat system has a harmony between farming and the local climate by providing solutions for overcoming drought hazards (e.g., Kharrat et al. 2019), while our results indicated that such findings have no certain role in the construction of Qanat water lines in the hydrological basin scales. Only an overall index of climate aridity showed good discrimination and an effective role in the construction of the Qanat system with an AUC index of 0.636. This fact verifies the distribution of Qanat water systems in the arid regions, as mentioned by Al-Ghafri et al. (2003) and Fattahi (2015).

Relations between the Qanat system and geological variables

The ROC plots between the Qanat water lines and geological landforms were analyzed to explore the efficiency of the different landforms of evaporate deposits, carbonate rocks, and alluvial sediments in the construction of the Qanat system with AUC indices of 0.392, 0.409, and 0.619, respectively. On this basis, the evaporate deposits and carbonate rocks have no discrimination level and no significant efficiency in the construction of the Qanat water system. Vice versa, results revealed acceptable discrimination and the effective role of alluvial sedimentation in the construction of the Qanat system in Iran with an exact AUC equal to 0.619 (Fig. 5b). This fact verifies the distribution of Qanat water systems over the Quaternary unconsolidated coarse alluvial and fluvial sediments with very high permeability, as mentioned by Beaumont (1971), Ahmadi et al. (2010), and Abbasnejad et al. (2016). Qanat systems lead water by the force of gravity and their mother wells are usually constructed on alluvial sediments at the bed of mountains and hills (Nasiri and Mafakheri 2015).

In the second step, the ROC plots between the Qanat water lines and soil units were assessed to explore the efficiency of two major units of inceptisols and aridisols with AUC indices of 0.205 and 0.795, respectively. On this basis, the aridisols have excellent discrimination and effective role in the construction of the Qanat system in Iran, revealing the expansion of Qanat water systems in the arid soils and regions (e.g., Salih 2006; Fattahi 2015). In the last step, the ROC plots between Qanat water lines and two geological factors of gravity index and fault density were assessed with AUC indices of 0.398 and 0.443, revealing no significant discriminations. This finding exposed that the basins with low gravity index, which have possible thick sedimentations and low density of crust rocks (Mansouri Daneshvar 2015), hosted the higher construction of the Qanat system in Iran. This fact again confirms the excellent discrimination and effective role of Quaternary aged sedimentation landforms in the expansion of Qanat water systems in arid soils, such as playa fields in NamakTehran, Kavir, Lut, and Siahkuh basins.

Meanwhile, the results revealed that the fault density in the major hydrological basins of Iran has no significant efficiency in the construction of the Qanat water system. This fact can reject the claim of Naghibi et al. (2015), regarding the strongest relationship between fault density and Qanat occurrence. Contrarily, our results confirm previous findings by Golkarian et al. (2018) and Naghibi et al. (2018), noting the high importance of elevation heights and less importance of fault density in the occurrences of the Qanat system.

Relations between the Qanat system and hydrological variables

The ROC plots between the Qanat water lines and land-cover map were analyzed to explore the efficiency of the different land covers of pastureland, farmland, and bareland in the construction of the Qanat system with AUC indices of 0.551, 0.409, and 0.614, respectively (Fig. 5c). On this basis, the pastureland and farmland have poor discrimination level (i.e., AUC < 0.6) in the construction of Qanat water system. Vice versa, very good discrimination and the effective role of bareland cover (i.e., AUC > 0.6) were identified in the construction of the Qanat system in Iran. This fact verifies that the distribution of Qanat water systems has been developed over the areas without crops or poor plant covers, e.g., poor rangelands and pasturelands. The Qanat system, as a horizontal, interconnected series of underground tunnels to accumulate and deliver groundwater from a mountainous source district along a water-bearing formation (aquifer) (Perrier and Salkini 1991; Naghibi et al. 2015), has been created for supplying water in the faraway agricultural areas and settlements. Hence, its nature of construction is developed in the regions, such as dry deserts and playas in the Kavir region (see Beaumont 1971), which are not supplied by surface waters without sufficient land covers of woodland, pastureland, or farmland. Qanat systems have enabled to create of new pastures and oases in the deserts and dominantly are constructed on deposits of the dried streambeds, outlets of the watersheds, and inside dry deserts of the central plateau of Iran (Ahmadi et al. 2010).

Moreover, the ROC plots between Qanat water lines and some hydrological indices of basin runoff coefficient, peak flood discharge, and drainage density were assessed with AUC indices of 0.574, 0.673, and 0.489, respectively. On this basis, only the peak flood discharge of the basins has acceptable discrimination and an effective role in the construction of the Qanat system in Iran, revealing the location of Qanat water systems in large-scale plains with high surface area and extreme flooding potential. One of the initial reasons for Qanat constructions in the arid regions of Iran depended on surface flooding control to organize groundwater utilization (Fattahi 2015). Although they reported direct relationship between drainage density and groundwater potential in the literature (Pourghasemi and Beheshtirad 2015), our results revealed poor discrimination and not enough controlling role of drainage density in the creation and construction of the Qanat system in Iran.

Relations between the Qanat system and anthropogenic characteristics

Ultimately, the ROC plots between the state variable of Qanat water lines and test variables of anthropogenic characteristics were analyzed to explore the efficiency of the different characteristics of settlement density, urban points, population rate, and total groundwater consumption in Qanat system with AUC indices of 0.693, 0.665, 0.571, and 0.847, respectively (Fig. 5d). Except population rate, all anthropogenic characteristics have the acceptable discrimination level and significant efficiency (i.e., AUC > 0.6) in the construction of Qanat water system. These results can be associated with previous research, noting Qanat water systems as the main supplier of water for rural and urban settlements of central Iran, for hundreds of years (Nasiri and Mafakheri 2015). Confirming the result of this research, the interrelations between the Qanat system and water consumption in the settlements have been studied previously in the systematic nexus (Khaneiki 2019).

Effective variables

From the physical viewpoint, Table 9 reveals that basin parameters, i.e., basin total area, basin total plains, total groundwater discharge, and mean elevation heights, have an effective role regarding the higher construction and extension of the Qanat system in Iran. From the climatic viewpoint, only an overall index of climate aridity showed acceptable discrimination. From the geological viewpoint, the Quaternary aged alluvial sedimentation class of geological landforms and arid soil class of soil units have an effective role in the construction of Qanat water lines, dominantly in the arid and semiarid regions (e.g., Al-Ghafri et al. 2003; Salih 2006; Ahmadi et al. 2010; Abbasnejad et al. 2016). From the hydrologic viewpoint, the bareland class of land covers and the peak flood discharge of the basins have acceptable discrimination, revealing the location and distribution of Qanat water systems in the plains with poor plant covers and extreme flooding potential to organize groundwater utilization (Fattahi 2015). From the anthropogenic viewpoint, the characteristics of seal-Ghsattlement density, urban points, and total groundwater consumption have significant efficiency in the construction of the Qanat system, as observed in earlier studies (e.g., Nasiri and Mafakheri 2015, and Khaneiki 2019).

Correlation coefficients

In this section, the correlation coefficients were calculated to reveal statistical evidence of the environmental effect on the Qanat water system (Table 10). This table confirmed the positive and significant relationships (with R from 0.304 to 0.796 at confidence 90%) between the Qanat system and 12 environmental parameters (i.e., basin total area, basin total plains, groundwater discharge, elevation heights, climatic aridity, alluvial sediments, aridisols, barelands, peak flood discharge, settlement density, urban points, and groundwater consumption). The strongest relationships are observed between the Qanat system and four parameters of groundwater consumption, groundwater discharge, basin total area, and settlement density with R > 0.7. In parallel, the negative or insignificant relationships between other parameters and the Qanat system were estimated. These coefficients certainly confirmed the results of AUC indices and the status of effective variables represented in Table 9.

Table 10 Correlation coefficients between the Qanat system and environmental parameters

Research limitation and future research

The main limitation of this research was the condensed tests, including the AUC index and Pearson test, for analysis of correlations and efficiency assessment. This kind of research can be analyzed based on other geo-referenced and geo-statistical approaches or models with spatial and numerical inputs and outputs instead of GIS software. In this regard, the use of other tests of Nash Sutcliffe Efficiency (NSE) and Kling Gupta Efficiency (KGE) in future research is seriously recommended to gain new insight into the comparative results and efficiency analysis of the environmental variables in the Qanat systems.

As mentioned in the literature, the ancient technology of Qanat systems should be reconsidered for the provision of irrigation water in arid regions to control the total groundwater consumption (Nasiri and Mafakheri 2015). Hence, as a future avenue of research, the possible programs to construct Qanat water lines should consider the certain efficiency of the environmental variables in detail at the basin scales. Furthermore, future studies could be assessed geographically in the different scales of hydrological watersheds to test the spatial correlations between dependent environmental parameters and new dependent classifications of the Qanat system, such as the length of the gallery, the depth of the mother well, and mean annual discharge.

Conclusion

Regarding the constructional details of the Qanat systems, the efficiency and performance of these water systems can be depended on environmental attributes. However, the comparative analysis and assessments within the hydrological basins of Iran and their possible efficiency in the Qanat construction have not been noted in previous studies. Hence, the main aim of this paper was to efficiency assessment of the environmental variables in the construction of the Qanat systems in Iran. Using the national and international databases in GIS and the statistical method of the AUC index, the efficiency of 20 environmental variables was assessed in the constriction of the Qanat water lines. For this purpose, divisions of 30 major hydrological basins were considered spatially as the study areas to correspond to the distribution of dependent and independent variables in GIS.

The ROC plots for the determination of AUC indices between the state variable of the Qanat system and 20 test variables indicated acceptable or excellent discriminations (i.e., AUC > 0.6) for basin total area (0.807), basin total plains (0.634), total groundwater discharge (0.790), mean elevation heights (0.830), climatic aridity index (0.636), a geological landform of alluvial sediments (0.619), soil units of aridisols (0.795), the land cover of barelands (0.614), settlement density (0.693), urban points (0.665), peak flood discharge (0.673), and groundwater consumption for agricultural uses (0.847). On this basis, the aforementioned variables had enough significant and effective roles in the construction of the Qanat system in Iran.

Overall, most physical parameters and anthropogenic characteristics revealed significant effects in the construction of Qanat water lines. The Pearson correlation test also confirmed the positive and significant relationships (with R > 0.7 at confidence 90%) between mentioned environmental parameters (e.g., groundwater consumption, groundwater discharge, basin total area, and settlement density) and the Qanat system. Other test variables with poor or no discrimination level (i.e., AUC < 0.6), such as annual temperature, annual precipitation, drought hazards, carbonate deposits, evaporate deposits, soil unit of inceptisols, fault density, pasturelands, farmlands and drainage density, had no any spatial relation with locations of Qanat water system in Iran. As a highlight, the results rejected the higher effects of climatic elements of temperature and precipitation, fault density, and drainage density in the spatial expansion of Qanat water lines in Iran.