Mapping Groundwater Prospective Areas Using Remote Sensing and GIS-Based Data Driven Frequency Ratio Techniques and Detecting Land Cover Changes in the Yellow River Basin, China

: Groundwater is an essential resource that meets all of humanity’s daily water demands, supports industrial development, inﬂuences agricultural output, and maintains ecological equilibrium. Remote sensing data can predict the location of potential water resources. The current study was conducted in China’s Yellow River region, Ningxia Hui Autonomous Region (NHAR). Through the use of a GIS-based frequency ratio machine learning technique, nine layers of evidence inﬂuenced by remote sensing data were generated and integrated. The layers used are soil characteristics, aspect, and roughness index of the terrain, drainage density, elevation, lineament density, depressions, rainfall, and distance to the river from the location. Six groundwater prospective zones (GWPZs) were found to have very low (13%), low (30%), moderate (25%), high (16%), very high (11%), and extreme potentiality (5.26%) values. According to well data used to validate the GWPZs map, approximately 40% of the wells are consistent to very high to excellent zones. Information about groundwater productivity was gathered from 150 well locations. Using well data that had not been used for model training, the resulting GWPZs maps were validated using area-under-the-curve (AUC) analysis. FR models have an accuracy rating of 0.759. Landsat data were used to characterize the study area’s changes in land cover. The spatiotemporal differences in land cover are detected and quantiﬁed using multi-temporal images which revealed changes in water, agricultural, and anthropogenic activities. Overall, combining different data sets through a GIS can reveal the promising areas of water resources that aid planners and managers.


Introduction
The ability to access water is crucial for maintaining life.Since the beginning of the twenty-first century ref. [1], population growth, environmental, climatic, and socioeconomic variables have all contributed to an increasing worry about the demand for fresh water in the world.Water resources are necessary for the growth of the agricultural, urban, and industrial sectors.In terms of total freshwater resources, groundwater makes up around 20% of them, placing it behind glaciers and lakes which make up only 1% and 79%, respectively, of the world's total freshwater resources.It is one of the water reserves that can be used to combat the problem of water shortage [2,3] as it can be used when necessary and is an effective substitute for limited surface water, which is getting harder to find [1].Groundwater is increasingly needed around the world for a variety of reasons [4,5] because it is more predictable and fresher than surface water, less likely to be polluted, always present, widely accessible, has great natural quality, is clean and accessible, and is inexpensive [6].It is crucial to remember that 80% of the world's rural population relies Land 2023, 12, 771 2 of 20 on groundwater for their clean drinking water.It is an essential water resource since it provides drinking water to more than half of the population [7].
Because groundwater resources are buried beneath strata, it is vital to use prediction models to investigate and uncover water resources [8][9][10] utilizing various methods.Groundwater discovery by traditional means is more time-and money-consuming.Groundwater exploration, prediction, and regional estimation can be done using remote sensing (RS) and geographic information systems (GIS) [11,12].A GIS technique can be used to combine and interpret large amounts of geographical data to forecast and locate new water sources [13].The RS and GIS can be used to identify potential groundwater resource locations, as demonstrated by several studies [9,14,15].For instance, multi-criteria decision-making is a quick and economical method [5,10].
In groundwater modeling, several knowledge-based (overlay, AHP, Boolean logic, index overlays, and analytical hierarchy process) and data-driven (e.g., WOE, machine learning models, logistic regression, weights of evidence, linear regression decision tree analysis, and neural networks) techniques are frequently used [16,17].The frequency ratio (FR) approach, a data-driven and bivariate statistical method, assesses spatial relationships between the dependent variable, e.g., water wells, and the independent variables, namely classes of thematic layers, to assign a rating (r) rate to each factor [18,19].It is widely applied in probabilistic processes such as geohazards [20][21][22][23] and in groundwater exploration in different environments [3,[24][25][26][27][28][29][30][31].The FR method provided better results than the evidential belief function (EBF) method [16,32]) and gave valuable information when coupled with the overlay analysis.In this technique, several criteria such as geologic, topographic, climatic, and hydrologic information are utilized for the prediction of groundwater occurrences by preparing, normalizing, and combining these elements [33].
The permeability and porosity of aquifer elements are influenced by soil parameters [34] and soil geometry, as the existence of sand and gravel deposits boosts porosity, which is an important factor in recharging groundwater aquifers [35].This is due to the runoff of water through time-induced soil erosion [36,37].When they flow in the same direction, elevation significantly affects groundwater penetration and motion [38][39][40] as runoff lowers groundwater potential in elevated regions, which also lowers the capacity for recharging [41].Additionally, the vegetation has gotten bigger on the faces to the north and east that promote the recharging of groundwater.Geomorphologic characteristics such as TRI have garnered a lot of interest, since it is essential for groundwater transport and management in any research region.Moreover, with increasing distance from rivers, the likelihood of encountering groundwater decreases [38,42].The hydrologic element of rainfall (Rf) is a significant source of recharge [43,44].Precipitation has a significant impact on percolation and recharging.
The LU/LC changes require thorough research in order to be properly planned, utilized, and regulated for challenging environmental studies.Researchers can assess and monitor the dynamics of natural resources using the LU/LC by utilizing remote sensing techniques, which are crucial for sustainable management.In addition, the LU/LC offers crucial evidence of the availability of water supplies.The present study aims to apply the GIS-based hydride FR and weighted overlay technique for optimization and revealing promising areas for water resources through analysis of multi-factors derived from satellite data/images.

Study Area
The Yellow River, Ningxia Hui Autonomous Area, in central China is the location of the current study (Figure 1).It covers a region with coordinates of 35 Oriented from north to south, it drains the Yellow River basin.The Ningxia Hui Autonomous Area in northwest China is the area under investigation [45].It is situated in the area of the inner desert.Warm continental arid and semi-arid weather and adequate light energy are present at the study site.The average annual sunlight hours are expected to Land 2023, 12, 771 3 of 20 range from 2194 to 3082; the mean annual temperature will be between 5 and 9 degrees Celsius; and the average annual precipitation ranges from 200 to 680 mm, with most of the precipitation occurring in the summer.Groundwater may be divided into two groups: the northern area's water and the southern area's water of the Quaternary deposits, depending on hydrodynamic characteristics, aquifer types, and occurrence parameters of groundwater karst in the southern area [46,47].Autonomous Area in northwest China is the area under investigation [45].It is situated in the area of the inner desert.Warm continental arid and semi-arid weather and adequate light energy are present at the study site.The average annual sunlight hours are expected to range from 2194 to 3082; the mean annual temperature will be between 5 and 9 degrees Celsius; and the average annual precipitation ranges from 200 to 680 mm, with most of the precipitation occurring in the summer.Groundwater may be divided into two groups: the northern area's water and the southern area's water of the Quaternary deposits, depending on hydrodynamic characteristics, aquifer types, and occurrence parameters of groundwater karst in the southern area [46,47].

Data and Methods
This study combined topography, hydrologic, and meteorological data with remote sensing information from several sensors to show potential water resource areas.GIS methods were utilized to combine various thematic maps created from these data, including those of the soil, elevation, aspect, terrain roughness index, depression, lineaments, drainage density, distance to a river, and rainfall intensity (Figure 2).

Data and Methods
This study combined topography, hydrologic, and meteorological data with remote sensing information from several sensors to show potential water resource areas.GIS methods were utilized to combine various thematic maps created from these data, including those of the soil, elevation, aspect, terrain roughness index, depression, lineaments, drainage density, distance to a river, and rainfall intensity (Figure 2).
The DEMs were derived from the SRTM data (90 m cell size) of NASADEM 1arc second WGS84 data (NASADEM 1arc second WGS84).The stream network was generated by the 8D approach [48], and the stream density map was produced via a GIS program.Estimating the depressions to illustrate where water accumulates can similarly be done using the "fill-difference" method [10].The OLI and TIRS sensors are carried by Landsat 8, which was launched on 11 February 2013.To map the land use and cover, VNIR and SWIR wavelength regions are used.Here, mosaiced obtained sceneries that featured the OLI bands 2, 3, 4, 5, and 7 underwent image alterations and enhancing procedures.Landsat-8 data "LC08_L1TP_129034_20220312_20220312_02_RT" was utilized to display vegetation and the signature of water resources.The DEMs were derived from the SRTM data (90 m cell size) of NASADEM 1arc second WGS84 data (NASADEM 1arc second WGS84).The stream network was generated by the 8D approach [48], and the stream density map was produced via a GIS program.Estimating the depressions to illustrate where water accumulates can similarly be done using the "fill-difference" method [10].The OLI and TIRS sensors are carried by Landsat 8, which was launched on 11 February 2013.To map the land use and cover, VNIR and SWIR wavelength regions are used.Here, mosaiced obtained sceneries that featured the OLI bands 2, 3, 4, 5, and 7 underwent image alterations and enhancing procedures.Landsat-8 data "LC08_L1TP_129034_20220312_20220312_02_RT" was utilized to display vegetation and the signature of water resources.
The TRMM rainfall measurements provided information on average rainfall.The data can be downloaded from Giovanni/NASA on the time period covered by the data on average rainfall, which runs from January 1998 to November 2015.The soil map was collected from the Geo Network Web Portal for the FAO Soil Map.
The GWPZs are produced using a data-driven FR modeled on a GIS.Numerous forecasting techniques utilized this kind of multi-criteria decision-making process.The relative weights of each observation in this model, which is based on data from remote sensing, hydrology, and geology, are chosen by the user [49].The GIS approach uses a raster combination in which each layer's pixel is matched with a certain geospatial site.The fusion processes are then more adapted to combining qualities from several data sets into one output layer.
The FR model, which indicates the probability of incidence for a given feature, is a straightforward geographic evaluation tool that is utilized to predict the consistency between GRW promising areas and the effective factors [3,27].The recharge occurrence ratio for each subclass of conditioning factors is calculated, and the frequency ratio is applied The TRMM rainfall measurements provided information on average rainfall.The data can be downloaded from Giovanni/NASA on the time period covered by the data on average rainfall, which runs from January 1998 to November 2015.The soil map was collected from the Geo Network Web Portal for the FAO Soil Map.
The GWPZs are produced using a data-driven FR modeled on a GIS.Numerous forecasting techniques utilized this kind of multi-criteria decision-making process.The relative weights of each observation in this model, which is based on data from remote sensing, hydrology, and geology, are chosen by the user [49].The GIS approach uses a raster combination in which each layer's pixel is matched with a certain geospatial site.The fusion processes are then more adapted to combining qualities from several data sets into one output layer.
The FR model, which indicates the probability of incidence for a given feature, is a straightforward geographic evaluation tool that is utilized to predict the consistency between GRW promising areas and the effective factors [3,27].The recharge occurrence ratio for each subclass of conditioning factors is calculated, and the frequency ratio is applied to the total recharge.The FR ratio for each predictor is computed and normalized regarding the total area of the watershed.Based on their degree of correlation with the GWR potential inventory, the FR numbers for each sub-feature of GWR potential influence factors were estimated.The scores of thematic maps were multiplied (Equation ( 1)) to determine the GWPZ [50], as shown in the following equation: where T m is thematic maps and F c is subclasses.

Soil
Due to geometric characteristics and the nature of the soil of the surface layer regulating how much precipitation seeps into the subsurface aquifer, lithological parameters are crucial.The size, shape, and arrangement of the soil grains, as well as the pore system that corresponds to them, can all have a major impact on the vertical and lateral flow of water.The development and management of groundwater resources depend mostly on the soil.Based on the arrangement of soil grains controlling the infiltration and recharge, each soil type was given a set of soil and textural characteristics based on the SWAT Soil Data classes for the various category levels in soil classification.The soil map of the study area is divided into Calcic xerosols (loam), Luvic xerosols (sandy clay loam), Lithosols (loam), Eutric Gleysols (clay loam), and Calcic cambisols (loam), which account for 26.13, 0.66, 53.06, 19.29, and 0.87% of the total area, respectively (Figure 3a).

Elevation
The flow of groundwater and surface discharge is influenced by topography.The link between infiltration and low-altitude areas is positive.Considering this, it is likely that low-lying places accumulate surface water following severe storms that drain downstream networks.Due to the substantial infiltration of river water, groundwater potential is often high in plain floodplain zones but decreases at high elevations.The topographical layer is crucial in determining groundwater availability, recharge capacity, and water movement across the land [4].Therefore, the elevation map of the study region varies from 1087 to 2941 m.The elevation map (Figure 3b) was divided into three classes, 1087 to 1276, 1276 to 1479, 1479 to 1705, and 1705 to 1998, encompassing 19, 31, 25, 18, and 7% of the total area, respectively.

Aspects
The DEM in ArcGIS 10.8 is used to extract aspects.There are 10 categories for aspect measurements: flat (or no aspect direction), N, NE, E, SE, S, SW, W, and NW.Water re-

Elevation
The flow of groundwater and surface discharge is influenced by topography.The link between infiltration and low-altitude areas is positive.Considering this, it is likely that low-lying places accumulate surface water following severe storms that drain downstream networks.Due to the substantial infiltration of river water, groundwater potential is often high in plain floodplain zones but decreases at high elevations.The topographical layer is crucial in determining groundwater availability, recharge capacity, and water movement across the land [4].Therefore, the elevation map of the study region varies from 1087 to 2941 m.The elevation map (Figure 3b) was divided into three classes, 1087 to 1276, 1276 to 1479, 1479 to 1705, and 1705 to 1998, encompassing 19, 31, 25, 18, and 7% of the total area, respectively.

Aspects
The DEM in ArcGIS 10.8 is used to extract aspects.There are 10 categories for aspect measurements: flat (or no aspect direction), N, NE, E, SE, S, SW, W, and NW.Water resources are more prevalent north of the Earth's equator on slopes that look north and east than on slopes that stand facing south and west.The mountain's eastern and northern slopes get less sunlight than its southern and western sides.The aspect map (Figure 3b) which was derived from the DEM was divided into three categories: Flat (−1), N (0-22.5),NE, E, SE, S, SW, W, NW, and N.These categories, collectively account for 0.26, 8.45, 14.54, 10.98, 10.29, 12.08, 12.92, 11.29, and 12.

Topography Roughness Index (TRI)
The TRI, a geomorphometric index that identifies and quantifies the geometry of landsurface terrain in the area under investigation, has an impact on groundwater occurrence as well.Afterward, using ArcGIS and the Natural Break Classifier (NBC), the TRI map was separated into four classes.The TRI map (Figure 4b) was divided into five categories: 0.614 to 0.889, 0.535 to 0.614, 0.461 to 0.535, 0.379 to 0.461, and 0.111 to 0.379 which, respectively, covered 10, 24, 31, and 10% of the overall region.

Topography Roughness Index (TRI)
The TRI, a geomorphometric index that identifies and quantifies the geometry of land-surface terrain in the area under investigation, has an impact on groundwater occurrence as well.Afterward, using ArcGIS and the Natural Break Classifier (NBC), the TRI map was separated into four classes.The TRI map (Figure 4b) was divided into five categories: 0.614 to 0.889, 0.535 to 0.614, 0.461 to 0.535, 0.379 to 0.461, and 0.111 to 0.379 which, respectively, covered 10, 24, 31, and 10% of the overall region.

Depressions/Sinks
Areas with low surface elevations that might be flooded during rainstorms are indicated by topographic depressions.These depressions can be filled with precipitated water until the water level increases to the point where the flow can exit.These areas are suitable for water resources because they flood the low topography.ArcMap version 10.5's spatial analysis can identify depressions using SRTM DEMs.The research region's 95.8, 2.7, and 1.5% of the depression map (Figure 5a) were categorized into three classes: 0-1, 1-3.8, and

Depressions/Sinks
Areas with low surface elevations that might be flooded during rainstorms are indicated by topographic depressions.These depressions can be filled with precipitated water until the water level increases to the point where the flow can exit.These areas are suitable for water resources because they flood the low topography.ArcMap version 10.5's spatial analysis can identify depressions using SRTM DEMs.The research region's 95.8, 2.7, and 1.5% of the depression map (Figure 5a) were categorized into three classes: 0-1, 1-3.8, and 3.8-72 (Table 1).

Lineaments
Lineaments are linear geometrical structures that represent underlying cracks in the earth's surface [51].These structures can be used to better define the existing aquifers and discover new reservoirs near densely populated areas that promote fluid migration to the bottom [52] and lineaments that can provide information on water penetration and circulation [53].When compared to other topographical features, they are the porosity of secondary sources and are visible as textural changes on satellite photos.A lineament might be anything from a crack to a fracture to a master joint.Long, linear geological formations, topographic linearity, and the straight routes that streams follow are further examples [54].They affect how deeply runoff water permeates below the strata and replenishes the aquifer's groundwater flow and storage.The lineament density map (Figure 5b) was obtained after digitizing lineaments from satellite images and then classified using Natural Breaks intervals into five classes, viz., 0 to 3.3, 3.3 to 8.04, 8.04 to 13.30, 13.30 to 21.10, and 21.10 to 40.87, covering 29, 31, 26, 12, and 3 % of the study area (Table 1; Figure 5b).

Distance to River
Alluvial deposits are typically found in river channels, especially in semi-arid regions; therefore, being close to hydrological systems is important when looking for water resources.The ideal circumstances for excellent penetration and, consequently, groundwater replenishment are provided by neighboring rivers.Utilizing data on rivers that were gathered from the OSM dataset, the distance from rivers was determined using ArcGIS' "Euclidean distance function".Spatial Analyst Tools were used to determine the distance from the river, which was then divided into five categories (Figure 6a): 1480 to 1750, 1110 to 1480, 740 to 1110, 370 to 740, and 0 to 370.These categories, which covered 4.64, 12.74, 20.99, 29.20, and 32.42 percent of the total area, respectively, each represented a different distance from the river.
water replenishment are provided by neighboring rivers.Utilizing data on rivers that were gathered from the OSM dataset, the distance from rivers was determined using ArcGIS' "Euclidean distance function".Spatial Analyst Tools were used to determine the distance from the river, which was then divided into five categories (Figure 6a): 1480 to 1750, 1110 to 1480, 740 to 1110, 370 to 740, and 0 to 370.These categories, which covered 4. 64, 12.74, 20.99, 29.20, and 32.42 percent of the total area, respectively, each represented a different distance from the river.

Rainfall Data
The distribution and volume of rainfall determine how much water is present in a hydrologic system.Rainfall intensity in a region allows for surface water accumulation which promotes infiltration, recharging, and runoff.The frequent precipitation through the years is the main source of rivers, lakes, and recharging the groundwater aquifers [26,55].Because rainfall is the main source of recharging the groundwater aquifers through the years, it is essential for this process [56,57].
The area that is exposed to storms that bring rain encourages groundwater recharge when it rains.Researchers can monitor, document, and measure the frequency of rainstorms in the watershed under study using information from the CRU data.Due to the high precipitation and flooding caused by this downpour, there was a scarcity of water supply, and infrastructure was severely damaged.
The study area's yearly mean rainfall from 2011 to 2020 was provided by the Climate Research Unit (https://crudata.uea.ac.uk/cru/data/hrg/ accessed on 10 January 2022).The precipitation point format was modified into a raster format using Multidimensional Tools of Spatial Analyst.We next converted the raster format to points by the Conversion Tools [49], which were interpolated by applying the Kriging method.

Drainage Density (Dd)
Drainage density, which is an effective indication for predicting infiltration rates, controls the relationship between water flow and water penetration in a terrain [58].By dividing the total number of streams and rivers in a drainage basin by its surface area, the drainage density of that basin can be determined [58,59].The kind of vegetation, the soil's ability to absorb precipitation, the slope gradient, and the bedrock nature and structure all have an impact on where rainfalls [1].Locations with less surface runoff and high drainage density have higher infiltration rates.A zone with high groundwater capacity is indicated by high drainage density values because they favor runoff [10].The Dd map (Figure 7a

Groundwater Potential Mapping and Validation of the Built Models
FR is the possibility that a phenomenon matches a given feature.[13].The locations of the validated wells and the training variables (the independent variables), e.g., the soil, and Dd were entered using the FR method [60].The frequency of wells in each class for all parameters was determined using FR [13].Using the below formula (Equation ( 2)), it was determined how each factor affected the predictor: A represents the percentage of points in each class, B is the total number of points across all classes, C is the number of classes, and D is the total number of pixels, where FR seems to be the ratio at which each class of each feature affects the parameter.The weights of each class of factors associated with the subject layers for delineating GWPZs are determined by the FR values generated for each class of conditioning factors.The new thematic layer is the input value to the hybrid model after each class has been subjected to FR.
In order to ascertain the spatial correlations between well sites and training factors,

Groundwater Potential Mapping and Validation of the Built Models
FR is the possibility that a phenomenon matches a given feature.[13].The locations of the validated wells and the training variables (the independent variables), e.g., the soil, and Dd were entered using the FR method [60].The frequency of wells in each class for all parameters was determined using FR [13].Using the below formula (Equation (2)), it was determined how each factor affected the predictor: A represents the percentage of points in each class, B is the total number of points across all classes, C is the number of classes, and D is the total number of pixels, where FR seems to be the ratio at which each class of each feature affects the parameter.The weights of each class of factors associated with the subject layers for delineating GWPZs are determined by the FR values generated for each class of conditioning factors.The new thematic layer is the input value to the hybrid model after each class has been subjected to FR.
In order to ascertain the spatial correlations between well sites and training factors, the FR model was applied (Table 1).FR values below 1 denote low correlation, while those above 1 denote higher correlation [61].The greatest FR (1.97) was recorded for elevation classes (Figure 8) 1087 to 1276, followed by 1479 to 1705 (0.96), 1705 to 1998 (0.84), 1276 to 1479 (0.64), and 1998 to 2941.(0.45).As a result, regions with low elevations between 1087 and 1276 are more likely to contain groundwater.Additionally, the range from 1 to 3.8 (FR = 1.46) has the largest potentiality in the context of depressions, while the ranges from 0 to 1 (FR = 0.99) and 3.8 to 72 (FR = 0.66) have low potentialities.
The lowest FR (1.21) for rainfall classes (Figure 8) was found in locations receiving between 234 and 288 mm, followed by 288 and 338 mm (1.01).The lowest FR values were found in areas with a lot of precipitation: 338 to 394 mm (0.92) and 394 to 477 mm (0.79).
The ROC curve for GWPZ is shown in Figure 10.The AUC emphasizes the importance of the prediction by demonstrating the system's capacity to predict "groundwater" and "no-existence of groundwater" with equal accuracy.The AUC is a 0-to-1 scale which scores less than 0.5, suggesting model coherence and higher simulations have greater accuracy.The results demonstrate great accuracy in creating a groundwater potential map by fusing statistical models with the RF model.The improved accuracy of the model (AUC = 0.759) is caused by the results' unknown existence of groundwater.

Changes Detection of in Land Use Land Cover
The analyses of land use and land cover (LULC) have the capacity to identify water resources that increase the quantity of agriculture and other anthropogenic activities [62].In order to plan, use, and regulate natural resources effectively for complicated environmental studies, LULC changes must be well studied.Through the use of remote sensing techniques, which are extremely essential for sustainable management, researchers are able to evaluate and track the dynamics of natural resources using the LULC [63,64].Additionally, the LULC provides important proof of the provision of water supplies.
Differences between two Landsat images acquired on 23/2/2010 (Landsat-5; bands 7, 4, and 2 in R, G, and B) and 12/3/2022 (Landsat-8; 7, 5, and 3 in R, G, and B) are assessed using the Difference tool of ArcGIS.To quantify the changes in LULC, the "Difference" function in ArcGIS is utilized.The differences between Landsat images from 2010 and 2022 are quantified (Figure 11).In this approach, the zero value reflects coherence among the processed images, and the more change between the multi-temporal images the farther the value divergence from zero.In Figure 11a,d,g, the differences in land cover/land use between Landsat 2010 and 2022 are detected in blue (positive); however, negative changes appear in dark brown.Such an approach detected changes in water, vegetation, and infrastructures.Based on this approach, the area experienced major changes regarding anthropogenic activities.There were more abundances of water in 2022 lakes than in 2010 as shown in Figure 11d-f, and more land reclamation in Figure 11h versus Figure 11g.The analyses of land use and land cover have the capacity to identify water resources that increase the quantity of agriculture and other anthropogenic activities.

Discussion
Based on the analysis of hydrologic, geologic, topographic, and climatic data, the downstream area has high groundwater potentiality as indicated by the spatial and statistical analysis of the input factors.The soil of "clay loam" and Calcic cambisols "loam" recorded high FR values are 2.23 and 1.15, respectively.This is because the rate of precipitation and groundwater infiltration is substantially controlled by the soil characteristics [49,65], as the arrangement of soil grains controlled the infiltration and recharge capability and has a major impact on the vertical and lateral flow of water [66].Additionally, porous

Discussion
Based on the analysis of hydrologic, geologic, topographic, and climatic data, the downstream area has high groundwater potentiality as indicated by the spatial and statistical analysis of the input factors.The soil of "clay loam" and Calcic cambisols "loam" recorded high FR values are 2.23 and 1.15, respectively.This is because the rate of precipitation and groundwater infiltration is substantially controlled by the soil characteristics [49,65], as the arrangement of soil grains controlled the infiltration and recharge capability and has a major impact on the vertical and lateral flow of water [66].Additionally, porous and permeable rocks facilitate water infiltration by subsurface fluxes [52].The existence of an equal proportion of sand, silt, and clay of lithosols facilitates the recharge potential due to watershed and soil erosion (Nguyen et al., 2021).Therefore, the amount and flow of groundwater in a specific area are determined by the soil underneath that location [67].Because the surface water can accumulate and drain downstream networks [4,68], the low-elevation areas recorded the highest FR (1.97) for elevation classes 1087 to 1276.Based on FR values, the N and NE have higher values of 2.37 and 1.37, respectively, increasing the vegetation in some areas which improves surface infiltration and groundwater recharge [11,69].On hillsides that look north and east, transpiration is low despite the significant soil moisture levels.The TRI factor was computed to assess how heterogeneous the landscape was, and it was utilized to search for groundwater [63,64].The areas of low TRI have a high FR value of 1.58 and are positively related to water accumulation.The groundwater aquifers would be able to receive water at topographic depressions [10,16], as they recorded high FR values of 1.46 and 0.66.Furthermore, high lineament density areas have a high FR value of 1.18.This is due to that location with high permeability and potential groundwater [67,[70][71][72] and, thus, groundwater occurrence.Geologic structures, such as drainage networks (rivers) and geological features (faults, fractures, and lithological limits), are rectilinear due to a phenomenon that takes place below the surface of the ground [52].Such structures increase secondary porosity and infiltration.The areas "0-370" km, as they are nearby rivers, are considered the more pronounced groundwater [64,73], as being close to hydrological systems is important when looking for water resources [5,52].The rainfall information may be useful in identifying regions that are likely to accumulate water and represent the potential recharge areas [9,10,49,56].Although the areas of high rainfall has the lowest FR value (0.79), the area with low rainfall has the highest FR (1.21) as the precipitation is driven by topography but water accumulates in low elevated areas in rugged terrains [74].Based on that Dd, classes 54.40 to 66.70 and 66.80 to 101 have the higher values 1.63 and 1.14.This is because the areas dissected by lineaments have higher Dd.It suggests that high-drainage-density locations are suited for the formation of groundwater [75][76][77][78].In similar studies concerning groundwater prospection which implemented bivariate statistical models [3,25,74], it was found that the present study provided more balanced results for the training and validation points.Although the present study presented a satisfactory output and validation performance, the results are controlled by the quality of input data.Moreover, detailed field investigations and very high-resolution images that reflect the hydrogeologic and geologic settings should be utilized to predict the groundwater potentiality in the future.

Conclusions
In this article, a part of the Yellow River basin that covers 35,915 km 2 was investigated to delineate the prospective locations of water resources.To achieve that, geological, hydrologic, topographic, and climatic data were combined with remote sensing and GIS data.Nine predictive GIS maps including soil, elevation, lineaments, rainfall intensity, Dd, Distance to the river, TRI, aspect, and topographic data were prepared and normalized using the FR technique, which ranked each sub-class based on its capability for holding groundwater.These layers were integrated using the overlay weight-based GIS technique.The resulting map was then classified using NBC into six groundwater prospection zones were found to have very low (~13%), low (~30%), moderate (~25%), high (~16%), very

Figure 1 .
Figure 1.(a) Location map of the present study area is marked in the blue polygon that covers the majority of Niangxia Hui Region; (b) watershed of the study area overlain by well data.

Figure 1 .
Figure 1.(a) Location map of the present study area is marked in the blue polygon that covers the majority of Niangxia Hui Region; (b) watershed of the study area overlain by well data.

Land 2023 , 23 Figure 3 .
Figure 3. (a) Soil map of the study area; (b) Digital elevation model (DEM) data of the present study displays the variations in elevation.

Figure 3 .
Figure 3. (a) Soil map of the study area; (b) Digital elevation model (DEM) data of the present study displays the variations in elevation.

Figure 4 .
Figure 4. (a) Aspect map of the study area; (b) Terrain roughness index of the present study.

Figure 4 .
Figure 4. (a) Aspect map of the study area; (b) Terrain roughness index of the present study.

Land 2023 , 23 Figure 5 .
Figure 5. (a) Topographic depression map of the study area; (b) lineaments of the present study.

Figure 5 .
Figure 5. (a) Topographic depression map of the study area; (b) lineaments of the present study.

Figure 6 .
Figure 6.(a) Distance to river map of the study area; (b) rainfall of the present study.Figure 6.(a) Distance to river map of the study area; (b) rainfall of the present study.

Figure 6 .
Figure 6.(a) Distance to river map of the study area; (b) rainfall of the present study.Figure 6.(a) Distance to river map of the study area; (b) rainfall of the present study.

23 Figure 7 .
Figure 7. (a) Stream-networks of the study area; (b) drainage density of the present study.

Figure 7 .
Figure 7. (a) Stream-networks of the study area; (b) drainage density of the present study.

Figure 8 . 23 Figure 8 .
Figure 8.A two-dimensional bar chart, for comparing the FR values among classes of the influencing factors (e.g., soil, elevation, aspect, TRI, depressions, lineaments, distance to rivers, rainfall, and drainage density (Dd)) and the areas covering the GWPZs.

Figure 9 .
Figure 9. (a) Groundwater prospective map of the study area; (b-d) Subsets of Landsat 753 band composites display the vegetation in green and water in dark blue color.

Figure 9 .
Figure 9. (a) Groundwater prospective map of the study area; (b-d) Subsets of Landsat 753 band composites display the vegetation in green and water in dark blue color.

Figure 9 .
Figure 9. (a) Groundwater prospective map of the study area; (b-d) Subsets of Landsat 753 ba composites display the vegetation in green and water in dark blue color.