Geological background of the study area
As shown in Fig. 6, the study area is located in Beishan, within the Hei Ying Shan (Black Hawk Mountain) region. It lies in the heart of the arid Beishan region of Inner Mongolia, bordered to the east by Dalaihubu Town (the seat of Ejina Banner) and to the north by the national border with Mongolia, making its geographical location relatively remote. The Hei Ying Shan region itself is situated in the northwestern part of Ejina Banner, Alxa League, Inner Mongolia, adjacent to the Sino-Mongolian border, and is a typical temperate continental arid desert zone.
The topography of the study area is dominated by low mountains, hills, and Gobi desert, with a general trend of being higher in the southwest and lower in the northeast. The elevation ranges from 972 to 1637 meters. The landscape is complex, characterized by widespread Gobi, gravel plains (bajada), paleochannels, and aeolian sand dunes, with significant wind erosion. The area has a mean annual temperature of 9.9 °C and a mean annual precipitation of only 37 mm, while potential annual evaporation is extremely high at 3842 mm. Consequently, the area suffers from chronic water scarcity. There is no perennial surface runoff; seasonal floods occur only briefly during the rainy season, making groundwater the primary water source.
The geological structure is controlled by NNW-trending faults and folds. The area features Mesozoic and Cenozoic sedimentary strata, along with some Paleozoic rocks, and abundant mineral resources. Vegetation coverage is extremely low, and the ecosystem is fragile. These complex geomorphological structures, arid conditions, and challenging hydrogeological settings pose significant challenges for groundwater exploration in this region, underscoring the practical significance of the present study.
Building on the foundation of prior hydrogeological surveys and geophysical exploration, this study conducted SNMR data acquisition in seven priority exploration zones (ZK1 to ZK7), identified by the hydrogeological investigation. Due to challenging field conditions, TEM soundings were performed only in zones ZK1 and ZK4. The specific distribution of all measurement points is shown in Fig. 6. To establish a baseline for interpreting the SNMR data in the work area, two additional SNMR measurements were conducted for calibration: one at the SK01 water well in Ejina Beishan and another at an abandoned dry well near the wind farm. The SNMR data were acquired using a GMR SNMR instrument, while the TEM data were collected with a GDP-32 Transient Electromagnetic system. The quality of the collected data was high and met the requirements for subsequent data processing and interpretation.
Establishment of benchmarks
Due to the arid nature of the study area, the subsurface water content is generally extremely low. Consequently, conventional water content thresholds for defining an aquifer (e.g., >15%) are not applicable. It was therefore necessary to establish a specific interpretation criterion tailored to this region. To define this benchmark, the SNMR data acquired at the SK01 water well were processed using both deterministic inversion (Fig. 7) and Bayesian inversion (Fig. 8). The deterministic inversion results indicated a water-bearing layer from 0-20 m depth with a water content of approximately 3%, and another from 20-30 m depth, also with a water content of about 3%. Below 40 m, the water content gradually decreased. The Bayesian inversion results similarly identified a water-bearing layer from 0-20 m with a water content of about 3%, and another from 30-40 m with a water content of approximately 5%. While the results from both inversion methods were largely consistent, the Bayesian approach provided a more sensitive and focused delineation of the water-bearing layers.
Subsequently, the data collected at the abandoned dry well near the wind farm were also processed using both deterministic (Fig. 7) and Bayesian (Fig. 8) inversion for comparison. The deterministic results showed a 1% water content variation in the 0-10 m depth interval and a 0.5% variation from 20-30 m. Although the water content at greater depths exceeded 1%, this was primarily bound water, with the free water content remaining below 1%. The Bayesian results indicated a 1% water content variation from 0-10 m and another 1% variation from 10-20 m, with a slight change noted at the 80-100 m depth, though the overall water content remained extremely low.
The consistent outcomes from both wells validated the accuracy of Bayesian SNMR inversion algorithm. By synthesizing the interpretation results from both the productive water well and the dry well, a definitive criterion for identifying aquifers in the study area was established: a water content exceeding 3% indicates a promising zone for water exploration, whereas a water content below 1% suggests an unfavorable zone. Based on this benchmark, all measured SNMR data were interpreted to identify promising targets for groundwater exploration.
Constraining geophysical data using borehole data
The SNMR and TEM response curves for the ZK1 key exploration area, as obtained from field measurements, are shown in Fig. 9a and b. Deterministic inversion was performed on the ZK1 data, and the results are presented in Fig. 9c. Bayesian inversion of the SNMR field data for the ZK1 exploration area was also conducted, utilizing a homogeneous half-space resistivity model of 100 as prior information. The Markov chain was sampled 300,000 times, with a burn-in period of 150,000 steps; the inversion results are shown in Fig. 10a and b. Additionally, Bayesian joint inversion of the SNMR and TEM data for ZK1 was performed under the same sampling conditions (300,000 samples, 150,000 burn-in), and the results are illustrated in Fig. 10c and d. By comparing the results from the single-method Bayesian inversion, the Bayesian joint inversion, and the deterministic inversion, a more comprehensive understanding of the aquifer distribution and water content variations can be achieved.
The individual Bayesian inversion results indicate an aquifer at a shallow depth of 0-5 m with a water content of approximately 8%, which is inferred to be surface water. Another aquifer is identified at a depth of 20 m with a water content of about 3%. A third aquifer is also recognized within the 50-70 m depth range, with a water content of approximately 2%. In comparison, the Bayesian joint inversion results also show an aquifer in the 0-5 m depth range with a water content of about 8%. The aquifer at 20 m is also present, but its water content increases to 4%. A third aquifer is identified in the 60-70 m depth range, where its water content significantly increases to 9%. The deterministic inversion results, on the other hand, show one aquifer in the 0-20 m depth range with a water content of about 4%, and a second aquifer is identified in the 40-80 m depth range, with a maximum water content reaching 7%. A comparison of the three methods reveals that the individual Bayesian inversion and the joint Bayesian inversion are largely consistent in identifying the positions of the aquifers, though they differ in the estimated water content values. Conversely, the results from the Bayesian joint inversion and the deterministic inversion show greater consistency in both the aquifer locations and their corresponding water contents.
Overall, the Bayesian joint inversion provides a more precise delineation of aquifer positions and demonstrates higher accuracy in estimating water content. The inversion results for ZK1 collectively indicate the presence of aquifers at both shallow and deep levels, with water content in both exceeding the 3% threshold for a viable target. This suggests that the ZK1 location is a promising target for water exploration. This conclusion not only offers a crucial reference for water resource exploration in the study area but also validates the reliability and accuracy of the Bayesian inversion method for aquifer identification.
The measured data from the ZK4 exploration area were processed using three different inversion strategies. The corresponding SNMR and TEM response curves are displayed in Fig. 11a and b, and the deterministic inversion results are shown in Fig. 11c. The individual SNMR Bayesian inversion results are shown in Fig. 12a and b, while the SNMR-TEM joint inversion results are presented in Fig. 12c and d. According to the individual Bayesian inversion results, an aquifer is present at a shallow depth of 10 m with a water content of approximately 4%. A second aquifer is identified in the 80-100 m depth range with a water content of about 1%, which gradually increases to 4% within the 100-120 m depth range. In comparison, the Bayesian joint inversion results also show an aquifer at 10 m with a water content of about 5%. A second aquifer is identified in the 60-80 m depth range with a water content of about 1%, which gradually increases to 4% in the 80-100 m depth range. The deterministic inversion results, on the other hand, indicate an aquifer in the 0-20 m depth range with a maximum water content of approximately 7%. A second aquifer is also present in the 40-120 m depth range, with a maximum water content of about 5%. A comparison of the three sets of results reveals a discrepancy between the individual Bayesian inversion and the joint Bayesian inversion in the delineation of the deep aquifer. In contrast, the Bayesian joint inversion and the deterministic inversion show greater consistency in both aquifer location and water content estimation. The Bayesian joint inversion provides a more precise delineation of aquifer positions and a more accurate estimation of water content, better reflecting the distribution characteristics of the subsurface aquifers.
Across the seven priority exploration areas, SNMR soundings produced high-quality data with minimal interference. Analysis of productive and dry-well measurements allowed us to establish an SNMR interpretation baseline for the Beishan region: intervals with water content> 3%, combined with appreciable aquifer thickness, are deemed promising groundwater targets.
Applying this criterion, together with the inversion results and local topography, identifies the vicinities of ZK1 and ZK4 as favourable targets: each shows water contents exceeding 3% and well-defined aquifer geometries. In the remaining areas, the likelihood of abundant groundwater above 150 m depth is low; domestic and industrial supply will probably require drilling deeper than 150 m.
These findings furnish a valuable reference for future groundwater exploration and development in the Beishan region and provide clear guidance for subsequent drilling operations.
Results and analysis
Based on the interpretation of the SNMR data, sites ZK1 and ZK4 were identified as promising targets for water exploration, and drilling was conducted at these locations for validation. According to the drilling results, water was successfully encountered at all three sites (ZK1 and ZK4), further confirming the accuracy and reliability of SNMR inversion algorithm. This demonstrates the high practical value of SNMR-based inversion and interpretation methods for aquifer identification and the delineation of promising exploration targets.
The borehole logs and drilling curves for ZK1 and ZK4 are shown in Fig. 13, respectively.These figures provide a direct visual representation of the aquifer distribution characteristics and the variations in water content.
An analysis of the borehole log for ZK1 reveals that it is fundamentally consistent with the inversion results for the same site, showing a high degree of correlation in both stratigraphic structure and hydrogeological properties?. Specifically, the borehole log indicates that the formation shallower than 20 meters consists primarily of a relatively loose gravel and sand layer. Due to its coarse grains and large pores, this layer exhibits good permeability and storage capacity, allowing it to readily accumulate groundwater (as an unconfined aquifer with a high proportion of free water), resulting in high water content. This characteristic makes the shallow zone the primary area for groundwater accumulation. Below 20 meters, the formation gradually transitions to gravelly sandstone. Although this layer retains some porosity and permeability, these properties are significantly reduced due to cementation, leading to diminished water storage capacity and lower free water content, which manifests as a weak aquifer. This change corresponds well with the decrease in water content at greater depths shown in the SNMR inversion results. The excellent consistency between the measured water content and the drilling results further validates the applicability of surface NMR technology and the reliability of its measurements in the Beishan area.
An analysis of the borehole log for ZK4 shows that the geological formation within the shallow 30-meter depth range is primarily composed of gravel/sand, mudstone, and gravelly sandstone. Within this structure, the mudstone layer acts as an intermediate aquitard; due to its fine particles, low porosity, and poor permeability, it functions as a barrier to groundwater flow. In contrast, the overlying and underlying layers of gravel/sand and gravelly sandstone exhibit higher porosity and better permeability, thus behaving as aquifers. This stratigraphic structure shows a high degree of correlation with the distribution of the two shallow aquifers identified in the ZK4 inversion results, further validating their accuracy. In the deeper 60-100 meter range, the formation consists mainly of gravelly sandstone. Although cementation has reduced its porosity and permeability to some extent, it retains a certain water storage capacity, manifesting as a deep aquifer. This characteristic corresponds to the trend of increasing water content at depth shown in the ZK4-1 results, indicating that the deep formation has strong hydrogeological properties and can effectively store groundwater. Overall, the excellent consistency between the ZK4 borehole log and the ZK4 inversion results, in terms of both stratigraphic structure and hydrogeological properties, provides a reliable basis for studying the regional distribution patterns of groundwater.