Novel Numerical Modeling of a Groundwater Level-Lowering Approach Implemented in the Construction of High-Rise/Complex Buildings

Abstract

Controlling groundwater levels is essential for the safe construction of complex or high-rise buildings. In México, dewatering regulations lack detailed references, and piezometric data are limited, making precise groundwater control a challenge. This study aimed to develop a numerical groundwater model by translating a conceptual hydrogeological model into a calibrated MODFLOW simulation using the graphical user interface ModelMuse, developed by the United States Geological Survey (USGS). For the project “Torre Tres Ríos”, field measurements recorded a water-table level of 33 m above sea level (masl) in July, rising to 35.74 masl in October due to rainy season recharge and the influence of the Tamazula River, then decreasing to 35.20 masl in November. The model, calibrated with a mean absolute error of 0.15 m and a standard deviation of 0.174 m, effectively represented steady and transient states. A spatiotemporal analysis based on the calibrated numerical model enabled the evaluation of different dewatering scenarios. Initially, deep wells with a pumping rate of 120 L per second (lps) were required for dewatering; however, a wellpoint system was proposed, showing improved performance with a reduced impact on groundwater flow and the surrounding environment during the critical August–November period. This study highlights the importance of numerical modeling in refining dewatering system designs, ensuring adaptability to fluctuating groundwater conditions. By providing a methodology for optimizing dewatering strategies, it contributes to more efficient and sustainable construction practices in regions with complex hydrogeological conditions.

1. Introduction

The effects of groundwater levels on a construction project can lead to numerous conflicts, for example, in the excavation process, foundation design, and the construction process itself, often resulting in significant delays and project redesigns. In México, the complexity of managing groundwater in construction is exacerbated by the lack of detailed regulations and methodologies for evaluating and controlling groundwater flow. However, the Building Regulation for México City [1] in Article 172 states that investigations must be conducted to verify that the construction does not cause damage nor affect adjacent buildings or service facilities. On the other hand, in the Technical Complementary Standards for Design and Construction of Foundations [2], in section 5.1.4 on the stability of adjacent structures, it recommends that, if necessary, structures adjacent to excavations must be reinforced to secure the area where work is being performed. Particularly, in section 7, it states that the construction procedure must be defined to prevent damage to nearby structures, installations, and public services due to vibrations or soil displacement. Section 7.2.2 also discusses groundwater flow control, but it does not reference any specific regulation in this regard. Moreover, the availability of piezometric data is often limited, making it difficult to develop precise hydrogeological models for dewatering strategies. Addressing this gap is essential for enhancing construction planning and promoting sustainable building techniques in projects affected by groundwater level fluctuations. One effective solution is to implement dewatering systems to lower the water table to the desired level through an appropriate method or technique, ensuring stable and workable conditions at the bottom of the excavation and guaranteeing the security of the project. Several methods are available to achieve this. Some of the commonly used methods are (1) wellpoint systems, (2) shallow wells with suction pumps, (3) gravity drainage, (4) deep wells, and (5) combination methods [3]. Scholars assert that groundwater control techniques can be broadly classified into two main categories—excavation-based and pumping-based [4]. Each technique involves a variety of systems or methods, making it essential to understand the advantages and disadvantages of each and their optimal application. Additionally, it is documented that the selection of the dewatering system depends on several factors such as the (1) size and depth of excavation, (2) nature and permeability of the foundation soil, (3) dewatered area, (4) depth of the water table and depth of required drawdown, (6) proximity to existing structures, and (7) proposed method of excavation [5]. Within this context, experts emphasize that it is essential to understand the characteristics of the project for which groundwater control is needed [3]. Understanding groundwater flow behavior requires analyzing the diverse soils, hydrogeological conditions, and project requirements that influence its dynamics. Addressing uncertainties in hydrogeological parameters and refining the conceptual model are essential steps in accurately translating these conditions into a reliable numerical model. However, it must be acknowledged that the dewatering process needs to be grounded in scientific principles to achieve optimal practical results.

Groundwater presents a significant challenge in construction, particularly when excavation occurs below the water table, as in the case study presented in this paper. In the “Torre Tres Ríos” project in Culiacán, Sinaloa, México, effective dewatering had to be performed since the groundwater flow behavior experienced recharge due to rainfall and the influence of the Tamazula and Humaya Rivers. Researchers emphasize the need for caution when using empirical formulas to estimate hydraulic conductivity, as their applicability depends on site-specific geological and hydrological conditions [6]. To address uncertainties in groundwater flow behavior, the observational method proposed by Terzaghi and further developed by Peck in 1969 is recommended [7]. According to CIRIA Report 185 [8], this method involves continuous monitoring and adaptive decision-making throughout the construction process, allowing for controlled modifications to optimize performance while maintaining safety. However, concerns remain regarding the potential negative impacts of dewatering, as studies have shown that groundwater extraction can induce subsidence in adjacent layers, leading to structural and environmental consequences [9]. Given these challenges, this study seeks to optimize dewatering designs by applying numerical modeling with MODFLOW and the ModelMuse interface, evaluating the effectiveness of the dewatering technique while minimizing impacts.

According to [10], numerical models serve as interpretative tools to study groundwater system dynamics; evaluation tools to analyze discharge, recharge, and storage processes; and predictive tools to forecast future conditions. For instance, researchers [11] mentioned that the application of groundwater flow models started in 1978, with the Regional Aquifer System Analysis, a program of the USGS. Over time, significant advances in regional flow system analysis were driven by the application of 3D groundwater flow models. Since its release in 1988, MODFLOW has become the industrial standard worldwide for groundwater modeling because of its flexible modular structure, complete coverage of hydrogeological processes, and public-domain free availability. MODFLOW-88 was progressively updated to MODFLOW-2000 [12] and MODFLOW-2005 [13]. Initially, MODFLOW was designed solely as a groundwater flow model, but the program’s designers saw the solution of adding related equations through packages, which optimizes the conceptualization of boundary conditions and other model elements, reducing the construction time and improving the interpretation of output data. The selection of the modeling approach should be proportional to the complexity of the project conditions. In past decades, models have been developed to estimate hydrologic parameters automatically by using computers, such as in [14,15,16,17], instead of tedious trial-and-error methods. In recent years, the application of numerical models to groundwater flows has gained significant importance. However, the focus is on the civil engineering sector, optimization, and the impact of the dewatering process. Pumping water from soil means that there will be subsidence. For instance, in México, academics [18,19,20] reported that groundwater extraction has induced subsidence in México City. In recent years, dewatering systems for groundwater flow control, along with numerical simulations, have been increasingly implemented worldwide. Notable examples include those by researchers who (1) evaluated the efficiency of wellpoint dewatering and the effects of high permeability on the excavation process [21]; (2) proposed that the dewatering process can be optimized by determining the optimal pumping rates and spacing between well locations, offering an effective alternative solution with a particular emphasis on the well locations [22]; (3) used integrated numerical models of groundwater and land subsidence caused by dewatering in California [23]; (4) produced a simulation of groundwater flow and land subsidence, also in California, using a package developed by the USGS [24]; (5) evaluated land subsidence considering underground structures to lower the groundwater level in China [25], although these structures, serving as an impermeable barrier blocking groundwater flow, can hardly be avoided [26]; (6) applied MODFLOW to solve issues related to groundwater inflows into mines [27]; (7) analyzed and compared settlement around pumping wells with existing analytical solutions and field observations [28]; and (8) reviewed the works of researchers who focused on the pumping-induced groundwater level drawdown [29] through field observation [30], an analytical approach [31], and numerical simulation [32]. Even though there are quite a few technical papers with relevant information, there is still much to analyze with the aim of optimizing these dewatering techniques and their impacts on civil engineering. This study contributes to the existing body of knowledge by emphasizing the importance of accurately defining the conceptual model, including the boundary conditions that govern groundwater flow in construction projects. Given that each project is unique, it is essential to evaluate the specific hydrogeological conditions and factors influencing groundwater behavior. A thorough assessment ensures that the numerical model accurately represents site-specific conditions, leading to more reliable simulations and effective dewatering techniques tailored to the particular challenges of each project. By ensuring a precise representation of hydrogeological conditions and implementing these parameters correctly in the numerical model, this research enhances the reliability of groundwater flow simulations. For example, for “Torre Tres Ríos”, based on hydrogeological [33,34] and soil mechanics studies [35], it was possible to idealize the behavior of the water table, which was initially at 33 masl in July and near 35.74 masl in October, in response to recharge from the rainy season and the influence of the Tamazula River. By November, it was already below 35.2 masl, indicating that the aquifer was in a state of discharge. The dewatering system used was deep wells with an extraction rate of 120 lps, different from wellpoints (the simulated and optimized scheme). The dewatering system and the inappropriate control of the groundwater level caused issues with excavation stability, and adjacent structures experienced settlement. This may have been due to poor practices with the deep wells, where proper planning and control could have been a determining factor. In a system, it is common for conditions to change over time, and the system must evolve to achieve equilibrium [36]. This underscores the importance of numerical modeling in ensuring system flexibility. Ultimately, this approach facilitates the evaluation of various dewatering scenarios, improves decision-making processes, and supports the development of more effective and adaptable groundwater control techniques in construction.

Based on the above discussion, the objective of this article is to analyze the numerical model and the response of the aquifer to an optimized dewatering system designed to improve groundwater flow control. This study evaluates how implementing a lower-pumping-rate system reduces drawdown impacts on surrounding structures. It also aims to contribute to the development of sustainable construction practices, shifting from the “conventional” approach to an optimized schemes for dewatering systems.

2. Materials and Methods

The process of a groundwater drawdown project consists of the stages presented in Figure 1.

In order to analyze the groundwater flow and propose an effective dewatering system for the Torre Tres Rios project, and to solve the differential equation that describes the water flow in the aquifer, a site-specific numerical model must be generated. To achieve this purpose, it is necessary to complete the following:

• Define the three-dimensional geometry of the geologic medium containing the groundwater;
• Identify the hydraulic properties of the materials through which the groundwater flow occurs;
• Specify boundary conditions congruent with the conceptual model of operation;
• Establish initial conditions of groundwater flow and hydraulic loads;
• Perform evaluations of the volumetric water flow term, which represents pumping extractions.

This methodology allows the analysis of different scenarios and varying conditions. Given the ongoing and substantial computational advancements in society, these methods are the most used and provide more reliable results. The choice of modeling approach should be proportional to the complexity of the project’s conditions. The stages for the modeling process are as follows:

• Development of the conceptual model;
• Numerical model representation;
• Verification and calibration;
• Prediction and evaluation.

2.1. Development of Conceptual Model

During the stage of developing the conceptual model, the project was contextualized and characterized topographically, geographically, hydrologically, climatologically, and geologically to understand what would be represented and the available data to ensure that they represent an approximation of the observed field conditions.

For the Torre Tres Ríos project, the sources for creating the model database were as follows:

• Definition of the geological and stratigraphy of the zone based on GIS data from the Instituto Nacional de Estadística y Geografía (INEGI), obtained through in situ geotechnical exploration;
• Topographic characterization using satellite imagery from INEGI with LiDAR digital models (5 m resolution), processed with QGIS (version 3.22) and Surfer 13 for a 3D model to integrate with the ModelMuse interface (version 4.1);
• Identification of hydraulic properties (hydraulic conductivity, transmissivity, and specific storage) of the aquifer using a pumping test and the Cooper–Jacob method, comparing the results with aquifer data from an available report [34];
• Hydrologic and climatologic characterization conducted by analyzing recharge and discharge factors, including field measurements of the water table level with a Solinst 101 P7 level sensor (Solinst Canada Ltd.), data from local meteorological stations to assess precipitation’s influence, and hydrological data from Banco Nacional de Datos de Aguas Superficiales (BANDAS) from CONAGUA to consider the impact of the Humaya and Tamazula Rivers.

2.2. Development of Numerical Model

As the manual [37] states that the results of a numerical model can only produce results based on the conceptual model, if the conceptual model does not reflect actual conditions, the results are unlikely to be useful.

Henry Darcy [38] presented a report on the flow of water through porous media, in which he found that the flow rate Q is determined as follows:

Q = k × i × A

(1)

where

• Q: flow rate (m³/s);
• k: hydraulic conductivity (m/s);
• i: hydraulic gradient;
• A: cross-sectional area (m²).

In a system, conditions often change, requiring adjustments to reach equilibrium over time [36]. To describe this process, fundamental equations such as Darcy’s Law, Laplace’s equation, and Poisson’s equation are utilized. The continuity equation for groundwater flow is as follows:

Ss ∂h/∂t = ∇ · (K∇h) ± W

(2)

where

• Ss: specific storage coefficient (1/m);
• ∂h/∂t: change in hydraulic head over time (m/s);
• ∇: divergence operator (1/m);
• ∇h: gradient of hydraulic head (m/m);
• K: hydraulic conductivity (m/s);
• W: volumetric flow rate per unit volume (1/s).

This equation represents transient flow conditions, accounting for temporal changes in hydraulic head due to storage effects, groundwater recharge, and extraction. In the case of steady-state flow, where hydraulic head does not change over time (∂h/∂t = 0), the equation simplifies to the following:

∇ · (K∇h) ± W = 0

(3)

Laplace’s equation is a special case of this expression when there are no sources or sinks (W = 0), describing purely conductive flow within a saturated porous medium. Since real problems often involve pumping and recharge, Poisson’s equation is typically more applicable.

Once the project has been characterized and described, and the most important characteristics and parameters of the conceptual model have been identified, it is time to translate that information into the mathematical/numerical model. This is where the equation governing the flow of the system is solved. The process is carried out by applying MODFLOW code (version 1.12.00) with the ModelMuse graphical interface.

2.2.1. Spatial and Temporary Discretization

The model was divided into a structured grid with 10 m cell sizes. A refined grid resolution was applied covering the Torre Tres Ríos project with 5 m discretization; contour lines from the project delimitations ensured the grid position to improve the accuracy of the groundwater flow model in areas of interest such as pumping wells. The grid generation was performed using ModelMuse, ensuring compatibility with the aquifer digital elevation model.

The temporal discretization was divided into stress periods, corresponding to changes in hydrological conditions (rainy season and river flux). Each stress period was subdivided into time steps, considering one day of the month with critical conditions, to ensure that the chosen dewatering technique would efficiently control the groundwater level. The model was initialized with a steady-state simulation, followed by transient-state simulations to analyze drawdown and recovery over time.

2.2.2. Steady and Transient States

It is important to mention that boundary conditions are mathematical expressions that include the geometry of the domain and specify the hydraulic load or the derivative of the flow at the model boundaries [39]. In groundwater applications, hydrogeologic boundaries are represented by three types of conditions:

• Constant load (Type I or Dirichlet’s) imposes the existence of piezometric heights or pre-described hydraulic potentials;
• Constant flow (Type II or Neumann) simulates a constant flow in the model, representing the main inputs and outputs of the system;
• Dependent load (Type III, Cauchy or Mixed) combines characteristics of the previous two, simulating interactions with other water bodies, like rivers.

The boundary conditions that were applied for the steady-state simulation were of constant load to simulate the static level of the zone. The conditions applied in the transient state were constant flow for precipitation recharge and pumping extraction, as well as dependent load representing the influence of the rivers in the study zone. The MODFLOW code includes packages for modeling different conditions of groundwater flow.

The following were some of the packages used:

• LMT (Layer-Property Flow) simulates flow in heterogeneous and anisotropic layers;
• RIV (River Package) models the exchange of water between rivers and aquifers;
• RCH (Recharge) simulates recharge (water input) to the aquifer from external sources like precipitation;
• WEL (Well) models water extraction from wells.

For verification in the steady state, regional flow data and flow directions [34] were compared with field measurements under stable conditions, considering the initial condition of the model. For the transient state, the accuracy was evaluated by assessing the model’s ability to simulate drawdown and recovery curves in response to observed groundwater level fluctuations, considering the influence of the mentioned boundary conditions.

2.3. Calibration and Validation

The calibration process consists of adjusting the values of hydraulic loads calculated by the model using those observed in the field for different stress periods, so that the groundwater dynamics comply with what is defined in the conceptual model [40].

The comparison of calibrated and modified loads was performed using two methods:

• The mean absolute error (MAE);
• The standard deviation (SD).

The equations defining the MAE (4) and standard deviation (5) are as follows:

$$\mathrm{MAE}={\displaystyle \frac{1}{\mathrm{n}}}{{\displaystyle \sum }}_{\mathrm{i}=1}^{\mathrm{n}}\left|\mathrm{hoi}-\mathrm{hi}\right|$$

(4)

$$\mathrm{SD}=\sqrt{{\displaystyle \frac{1}{\mathrm{n}}}{{\displaystyle \sum }}_{\mathrm{i}=1}^{\mathrm{n}}{\left[\left|\mathrm{hoi}-\mathrm{hi}\right|\right]}^2}$$

(5)

where

• MAE = mean absolute error;
• SD = standard deviation;
• n = number of observations;
• hoi = observed hydraulic load (m);
• hi = calculated hydraulic load (m).

The MAE and SD are important statistical metrics for assessing model performance in numerical groundwater modeling. A lower MAE indicates a better fit between the model and field measurements, providing a measure of overall model accuracy. A lower standard deviation suggests that errors are evenly distributed.

Finally, this model was calibrated using field-observed data to ensure its fidelity to observed behavior.

2.4. Sensitivity and Uncertainty Analysis Method

To evaluate the reliability of the numerical model, a sensitivity and uncertainty analysis was conducted on hydrogeological parameters, including piezometric levels, recharge rates, and river flow variations. The sensitivity analysis was performed by assessing the correlations between model outputs and input variables, identifying the most influential parameters affecting groundwater behavior. For this study, one-at-a-time (OAT) sensitivity testing was performed, where each parameter was modified within a reasonable range while keeping the others constant.

For the uncertainty analysis, a Monte Carlo simulation was implemented, considering the statistical distribution of the input parameters. The hydraulic conductivity and transmissivity were varied within their reported ranges, while river flow data and recharge rates were incorporated from historical records.

This approach enhances model robustness by quantifying the impact of parameter variability and reducing potential errors in groundwater flow predictions, contributing to a more reliable assessment of dewatering strategies for construction applications.

2.5. Dewatering System

Information obtained from the deep well technique, including the pumping rate and the piezometric record, served as reference data for comparison.

Once calibrated in the transient state, the model was subjected to the dewatering system to evaluate its response to the dewatering process. The objective was to achieve an optimized and controlled dewatering technique using a wellpoint system instead of deep wells. The radius of influence of the drawdown cone was considered as a potential factor that may affect the stability of the excavation and settlement with adjacent structures. The Theim formula was applied to evaluate the radius of influence:

r = Q/(2·π·K·∆h)

(6)

where

• r = radius of influence (m);
• Q = pumping rate (m³/s);
• K = hydraulic conductivity (m/s);
• ∆h = drawdown (m).

3. Case Study

This case study was selected because it highlights the need for a predictive approach to manage groundwater in construction and provide an optimized strategy for controlled drawdown. The case study is based on a construction project carried out in 2010. The building included two levels of underground parking below the natural ground level. The building foundation was estimated to reach a depth of 6.50 m. Figure 2 illustrates the “Torre Tres Ríos” (one of the highest towers in Culiacan). The building is a 21−story structure with commercial areas and office spaces.

At the time of planning and early construction, the groundwater level was not a concern, as it was 1 m below the excavation depth. However, as construction progressed, the water table rose unexpectedly, as can be seen in Figure 3a. This required the implementation of dewatering techniques, including deep well pumping at rates of 120 lps, to maintain the water table level at desired depth, as shown in Figure 3b. Unfortunately, the dewatering process was not adequately controlled, leading to unintended consequences with adjacent structures.

The deep well shown in Figure 3b is located about 10 m from the boundary of the adjacent structure, as depicted in Figure 4. This information is valuable for assessing the impact of deep well pumping on the surrounding environment.

The project area is located at 1695 Blvd. Francisco Labastida Ochoa, Desarrollo Urbano Tres Ríos, in the city of Culiacán, Sinaloa, México. Its geographic coordinates are 24.8158, −107.4019. The urban zone is called “Desarrollo Urbano Tres Ríos”, where the Tamazula and Humaya Rivers converge to form the Culiacán River. The average elevation of the study area is 40 masl.

Based on the information provided, the project context is understood: The groundwater level increased due to recharge from precipitation and the influence of nearby rivers. These conditions also favored the aquifer’s response to recharge but delayed the construction process of the excavation, as the foundation level was below the depth of the groundwater table. As a result, a deep well had to be implemented, extracting between 85 and 120 L per second under the most critical conditions. However, the drawdown was not controlled, which led to issues in the adjacent building.

3.1. Conceptual Hydrogeological Model

The project in the study area involved the dewatering of the groundwater level to below the maximum excavation depth (6.5 m). During the project’s study phase, based on the available data from the soil mechanics study and the pumping test, the groundwater level exhibited the following behavior:

In July, the groundwater level was at 33 masl, which was initially considered not to pose a problem for construction processes. However, by October, when a dewatering technique was needed, it had risen to 35.74 masl, indicating the aquifer’s response to recharge effects from the rainy season and river influences. Also, in November, the level dropped below 35.74 masl as the aquifer entered a discharge period.

Table 1. Field data—maximum piezometric level.

Month	Observed Water Table (masl)	Depth (m)
June	33.00	7.00
July	35.55	4.45
August	35.65	4.35
September	35.75	4.25
October	35.74	4.26
November	35.20	4.80

shows field data of the maximum piezometric level from June to November.

The study area is influenced by its inclusion in Irrigation District 010. The agricultural irrigation period extends from October to May, and the end of the irrigation season leads to reduced recharge, causing a decline in the static water level. However, as shown in Figure 5, the rainy season began in July, recharging the aquifer, resulting in the observed seasonal behavior.

It was important to climatologically characterize the study area; for this purpose, the Simulator of Water Flows in Hydrographic Basins (SIATL) was used, which is a hydrographic network generated by the National Institute of Statistics and Geography (INEGI) in México. It is a useful tool since it shows surface drainage and important database information for the study area. Once the area of interest was located by its geographic coordinates in SIATL, the sub-basin of which it forms part was selected, as shown in Figure 6.

Based on the network functions of the SIATL tool, Figure 7 shows the rainfall period from July to October.

When the rainy season ends, the cycle of irrigation for the spring–summer season in the mentioned months resumes. This behavior confirms the aquifer’s recharge process, influenced significantly by factors such as rainfall and river effects during the agricultural irrigation seasons. During the discharge process, water exits through horizontal flow within the same system, along with some evapotranspiration. This information was used to conceptualize the mathematical flow model. According to a study by CONAGUA titled “Determination of Water Availability in the Río Culiacán Aquifer”, geological, geophysical, and hydrogeological evidence allows for the definition of a free-type, heterogeneous, and anisotropic aquifer. It is composed of alluvial and fluvial sediments with varying grain sizes, as well as conglomerates and lacustrine sediments, all of which are situated beneath volcanic rocks [34].

To consider the dynamic response of the hydrogeological system, it was necessary to consider river interactions. The study area is influenced by the Humaya and Tamazula Rivers. The average monthly discharge rates of these rivers are as shown in

Table 2. Average discharge rates from Tamazula and Humaya Rivers.

Month	Tamazula River Discharge Rate (m3/s)	Month	Humaya River Discharge Rate (m3/s)
June	20.00	June	89.56
July	19.83	July	100.27
August	25.56	August	99.34
September	31.52	September	97.50
October	26.31	October	98.94
November	19.85	November	72.39

.

In the study area of Torre Tres Rios, with a pumping test (see Appendix A, Figure A1), an analysis and interpretation of transmissivity was carried out with the Cooper–Jacob method, which is also known as Jacob’s logarithmic approximation. This method was chosen because it is applicable in aquifers of this type (free, of variable regime, with a water table drawdown test and with drawdowns smaller than the saturated thickness of the aquifer).

The calculation was made using an ”xcel’spreadsheet programmed with the criteria of the Cooper–Jacob method and its adjustment to determine the values of transmissivity (T) and hydraulic conductivity (K). The results are shown in Figure A2 from Appendix A, and the values were 470 m²/day (5.44 m²/s) and 28 m/day (3.24 × 10⁻⁴ m/s), respectively. Specific storage of 0.11 was considered. The obtained value of “T” is within the range for the Culiacán River aquifer, and the value of “K” refers to the sand and gravel (alluvial soil) coefficient; both values are consistent and, therefore, are considered good hydraulic parameters for the study area.

The Torre Tres Rios project is in a Quaternary formation current floodplain composed of gravels, sands, silts, and clays deposited by fluvial processes. The stratigraphy of the study area was considered from a study of soil mechanics [35]; from this exploration, it was observed that the stratigraphy corresponds to fill material for the first 4 m, found above sandy silt and gravels of different granulometry down to 18 m, all of which rests on a rocky stratum, as shown in Figure 8.

The topographic characterization of the study area, as shown in Figure 9, was conducted using satellite imagery and software tools like QGIS (version 3.22) and Surfer 13, achieving a resolution of 5 × 5 m.

As a matter of fact, the recharge process has important factors such as rainfall, the static water table, and the influence of the rivers in the agricultural irrigation seasons, while the discharge process involves outflows by horizontal flow within the same system.

Table 3. Hydrogeological conceptual model.

Torre Tres Ríos
Aquifer	Unconfined, heterogeneous, anisotropic
Aquifer depth	40 m
System outflows	Rivers and pumping
System inflows	Rivers and precipitation
Water table level	Variable (33–35.74 masl)
Excavation level	33.5 masl
Stratigraphy	Fill material, alluviofluvial, rock

summarizes the hydrogeological conceptual model.

3.2. Numerical Model

3.2.1. Spatial and Temporary Discretization Model

The proposed grid for the regional model was 1.8 km long by 1.4 km wide and was aligned with the main direction of the groundwater flow, in a north–south direction. Once the dimensions of the model were known, a suitable cell size was defined. The mesh coordinates were adjusted based on the coordinates of the DEM points, using the upper-left corner of the proposed grid as a reference. General horizontal dimensions of 10 m, which would generate a mesh of 180 columns and 140 rows, were considered. In addition, the mesh of the Tres Rios Tower project was refined to a local cell size of 5 m, obtaining Figure 10.

According to the hydrogeological conditions and for the purpose of simulating the groundwater flow of the aquifer, the conceptual model considered, in principle, the stratigraphy of the area of interest composed of three layers, as shown in Figure 6; for practical purposes and to avoid convergence errors in the calculation of the groundwater flow behavior, only two layers were considered. Also, at no time did the groundwater flow reach above a depth of 4 m; thus, the layer of fill material was never influenced by the groundwater flow. In this way, the depth of the first layer corresponding to “fill material” became part of the second layer containing alluvial–fluvial material, having as a final result the first layer of sandy silt and gravels (alluvial–fluvial material) and the second layer of rock material.

A two-layer discretization was performed, composed of three dividing lines. One of them, called “Model Top”, corresponds to the topography of the area, where the surface (Figure 11) created in the Surfer program is compatible with loading the topography. It also helps to accurately represent the heights of the local cells of the model. The other two dividing lines mark the separation between strata. In this way, by feeding the ModelMuse interface with the data for the spatial discretization, a three-dimensional model was obtained.

The model was simulated over a period of one year, from January to December, considering that the dewatering system was temporary work to facilitate the excavation and foundation processes. The runs were carried out with monthly stress periods that allowed us to identify the behavior of the aquifer.

3.2.2. Steady-State Flow Model

Although the parameters assigned at this stage of the modeling were subsequently modified in the calibration process, it was necessary and convenient to introduce reasonable initial values that were consistent with the aquifer for the model output, such that the model could estimate a distribution of hydraulic loads corresponding to the static-level conditions for the area. Therefore, according to a prior report [34], the study area was evaluated in the steady state with respect to the hydraulic loads of the groundwater, with flow ranging from 35 to 30 masl with a flow direction from north to south.

The horizontal hydraulic conductivity value of the first layer was 3.24 × 10⁻⁴ m/s (according to K = 28 m/day from the Cooper–Jacob analysis), and for the second layer, being less permeable (for the rocky stratum), a value of 1 × 10⁻⁵ m/s was considered. MODFLOW code automatically considers 10% of the horizontal components for the vertical component (Kz) as a “drip factor” where the capacity to transmit the flow vertically is much lower. The value used for Kz was 3.24 × 10⁻⁵ m/s for the first layer and 1 × 10⁻⁶ m/s for the second layer.

3.2.3. Transient-State Flow Model

Given the seasonal variability of the water table, the transient simulation incorporated monthly recharge from precipitation and river–aquifer interactions to reflect real hydrogeological conditions. The initial conditions were established based on piezometric data and reports from CONAGUA, ensuring consistency with observed groundwater levels before external influences were introduced. The MODFLOW model, implemented through ModelMuse, integrated important hydrological components, including precipitation recharge (10% infiltration rate) and interactions with the Humaya and Tamazula Rivers, using the RIV package to simulate exchange processes. The model was evaluated from June to November.

3.2.4. Sensitivity and Uncertainty Analysis

In the Torre Tres Ríos case study, sensitivity and uncertainty analyses were essential to assess the reliability of the numerical model under transient-state conditions, ensuring that the model was representative of real conditions. One of the primary sources of uncertainty was the piezometric measurement data, as variations in recorded groundwater levels could result from instrument precision limitations or reading errors. To address this, the model was calibrated against observed piezometric levels (Xi) from June to November, ensuring that simulated water table fluctuations remained within the expected range of ±1.072 m. The analysis results are shown in Figure 12.

The analyzed parameters included the hydraulic conductivity, recharge from precipitation, and river conditions. A sensitivity analysis was performed by systematically varying each parameter within its reported range and observing the impact on groundwater levels. The results are summarized in

Table 4. Sensitivity of groundwater level.

Parameter	Baseline Value	Variation Range	Sensitivity (∆h)	Influence
Hydraulic conductivity (Kx, Ky)	3.24 × 10−4 m/s	1 × 10−5 to 5 × 10−4 m/s	±1.5 m	High
Transmissivity	470 m2/day	280.1 to 843.3 m2/day	±1.2 m	Moderate–High
Recharge from precipitation	10% of rainfall (Figure 7)	5 to 15%	±1.8 m	High
River flow influence	BANDAS database (Table 2)	±15% variation in average flow	±0.7 m	Moderate

, which shows the change in simulated piezometric levels relative to baseline conditions when each parameter was varied.

A Monte Carlo uncertainty analysis was performed. The range of simulated groundwater was analyzed to determine the 95% confidence interval for model predictions.

Table 5. Uncertainty analysis of groundwater level.

Month	Mean Water Table (masl)	Standard Deviation (m)	95% Confidence Interval (masl)
June	33.00	0.72	32.3–33.7
July	35.55	0.80	34.8–36.3
August	35.65	0.84	34.9–36.5
September	35.75	0.88	34.9–36.6
October	35.74	0.85	34.9–36.6
November	35.20	0.92	34.3–36.1

presents the mean water table, standard deviation, and uncertainty range.

3.3. Dewatering System Model

As shown in Figure 3b, the deep well system, along with its components, was represented using a graphical system. Figure 13a illustrates the components and configuration of the deep well system used in the study, and Figure 13b shows the wellpoint system (the proposed technique to evaluate).

The proposed pumping system for the project consists of a wellpoint system of a total of 28 wells located in the area along the direction of flow and distributed as shown in Figure 14. The wells are located 3 m apart and 2 inches in diameter at a depth of approximately 6.5 m above the natural surface and 3 m from the property boundary. On the surface, the riser pipe of the top well is connected to the main collector by flexible pipes, with the addition of a valve that allows each well to be treated individually so that it can be turned off if it is sucking air or in case it is not necessary to use all the wells. The main collector has a diameter of 6″ and is connected to a rotary suction pump with an air extractor with a power of 20 HP, which is able to control the extraction rate.

The proposed controlled dewatering technique aims to optimize groundwater level reduction while minimizing impacts within the influence area. By utilizing a wellpoint system instead of deep wells, the model demonstrated a more efficient and evenly distributed drawdown, reducing excessive pumping rates and preventing instability in adjacent structures. The importance of numerical modeling in designing site-specific dewatering strategies is that it allows engineers to predict groundwater behavior, adjust dewatering parameters dynamically, and implement solutions for sustainable construction practices.

4. Results and Discussion

4.1. Calibration of the Model

The calibration process aimed to minimize discrepancies between observed and simulated piezometric levels by adjusting hydrogeological parameters. Calibration was performed for both steady-state and transient-state conditions using in situ piezometric measurements as reference points. The MAE and SD were used as statistical metrics to assess model accuracy.

The calibration process was performed with field data and model results, considering the aquifer in a transient state without pumping.

Table 6. Observed vs. modeled water tables (masl).

Month	Observed Water Table	Modeled Water Table	Residual
June	33.00	32.725	−0.275
July	35.55	35.675	0.125
August	35.65	35.850	0.200
September	35.75	36.025	0.275
October	35.74	35.676	−0.064
November	35.2	35.438	0.238

presents a comparison between observed water levels and model results, along with the corresponding residuals.

The model calibration had a mean absolute error of 0.15 m and a standard deviation of 0.174; these results reproduce the groundwater flow behavior in a congruent and consistent manner, although if it had been possible, with a larger number of observations, the calibration could have been more refined.

The final transient-state model demonstrated a strong correlation between observed and simulated water levels. The results confirmed that the model accurately captured groundwater fluctuations and provided a solid foundation for predicting future dewatering scenarios. The calibration process highlights the importance of defining accurate boundary conditions and integrating field measurements into the numerical modeling framework.

4.2. Sensitivity and Uncertainty Analysis Results

Table 4 indicates that the model was highly sensitive to variations in hydraulic conductivity, where changes within the reported range (1.0 × 10⁻⁵ to 5.0 × 10⁻⁴ m/s) resulted in up to 1.5 m deviations in simulated water table elevations. Similarly, transmissivity also showed a significant influence, reinforcing the necessity of accurate in situ pumping tests to obtain precise hydrogeological parameters. Ensuring well-defined values for these parameters is crucial to minimizing model discrepancies and improving the reliability of dewatering strategies.

Recharge from precipitation also exhibited a substantial impact on the simulated water table, with uncertainties in the infiltration rates contributing to deviations of ±1.8 m. This highlights the need for continuous and precise monitoring of recharge rates to refine model predictions. Despite these uncertainties, the model effectively constrained piezometric level variations to within ±3 m of observed field measurements, maintaining consistency with field conditions.

These findings underscore the importance of accurate hydraulic characterization and recharge estimations in improving numerical model performance. By integrating sensitivity and uncertainty analyses, this study provides a reliable framework for optimizing dewatering strategies in complex hydrogeological conditions, reducing the risks associated with over-extraction or ineffective groundwater control in construction projects.

4.3. Steady-State Flow Model

The steady-state flow model (Figure 15 and Figure 16), considering initial hydraulic loads with values from 35 to 30 masl in a north–south direction, was consistent with the water table field data. The piezometric lines were represented in the model by red, orange, yellow, green, blue, and light blue colors. For a general understanding of the following figures, these colors indicate different risk levels: red represents a potential problem, while green and beyond (the target zone below) indicate a safe area where the desired water level is achieved.

4.4. Transient-State Flow Model

The transient-state model considered the hydraulic loads obtained from the previous steady-state exercise as initial conditions, and recharge by precipitation and rivers were also considered. The discretization periods for the results were the individual months with their respective characteristics. August and September are the months with the most critical conditions for recharge in the aquifer, and October and November are the months in which the dewatering technique would be implemented; Figure 17 shows the behavior of the transient-state flow.

An observation well was added to the study area, located at the site where measurements were taken to validate and compare against the results obtained from the numerical model. This well was designed to monitor the groundwater flow behavior in the specified space and time under the conditions outlined in the project. The inclusion of this well enabled an evaluation of the accuracy and representativeness of the model against the actual conditions of the study area. Figure 18 illustrates the location of the observation well within the model.

4.5. Dewatering System Performance

With the model calibrated for transient-state flow, it was necessary to model the well package to evaluate it and thereby obtain an approximation of the flow rate that the system must have to lower the water table to the desired level to perform the construction procedures in optimal conditions. It is important to mention that the model was evaluated with a traditional technique of deep wells where 120 lps had to be pumped to lower the water level to the excavation level; however, there were problems with overexploitation of the aquifer and inefficiency of the system, which could also lead to settlement of nearby buildings. As an important point,

Table 7. Comparison of pumped flow rates.

Month	Wellpoint Flow Rate (lps)	Deep Well Flow Rate (lps)
July	22.4	89.3
August *	28	120
September *	25.76	99.52
October	22.4	89.3
November	19.6	85
December	19.6	85

Note: * Months with the most critical conditions.

below shows the flow rates that would have to be pumped with the wellpoint system in each month to achieve the excavation level.

The water table was in the order of 34 m above sea level (6 m depth, reaching the excavation level), as shown in Figure 19, for August and September, which had the most critical conditions; it was in the same order for October and November, when the dewatering technique was also employed. The results show that the wellpoint system has good control over the water table, with less stress from the pumping flow rate when compared with the deep well system.

Although the objective of reducing the pumping rate by more than 75% using wellpoints was achieved, the goal of reaching the excavation depth was also successfully met with this reduction. Additionally, minimizing the impact on excavation stability and settlement of adjacent structures remained a priority. As mentioned in the methodology, the radius of influence of the drawdown cone was used as a reference to evaluate its impact on the surroundings.

Table 8. Radius of influence for dewatering technique.

Month	Wellpoint Flow Rate (lps)	Radius of Influence (m)	Deep Well Flow Rate (lps)	Radius of Influence (m)
August	28	14	120	60
September	25.76	12.26	99.52	47.43

presents the results under the most critical conditions.

Excessive pumping creates drawdown cone influence radii that extend beyond the distance of the extraction well, which is located 10 m from the boundary with the adjacent property. Although the formula used herein applies to radial flow under steady-state conditions, the objective is to demonstrate whether the variables are functioning as expected. As a matter of fact, the most critical aspect would be the lack of recharge during this period, and maintaining the extraction rate would be the most critical factor to assess.

Although the results are satisfactory, the model is an interpretation of the functioning of the aquifer; therefore, the results should not be accepted as unique. The model should be used as a support tool for management and decision making but should always be based on knowledge. The control system must be flexible and capable of incorporating corrective measures for possible singularities.

It is important to mention that there are many applications where this methodology could be useful to determine infiltration processes by natural or induced recharge and discharge, to achieve not only the objectives of this paper but also strategic objectives such as the management, use, and analysis of groundwater, as well as the recovery of exploited aquifers.

4.6. Discussion

The uncontrolled dewatering process implemented at the Torre Tres Ríos project, using deep wells with a pumping rate of 120 lps, resulted in excessive drawdown, leading to differential settlement of adjacent structures considering the radius of the conus depression with groundwater over-extraction. This underscores the necessity of adopting controlled dewatering strategies that minimize disturbances and optimize pumping flow rates while maintaining stable excavation conditions. This finding aligns with previous research [9] emphasizing that the optimization of construction dewatering has important value in protecting the surrounding environment and reducing engineering costs. Previous studies [24,41,42] confirmed that dewatering techniques cause settlement, broadly affecting the project and its surroundings. By incorporating numerical modeling, this study provides a systematic approach to designing optimized dewatering systems, ensuring safer and more predictable outcomes in similar construction scenarios.

While this study provides valuable insights into optimized dewatering strategies, several limitations must be acknowledged. First, the model relies on hydrogeological parameters derived from pumping tests and comparisons with regional reports, which may have introduced uncertainty. Second, this study assumed homogeneous soil conditions, whereas actual field conditions may exhibit localized heterogeneities that affect water movement. Third, additional field data—such as high-resolution time-series groundwater measurements—could further refine the model accuracy.

In this study, a sensitivity analysis revealed that hydraulic conductivity and transmissivity were the most influential parameters, significantly affecting simulated water table elevations. Variations in hydraulic conductivity within the reported range (1.0 × 10⁻⁵ to 5.0 × 10⁻⁴ m/s) resulted in deviations of up to 1.5 m, highlighting the necessity of accurate in situ measurements to reduce uncertainty. Similarly, transmissivity exhibited a moderate-to-high influence on model predictions, reinforcing the importance of detailed site-specific pumping tests as applied to refine these estimates.

It is important to mention that the excessive pumping rate was found to cause the drawdown cone’s radius of influence to exceed the distance from the extraction well. While the formula used for this calculation assumes radial flow and steady-state conditions for practical use, the uncertainties of parameters involved in the system could lead to a misrepresentation of the aquifer behavior. Moreover, maintaining a constant extraction rate throughout this period would exacerbate the potential impacts on surrounding environments and structures. This highlights the need for further evaluation of these variables, especially regarding their sensitivity to changes in hydraulic conditions and the long-term effects on adjacent areas.

Finally, researchers [43] emphasize that numerical modeling is an important tool and, in some cases, the only available option for designing dewatering systems, especially in projects planned under complex hydrogeological conditions with limited data.

The broader implications of this research extend to groundwater management policies and sustainable construction practices. Construction specialists can benefit from this study by ensuring that excavation processes achieve optimal conditions while minimizing risks to urban infrastructure. Also, by evaluating the impact of key parameters, engineers and hydrogeologists can make informed decisions, reducing the risks of over-extraction or ineffective groundwater control.

Future research should focus on enhancing model accuracy through high-resolution field monitoring, including automated piezometric sensors to capture real-time groundwater fluctuations. By continuously refining numerical modeling techniques and integrating field data, future studies can contribute to more effective, sustainable, and safer groundwater control strategies.

5. Conclusions

Based on the results presented in this paper, the following conclusions were reached:

• The analysis of a numerical model of the “Torre Tres Rios” project in the city of Culiacan, Sinaloa, indicated that the main recharges are induced by the constant load of the water table in the area, as well as the influence of the Tamazula River and precipitation. Additionally, the main discharges include the river and the same constant load of the water table.
• The configuration of the water table suggests that the contribution of the Tamazula River discharges into the Humaya River and is influenced in the area where the project was carried out.
• The wellpoint system model consisted of 28 wells positioned along the flow direction. The wells were spaced 3 m apart with a diameter of 2 inches, as shown in Figure 14. They were connected to a main 6-inch collector and linked to a rotary suction pump with a capacity of 20 HP. In the model, the system reached the excavation level without facing localized overexploitation of the aquifer.
• The observed data from the field, compared with the simulated data, showed that the system controlled the water table at a depth of 6.5m by pumping 28 lps under critical conditions; in contrast, with traditional methods such as deep wells, pumping of 120 lps was needed. In this way, the process was efficient, since the pumping required was considerably reduced (by 75%); also, the pump power was lower, so the hydraulic design had better performance. It also reduced the radius of influence of the drawdown cone and the potential impacts, including overexploitation (excessive pumping rates), which could cause instability in the excavation slopes or settlement of nearby structures.
• The sensitivity and uncertainty analyses confirmed that hydraulic conductivity, transmissivity, and recharge rates were the primary factors influencing groundwater fluctuations in the study area.
• Finally, a model was obtained that achieved the objective of this paper: to evaluate a dewatering system that enabled the construction of the project under optimal conditions and a reduced pumping flow rate. Additionally, the model represents the behavior of the aquifer flow regime, which is altered by the dewatering system. These results provide a reliable basis for informed decision-making in dewatering optimization, allowing for safer and more efficient and sustainable practices in construction.