Analysis of the Influence of Moisture Variation on the Behavior of Tropical Soils of Carajás Railway

Abstract

The criteria for selecting materials for railway pavement (particle size distribution, California Support Index, and physical indexes) are not suitable for evaluating the hydraulic behavior of tropical soils under unsaturated conditions, often resulting in the rejection of soils with good mechanical and hydraulic performance. The railway trackbed is exposed to precipitation, allowing water infiltration into the pavement layers and consequently leading to moisture variation. Therefore, the present study aims to assess the moisture variation over time in the railway trackbed when subjected to the action of rainfall. For this purpose, five soils were analyzed as constituent materials of the subballast, with the infrastructure of the Carajás Railway in Brazil being used as a reference. Water retention curves and conductivity tests were conducted using the HYPROP (Hydraulic Property Analyzer) and WP4-C (Water Potential Four) equipment, and numerical modeling was developed using the IVFlow software. The numerical modeling revealed that sample 1 (lateritic sand) performed better compared to the other analyzed soils, as it exhibited lower moisture variation (3.25% above the optimum content) under the influence of rainfall and it also had low permeability.

1. Introduction

Several factors influence the performance of a railway pavement throughout its lifespan, such as traffic loads, properties of the materials used, and environmental factors.

One relevant environmental factor is the variation of moisture in the ballast and subballast layers, which can occur due to precipitation, surface runoff, and groundwater rise, as indicated by Indraratna et al. [1].

In conventional pavement designs, the determination of the mechanical properties of materials assumes their optimum moisture content. However, it is important to consider that the railway platform is constantly exposed to rainfall, allowing water infiltration into the pavement layers.

Therefore, it is necessary to consider that the mechanical properties of the materials used in the pavement layers, such as the resilient modulus and permanent deformation, are influenced by moisture content variation, as highlighted by Guimarães et al. [2]. Consequently, the initial design may not always reflect the actual conditions to which the railway pavement will be subjected throughout its lifespan.

The resilient modulus is an indicator of the material’s elastic behavior and plays a crucial role in evaluating its suitability as one of the pavement layers. Therefore, it is essential to examine the soil’s response to water to make well-informed decisions in selecting materials for railway pavement construction [2].

In the design of railway pavements in tropical regions, the use of lateritic soil is common. However, the guidelines provided by the American Railway Engineering and Maintenance-of-Way Association (AREMA) [3], which govern railway pavement design, do not offer specific criteria for this type of material.

The Carajás region in Brazil is known for its abundance of lateritic soils, which are frequently used in railway construction. However, the behavior of these soils under variable moisture conditions is not fully understood. Unlike the stone materials used in international railways, lateritic soil is susceptible to water impact, and infiltration can compromise the railway platform, leading to a reduction in resilient modulus and an increase in permanent deformations.

When it comes to the characteristics and main functions of the subballast layer, there is some discordance among authors regarding the functions that it should fulfill, as per Figure 1.

Stopatto [5] and Brina [6] proposed that the subballast layer should enhance the bed’s erosion resistance and impede water penetration, thereby maintaining the subgrade’s properties.

In contrast, Selig and Waters [7] proposed that the subballast layer should serve as a drainage layer to facilitate the removal of water that rises through the subgrade by capillarity.

Furthermore, it is stated that the subballast must adhere to the Terzaghi filter criterion by possessing a certain level of permeability that enables drainage, while also preventing the migration of fines into the underlying layers.

In contrast, Guimarães et al. [2] suggest that the subballast should be impermeable to safeguard the subgrade from erosion. However, in agreement with Selig and Waters [7], Indraratna et al. [1] assessed that the layer should be permeable enough to enable water to flow and channel it towards the lateral drainage devices. It should also permit the rise of water from the subgrade, alleviating excessive pore pressures when saturated. To achieve this, the subballast must meet the Terzaghi filter criterion [8].

AREMA [3] recommends that the subballast layer should be both impermeable to prevent saturation of the subgrade and permeable enough to allow for a capillary rise of water, which can help prevent saturation.

However, it is worth noting that the authors who advocate for the subballast layer to adhere to the Terzaghi filter criterion have focused on studying stony materials. Given that this study employs tropical soils as a subballast material, it is important to acknowledge that, as pointed out by Nogami and Villibor [9], lateritic soils classified as tropical soil have promising potential for pavement use due to their high resistance and low permeability, qualities of great interest in permanent way projects. Indeed, according to some authors ([10,11,12,13,14,15,16,17,18,19]) the use of lateritic soils in the subballast layer has demonstrated satisfactory performance.

Delgado [10] conducted a study using fine lateritic materials collected in western Maranhão, which yielded excellent results and was validated using this material for the subballast layer. The layer tended to accommodate the permanent deformation shakedown, indicating that the material classified as LG’ (laterite clayey soil) is suitable for use in this layer.

Von der Osten [11] assessed the mechanical performance of tropical soils from the Carajás railway (EFC) region, which were previously considered unsuitable for railway pavement according to conventional criteria. However, his study showed that these soils exhibited favorable behavior when used in the subballast layer.

Sousa [12] conducted a study where he measured the resilient modulus of 53 soil samples collected along the Estrada Ferro Carajás. The lateritic samples exhibited a predominantly medium-to-high resilient modulus, indicating their potential suitability for use as subballast material.

Lopes [13] conducted a study to assess the suitability of four types of tropical soils (fine and granular, lateritic and non-lateritic) as subballast material for a permanent railway track, taking into account their hydraulic and mechanical properties. Based on laboratory tests and numerical modeling of infiltration using the IVFlow model, the author concluded that the lateritic soils under investigation had high suction, low permeability, and were prone to water-induced damage.

Menezes [14] conducted a comparative analysis of the performance of five soil types used in the subballast layer, as a capillary barrier to protect the subgrade of a railway platform under the impact of rainfall. The study revealed that sandy lateritic soil and lateritic sand exhibited a superior performance compared to other soil types.

Zhang [15] carried out a study that involved laboratory testing on cohesive soil samples collected from different locations in South China. The researchers measured the resilient modulus and soil suction under various moisture conditions using specialized equipment and testing procedures. The findings indicate a significant correlation between the resilient modulus and soil suction for cohesive soils. As soil suction increased, the resilient modulus decreased, indicating a softer and more deformable material. This relationship has implications for pavement design and performance, as it helps in predicting the resilient behavior of cohesive soils under different moisture conditions.

Guimarães [16] and Guimarães et al. [2] observed in the field the satisfactory performance of a fine sandy lateritic soil used as a subballast material in a railway section built in 2015. The section has been monitored and is still in good condition.

López-Pita et al. [17] declare that when the subgrade lacks adequate mechanical properties, it contributes significantly to the overall deterioration of the geometry. Therefore, it is important for the subballast layer to function as a capillary barrier to safeguard the subgrade.

Supporting this argument, the climate is considered one of the most relevant factors that lead to issues in railway subgrade. The adverse weather conditions, coupled with the high axle loads of heavy haul railways, significantly impact the soil saturation levels, resulting in fluctuations in its bearing capacity, as highlighted by Li and Selig [18].

Considering the above, it is evident that conducting an in-depth study on the utilization of tropical soil as a subballast layer is necessary. This layer can act as a protective shield against water’s detrimental effects on the subgrade, particularly because there is a vast abundance of tropical soils available in Brazil.

Moreover, while the subballast of the Carajás railway serves as a case study, exploring the suitability of tropical soils is a topic of interest for other regions, including Africa and Australia, which share similar soil characteristics [19].

In this context, this study aims to contribute to the understanding of moisture variation over time in the railway platform when exposed to rainfall. This study presents the following hypothesis: the action of rainfall significantly impacts the moisture variation of the layers that comprise the railway pavement and, consequently, its mechanical behavior. Thus, this research is of great relevance to the railway sector since there are few studies evaluating the impact of rainfall on the mechanical behavior of railway pavement layers.

2. Materials and Methods

2.1. Site Description

Carajás railway (Estrada de Ferro Carajás—EFC) is the highest transport capacity railway in Brazil, covering 972 km, as seen in Figure 2. The EFC, located in Carajás, Brazil (Southeast Pará), connects the largest surface mining iron ore operation in the world with the Port of Ponta da Madeira in São Luís, Maranhão Figure 2, and currently tops the list of most productive railways in Brazil according to National Land Transportation Agency (ANTT) [20].

EFC has a throughput of more than 120 million tons and transports 350,000 passengers per year. Among the largest trains in operation in the world, about 35 of them run in parallel along its tracks, including one comprising 330 wagons and extending for 3.3 km. The construction of 559 km that will connect the 54 switching yards of the EFC is still underway, which will further increase the transportation volume of the EFC, requiring platforms with higher performance.

The EFC is in a region of high precipitation and elevated temperatures. The railway track is continuously exposed to high precipitation, which leads to the premature deterioration of the permanent way. The rising humidity levels accelerate the degradation process, expediting the deterioration of the railway track.

Due to the abundance of humid tropical climates, Brazil has a high prevalence of tropical soils. According to Nogami and Villibor [9], these soils possess distinctive characteristics and properties that are notably distinct from non-tropical soils due to the operation of geological and/or pedological processes that are common in regions with humid tropical climates.

Given that laterite soil is common as a substructure material in tropical regions, the standard that guides the sizing of railway pavements soil, the American Railway Engineering and Maintenance-of-Way Association (AREMA) [3], does not provide criteria for this type of material, thus it is necessary to carry out studies focused on the behavior of tropical regions.

Unlike the stone materials used in international railways, lateritic soil is susceptible to the impact of water, and seepage can jeopardize the railway platform, leading to a decline in the resilient modulus, culminating in permanent deformations.

The resilient modulus is an indicator of the elastic behavior of the material, which plays an important role in assessing its suitability for use as one of the layers of the pavement. Thus, examining the soil’s water response is imperative to make a well-informed choice of materials for railway pavement construction.

Given this context, the objective of this research is to evaluate the variation of humidity over time on the railway platform when subjected to the action of rainwater.

2.2. Research Description

This research is a part of a collaborative agreement named “Studies for the Review of Design Criteria for Permanent Way”, which was established between the IME (Military Institute of Engineering) and VALE SA. The primary objective of this agreement is to evaluate the material selection criteria and dimension assumptions currently implemented in railway engineering projects. Broadly speaking, these criteria are derived from international studies, and are thus not tailored to the specific conditions and peculiarities of Brazil, as is the case with the AREMA [3] standard.

To evaluate the suitability of these materials as a railway pavement layer, 53 samples were collected along the EFC. As stated by Sousa [12], the collection process aimed to identify materials that, from a visual standpoint, would not meet the design criteria recommended by traditional selection methodologies, such as those prescribed by the AREMA standard. Nonetheless, using more contemporary methods could reveal their potential for good performance.

The following methodology was employed in this study: sample collection, experimentation, and numerical modeling as shown in Figure 3.

2.3. Soil Characterization

Five soil types were analyzed to evaluate their suitability for use as the subballast layer of the railway platform, with regards to their hydraulic behavior as shown in

Table 1. Characterization of soils from EFC.

Characterization
Compactation/Intermediate Energy					Atterberg Limits
Sample	MCT	USCS	W Optimum (%)	MEAS (g/cm3)	DR	WI (%)	Wp (%)	PI (%)
1	LA	SP-SM	6.00	2.07	2.65	17.99	NP	17.00
2	NS’	CL	15.40	1.67	2.72	35.00	0	35.00
3	LG’	CL	6.00	1.92	2.62	29.00	19.00	9
4	NA’	CL	16.50	1.74	2.66	NL	NP	NP
5	NG’	SC	16.00	1.76	2.64	27.00	10.00	17.00

where LA = laterite sand; NS’ = non-laterite silty soil; LG’ = laterite clayey soil; NA’ = non-laterite sand soil; NG’ = non-laterite clayey soil; Woptimum = optimal humidity; MEAS = specific mass of dry aggregate.

.

The MCT (Miniature, Compacted, Tropical) methodology was employed for soil classification, as this system takes into consideration its mechanical properties such as particle size distribution, plasticity, and compaction characteristics. According to Nogami and Villibor [9], authors of the MCT methodology, soils classified as LA (laterite sand) and LA’ (non-laterite sand) exhibit desirable mechanical behavior for pavement applications due to their high cohesion, which results in high stiffness when moistened and compacted. Furthermore, when used in railway subballast, the LA soil type, as described in Section 2, showed excellent performance.

However, employing the criteria established by AREMA [3] for selecting materials for railway pavement, sample 1 (LA) would be considered inadequate as it falls outside the recommended ranges for particle size distribution and the values of physical indices traditionally adopted by the standards.

2.4. Retention Curves and Conductivity Curves

To obtain the required parameters for numerical simulation, the first step was to obtain the characteristic curve and soil conductivity of the samples under study. Therefore, the material available was initially sieved, and the fraction that was retained on the no. 10 sieve (2 mm opening) was discarded as the HYPROP (Hydraulic Property Analyzer) and WP4-C (Water Potential Four) equipment utilized for obtaining the hydraulic functions does not allow for working with material with larger granulometry.

The HYPROP is an instrument used to measure the soil water potential at different levels of saturation, employing two tensiometers. During the experiment, the soil sample is placed on a laboratory balance, while the HYPROP records the changes in water potential resulting from water evaporation over time. Additionally, the instrument monitors the weight variations of the sample, allowing for the determination of soil moisture content. It is worth noting that the HYPROP exclusively measures the matric potential in the wet range of the soil. Therefore, to obtain measurements of water potential in the dry range, it is necessary to complement the use of HYPROP with the WP4-C [21].

The HYPROP-FIT software was selected for curve fitting due to its high accuracy and versatility in fitting retention data using various methods. It is particularly effective in fitting both traditional and bimodal (tropical soils) retention curves, accurately capturing the bimodality of the measured retention curves and effectively describing the hydraulic behavior across the entire measurement range [22,23].

Medina and Motta [24] claimed that the nature of the fine fraction plays an important role in determining the soil’s resilience behavior, and this was taken into consideration. The HYPROP equipment has an optimal measurement range limited to tension values ranging from 0 cm to 1000 cm of the water column. Thus, additional data points were necessary throughout the tests performed using the WP4-C equipment to obtain a more precise delineation of the retention curve obtained with the HYPROP, since the optimal measurement range of the equipment WP4-C varies from 1000 cm to 1,000,000 cm of the water column.

The unsaturated hydraulic conductivity of soil is related to its saturated hydraulic conductivity value as well as its water retention behavior, according to Wang [25]. The determination of the permeability coefficient k involves various methods, which can range from simple calculations to more intricate laboratory and field tests. The selection of the appropriate method for evaluating parameter k primarily depends on the soil type. Eurocode 7 identifies four methods for assessing parameter k, that are crucial to conducting a comprehensive analysis of the soil’s properties: laboratory tests, field tests, estimation based on oedometer tests, and empirical correlations [26]. In the present work, the permeability was obtained through laboratory testing using the HYPROP.

The tests with HYPROP were carried out at the Geotechnical Laboratory of COPPE/UFRJ—Environmental Geotechnics Section and the tests with WP4-C were carried out at Embrapa Solos at Rio de Janeiro-RJ.

2.5. Numerical Simulation

The second stage of this study involved the utilization of numerical modeling through the software IVFlow, to simulate water infiltration on the railway platform. IVFlow was jointly developed by the Military Institute of Engineering (IME) and VALE SA [27]. The software uses the finite element method to solve the Richards equation [28], which governs the transient process of groundwater infiltration. The purpose of this stage was to comprehend the process of infiltration which takes place on the railway platform.

The model construction involved incorporating parameters that accurately represented the EFC infrastructure, while the selection of the study region facilitated the adoption of precipitation conditions from the program’s extensive database. Utilizing the multilinear option, the hydraulic model was meticulously constructed, and the experimentally derived hydraulic functions were precisely inputted, referencing the relevant information provided in

Table 2. Hydraulic properties of samples tested in HYPROP.

Hydraulic Properties
Sample	1	2	3	4	5
Conductivity saturation Hydraulics (cm/h)	0.0099	0.0011	0.0013	0.0074	0.017
Saturation humidity (m3/m3)	0.2946	0.4851	0.4907	0.3662	0.3849
Residual moisture (m3/m3)	0.1817	0.1822	0.0685	0.0621	0.1235

,

Table 3. Data from the Multilinear constitutive model of sample A1 (LA).

Retention Curve Points		Conductivity Curve Points
Pressure (cm)	Volumetric Humidity (m3/m3)	Pressure (cm)	Hydraulic Conductivity (cm/h)
0	0.29	0	0.0099
−3	0.29	−30	0.0099
−8	0.29	−36	0.0093
−16	0.28	−44	0.0067
−21	0.27	−57	0.0042
−26	0.26	−80	0.0019
−31	0.24	−126	0.0020
−1000	0.13	−245	0.0016
−5370	0.07	−515	0.0009
−12,022	0.03	−805	0.0021

and

Table 4. Data from the Multilinear constitutive model of sample 4 (NA’).

Retention Curve Points		Conductivity Curve Points
Pressure (cm)	Volumetric Humidity (m3/m3)	Pressure (cm)	Hydraulic Conductivity (cm/h)
0	0.37	0	0.0074
−5	0.36	−5	0.0077
−11	0.36	−11	0.0083
−25	0.35	−41	0.0081
−49	0.33	−63	0.0045
−67	0.31	−72	0.0051
−76	0.29	−89	0.0016
−2570	0.12	−128	0.0004
−3019	0.15	−264	0.0001
−645	0.06	−642	0.0000

.

3. Results

3.1. Laboratory Results

The present article focuses solely on the methodology employed for sample 1, which yielded the most significant results.

Sample 1 is a coarse sandy soil that was collected from the foot of a slope at the site of the railway duplication. It was visually identified as LA (lateritic sand) according to the MCT classification, and its characteristics are presented in Table 1, as reported by Sousa [12].

The retention curve of sample 1 was obtained from the test carried out in the HYPROP equipment and was adjusted by the bimodal Fredlung–Xing model [29] using the HYPROP-FIT program, which was found to provide the best fit among the five available models (Figure 4).

By including the complementary points from the WP4-C test in the HYPROP-FIT program, a new curve with the appearance shown in (Figure 5) was obtained.

The addition of supplementary points did not substantially alter the appearance of the retention curve for sample 1. The sample exhibited the typical curve shape of sandy soils, with a low air entry value followed by a sharp decline.

However, despite being identified as lateritic sand, the retention curve of the sample does not exhibit the typical bimodal shape of lateritic soils. It only shows a single air entry value, around 1 pF (10 cm of water column), after which the soil behaves like an unsaturated soil, with moisture content decreasing as the suction increases. At the optimum moisture content (12.42 m³/m³), the suction reaches a value of approximately 1.7 pF (50 m of water column).

The hydraulic conductivity curves, obtained using the bimodal Fredlung–Xing model to adjust the retention curve (Figure 6) in HYPROP, indicated a saturated conductivity of 4.6 × 10⁻⁶ cm/s and 1.2 × 10⁻⁷ cm/s at the optimum moisture for sample 1. These values are considered low for sandy soils. The same process was carried out to obtain the retention and conductivity curves of sample 4 (Figure 7 and Figure 8).

3.2. Numerical Modelling

The program IVFlow generated a temporal graph of volumetric humidities (t = 0) as an initial step. By analyzing the graph, it was observed that the subballast layer had a moisture content of 22.65 m³/m³ before being subjected to the weather variation. This value was 3.3% below the saturation moisture, and 4.95% higher than the optimum moisture content of the material.

The elevated initial humidity can be attributed to capillary rise caused by the infiltration of groundwater. This is because in the model construction, a water level of 2.6 m was assigned, and before precipitation, this was the only probable water source for the pavement. After identifying the issue, the simulations were repeated by assigning a higher water level of 10 m. However, the subballast layer continued to exhibit the same initial moisture content.

The rainfall regime adopted was that of Otto Pfastetter, given that the parameters of his equation for the study region were already in the program’s database (a:0.156; b:0.08; a:0.4; b:42; c:10). In addition to existing data, the following were also used: start: 0 h; duration: 1 h; ramp-up: 5%; TR: 30 years. The simulation considered a period of three hours of rain.

The program’s outputs revealed that the subballast layer, composed of sample 1, attained a maximum moisture content exceeding the optimal level by 8.2%. This indicates that rainfall caused a 3.3% increase in the moisture content, considering that the layer already had a moisture content exceeding the optimum by 4.95% prior to the simulation. Furthermore, the results showed that the entire layer underwent a moisture variation greater than 3% compared to the optimal moisture content (see Figure 9 and Figure 10).

The obtained result confirms the conclusions of previous studies such as Delgado [10], Von der Osten [11], Sousa [12], and Lopes [13], which verified the good performance of fine sandy lateritic soils in railway pavement in terms of mechanical behavior. Specifically, Sousa [12], when evaluating the soil from sample 1 in terms of its mechanical behavior, suggests the application of the material as a subballast layer. Guimarães et al. [2] also verified the good field behavior of the soil classified as LA (laterite sand) in a specific section, which once again confirms the results obtained in this research.

Sousa [12] concluded in her study that LA-type soils are suitable for pavement applications. The author also mentioned that NA-type soils exhibit good mechanical behavior. However, when in contact with water, these soils experience compromised performance, making them suitable for highways that have surface coatings providing protection against moisture. On the other hand, they may not be recommended for railway pavements that lack such protection. These findings align with the hydraulic behavior observed in this study.

In her research, Lopes [13] obtained a bimodal curve for LA soils and a unimodal curve for NA soils, similar to the curves obtained in this study. The hydraulic conductivity value found by the author for saturated LA soil of 0.0758 cm/h is of the same magnitude as the value of 0.009 cm/h found for the LA soil studied in this research. The author also found a value of 0.066 cm/h for NA soil, which is consistent with the value of 0.0074 cm/h found for the same soil type in this study. Both conductivity values are also within the range typically observed for sandy soils.

4. Conclusions

The numerical analysis revealed that at the onset of simulations, all samples displayed moisture levels above the optimum, except for sample 5 which initially had a moisture content 4.9% lower than the optimal level. This pattern can be attributed to the impact of the water level, as the pavement was initially affected only by groundwater infiltration.

Samples 1 (LA) and 4 (NA’) were found to be less vulnerable to the impact of the water level. Conversely, the remaining samples exhibited initial moisture contents near the saturation level, indicating a higher probability of notable variations in the resilience modulus and consequently, permanent deformations.

Based on the initial humidity values provided by the program and using the increase in humidity as a criterion, it was observed that sample 1 performed better with a lower increase in humidity after exposure to rain, compared to sample 4. Sample 4 experienced an increase of 9.55%, while sample 1 showed a variation of only 3.25% in relation to the optimum moisture content.

Sample 1 also exhibited low permeability, which is a desirable property for the subballast to act as an impermeable layer and protect the subgrade. The lateritic nature of this sample further enhances its applicability due to the formation of rigid blocks, typical behavior of laterites, providing greater support capacity to the pavement. Based on the hydraulic behavior analysis, it can be concluded that sample 1 can still be used to compose the subballast layer of the railway pavement, even if it does not meet the traditional criteria currently in practice.

Thus, through this study, it was possible to verify the need for the development of a more appropriate material selection method that considers the hydraulic behavior of the soil in addition to its mechanical behavior. This is because current criteria often result in the exclusion of soils that have demonstrated, in this study and in various cited studies, good performance when employed in the subballast layer.