Existing pavement performance and pothole prediction models address pavement deterioration primarily through indirect indicators rather than explicitly modeling pothole formation [14]. Mechanistic-empirical design approaches and AASHTO-based procedures provide robust frameworks for structural pavement design; however, they largely focus on long-term fatigue cracking and rutting performance and may not adequately capture early surface failures driven by localized moisture ingress and drainage deficiencies [13][15][19]. Similarly, network-level models such as HDM-4 and pavement management systems are effective for maintenance prioritization but often rely on aggregated condition indices, limiting their ability to represent site-specific interactions between traffic loading, rainfall exposure, material variability, and construction quality [20]. Furthermore, many existing models consider traffic, environmental, and material factors independently, despite evidence that pothole formation results from their combined and interacting effects [21]. These limitations highlight the need for an integrated modeling approach capable of capturing the coupled influence of traffic loading, environmental exposure, subgrade strength, and construction practices, particularly under tropical conditions where moisture-related deterioration is dominant.

The Machakos-Kangundo Road in Kenya, upgraded to bitumen standard in 2018, provides a representative case of premature pothole formation in a newly constructed flexible pavement. Despite being designed for an expected service life consistent with national standards, the road has experienced widespread pothole occurrence well before the end of its design period. This study addresses this challenge by developing a comprehensive road design model that integrates traffic loading, environmental conditions, material properties, and construction practices to mitigate pothole formation in newly constructed flexible pavements.

A mixed-methods approach combining qualitative and quantitative techniques was adopted to investigate pothole formation in flexible pavements. Qualitative data captured road users’ perceptions of pavement condition and safety, while quantitative data were used to analyse pothole characteristics, traffic loading, pavement materials, and environmental conditions.

A case study design was employed using the Machakos-Kangundo Road in Machakos County, Kenya. The corridor carries mixed traffic dominated by public transport vehicles and freight traffic. The road is exposed to variable climatic conditions. Meteorological data, including rainfall and temperature, were obtained from a weather station situated at Machakos, which were aggregated into monthly averages. The temperature and rainfall data were considered for the year between 2017 and June 2024.

The road corridor was divided into 20 m chainages to enable accurate localization of potholes and sampling points. Field surveys conducted at each of the chainages documented pothole dimensions, frequency, and spatial distribution using a standardised checklist. Potholes of a minimum size of 500 × 200 × 50 mm (length × width × depth) were included in the survey.

Traffic counts and environmental measurements were conducted at three representative locations: Machakos (0 + 200), Kenol (11 + 000), and Kangundo (20 + 000). Traffic and axle load measurements involved capturing the volume, classification, and physical weight of vehicles. In the representative locations, three stations were set up, and the vehicles were predefined into classes of cars, vans, small trucks, buses, medium trucks (2 axles), heavy trucks (4 or more axles), and oil tankers. The traffic counts were done in 24hours for 4 consecutive days by traffic enumerators. This enabled the calculation of the Average Daily Traffic (ADT). In order to obtain the Average Annual Daily Traffic (AADT), the ADT values were multiplied by passenger car equivalent together with monthly expansion factors specified in the Kenya Road Design Manual [22].

The Equivalent Standard Axle Load (ESAL) was determined using the axle load equivalency factors prescribed in the design manual [22]. To determine the Cumulative Equivalent Standard Axles (CESA), three traffic growth scenarios (low, medium, and high) were considered and projected for a 15-year design period.

Laboratory tests were conducted to determine material properties influencing pavement performance, including Atterberg limits, California Bearing Ratio (CBR), bitumen penetration, Marshall stability, and aggregate impact value. Testing followed relevant standards to ensure consistency and reliability of results [23]-[26].

GIS-based analysis was used to evaluate road classification, pavement type, and condition. Pavement coring was undertaken to determine actual layer thickness and material conformity according to [25]. Pavement performance was assessed using standard indicators, including International Roughness Index (IRI), Present Serviceability Index (PSI), Pavement Condition Index (PCI), rutting depth, cracking index, and pothole density in line with established pavement performance evaluation frameworks [27].

A comprehensive pothole formation model was developed using a hybrid analytical approach that integrates regression modeling and a mechanistic-empirical framework. Regression modeling was applied to quantify relationships between traffic load, subgrade strength, rainfall, and pavement condition. The mechanistic approach enabled the inclusion of material properties and structural behavior of pavement to loading. The general form of the model was expressed using Equation (1):

where P d is pothole density, X 1 axle load, X 2 subgrade CBR, X 3 rainfall intensity, X 4 pavement age, β i regression coefficients, and ε the error term.

ModelArchitecture:

The flowchart shown in Figure 1 has three main sections, namely, the input layer, the Mechanistic-Empirical modeling, and the formation of the comprehensive model. At the input stage, raw variables were input as collected from the road chainages. The mechanistic-empirical (ME) modeling computed the structural damage caused to the road corridor from the effects of traffic and material stiffness. The damage predicted by the ME was amplified by rainfall intensity and drainage inefficiency. The construction defect factor adds a consideration for under-compaction, uneven layer thickness, poor tack coat, and the observed quality assurance and control practices. Finally, in order to get a numerical score representing the likelihood of pothole formation, the construction quality was integrated into the ME modeling.

The Machakos-Kangundo road comprises a two-lane flexible pavement carrying mixed traffic, with the dominant being 14-seater matatus, heavy commercial trucks, and motorcycles. Table 1 summarizes the observed site-specific factors.

Figure 1. Flowchart for the comprehensive pothole formation model.

Table 1. Observed site-specific factors along the Machakos-Kangundo road.

As illustrated in Figure 2, pothole frequency peaks sharply between chainages 18 + 000 and 24 + 000 km, where frequencies exceed 70 potholes per 20 m. These sections coincide with poor surface drainage, soft subgrade conditions, and visible structural degradation. These findings demonstrate that pothole formation is governed by localised structural and environmental conditions rather than uniform pavement ageing.

Figure 2. Pothole frequency distribution along the Machakos-Kangundo road chainages.

As observed by Hamzah et al. [28], temperature and rainfall variations affect asphalt stiffness, moisture infiltration, and subgrade performance. The results presented in Table 2 indicate bimodal rainfall patterns (April and December peaks), with temperature ranging from 16.9˚ to 21.5˚. These variations, though not extreme, are sufficient to influence asphalt binder stiffness and flexibility as noted by [28]. High temperature has been argued to cause weakening of flexible pavements and cause cracking of the asphalt through thermal expansion and contraction [29]. Conversely, rainfall saturates through the existing cracks, creating a weakened base layer that is susceptible to damage from continuous loading by the traffic. It can therefore be inferred from these results that this weather pattern provides a conducive environment for pothole formation and propagation along the Machakos-Kangundo road.

Table 2. Average monthly temperature and rainfall along the Machakos-Kangundo road.

To quantify the relationship between pothole formation and influencing variables, a correlation and regression analysis were performed using data on pothole frequency, drainage condition, rainfall, and construction deficiencies as presented in Table 3.

Table 3. The contributing effect of hydrological and construction deficiencies in pothole formation.

Pavement condition indices showed progressive deterioration from Machakos towards Kangundo, as shown in Table 4. The progressive increase in International Roughness Index (IRI) values from Machakos to Kangundo section suggests a consistent decline in surface smoothness due to accumulated pavement distress, particularly rutting and cracking. This pattern aligns with the field observations where increased traffic loading and inadequate drainage around the Kangundo area contributed to surface unevenness and reduced comfort levels.

Table 4. Pavement performance indicators for the Machakos-Kangundo road.

According to [27], Present Serviceability Index (PSI) values below 2.5 indicate the need for immediate maintenance or rehabilitation. Therefore, the Kangundo section requires priority intervention to restore it to acceptable serviceability levels. The results demonstrate a strong negative correlation between PSI and IRI. Similarly, the cracking index’s sharp increase from 12% to 34% across the same sections illustrates growing loss of pavement integrity.

A critical finding of the volume characteristics along the corridor indicates that the Kangundo section, as shown in Figure 3, carries 8.52% more heavy goods vehicles compared to the Machakos section. This explains the heightened distress, higher rut depths, and reduced pavement performance indices along this section.

Figure 3. Average daily traffic by vehicle category along the Machakos-Kangundo road.

In order to allow for evaluation of cumulative pavement loading under different traffic growth conditions, the Equivalent Standard Axle Loads (ESAL) were determined using the axle load equivalency factors prescribed in [30].

The Daily Equivalent Standard Axles (DESA) for the Machakos-Kangundo road represented a moderately trafficked Class T3 roadway according to [30]. Importantly, as shown in Table 5, projections under various growth scenarios (low 4%, medium 5%, and high 6%) showed a steady rise in DESA to the year 2027. This growth pattern demonstrates not only the increasing traffic burden but also the accelerating deterioration rate expected over the Machakos-Kangundo corridor.

Table 5. Traffic loading and cumulative equivalent standard axles for the Machakos-Kangundo road.

The Atterberg Limit values indicate that the subgrade soils along the Machakos-Kangundo road predominantly fall within the medium to high plasticity range, consistent with the AASHTO classification system, as shown in Table 6.

The trend depicted in Figure 4(a) confirms the presence of highly plastic clayey soils between chainages 18 + 000 and 24 + 000 km.

Subgrade soils exhibited poor to fair strength as expressed in Figure 4(b), with CBR values ranging between 5% and 14%. This indicated limited structural support by the subbase. The variability in compaction quality and pavement layer thickness (Figure 5(a) and Figure 5(b)) indicated further reduced pavement structural capacity. These deficiencies, when combined with moisture ingress and heavy traffic loading, created favourable conditions for premature pothole formation, in agreement with the findings of [8].

Table 6. Atterberg limits at various sections of the Machakos-Kangundo road.

Figure 4. Variation of Atterberg limits and CBR on Machakos-Kangundo road.

Figure 5. Variation of maximum dry density, optimum moisture content, and degree of compaction along the Machakos-Kangundo road.

The General Marshall Mix (GMM) value of 2.37 g/cm³ indicates a moderately dense and well-designed asphalt mixture that conforms to the requirements of conventional bituminous surfacing [22]. This suggests that, at the material design stage, the asphalt mixture was structurally sound and suitable for use under medium to heavy traffic loading.

The core density values shown in Table 7 revealed that the degree of compaction and air voids fall within the acceptable limits recommended by [27]. Confirming that most pavement sections achieved satisfactory field compaction.

The assessment of construction methods established a direct relationship between workmanship quality and pavement deterioration patterns (Table 8). The most significant deficiencies—namely, undercompaction, inadequate drainage management, and temperature mismanagement during asphalt laying—accelerated the degradation process, consistent with the observed pothole distribution and core density results.

Table 7. Core density results and degree of compaction.

Table 8. Summary of construction aspects evaluated.

Using the field and laboratory findings, the developed model helps to estimate pothole susceptibility and expected pothole density by integrating environmental conditions, traffic loading, material properties, road performance indicators, and construction methods. The Mechanistic-Empirical damage modeling enabled the prediction of pavement performance through consideration of two types of damage: fatigue damage and rutting damage.

FatigueDamage

The ME damage modeling predicted the damage fatigue as a function of cumulative equivalent standard axles (CESA) and the allowable load repetitions on the road corridor. The allowable load repetition values were calibrated using the CBR and the PCI trends observed in the results analysed. The damage was therefore calculated using Equation (2).

where,

CESA—cumulative equivalent standard axles;

N f —number of repetitions up to fatigue cracking (calibrated as a function of CBR and PCI trends).

RuttingDamage

The road performance indicators expressed the condition of the road in relation to the comfort of use. The ME modeling predicted the permanent deformation due to cracking and the vertical ruts created on each layer. The empirical rut depth takes into account the resilient modulus of each layer by incorporating the road performance indicators as shown in Equation (3).

where,

R—predicted rut depth (mm);

CESA—cumulative equivalent standard axles;

CBR—soaked California Bearing Ratio (%) of the subgrade;

α , β , γ —empirical parameters to be calibrated (rut depth data (6 - 15 mm) were used to calibrate coefficients).

3.7.1. Quantification of Drainage and Construction Quality Variables

DrainageCondition Score

The three road sections were assigned a drainage condition score (DCS) at intervals of 0.25. The drainage condition score was then incorporated into the moisture-environment multiplier to reflect the role of water infiltration in pothole development.

Moisture-Environment Multiplier

In order to quantify the combined effect of rainfall and drainage performance on pavement deterioration, a moisture-environment multiplier (MEM) was developed using Equation (4). This multiplier combined the effect of recorded average monthly rainfall and the drainage condition score.

where,

MEM—moisture-environment multiplier;

R—average monthly rainfall (mm);

R_max—maximum rainfall recorded during the study period;

DCS—drainage condition score;

a—rainfall coefficient (calibrated to 0.40);

b—drainage coefficient (calibrated to 0.60).

On replacing the calibrated values of rainfall and drainage coefficient, Equation (4) can be written as Equation (5).

3.7.2. Construction Defect Factor

A construction defect factor (CDF) was established in order to combine the degree of compaction, layer thickness variation, tack coat quality, temperature control during asphalt laying, and construction quality assurance practices. First, a construction quality index (CQI) was calculated using Equation (6).

where,

S_i—score for each construction parameter;

n—number of parameters evaluated.

Then, the construction defect factor (CDF) was determined using Equation (7).

Integration into the effective damage model, the overall structural deterioration index becomes (Equation (8)).

where,

D_fatigue—traffic-induced structural damage;

MEM—environmental and moisture effect;

CDF—construction-related defects.

3.7.3. Weighting of the Variables

All variables incorporated in the pothole susceptibility index (PSI) model were normalized using min-max scaling to eliminate unit inconsistencies and ensure comparability. The weighting factors were derived through a hybrid approach combining observed statistical relationships between pothole density and explanatory variables with established pavement engineering principles. Structural damage, subgrade strength, and pavement condition exhibited the strongest influence on pothole formation and were therefore assigned higher weights. This approach converts previously descriptive variables into measurable and repeatable model inputs.

Considering the weighted factors, the comprehensive pothole formation model is therefore estimated using Equation (9). The equation estimates the PSI, which integrates the parameters measured and observed in the study.

The comprehensive pothole formation model successfully captures the combined effects of traffic loading, soil loading, soil plasticity, moisture susceptibility, pavement condition, and construction quality. The implementation of the model can be integrated with the suggested classification of PSI values summarized in Table 9.

Table 9. Classification values for the pothole formation model.