Past studies developed pavement performance models using data from all or multiple states across the United States. This study hypothesized that due to variation in agency practices and work activity profiles, individual pavement performance models should be estimated for each state, using data from its own roadway network, for use in its pavement management system. To test this hypothesis, this study used condition data of Interstate flexible pavements from eight Midwestern states to estimate three models: fixed-parameters regression, random-parameters regression, and artificial neural networks (ANNs). The ANNs model was found to statistically outperform the regression counterparts when estimating pavement roughness across all states. In contrast, the random-parameters model was statistically superior to the ANNs model in some cases when exploring the performance of these models for individual states. The statistical performance of models did not have a consistent trend across all states. Therefore, the application of models, based on data from multiple jurisdictions, could lead to erroneous/nonoptimal maintenance and rehabilitation decisions. Highway agencies are recommended to rely on their own jurisdictional data when developing their pavement performance models.

Various methodologies are being developed to build and improve probabilistic pavement performance models that have high prediction capabilities. However, the effectiveness of preventive maintenance (PM) has not been considered in such models due to the lack of historical PM data. Consequently, the predicted pavement condition is erroneous and often biased, which leads to nonoptimal maintenance and rehabilitation (M&R) decisions. This paper introduces and validates a hybrid approach to incorporate the impact of PM into probabilistic pavement performance models when historical PM data are absent. The types of PM treatments and their times of application are estimated using two approaches: (1) analysis of the state of practice of pavement maintenance through literature and expert surveys, and (2) detection of PM times from probabilistic pavement performance curves. Using a newly developed optimization algorithm, the estimated times and types of PM treatments are integrated into pavement condition data. A nonhomogeneous Markovian pavement performance model is developed by estimating the transition probabilities of pavement condition using the ordered probit method. The developed hybrid approach and performance models are validated using cross validation with out-of-sample data and through surveys of subject matter experts in pavement engineering and management. The results show that the hybrid approach and models developed predict probabilistic pavement condition incorporating PM effects with an accuracy of 87%.

Various types of Markov chain methodologies have been used in past research to develop stochastic performance models for infrastructure systems—the staged-homogeneous, nonhomogeneous, homogeneous, semi-, and hidden Markov. The primary components of a Markov model are the transition probability matrix, condition states, and duty cycle. This paper hypothesizes that the number of condition states (NCS) and length of duty cycle (LDC) significantly influence the prediction accuracy of a Markov model and that the nature of such influence varies across the different Markov methodologies. In previous studies on Markovian performance modeling, not only did this hypothesis go unanswered, but also there is a lack of comparison across the different Markov methodologies. In an attempt to shed light on this issue, this paper describes the development and comparison of performance models using these Markov methodologies and empirical data. The paper also investigates the sensitivity of Markovian model prediction accuracy to NCS and LDC. The results indicate that the semi-Markov techniques yield performance models that are generally statistically superior to those of the homogeneous and staged-homogeneous Markov (except in a few cases of NCS and LDC combinations) and that the prediction accuracy of Markovian models is very sensitive to NCS and LDC: an increase in NCS improves the prediction accuracy up to a certain NCS threshold, after which the accuracy diminishes, likely due to data overfitting. In addition, an increase in LDC generally enhances prediction accuracy when the NCS is small. The results can help guide highway agencies and researchers not only in selecting the appropriate Markovian methodology for a given NCS and LDC but also in selecting the appropriate NCS and LDC for a given Markovian methodology.

Effective pavement management systems are essential for accurately predicting pavement conditions and efficiently planning and scheduling maintenance, rehabilitation, and reconstruction activities. Significant efforts are dedicated to developing accurate pavement condition prediction models using machine learning (ML) at the state level. Conversely, insufficient investment, poor quality, and large variations in local roads data have resulted in less attention to modeling local pavement conditions. This study develops eight Bayesian-optimized single-estimator and ensemble ML classification models to predict local pavement roughness. Moreover, the classification granularity of pavement condition was investigated to assess its impact on the predictive power of various ML models. The results reveal that ML classification models with fewer classes exhibit higher accuracy and more stability in precision over recall values, in contrast to models with larger number of classes. The ensemble ML models surpass their single-estimator counterparts, with the category boosting algorithm demonstrating the highest performance, achieving testing accuracies of 0.77 and 0.65 for the three-level and five-level classifications, respectively. Hence, it is recommended to employ ensemble ML algorithms and a smaller number of classes to develop reliable, accurate, and stable predictive models for local roads with imbalanced condition data. This research helps transportation agencies improve their pavement condition prediction, thereby optimizing pavement management and resource allocation.

Radio Occultation (RO) is pivotal for profiling the neutral and ionized atmosphere, with the PlanetiQ mission, via its GNOMES satellites, striving to establish an advanced atmospheric observing system. However, an assessment of the spatiotemporal distributions of PlanetiQ observations and comparisons with reliable datasets are lacking. This study addresses this gap by examining the temporal and spatial distribution of RO observations from PlanetiQ during its initial 198 operational days in 2023, alongside comparisons with COSMIC and Numerical Weather Prediction (NWP) models. Data from GN02, GN03, and GN04 satellites, yielding 1099, 1313, and 1843 RO events per day, respectively, were analyzed. The satellite constellation’s observations demonstrate a generally well-distributed pattern, albeit minor deficiencies in equatorial and polar regions. Single-profile comparisons with COSMIC data reveal strong correlations for pressure, temperature, Water Vapor Pressure (WVP), and refractivity profiles, with temperature exhibiting larger variations (RMSE = 1.24 °C). Statistical analyses confirm statistically insignificant differences between the PlanetiQ and COSMIC profiles at the same spatio-temporal coordinates. Comparisons with NWP models show slight differences with GFS, with overall RMSE values of 0.23 mb (WVP), 0.6 mb (pressure), 1.3 (refractivity), and 1.5 °C (temperature). However, assessments against GFS/ECMWF models indicate overall compatibility, with insignificant differences between PlanetiQ profiles and model observations.