Transportation infrastructure plays a crucial role within major cities around the world. Within the GCC countries, there has been a major development in transportation infrastructure over the past few decades. Bridges are considered one of the most important infrastructure projects that require a significant budget and a high degree of safety. Moreover, the number of bridges within the GCC region is increasing. This increase and the age of existing bridges highlight the need for plans to maintain these bridges in excellent operational conditions, which requires huge funds and efforts. Bridge deterioration is one of the major threats that may cause the sudden failure of some bridges. A bridge management system (BMS) is one of the effective tools that has been proposed to improve the efficiency in managing the inspection, maintenance, condition prediction, and budget allocation for bridges at both project and network levels. A BMS usually contains four basic modules, database, maintenance cost, deterioration, and decision making and optimization. The deterioration module is a very important module because it predicts the future condition of the bridge to assess proper planning for maintenance strategies and allocation of the budgets. Concrete bridges are widely constructed and preferred in the GCC countries due to their durability and the availability of raw material and workmanship. However, concrete bridges are exposed to many external deterioration mechanisms originating from environmental influences that decrease their serviceability and may compromise their integrity. Identifying the factors that affect bridge deterioration will help in developing reliable deterioration models.

Deterioration models are one of the main modules in any BMS as they facilitate the prediction of the maintenance requirements to preserve the bridge’s integrity and the safety of the public. Identifying the key factors that should be considered, when developing deterioration models within the GCC, is an essential step towards the development of realistic models. Identifying these factors can be of great benefit for the authorities managing bridges in the GCC and could be utilized to develop deterioration models that reflect the local conditions.

A bridge management system (BMS) is considered one of the best tools used to monitor bridge activities and the health condition of the bridges. It is a process to help in making decisions regarding the maintenance, rehabilitation, and replacement of bridges in the most efficient approach ^[1][2][1,2].

The need for BMS was first realized in the United States after the “Silver Bridge” incident. On 15 December 1967, the Silver Bridge collapsed under the weight of rush-hour traffic. Accordingly, recommendations were made to the US Secretary of Transportation to begin a research program aimed at developing new inspection procedures. From that perspective, the Federal Highway Administration (FHWA) introduced the National Bridge Inventory (NBI), where all the states are required to report all inspection data, which eventually became the inventory of all BMS in the USA [3]. The BMS helps in maintaining the health and well-being of the bridges ^[4][5][6][4,5,6].

The four major modules of BMSs were defined by the American Association of State Highway and Transportation Officials (AASHTO) guidelines in 1993 and were re-defined by Czepiel [7]. These modules are the database, the maintenance cost, the deterioration, and the decision making and optimization.

The FHWA through the National Cooperative Highway Research Program (NCHRP) initiated a research project for the development of a BMS and came up with the first BMS software called PONTIS 1.0. The software was first developed in 1991 and then licensed by AASHTO in 1994 ^[5][8][5,8]. Approximately 50% of the U.S. transportation agencies have adopted PONTIS for their bridge management activities [9]. However, the accuracy of the predicted conditions of bridges might not be reliable enough if the input data are based on visual inspections only [10].

Another BMS software that is commonly used in the USA is BRIDGIT 1.0. It was developed by NCHRP in 1998 and has similar functions to PONTIS. The major difference is that BRIDGIT has an optimization module, called OPBRIDGE, which uses a “bottom-up” analysis method for cost planning [10].

There are currently numerous BMS packages in service around the world. Some of these systems were genuinely adopted from the PONTIS or BRIDGIT systems but were modified to fit the local conditions. A brief discussion on the common BMSs around the is summarized in Table 1. The information in Table 1 provides some observations on the developed BMSs in different parts of the world. In addition, it discusses the historical development of each system. Development of BMSs is discussed in detail in several previous studies ^{[1][5][6][7][9]}[1,5,6,7,9].

Region	Comments
North America	Although PONTIS and BRIDGIT were mainly used in the U.S., some other states and provinces in North America such as New York, Indiana, Pennsylvania, North Carolina, Alabama, Florida, Denver, and Ontario have developed their own bridge management systems [3][5][7][11][12][13][14][15][16][17][18][19]][3,5[,720],11[21],12[,1322,14,15,16,17,18,19,20,21,22].
Europe	Denmark developed a BMS, which includes six modules, called DANBRO (DANish Bridges and Roads) in 1988 [13][23][13,23]. Although DANBRO does not include a condition deterioration module, it has been implemented in Saudi Arabia, Mexico, Colombia, Honduras, Croatia, and Malaysia [24]. A bridge management software named SIHA was developed in Finland. At the beginning, the system included inventory data only [25]. The latest version of the system included a deterioration module that optimizes the maintenance and repair costs using a probabilistic Markov Chain model [3][26][3,26]. Another BMS named Highway Structural Management Information System (HiSMIS) was developed in the UK [27]. Belgium, Norway, and Sweden operated a functionally complete BMS that included inventory, inspection, and maintenance modules. However, only Belgium’s BMS included a deterioration model [28]. The BMS applied in Belgium and Sweden lacks a life-cycle cost analysis module to plan for optimal maintenance planning [13][29][13,29]. Finally, France, Germany, Hungary, and Italy have developed basic BMSs to manage the bridge activities. Their BMSs basically involve inspection and condition ratings [30][31][30,31]. A maintenance decision support system is implemented in Germany and Italy [28][32][28,32].
Africa and Asia	The South African National Roads Agency Limited (SANRAL) developed a BMS named STRUMAN by the Council for Scientific and Industrial Research (CSIR) [33]. The first BMS in Japan was developed in 1995 and it was mainly for bridge condition ratings and rehabilitation strategies [3]. Miyamoto et al. [34] proposed a comprehensive bridge management system for Japan called J-BMS. The Indonesian Directorate General of Highways developed a bridge management system that contains modules to store inspection data, rank the bridges, prepare a report with annual and five-year programs of bridge work, and optimize the required repair works [13].
Australia and New Zealand	A report was initiated with the proposed BMS by Steele et al. [35] that included four modules: activities, engineering inputs, management inputs, and outputs. The engineering inputs module provides a set of feasible actions that can be taken [12]. The output module provides data on the bridge condition prediction, options for maintenance, and estimated costs [13].

2.2. Deterioration Models

Concrete bridges tend to deteriorate with time. Thompson et al. [36] emphasize that an effective BMS should be equipped with a reliable deterioration model. Deterioration models are models designed to predict the future condition of the bridge to assess proper planning for maintenance strategies and allocation of the budgets. Many deterioration models have been developed to help in predicting the condition of the bridge elements. The models are developed through different techniques that can be classified into deterministic, stochastic, and artificial intelligence. Deterministic deterioration models are obtained through statistical and regression analysis based on the historical data on the structural deterioration [37]. These models do not provide enough accuracy for prediction of the bridge performance and may underestimate or overestimate the bridge condition at a specific time [38]. Stochastic models are probabilistic models that consider the uncertainty and randomness within deterioration conditions. A stochastic approach can also be used to incorporate environmental influences and material characteristics into the deterioration model [39]. Models developed using this approach can be state-based models or time-based models [40]. Artificial intelligence (AI) models are computational models that use AI techniques like Artificial Neural Networks (ANNs) [41]. The ANN consists of networks of an input layer, hidden layers, and an output layer, which are parallelly interconnected [42].

2.3. Factors Included in Deterioration Models

2.3.1. Factors Considered in North America

Although some of the states used the same methods and types in developing the deterioration models, different factors were considered in each model, based on the local conditions. Age, average daily traffic (ADT), and average daily truck traffic (ADTT) were found to be the most common factors adopted in the deterioration models developed in North America ^{[8][38][42][43][44][45][46][47][48][49][50][51][52][53][54][55][56][57][58][59][60][61][62]}[8,38,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62]. Madanat et al. [44], Moomen et al. [14], and Saeed et al. [53] have developed different deterioration models for Indiana. The common factors found in the three studies were the following: age, ADT, state or non-state type of bridges, freeze and thaw cycles, number of spans, bridge length, skewness, and type of services under the bridge. Deck protection was adopted only by Moomen et al. [14]. Madanat et al. [44] were the only ones to consider the humidity, wearing surface type, bridge width, and structure type in their model. In Ohio, Ramani [56] considered geographical location and bridge material type, while AlThaqafi and Chou [57] included skewness, ADT, design type, number of spans, bridge length, deck width, and design load in their model. The age of the bridge was the only common factor between the two models. Winn [42] and Chyad [58] used ANN, deterministic, and stochastic approaches to develop a deterioration model in Michigan. Common factors used in these studies were age, ADT, and structure type. Weissmann et al. [60] developed a deterioration model in Texas utilizing inventory data, which considered climate factors such as rainfall. Other studies ^{[10][59][60][61]}[10,59,60,61] have also considered the environmental effects on bridge deterioration when developing different deterioration models. Shengzhi et al. [62] categorized the US into five zones: Hot-Humid, Mixed-Humid, Hot-Dry/Mixed-Dry, Cold/Very Cold, and Marine. There are several states that considered the effect of the environmental parameters in their deterioration models ^{[14][43][44][48][53][54]}[14,43,44,48,53,54]. Foster [63] conducted a survey of the state departments of transportation (DOTs) to investigate their experience with bridge deterioration models. Twenty-nine states responded, with seventeen states that have a deterioration model in their BMS and seven states that do not use bridge deterioration models in their BMS. The most common factors in the deterioration models reported were age, superstructure’s material, condition rating, average daily traffic (ADT), and deck wearing surface. In Canada, Marcous [45] considered highway class, region, age, ADT, and span length in the deterioration model. Morcous et al. [55] included more factors in their model such as ADTT, skew angle, structural system, wearing surface, and girder material and spacings. Another study in Ontario by Martinez et al. [59] was conducted to predict the condition rating of the province’s bridges using different types of models and a set of variables for the models. The set of variables includes rehabilitation time, number of spans, bridge length and width, region, age, and bridge material.

2.3.2. Factors Considered in Other Countries

To forecast the bridge condition rating using condition rating data, Khairullah and Roszilah [64] developed an ANN deterioration model in Malaysia. Another study by Al Hussein [61] utilized the inventory data of London bridges, where the defect type, exposure condition, and bridge component type were used as input parameters for the ANN deterioration model development. A study in Japan was conducted by Miao [65], where an ANN deterioration model was developed using age, ADT, ADTT, and bridge geometry as factors included in the model. The study also involved climate-related factors such as snow and rain precipitation, temperature change, chloride, and carbon dioxide content. The same environmental factors were used in a study in China but without considering the rain precipitation, chloride, and carbon dioxide content [66]. Environmental factors were also considered in deterioration models in Australia. In addition to age, ADT, ADTT, and structure type were adopted by Callow [11], and the chloride and sulphate content were included in the ANN deterioration model developed by Hasan [67]. In Europe, a deterioration model was developed in Finland based on age, freeze–thaw cycles, carbonation depth, and reinforcement corrosion [25]. In Serbia, the deterioration model included structural components and inspection gap [68]. In the Netherlands, Kallen and Noortwijk [69] considered age and condition data of the bridges to develop a Markov deterioration model. Santos et al. [70] in Brazil categorized the bridge data according to geographic location, material type, and superstructure type to develop Markovian and ANN deterioration models. For each bridge category, age, deck width, and bridge length were considered as parameters in both models. Another study in Brazil was conducted by Furtado and Ribeiro [71] in which they used age, bridge length, number of spans, bridge material, traffic volume, and different classes of environmental aggressiveness as variables for the semi-Markovian model. A stochastic model that considered age and environmental conditions was also developed in Japan [72].

2.3.3. Discussion of the Factors

Figure 1 summarizes the thirty-three factors found, and these could be classified into six groups; environmental factors, structural-related factors, dimensional factors, factors related to geographic location, traffic, and other factors. In addition, the structural-, traffic-, and dimensional-related factors are the most included in the development of deterioration models. Figure 2 shows several structural-related factors which were included in different deterioration models. The age of the bridge and the superstructure type are the main factors that were included in the development of DMs. Furthermore, snow, rain, temperature, humidity, freeze index, number of freeze–thaw cycles, number of cold days, carbon dioxide, chloride, sulphate, and salt usage are some of the environmental conditions that are included in DMs, as shown in Figure 3.