Identifying influencing factors on paved roads silt loading.

ABSTRACT

The factors that influence the increase or decrease of silt loadings on paved roadways have not been fully quantitatively investigated. They were identified in this study based on the quarterly silt loading sampling data collected from 20 sites by the Clark County Department of Air Quality and Environmental Management in Southern Nevada for the period from 2000 to 2003. The silt loading and associated data collected over these years at one sampling site may inherently possess site-specific characteristics that can be better incorporated by using panel data models. The factors that are identified as significant are the presence of curbs and gutters, shoulder type, pavement conditions, and the presence of construction activities in the vicinity of roadways. The presence of curbs and gutters, stabilized shoulders, and good pavement conditions would result in decreased silt loadings. Conversely, the presence of construction activities within the immediate vicinity of sampled areas would result in increases of silt loadings on the roadway surfaces. Based on the analysis of the results, it was recommended that constructing curbs, gutters and stabilized shoulders, preventing or reducing construction track-out from construction activity, and improving pavement conditions be the preferred control measures to reduce silt loading on paved roadways.

INTRODUCTION

In 1993, the U.S. Environmental Protection Agency (EPA) designated the Las Vegas Valley (Hydrographic Area 212) as a serious coarse particulate matter (PM; [PM.sub.10]) nonattainment area. PM includes dust, dirt, soot, smoke, and liquid droplets directly emitted into the air by sources such as factories, power plants, cars, construction activities, fires, and windblown dust. It is one type of silt that is any aggregate material with a particle size <75 [micro]m in diameter. There are two typical PMs: fine PM ([PM.sub.2.5]; particles with aerodynamic diameter of <2.5 [micro]m) and [PM.sub.10] (particles with an aerodynamic diameter of <10 [micro]m). These particles can be transported from one location to another by different mechanisms, such as wind, traffic activities, and so forth. For example, vehicles driving off roadway shoulders, especially on narrow roads, can carry loose soil and deposit it on the pavement when driven back on the roadway. Because of health concerns, EPA published health-based PM standards. (1) Areas that do not meet the required health-based standards are declared to be nonattainment areas and are required to implement control measures that can lead to attainment of the required health-based standards. The concentrations of [PM.sub.10] in the Las Vegas Valley have ranged from 155 to 254 [micro]g/[m.sup.3], which exceeds the required health-based 24-hr standard of 150 [micro]g/[m.sup.3].

To develop control measures for reducing [PM.sub.10], it is necessary to identify the factors that influence [PM.sub.10]. Intuitively, more [PM.sub.10] would be generated if more silt is present on a roadway. With data of silt loading available from the [PM.sub.10] monitoring program in the Clark County Department of Air Quality and Environmental Management, this study investigated the relationship between the increases or decreases of silt loading on a roadway and the influencing factors, such as traffic volume, roadway classification, geometric features of roadway (e.g., shoulder, curb, and gutter), and any construction activities. Efforts were made in this study to extract the data representing the influencing factors and the data of silt loading from reading the quarterly reports provided by Clark County in Southern Nevada. These reports describe the data collection efforts that were conducted on a quarterly basis. The number of sampling sites for collecting the silt loading data and the specific locations of these sampling sites vary between quarters. The quarterly reports include the descriptions of the sampling sites and the data collected from them. The data for the influencing factors were not in a format for statistical analysis. They were extracted through reviewing the reports. Because not every location of the sampling sites was chosen in each quarter (see Table 1), there are silt loading data available for more than one time during the 3-yr period at some sampling locations. The silt loading data and the associated data of possible influencing factors for these sampling locations can be seen as panel data, which makes it possible to calibrate a panel data model. In this study, from the results of the calibrated panel data model, the factors (or variables) that show statistical significance were identified. The relative extent of the factors/variables that influence silt loading was estimated. Based on the identification of the influencing factors, recommendations on control measures for reducing silt loading and [PM.sub.10] were developed.

The following sections of the paper are organized as described here. The first section provides a literature review, which is followed by an introduction to the panel data model. In the third section, the silt loading and associated data collected by the Clark County Department of Air Quality and Environmental Management are presented. The fourth section is devoted to the presentation of the development of the panel data model with the results of the estimation discussed. The last section presents the conclusions for this study and provides recommendations for future study needs.

LITERATURE REVIEW

According to the study by Muleski and Cowherd, (2) the most important factors for dust emissions are the mean speed of vehicles traveling the road, the average daily traffic, the number of lanes, the fraction of heavy vehicles (buses and trucks), and the presence/absence of curbs, storm sewers and parking lanes. However, the study recommended a formula for calculating [PM.sub.10] emission that includes only silt loading and vehicle weight.

The study by Dames and Moore (3) is one of the few studies that quantitatively identifies the factors that influence paved road dust emissions. The study determined that there is no significant difference in silt loading for local roads with gravel shoulders versus paved shoulders. In addition, for roadways with narrower lanes (<10 ft), the effect of paved versus unpaved shoulders was significant. It was surmised that, when a vehicle was passing another one moving in the opposite direction, it tended to move off the roadway shoulder and then back to the middle of the roadway carrying loose soil from the shoulders with them. The same study investigated the effect of paved travel lane width on silt loading. It was determined that wider travel lanes reduce the amount of silt loading on the paved roadway.

It is noted that the dataset used in Muleski and Cowherd (2) is the result of synthesizing data from different studies, which may cloud the accuracy of the results. The number of samples in Dames and Moore (3) is small. With the availability of dust monitoring data from the Clark County Department of Air Quality and Environmental Management for the years of 2000-2003, it is possible for the study presented in this paper to investigate the possible influencing factors, such as the number of lanes, shoulder type, presence/absence of curbs and gutters, pavement conditions, and presence/absence of construction activities.

METHODOLOGY

Panel Data Model

Panel data are a special form of longitudinal data in which observations are collected on individuals over a period of time. They are, therefore, a combination of time series and cross-sectional data. There are several advantages associated with using panel data over cross-sectional or time series analysis. For example, dynamics of changes can be studied by studying the repeated cross-section of observations. Because panel data relate to individuals, the approach can take care of heterogeneity by allowing for individual-specific variables. In addition, by combining time series and cross-section observations, the approach can give more informative data and more degrees of freedom by boosting the number of observations. (4)

[FIGURE 1 OMITTED]

Generally, a panel data model can be written as follows:

[y.sub.it] = [beta]'[x.sub.it] + [[epsilon].sub.it], t = 1,..., T (i), i = 1,..., N, (1)

E[[[epsilon].sub.it]|[x.sub.i1], [x.sub.2t],..., [x.sub.iT(i)]] = 0, (2)

Var[[[epsilon].sub.it]|[x.sub.i1], [x.sub.2t],..., [x.sub.iT(i)]] = [[sigma].sub.it.sup.2], (3)

where [y.sub.it] is the dependent variable for group i at time t, [beta] is the constant vector of parameters, [x.sub.it] vector of independent variables for group i at time t, [[epsilon].sub.it] is the error term for group i at time t. Note that different groups of observations could have different length of time series, which is the case in this study.

In the case of silt loading in this study, the measurements of silt loadings from one sampling site can be viewed as a group. Corresponding to each measurement of silt load, there is a vector of influencing factors or variables such as roadway function classification (FUNC), number of lanes (NUMLANES), presence of curbs and gutters (CUGU), shoulder types (SHOULT), pavement condition (PAVEC), and presence of construction activities close to the roadway (CONSTR). Then, the panel data model for the silt loading can be written as follows:

(Log_SiltLoad)[.sub.it] = [[beta].sub.0] + [[beta].sub.1](FUNC)[.sub.it] + [[beta].sub.2](NUMLANES)[.sub.it], + [[beta].sub.3](CUGU)[.sub.it] + [[beta].sub.4](SHOULT)[.sub.it] + [[beta].sub.5](PAVEC)[.sub.it], + [[beta].sub.6](CONSTR)[.sub.it] + [[epsilon].sub.it]. (4)