Guidelines for Informing Decisionmaking to Affect Pavement Performance Measures: Final Report

CHAPTER 2. Development of Performance Measure Drivers

Background

The purpose of developing the performance measure drivers is to understand the metric or metrics that are affecting the overall pavement condition. Understanding the performance measure drivers is necessary in the treatment selection process so that the treatments selected address the cause of the pavement condition, improving the individual metrics and, ultimately, the overall pavement condition.

To develop the performance measure drivers, there needs to be an understanding of the metrics for the various pavement types. The metrics according to pavement type are as follows:

• AC pavements.
  • Roughness, percent cracking, and rutting.
• JPCC pavements.
  • Roughness, percent cracking, and faulting.
• CRCPs.
  • Roughness and percent cracking.

The thresholds presented in table 1 are used to assign the metric condition, as illustrated in the following examples:

• AC pavements example. A given AC pavement has a roughness of 101 inches/mi, 4-percent cracking, and 0.15 inch of rutting. The metric conditions according to the thresholds in table 1 are fair, good, and good for roughness, percent cracking, and rutting, respectively. The overall condition is determined based on the Final Rule as detailed in chapter 1 of these guidelines. For the AC pavement, since two metrics are good and one metric is fair, the overall condition is fair.
  
  
• JPCC pavements example. A given JPCC pavement has a roughness of 182 inches/mi, 3-percent cracking, and faulting of 0.20 inch. The metric conditions according to the thresholds in table 1 are poor, good, and poor for roughness, percent cracking, and faulting, respectively. The JPCC pavement overall condition is poor, since two metrics are poor and one metric is good.
  
  
• CRCPs example. A given CRCP has a roughness of 120 inches/mi and 11 percent cracking. The metric conditions according to the thresholds in table 1 are fair and poor for roughness and percent cracking, respectively. The CRCP overall condition is fair, since both metrics are neither good nor poor.

The goal of the guidelines is to inform decisionmaking that affects the pavement performance measures, which are the percent of pavements on the IHS and NHS in good and poor condition. While the discussion in this chapter and throughout the guidelines focuses on the impact of drivers on overall condition, it is that overall condition that is ultimately going to drive the performance measures of the network. For instance, by improving the overall condition to good, the performance measure (percent good) is improved. The overall condition feeds directly into the performance measures.

Another factor that can affect pavement condition and therefore needs to be recognized and considered is measurement error. For example, it is expected that rut depth will increase over time, but the rate of change (i.e., increase) in rutting is likely to be less than the measurement error. The Guide for the Local Calibration of the Mechanistic-Empirical Pavement Design Guide reports a reasonable standard error of the estimate for rutting to be 0.10 inch, while the average rate of change for rutting can be less than 0.01 inch per year.⁽⁵⁾ Therefore, it is possible for the data to show a decreasing trend in the rutting as a result of measurement error, as discussed next.

Figure 2 illustrates two possible relationships for growth rates depending on the measurement errors—one with a positive trend and one with a negative trend—as well as the “true” rutting values, which show a positive growth rate. As shown, for every measurement, there is a “true” value that represents the actual value. Each measurement also has a range of likely measurable values, which can be attributed to errors such as measurement errors, errors in linear referencing from year to year, and errors introduced by averaging many measurements into one representative value for a segment. Due to this plausible interval of measurements, it is conceivable to report a variety of growth rates from the measured data depending on where in the interval the measured value falls in comparison to the true value. This example helps illustrate the possible impact of measurement error on pavement condition trends (e.g., improving pavement condition with time in the absence of M&R). State DOTs need to recognize the potential impact of measurement error on the data and the performance measures and should continuously strive to improve the accuracy of the data collection through improved technology with increased data collection accuracy and precision, training of personnel, and implementation of data quality management plans.

Source: FHWA.
Figure 2. Graph. Rutting growth.

Development of Drivers

The first step in developing performance measure drivers is to understand the metric condition combinations that compose the overall condition. It is important to differentiate between metric condition combinations because different metric condition combinations require different treatments to have an effect on the overall condition. The combination of metric conditions to consider for AC and JPCC pavements include the following:

• Good-good-fair (G-G-F).
• Good-good-poor (G-G-P).
• Good-fair-fair (G-F-F).
• Good-fair-poor (G-F-P).
• Fair-fair-poor (F-F-P).
• Good-poor-poor (G-P-P).
• Fair-poor-poor (F-P-P).

Three other possible metric condition combinations include those when all metrics are either good, fair, or poor. However, these are not included in this analysis, as there are no drivers of these conditions, since all three metrics contribute to the all good, all fair, and all poor conditions.

For CRCPs, the metric condition combinations considered are as follows:

• Good-fair (G-F).
• Good-poor (G-P).
• Fair-poor (F-P).

Emphasis should be given to the drivers of the borderline conditions, which are those conditions where one change (e.g., good to fair, fair to good, fair to poor, poor to fair) in metric condition would result in a change in overall condition. The borderline conditions for AC and JPCC pavements include the following:

• Good-good-fair (G-G-F).
• Good-fair-poor (G-F-P).
• Fair-fair-poor (F-F-P).
• Good-poor-poor (G-P-P).
• Fair-poor-poor (F-P-P).

For CRCP, the borderline conditions include the following:

• Good-fair (G-F).
• Good-poor (G-P).
• Fair-poor (F-P).

To develop the performance measure drivers for each metric condition combination, the following steps are required:

1. Separate data based on pavement type (e.g., AC, JPCC, and CRCP).
   
   
2. Assign condition (good, fair, poor) for metrics for each pavement segment according to table 1.
   
   
3. Assign overall condition according to the pavement segment metric condition combinations as explained in chapter 1 of these guidelines.
   
   
4. Assign each pavement segment a metric condition combination (G-F-P, G-G-F, etc.).
   
   
5. Identify the metric or metrics that are driving the overall condition for each pavement segment. The driver is defined as the metric or metrics that are most responsible for the overall condition. The driver(s) are identified as follows:
   
   
   1. G-G-F—metric in fair condition.
   2. G-G-P—metric in poor condition.
   3. G-F-P—metrics in fair and poor condition.
   4. G-F-F—metrics in fair condition.
   5. F-F-P—metrics in fair condition.
   6. G-P-P/F-P-P—metrics in poor condition.
      
      
6. Calculate percentage that each metric is identified in step 5 for the metric grouping to determine the performance measure driver.

Although the emphasis of this chapter is on identifying the metric or metrics that are driving the overall condition, it should be noted that there are other possibilities that affect the condition besides the metrics and drivers identified in this chapter. The drivers identified are those metrics that are most responsible for the overall condition. However, there are other metrics that can also affect the condition and various scenarios that could change the overall condition. For instance, for the F-F-P grouping, the two metrics that are in fair condition are identified as the drivers. That is not to say that the metric that is in poor condition should be ignored. By improving the condition of the metric in poor condition, the overall condition would remain fair assuming the other metrics remained fair as well, but by improving the poor metric to fair, the overall condition may be less likely to become poor, as now two metrics would need to deteriorate to poor condition. Also, if a metric is identified as the driver, this does not mean that other metrics do not affect the condition but that it is not as likely. For example, if the driver of the G-G-F grouping is rutting, rutting is the metric that is most often the metric in fair condition. That is not to say that roughness and cracking are never in fair condition but that it is less likely than rutting being the metric in fair condition.

Examples

The following examples—one for each pavement type—were developed using LTPP data to illustrate the steps to develop the performance measure drivers. The metric conditions and overall conditions were assigned according to steps 2 and 3. The metric condition combinations were then assigned according to the metric conditions assigned in step 2.

• AC pavements example. This example illustrates the G-G-P and F-F-P metric condition combinations. The G-G-P and F-F-P groupings represent 12 percent and 8 percent of the AC pavement sections in fair condition, respectively. For the G-G-P grouping, the driver is the metric in poor condition. Table 3 presents the number of pavement sections in the G-G-P grouping where the metrics are in poor condition. The percentages in table 4 represent the proportion of all the pavement sections in the G-G-P grouping where the metric is in poor condition. The table shows that cracking, with 67 percent, is the metric that is most likely to be in poor condition and therefore is the driver of the overall condition for the G-G-P grouping. For the F-F-P grouping, the drivers are the metrics in fair condition. Table 5 presents the number of pavement sections in the F-F-P grouping where both metrics are in fair condition. The percentages in table 6 represent the proportion of all the pavement sections in the F-F-P grouping where the metrics are in fair condition. The table shows that rutting and roughness, with 90 percent, are the metrics that are most likely to be in fair condition and therefore the drivers of the overall condition for the F-F-P grouping.

• JPCC pavements example. The G-G-F and G-P-P/F-P-P groupings are illustrated in this example. The driver of the G-G-F metric grouping will be the metric that is in fair condition. For an LTPP dataset of JPCC pavements, table 7 presents the number of sections where the metric is in fair condition and the metric grouping is G-G-F. The percentages in table 8 represent the proportion of all pavement sections in the G-G-F grouping where the metrics are in fair condition. The results presented in table 8 show that the driver of the G-G-F metric grouping is roughness, since roughness is the metric in fair condition 66 percent of the time.

The drivers of the G-P-P/F-P-P grouping are the metrics that are in poor condition. Table 9 presents the number of segments where the two metrics listed are both in poor condition and the metric grouping is G-P-P/F-P-P. The percentages in table 10 represent the proportion of all pavement sections in the G-P-P/F-P-P grouping where the metrics are in poor condition. The drivers of the G-P-P/F-P-P grouping shown in table 10 are roughness and cracking, since roughness and cracking are both in poor condition 62 percent of the time.

• CRCPs example. The LTPP dataset only had sufficient data to analyze the drivers of the G-F metric condition combination. The performance measure driver of this grouping is the metric in fair condition. Of the 178 pavement sections in the G-F grouping, 168 of them, or 90 percent, had roughness as the metric in fair condition, making roughness the driver of the G-F metric condition combination.

The process described in this chapter for determining the performance measure drivers was completed for all the metric condition combinations, and they are presented in the companion report.⁽⁴⁾ The performance measure drivers identified in this companion report are summarized in table 11.