U.S. Department of Transportation
Federal Highway Administration
1200 New Jersey Avenue, SE
Washington, DC 20590
202-366-4000


Skip to content
Facebook iconYouTube iconTwitter iconFlickr iconLinkedInInstagram

Federal Highway Administration Research and Technology
Coordinating, Developing, and Delivering Highway Transportation Innovations

Report
This report is an archived publication and may contain dated technical, contact, and link information
Publication Number: FHWA-HRT-05-054
Date: September 2005

Quantification of Smoothness Index Differences Related To Long-Term Pavement Performance Equipment Type

Chapter:6 Differences Between Dipstick and Profiler Iri

INTRODUCTION

In this chapter, a description of the factors that cause IRI obtained from the Dipstick data to differ from IRI obtained from the profiler data is presented. The range of the difference in IRI that can occur between Dipstick and profiler IRI, obtained from an analysis of data from past LTPP comparison studies, also is presented in this chapter.

FACTORS CONTRIBUTING TO DIFFERENCES BETWEEN DIPSTICK AND PROFILER IRI

There are a variety of factors that can cause IRI obtained from the Dipstick data to differ from IRI obtained at the same section by an inertial profiler. These factors are:

  • Sampling qualities of Dipstick.
  • Variability in the path followed by a profiler.
  • Features recorded by the profiler that are missed or underestimated by Dipstick.
  • Averaging effects of profiler data.
  • Dipstick data errors.
  • IRI computation procedures for Dipstick data.

A description of how each of these factors can contribute to differences in the IRI values between Dipstick and the profilers is presented in the following sections.

Sampling Qualities of Dipstick

The footpad spacing (i.e., the distance between the center of the footpads) for Dipstick and the footprint of Dipstick (the area covered by a footpad) affect the measurements obtained by Dipstick.

The diameter of the footpads in the Dipsticks that are used in the LTPP program is approximately 32 mm (1.25 inches). The footpad spacing in Dipstick can be adjusted, with the maximum spacing usually being 304.8 mm (12 inches) or 300 mm (11.8 inches) for devices with a base plate set to metric units or U.S. customary units, respectively. The footpad spacing of the Dipsticks that are used in the LTPP program is 304.8 mm (12 inches). In the following discussion, the footpad spacing of Dipstick is assumed to be set to this value.

Gain Characteristics of Dipstick

Dipstick will have a varying response to sinusoids of different wavelengths. Consider the response of Dipstick to a sinusoidal road feature located on a road with zero slope that has a wavelength equal to the footpad spacing of Dipstick as shown in figure 52. No matter where Dipstick is placed on this sinusoid, Dipstick will give a reading of zero, because the two supporting feet will have the same elevation. Dipstick measurements taken along a roadway that has such a feature will not capture the feature, and will simply give a straight line for the elevation profile. In the case of a road with a slope that has such a feature, the Dipstick measurements will only show the slope of the road, and will not show the sinusoidal feature present on the road.

This figure shows a sketch of Dipstick placed on a sinusoid that has a wavelength equal to the Dipstick footpad spacing. The Dipstick footpads rest on the highest point of the sinusoid.

Figure 52. Dipstick response to a sinusoid with a wavelength equal to the footpad spacing of Dipstick.

The response of Dipstick to different wavelengths can be studied by using a gain plot. The following procedure is used to generate the gain plot:

  1. Select a sinusoid that has a specific wavelength and specific amplitude. Assume that the amplitude of the sinusoid is A.
  2. Simulate placing Dipstick at the start of the sinusoid and compute the reading that would be obtained by Dipstick (i.e., the difference in elevation between the front and back footpads).
  3. Again place Dipstick on the sinusoid so that the current position of Dipstick is slightly ahead (e.g., 25 mm (1 inch)) of the previous location. Compute the reading that would be obtained by Dipstick. Repeat this procedure until Dipstick has been placed for all positions in the sinusoid, covering an entire wavelength.
  4. After completing this exercise, use the readings that Dipstick obtained for each Dipstick position to generate the profile (sinusoid) that was recorded by Dipstick.
  5. Divide the maximum amplitude of the generated profile by the amplitude of the sinusoid that was used to generate this profile (i.e., A) to obtain the gain for that wavelength.
  6. Repeat steps 1 through 5 for different wavelengths.

Note: This is not the procedure that is used when Dipstick is used to obtain measurements. The exercise was performed to obtain the gain characteristics of Dipstick.

The gain plot that was obtained by the previously described procedure is shown in figure 53. As discussed previously, the gain is zero for a wavelength of 0.305 m (1 ft). For wavelengths above 0.305 m (1 ft), the gain gradually increases as the wavelength increases. For a wavelength of 0.61 m (2 ft), which is approximately twice the footpad spacing of Dipstick, the gain is 0.63. For Dipstick to measure an amplitude that is more than 95 percent of the correct amplitude of a sinusoid, the wavelength of the sinusoid should be more than 2 m (7 ft). As shown in figure 53, the footpad spacing of Dipstick will also cause it to underestimate wavelengths that are shorter than the footpad spacing.

This figure shows a gain plot for Dipstick, with the X-axis showing the wavelength and the Y-axis showing the gain. The gain is 0 for a wavelength of 0.305 meter (1 foot). For wavelengths greater than 0.305 meter (1 foot), the gain gradually increases as the wavelength increases. For a wavelength of 0.61 meter (2 feet) the gain is 0.63. The gain increases to about 0.95 for a wavelength of 2 meters (7 feet). After that, there is very little change in gain with an increase in wavelength

1 m = 3.28 ft

Figure 53. Gain plot of Dipstick.

Sampling Interval

Sampling data at discrete intervals limits the range of wavelengths that may be recognized. A common rule that is used to characterize the sampling effect is the Nyquist Sampling Theorem. This theorem states that the highest frequency that can be accurately represented from discretely sampled data is less than one-half of the sampling rate. For this statement of the theorem, the sampling rate is expressed in samples per second. When it is restated for road profiles, the theorem indicates that the shortest wavelength that can be represented by discretely sampled data is longer than twice the sampling interval. Applying this rule to a Dipstick with a footpad spacing of 0.305 m (1 ft) indicates that wavelengths shorter than 0.61 m (2 ft) will not be represented in the measurements obtained with Dipstick.

The sampling interval of Dipstick also may cause features with wavelengths shorter than twice the sampling interval to contaminate the measurement of longer wavelength features through a process called aliasing. Aliasing occurs when short-wavelength features are inadvertently interpreted to be longer wavelength features. For example, consider a sinusoid with a wavelength that is 11 percent longer than the sampling interval of Dipstick, as shown in figure 54. The first sample detects the highest point in the sinusoid; however, each successive point misses the peak by a progressively greater distance. The consequence is a gradual transition from the peak level to the valley over a long distance. Thus, a Dipstick having a sampling interval of 0.305 m (1 ft) that is measuring a sinusoid with a wavelength of 0.339 m (1.1 ft) would obtain a sinusoid of roughly equal amplitude, but having a wavelength of 3.05 m (10 ft).

This figure illustrates how aliasing can occur. The figure shows a sinusoid that has a wavelength that is 11 percent longer than the sample interval of Dipstick. Measurements are obtained on this sinusoid using a Dipstick. When Dipstick obtains the first measurement, the back footpad of Dipstick is resting on the peak of the sinusoid, while the front footpad is resting at a location that is slightly before the next peak of the sinusoid. As the sampling progresses, the readings miss the peak by a progressively larger distance. Hence, the readings taken by the Dipstick define a sinusoid that has a much larger wavelength than the measured sinusoid.

1 m = 3.28 ft
1 m/km = 5.28 ft/mi

Figure 54. An example of aliasing.

In cases where the wavelength of the sinusoid is shorter than the sampling interval of Dipstick, a much longer wavelength than that actually present also will appear in the measurement. For example, a sinusoid with a wavelength that is 10 percent shorter than the sampling interval would also be misinterpreted as a feature with a wavelength of about 3.05 m (10 ft). On the other hand, features with wavelengths that are just shorter than double the sampling interval will be aliased into features with wavelengths that are just longer than twice the sampling interval.

The way to avoid aliasing errors is to sample much more often than the shortest wavelength of interest, then apply a low-pass filter to remove the contents within the signal that are not of interest. However, such a procedure cannot be used with the Dipstick. The overall consequence of aliasing by Dipstick is that contents within the profile in a wavelength range shorter than 0.61 m (2 ft) are “folded” into the longer wavelength range. This causes an upward bias in the IRI value by artificially increasing the contents within the profile that fall within the wavelength range that affects IRI. The precise level of upward bias depends on the properties of the road surface, and this effect is much greater on pavement with a high level of megatexture or coarse macrotexture.

The probable effect of aliasing on IRI for a sampling interval of 0.3 m (1 ft) was estimated in a recent study of profile sampling procedures.(1) This study showed that the upward bias in IRI because of aliasing is probably on the order of 7 to 9 percent. The probable error level in this study was estimated by decimating profiles collected with the FHWA PRORUT profiler at two test sections. It should be noted that the treatment of very short road features by the feet of Dipstick might be different than the procedure used by the sensor footprint of the profiler. The anti-aliasing that had been used for the profile that was used in the study may impact the conclusions of that study. It should also be noted that data from only two sites were used in this analysis. Therefore, the 7 to 9 percent error level is only a rough estimate.

Dipstick Footprint

Dipstick contacts the pavement at two locations (spaced 304.8 mm (12 inches) apart) with rigid circular feet that are approximately 32 mm (1.25 inches) in diameter. The feet are attached to the base of Dipstick with a ball joint so that the footpads are most likely to rest upon the three highest points within the footprint of the footpad. An important property of this type of footprint is the ability to bridge over narrow cracks and rest on small pieces of protruding aggregate. If the macrotexture is not very deep, the footprint may serve as a mechanical filter that removes the potential for aliasing error caused by wavelengths on the order of 32 mm (1.25 inches) and shorter.

Variations in the Path Followed by the Profiler

During profiler comparison studies, Dipstick measurements at test sections are performed along the two wheelpaths. In LTPP comparison experiments, the wheelpaths are laid out so that each wheelpath is at a distance of 0.826 m (2.7 ft) from the center of the lane.(6) Where wheelpaths are easily identified, the midway point between the two wheelpaths is defined to be the center of the lane. If the wheelpaths cannot be clearly identified, but the two lane edges are well defined, the center of the travel lane is considered to be midway between the lane edges. Chalk lines are laid out along the wheelpaths when performing Dipstick measurements.

The wheelpaths were marked with paint dots at all sections during the 1991 and 2003 LTPP profiler comparisons, while two sections used for the 2000 LTPP profiler comparison had the wheelpaths marked. The paint dots provided a guide so that the driver could align the profiler along the wheelpaths when profiling the test sections. During other LTPP profiler comparisons, the drivers had to judge the location of the wheelpath when profiling the test sections.

If the profiler driver does not align the profiler along the wheelpaths, the longitudinal path that is followed by the sensors in the profiler will be different from the path where Dipstick measurements were obtained. Also, if the profiler driver does not consistently follow the wheelpaths within the section, but has some lateral wander, this will result in differences in the paths measured by the profiler and Dipstick. Both of these conditions can cause IRI obtained from profiler measurements to differ from that obtained from Dipstick measurements. A discussion of lateral wander and the effects on IRI were presented in chapter 5. As shown in table 3, at some sections, lateral wander can have a significant effect on IRI.

Figure 55 shows the 10-m (33-ft) base-length roughness profiles for the left wheelpath obtained for nine repeat runs performed by the Western profiler at test section 1, which is a smooth AC section, during the 2003 LTPP profiler comparison in Minnesota. The roughness profiles overlay very well, with the IRI for the entire section for the nine runs ranging from 1.29 to 1.33 m/km (82 to 84 inches/mi). The roughness profiles imply that either the profiler driver followed the same path for all nine runs or that minor variations in the wheelpath during profiling did not affect IRI.

This figure shows the 10-meter (33-foot) base length roughness profiles for the left wheelpath obtained for nine repeat runs performed by the Western profiler at test section 1 during the 2003 LTPP profiler comparison. The X-axis of the plot shows distance, while the Y-axis shows the IRI. The roughness profiles of the nine runs overlay extremely well with each other

1 m = 3.28 ft
1 m/km = 5.28 ft/mi

Figure 55. Roughness profiles for nine runs that show good agreement.

Thus, if the driver was tracking the wheelpath correctly, close agreement between profiler IRI and Dipstick IRI is expected at this site. However, the Dipstick IRI for the left wheelpath was 1.17 m/km (74 inches/mi), which is less than the IRI obtained from the profiler data. A comparison of the roughness profiles obtained by the profiler and Dipstick showed that the roughness profiles were different in a few localized areas. An evaluation of the profiles indicated that the profiler data had a deep narrow feature that was not captured by the Dipstick measurements, and omission of this feature in the Dipstick data was the primary reason why the IRI from the Dipstick data was lower than the IRI from the profiler data for this section.

Figure 56 shows the 10-m (33-ft) base-length roughness profiles along the right wheelpath for two runs of the Western profiler at test section 1, which is a smooth AC section, during the 2003 LTPP profiler comparison that was held in Minnesota. The two roughness profiles show good agreement, except at three high roughness locations: (1) between 40 and 50 m (131 and 164 ft), (2) between 60 and 75 m (197 and 245 ft), and (3) between 85 and 100 m (279 and 328 ft). IRI for the two runs shown in this figure are 1.62 m/km (103 inches/mi) for run 4 and 1.79 m/km (113 inches/mi) for run 2. The difference in the IRI between the two runs occurs because of differences in the pavement features that are measured during the two runs within the three rough areas that were described previously. Therefore, variations in the path profiled at this site can have a significant influence on the IRI obtained for the right wheelpath. At sites having such characteristics, variations in the path profiled by the profiler compared to the path where Dipstick data were collected can cause significant differences between profiler IRI and Dipstick IRI.

Figure 56 shows the 10-m (33-ft) base-length roughness profiles along the right wheelpath for two runs of the Western profiler at test section 1, which is a smooth AC section, during the 2003 LTPP profiler comparison that was held in Minnesota. The two roughness profiles show good agreement, except at three high roughness locations: (1) between 40 and 50 m (131 and 164 ft), (2) between 60 and 75 m (197 and 245 ft), and (3) between 85 and 100 m (279 and 328 ft). IRI for the two runs shown in this figure are 1.62 m/km (103 inches/mi) for run 4 and 1.79 m/km (113 inches/mi) for run 2. The difference in the IRI between the two runs occurs because of differences in the pavement features that are measured during the two runs within the three rough areas that were described previously. Therefore, variations in the path profiled at this site can have a significant influence on the IRI obtained for the right wheelpath. At sites having such characteristics, variations in the path profiled by the profiler compared to the path where Dipstick data were collected can cause significant differences between profiler IRI and Dipstick IRI.

Figure 57 shows 10-m (33-ft) base-length roughness profiles for a single run by the North Atlantic, North Central, and Western profilers along the left wheelpath of section 4, which is a PCC section that was used for the 2003 LTPP profiler comparison.

This figure shows the 10-meter (33-foot) base length roughness profiles along the right wheelpath for two runs of the Western profiler at test section 1 during the 2003 LTPP profiler comparison. The X-axis of the plot shows distance, while the Y-axis shows the IRI. The roughness profiles for run 2 and 4 are shown in the figure. The two roughness profiles show good agreement with each other, except at three high roughness locations located approximately between 40 and 50 meters (131 and 164 feet), 60 and 75 meters (197 and 245 feet), and 85 and 100 meters (279 and 328 feet). At these three locations, the roughness profile of run 2 shows higher roughness than that for run 4.

1 m = 3.28 ft
1 m/km = 5.28 ft/mi

Figure 56. Roughness profiles for two profile runs that show variations.

This figure shows 10-meter (33-foot) base length roughness profiles for a single run by the North Atlantic, North Central, and Western profilers along the left wheelpath of section 4 at the 2003 LTPP profiler comparison. The X-axis of the plot shows distance, while the Y-axis shows the IRI. The roughness profiles from the three profilers agree well with each other throughout the section, except between 25 and 40 meters (82 and 131 feet). Between these limits, major differences in roughness profiles are noted between the three devices; the North Central profiler has the highest roughness, while the Western profiler has the lowest roughness

1 m = 3.28 ft
1 m/km = 5.28 ft/mi

Figure 57. Roughness profiles along the left wheelpath for three profilers.

The roughness profiles from the three profilers agree well with each other throughout the section, except between 25 and 40 m (82 and 131 ft). This indicates that variations in the path followed by the three profilers within these limits are causing differences in the pavement features that are measured. This will cause different IRI values to be obtained for this wheelpath by the three profilers. Thus, IRI obtained by the three profilers will differ from that obtained by Dipstick.

Features Recorded by the Profiler That Are Missed or Underestimated by Dipstick

Dipstick obtains measurements at 304.8-mm (12-inch) intervals and is equipped with footpads that are 32 mm (1.25 inches) in diameter. The profilers that have a recording interval of 25 mm (1 inch) will record 12 measurements within the distance between the Dipstick footpads. Thus, Dipstick can miss features that are measured by a profiler.

In addition, Dipstick will not measure narrow cracks or joints in a PCC pavement when a Dipstick footpad is placed over such a feature, since the footpad will bridge over such features. However, a profiler will measure such features. When computing IRI, the IRI algorithm will treat a narrow downward feature the same as if that feature were upward. Features missed because of the footpad spacing and bridging over of narrow downward features can cause IRI obtained from Dipstick measurements to be less than that obtained from profiler measurements.

Figure 58 shows profile plots for (1) data collected along the left wheelpath at section 5 (chipseal section) during the 2003 LTPP profiler comparison by the North Atlantic ICC profiler at 25-mm (1-inch) intervals, (2) the same data after being processed using ProQual (i.e., moving average applied), and (3) data collected by Dipstick after all data sets have been subjected to a 3-m (10-ft) high-pass filter. When the 25-mm (1-inch) profile data are processed using ProQual, the depth of the cracks in the resulting averaged profile will be attenuated. Also, narrow cracks that appear in the 25-mm (1-inch) data will be more spread out, and can appear as a slight dip rather than a narrow crack.

This figure shows profile plots for data collected along the left wheelpath at section 5 during the 2003 LTPP profiler comparison by the North Atlantic ICC profiler at 25-millimeter (1-inch) intervals, the same data after being processed by ProQual (I.E., moving average applied), and the data collected by the Dipstick, after all data sets have been subjected to a 3-meter (10-foot) high-pass filter. The X-axis of the plot shows distance, while the Y-axis shows the IRI. At a distance of 120.5 meters (395 feet) there is a crack in the pavement that is seen in the 25-millimeter (1-inch) data as well as in the averaged 150-millimeter (5.9-inch) data. However, this feature is not seen in the Dipstick data. The Dipstick profile does show a slight dip near many of the other crack locations. However, the depth of the dip seen in the Dipstick profile is much less than that seen for the profiler data.

25.4 mm = 1 inch
1 m = 3.28 ft

Figure 58. Measurement of cracks by a profiler and Dipstick.

At a distance of 120.5 m (395 ft), there is a crack in the pavement that is seen in the profile data; however, this crack is not seen in the Dipstick data, either because the Dipstick footpad bridged the crack or the crack was between the contact points of the Dipstick footpad. Sometimes the pavement area adjacent to a crack has a slight dip because of settlement close to the crack. If the footpad of Dipstick falls within this settled area, the Dipstick profile will show a slight dip at such locations. The Dipstick profile shown in figure 58 does show a slight dip close to many crack locations. However, as seen at a distance of 135 m (443 ft), the depth of the dip that is seen in the Dipstick profile is much less than that seen in the profile data.

Figure 59 shows 3-m (10-ft) high-pass filtered profile plots for data collected along the right wheelpath at section 1 (smooth AC section) during the 2003 LTPP profiler comparison by Dipstick, the Western ICC profiler at 25-mm (1-inch) intervals, and the 25-mm (1-inch) data after being processed using ProQual (i.e., moving average applied to the data). The 25-mm (1-inch) profile data show a narrow downward feature that has an approximate depth of 10 mm (0.4 inches). The averaged 150-mm (5.9-inch) data show that applying the moving average onto the 25-mm (1-inch) data reduces the depth of this feature, and causes the feature to spread out over a much wider distance. The Dipstick data does capture this feature; however, the depth that it records for the feature is less than the depth of the feature that is seen in both the 25-mm (1-inch) and 150-mm (5.9-inch) data.

This figure shows 3-meter (10-foot) high-pass filtered profile plots for data collected along the right wheelpath at section 1 during the 2003 LTPP profiler comparison by Dipstick, the Western ICC profiler at 25-millimeter (1-inch) intervals, and the 25-millimeter (1-inch) data after being processed by ProQual (I.E., moving average applied on the data). The X-axis of the profile plots show distance, while the Y-axis shows elevation. Profile data between 119 and 125 meters (390 and 410 feet) are shown in the figure. The 25-millimeter (1-inch) profile data show a narrow downward feature that has an approximate depth of 10 millimeters (0.4 inches). The averaged 150-millimeter (5.9-inch) data shows this feature to have much less depth and to be spread out over a much wider distance than that shown in the 25-millimeter (1-inch) data. The Dipstick data does capture this feature, but it records a depth for the feature that is less than the depth of the feature seen in both the 25-millimeter (1-inch) and 150-millimeter (5.9-inch) data.

25.4 mm = 1 inch
1 m = 3.28 ft

Figure 59. Measurement of a downward feature by a profiler and Dipstick.

The missing of features by Dipstick as illustrated in figure 58, and the underestimation of the depth of downward features by Dipstick as illustrated in figures 58 and 59, will cause IRI obtained from Dipstick data to be lower than IRI obtained from the profiler data.

Averaging Effects of Profiler Data

In the LTPP profiler comparison studies, the IRI values for profile data obtained from the profilers were computed using the ProQual software. Data from the DNC 690 profilers we recorded at 152.4-mm (6-inch) intervals, and a moving average had been applied to this data b the profiler software before the data were recorded. When computing IRI values for the DNC 690 data, the data available at 152.4-mm (6-inch) intervals was used by ProQual. D obtained at 25-mm (1-inch) intervals are available for both the T-6600 profilers and the ICC profilers. When ProQual computes IRI values for data obtained by these profilers, a 300-mm (11.8-inch) moving average is first applied to the data, then profile data points that are at 150-mm (5.9-inch) intervals are extracted and the IRI is computed using this averaged data.

The effects of application of the moving average onto the 25-mm (1-inch) data were discussed in chapter 5. The averaged profile obtained from the profiler data that are used for computation of IRI by ProQual is an artificial profile, and this profile may not actually exist on the road. However, when measurements are performed with Dipstick, readings will be obtained on the actual profile of the road. For example, consider the profiles shown in figure 60 that show the 25-mm (1-inch) profile data, and the profile obtained after ProQual has processed the 25-mm (1-inch) data. When computing IRI for profile data, the 150-mm (5.9-inch) interval profile, which is an artificial profile, is used. However, when Dipstick obtains readings, the footpads of Dipstick will rest on the actual profile of the pavement, and not on the artificial profile that is defined by the 150-mm (5.9-inch) interval profile. Thus, this difference in the profiles between the profiler data and the Dipstick data can result in differences in the IRI values.

This figure shows the 25-millimeter (1-inch) profile data, and the profile obtained after ProQual has processed the 25-millimeter (1-inch) data. The X-axis of the plot shows distance, while the Y-axis shows elevation. Profile data between 136 and 137.5 meters (446 and 451 feet) are shown in this figure. The 25-millimeter (1-inch) data show a sharp drop in elevation just after 136.5 meters (448 feet). The 150-millimeter (5.9-inch) averaged data also show this drop, but the magnitude of the drop is lower than that seen for the 25-millimeter (1-inch) data. In the 150-millimeter (5.9-inch) data, the change in elevation over the fault is about 7 millimeters (0.3 inches), and it occurs over a distance of 0.3 meters (1 foot).

25.4 mm = 1 inch
1 m = 3.28 ft

Figure 60. Illustration of artificial profile used by ProQual for computing IRI.

For asphalt-surfaced pavements without distress, the difference between the averaged profile computed using ProQual and the profile defined by the 25-mm (1-inch) data will be very minor, and the two profiles may actually coincide. However, in new PCC pavements, there could be differences between these two profiles at the joints. The 25-mm (1-inch) profile will show the joint as a downward feature; however, this feature may not be seen in the averaged profile computed using ProQual, since it is smoothed out when the moving average is applied. On pavements that have distress, there could be significant differences between the 25-mm (1-inch) interval profile and the 150-mm (5.9-inch) interval averaged profile.

Dipstick Data Errors

In the LTPP program, when longitudinal Dipstick measurements are performed, the Dipstick readings are recorded on a form. Afterwards, these readings are entered into ProQual to computethe IRI. After Dipstick measurements are performed in the field, a closure error computation is performed as a data quality check.(6) Although this procedure does provide a check on the data quality, it is always possible for errors during measurement to occur, yet the closure error may be within the acceptable value. Also, multiple errors may occur that compensate for each other and cause the closure error to be within the acceptable value.

Figure 61 shows the 10-m (33-ft) base-length roughness profiles obtained along the left wheelpath at site 5 (chip-seal section) during the 2003 LTPP profiler comparison for a run by the Western profiler and Dipstick. There are differences between the two roughness profiles, with Dipstick showing much higher roughness than the profiler between 20 and 30 m (66 and 100 ft). This indicates that Dipstick is capturing a feature between these limits that is causing a high roughness, and this feature is not appearing in the profile data collected by the profiler.

This figure shows the 10-meter (33-foot) base length roughness profiles obtained along the left wheelpath at site 5 during the 2003 LTPP profiler comparison for Dipstick and a run by the Western profiler. The X-axis of the plot shows distance, while the Y-axis shows roughness. The two roughness profiles do not overlay with each other perfectly, and differences between the two profiles are seen throughout the section. The Dipstick roughness profile shows significantly higher roughness than the profiler between 20 and 30 meters (66 and 100 feet).

1 m = 3.28 ft
1 m/km = 5.28 ft/mi

Figure 61. Left-wheelpath 10-m (33-ft) base-length roughness profiles for profiler and Dipstick at site 5.

Figure 62 shows the 10-m (33-ft) high-pass filtered profiles of the profiler and Dipstick at this site from the start of the section to a distance of 50 m (164 ft). This figure shows that at an approximate distance of 22 m (72 ft), a sharp upward feature is seen in the Dipstick profile; however, this feature does not appear in the profile recorded by the profiler. This is the feature that caused the Dipstick roughness profile to show a higher value than the profiler roughness profile between the limits of 20 and 30 m (66 and 100 ft). This particular feature did not appear in any of the profile data collected by all of the profilers at this site.

Since this feature did not appear in the profile data collected by any profiler, the sharp upward nature of the feature indicates that it is most likely an erroneous data point. The possible reason for this feature appearing in the Dipstick data is either: (1) an incorrect reading being recorded at that location in the field, (2) an incorrect reading at that location being entered into ProQual, or (3) a Dipstick malfunction occurring at that location.

Figure 63 shows a portion of the right-wheelpath profile that was recorded by Dipstick and a profiler at site 1 (smooth AC section) during the 2003 LTPP profiler comparison. Both profiles have been subjected to a 10-m (33-ft) high-pass filter, and the profiles have been offset for clarity. At a distance of about 145 m (476 ft), the Dipstick profile shows a sudden drop in elevation; however, this feature is not seen in the profile recorded by the profiler. None of the profile runs for any of the profilers showed this feature in the profile. It is most likely that this feature in the Dipstick profile was caused by one of the factors that were described previously for the previous example. The inclusion of this feature in the Dipstick profile will cause an increase in the IRI for the Dipstick data.

This figure shows the 10-meter (33-foot) high-pass filtered left-wheelpath profiles of Dipstick and a run from a profiler at site 5 in the 2003 LTPP profiler comparison. The X-axis of the plot shows the distance, while the Y-axis shows the roughness. Profile data for a distance of 50 meters (164 feet) are shown in this figure. At an approximate distance of 22 meters (72 feet), a sharp upward feature is seen in the Dipstick profile, while this feature does not appear in the profile recorded by the profiler.

25.4 mm = 1 inch
1 m = 3.28 ft

Figure 62. Left-wheelpath profiles for profiler and Dipstick at site 5.

This figure shows a portion of the right-wheelpath profile that was recorded by the Dipstick and a profiler at site 1 during the 2003 LTPP profiler comparison. The X-axis of the plot shows the distance, while the Y-axis shows the elevation. The profile data between 130 and 150 meters (426 and 492 feet) are shown in the plot. Both profiles have been subjected to a 10-meter (33-foot) high-pass filter, and the profiles have been offset for clarity. At a distance of about 145 meters (476 feet), the Dipstick profile shows a sudden drop in elevation, but this feature is not seen in the profile recorded by the profiler.

25.4 mm = 1 inch
1 m = 3.28 ft

Figure 63. Right-wheelpath profiles for profiler and Dipstick at site 1

At the start of the LTPP program, the data collection procedures that were employed for Dipstick data collection were different from the current procedures. In those procedures, measurements along a wheelpath were performed from the start of the section to the end of the section and, thereafter, measurements were made back along the same path to end at the start of the section These procedures provided two profiles for a wheelpath, and if an error was suspected, the forward and return runs could be compared to determine whether there were differences. Th current Dipstick data collection procedures where measurements are performed along a loop were adopted to save the time required to perform Dipstick measurements (and the time requi was cut in half). However, the disadvantage of these procedures is that since only one profile is available for a wheelpath, there is no way to check whether a potentially incorrect data point is indeed incorrect.

collection. Such errors can occur because of a data recording error, an error during data entry a malfunction of Dipstick. If such errors occur, they will cause an upward bias in the IRI computed from the Dipstick data. The two examples shown in this section illustrate errors that can occur during Dipstick data

IRI Computation Procedure for Dipstick Data

In the LTPP program, the IRI values of a profile obtained from Dipstick measurements are computed using the ProQual software.(7,8) The procedure used by the ProQual software to compute the IRI values from Dipstick elevation data was described in chapter 2. As describ that chapter, the Dipstick elevation profile is rotated to obtain an additional distance of 152.4 m (500 ft) before the section, the entire profile is filtered with the upper-wavelength cutoff filter used in the profiler, and then the portion of the profile corresponding to the test section is extracted from the filtered profile and the IRI is computed using this extracted profile.

The correct method to accurately compute the IRI from the Dipstick data is to apply the IRI algorithm to the actual elevation profile that is obtained from the Dipstick data, and not to a filtered Dipstick profile. The analysis of the data from the LTPP profiler comparison conducted in 2003 indicated that there was a slight difference in IRI values obtained when the same Dipstick elevation profile was processed using ProQual and RoadRuf. The RoadRuf software uses the IRI computation procedure documented in American Society for Testing and Materials (ASTM) Standard E1926-98 (2003).(29) RoadRuf does not perform any prefiltering of the Dipstick data before computing the IRI. Table 8 presents the IRI values obtained for four test sections in the 2003 LTPP comparison using ProQual and RoadRuf. The IRI computed using ProQual were slightly higher than that computed using RoadRuf for all of the cases by amounts varying from 0.02 to 0.08 m/km (1.3 to 5.1 inches/mi). For all practical purposes, differences of these magnitudes can be considered to be negligible. However, for LTPP comparison studies where one criterion that is being evaluated is to determine whether the profiler IRI is within 0.16 m/km (10 inches/mi) of the IRI obtained from Dipstick, the magnitude of the IRI differences shown in table 8 can make the difference between either satisfying or failing the criterion. There was perfect agreement in the IRI values that were computed for the profiler data using ProQual and RoadRuf. Therefore, a possible reason for the discrepancy in Dipstick IRI values between RoadRuf and ProQual may be the filtering that is performed on the profile data before ProQual calculates IRI.

 

Table 8. IRI values for Dipstick data computed using ProQual and RoadRuf.
Section Wheelpath Dipstick IRI (m/km) Difference in IRI1
ProQual RoadRuf
2 Left 2.88 2.80 0.08
3 Left 0.90 0.88 0.02
4 Left 1.35 1.32 0.03
5 Left 2.29 2.24 0.06
2 Right 2.87 2.79 0.08
3 Right 1.00 0.99 0.00
4 Right 1.68 1.64 0.03
5 Right 2.70 2.63 0.07
1 Difference in IRI = ProQual IRI — RoadRuf IRI

1 m/km = 5.28 ft/mi

DISCUSSION OF DIFFERENCES IN IRI BETWEEN DIPSTICK AND THE PROFILERS

As described in the previous section, differences between profiler IRI and Dipstick IRI can occur because of a variety of factors. In some cases, the different errors compensate for each other and cause the profiler IRI and Dipstick IRI to agree well with each other.

Figure 64 shows the 10-m (33-ft) base-length roughness profiles along the right wheelpath at section 2 (rough AC section) used in the 2003 LTPP profiler comparison for Dipstick and one run from the Western profiler. IRI from the profiler run and the Dipstick measurements were 2.77 and 2.79 m/km (176 and 177 inches/mi), respectively. When the overall IRI value for the section is considered, the profiler IRI and Dipstick IRI are virtually identical. As seen in figure 64, the roughness profiles for the profiler and the Dipstick overlay very well, with only some minor deviations noted at some localized area. The roughness profiles show that both the profiler and Dipstick are sensing the same roughness in terms of IRI throughout the section.

This figure shows the 10-meter (33-foot) base length roughness profiles along the right wheelpath at section 2 used in the 2003 LTPP profiler comparison for Dipstick and one run from the Western profiler. The X-axis of the plot shows the distance, while the Y-axis shows the IRI. The roughness profiles for the profiler and the Dipstick overlay well, with only some minor deviations noted at some localized locations.

1 m/km = 5.28 ft/mi
1m = 3.28 ft

Figure 64. Roughness profiles for a profiler and Dipstick showing good agreement in roughness distribution.

Figure 65 shows the 10-m (33-ft) base-length roughness profiles for the same wheelpath of the same section for Dipstick and one run for the Southern profiler. The IRI for the profiler run and the Dipstick measurements were 2.70 and 2.79 m/km (171 and 177 inches/mi), respectively. When the overall IRI value for the section is considered, the profiler IRI and Dipstick IRI agree very well with each other. However, the agreement in the roughness profiles for the profiler and Dipstick is not as good as the agreement that was seen for the previous example. There are several locations where noticeable differences in the roughness profiles are seen. At some locations, the profiler sees higher roughness than Dipstick, while at other locations, Dipstick sees higher roughness than the profiler. However, when the overall roughness for the section is evaluated, these differences compensate for each other, and the overall roughness as measured by the profiler and Dipstick agree very well with each other.

The cross-correlation technique is a method that can be used to judge the agreement in IRI and the agreement in spatial distribution of IRI between two devices. The data collected for the 2003 LTPP profiler comparison by the North Atlantic and Western profilers were used with the Dipstick data to calculate cross correlation for IRI between the profilers and Dipstick. In this analysis, Dipstick was considered to be the correct device, and the analysis will indicate how well each of these profilers was able to reproduce the IRI obtained from the Dipstick measurements. The five ProQual-processed profile runs submitted by the North Atlantic and Western regions were used to calculate the cross correlation with Dipstick. The results of this analysis are presented in tables 9 and 10 for the left and right wheelpaths, respectively.

This figure shows the 10-meter (33-foot) base length roughness profiles along the right wheelpath at section 2 used in the 2003 LTPP profiler comparison for Dipstick and one run from the Southern profiler. The X-axis of the plot shows the distance, while the Y-axis shows the IRI. The agreement in roughness profiles for the profiler and the Dipstick is not as good as the agreement that was seen in figure 64. There are several locations where noticeable differences in the roughness profiles are seen. At some locations, the profiler sees higher roughness than the Dipstick, while at other locations, the Dipstick sees higher roughness than the profiler.

1 m/km = 5.28 ft/mi
1m = 3.28 ft

Figure 65. Roughness profiles for a profiler and Dipstick showing moderate agreement roughness distribution.

 

Table 9. Results of cross-correlation analysis: Left wheelpath
Site Profiler Cross Correlation With Dipstick
RUN NUMBER Minimum Maximum Average
1 2 3 4 5
1: Smooth AC North Atlantic 0.84 0.84 0.86 0.86 0.88 0.84 0.88 0.86
Western 0.82 0.83 0.83 0.84 0.84 0.82 0.84 0.84
2: Rough AC North Atlantic 0.82 0.82 0.82 0.83 0.87 0.82 0.87 0.83
Western 0.85 0.89 0.89 0.89 0.90 0.85 0.90 0.89
3: Smooth PCC North Atlantic 0.78 0.78 0.79 0.80 0.82 0.78 0.82 0.79
Western 0.82 0.82 0.83 0.84 0.84 0.82 0.84 0.83
4: Medium PCC North Atlantic 0.79 0.80 0.86 0.87 0.89 0.79 0.89 0.84
Western 0.82 0.83 0.84 0.84 0.84 0.82 0.84 0.83
5: Chip Seal North Atlantic 0.68 0.70 0.71 0.72 0.72 0.68 0.72 0.71
Western 0.69 0.70 0.72 0.73 0.73 0.69 0.73 0.71
Table 10. Results of cross-correlation analysis: Right wheelpath
Site Profiler Cross Correlation With Dipstick
RUN NUMBER Minimum Maximum Average
1 2 3 4 5
1: Smooth AC North Atlantic 0.68 0.68 0.69 0.74 0.74 0.68 0.74 0.70
Western 0.62 0.65 0.70 0.72 0.77 0.62 0.77 0.69
2: Rough AC North Atlantic 0.76 0.82 0.82 0.86 0.87 0.76 0.87 0.83
Western 0.67 0.70 0.70 0.74 0.75 0.67 0.75 0.71
3: Smooth PCC North Atlantic 0.80 0.81 0.81 0.82 0.82 0.80 0.82 0.81
Western 0.80 0.81 0.81 0.82 0.83 0.80 0.83 0.81
4: Medium PCC North Atlantic 0.85 0.85 0.85 0.87 0.87 0.85 0.87 0.86
Western 0.87 0.88 0.88 0.89 0.90 0.87 0.90 0.88
5: Chip Seal North Atlantic 0.75 0.77 0.79 0.79 0.79 0.75 0.79 0.78
Western 0.75 0.77 0.78 0.78 0.79 0.77 0.79 0.78

Along the left wheelpath, the average cross-correlation values for the five sites ranged from 0.71 (site 1) to 0.89 (site 2). For a specific profiler and a specific wheelpath, the difference between the maximum and minimum cross-correlation values for the different runs was within 0.05 for all of the cases, except for the North Atlantic profiler at site 4, which had a difference of 0.10. This indicates that the profilers are tracking a consistent path during the repeat runs or that minor variations in the wheelpaths for repeat runs are not significantly influencing the spatial distribution of roughness. However, this does not necessarily mean that the runs of the profiler followed the path that was measured by Dipstick.

Along the right wheelpath, the average cross-correlation values for the five sites ranged from 0.69 (site 1) to 0.88 (site 3). For a specific profiler and a specific wheelpath, the difference between the maximum and minimum cross-correlation values for the different runs was within 0.05 for six cases, between 0.05 and 0.10 for two cases, and exceeded 0.10 for two cases. The two cases where the value exceeded 0.10 were for the Western profiler at site 1 (a difference of 0.15) and the North Atlantic profiler at site 2 (a difference of 0.11). The cross-correlation values along the right wheelpath also generally indicate that the profilers are tracking a consistent path during the repeat runs or that minor variations in the wheelpaths for repeat runs are not significantly influencing the spatial distribution of roughness.

Figure 66 shows the roughness profiles along the right wheelpath at section 1 for Dipstick and run 1 from the Western profiler. The IRI cross correlation for these two profiles was 0.62, which was the lowest cross correlation for all of the cases considered in this study. The IRI values for the profiler and Dipstick for this case were 1.57 and 1.81 m/km (100 and 115 inches/mi), respectively. Most of the roughness at this site occurs at four localized areas, as shown in figure 66. However, the profiler and Dipstick are measuring the features within these limits differently, and this results in a low cross-correlation value for the two devices.

Figure 67 shows the roughness profiles along the left wheelpath at section 2 for Dipstick and run 5 for the Western profiler. The IRI cross correlation for this case was 0.90, which was the highest cross correlation for all of the cases considered in this study. The IRI from the profiler and Dipstick for this case were 2.76 and 2.80 m/km (175 and 178 inches/mi), respectively. Roughness profiles for the two devices overlay well within the section, except for some localized areas. At these locations, the profiler and Dipstick recorded different features.

This figure shows the roughness profiles along the right wheelpath at section 1 for Dipstick and run 1 from the Western profiler. The X-axis of the plot shows distance, while the Y-axis shows the IRI. Most of the roughness at this site occurs at four localized locations that are between 35 and 50 meters (115 and 164 feet), 55 and 70 meters (180 and 230 feet), 85 and 95 meters (279 and 312 feet), and 115 and 130 meters (377 and 426 feet). At all four of these locations, the roughness profile of the Dipstick shows higher values than that obtained for the profiler.

1 m/km = 5.28 ft/mi
1m = 3.28 ft

Figure 66. Roughness profiles for the case with the lowest cross correlation.

This figure shows the roughness profiles along the left wheelpath at section 2 for the Dipstick and run 5 from the Western profiler. The X-axis in the plot shows distance, while the Y-axis shows the IRI. The roughness profiles for the two devices overlay well, except for some localized locations where slight differences in roughness profiles are seen.

1 m/km = 5.28 ft/mi
1m = 3.28 ft

Figure 67. Roughness profiles for the case with the highest cross correlation.

As shown in the cross-correlation study, perfect agreement in the magnitude of the roughness and spatial distribution of the roughness did not occur between the profiler and Dipstick. However, a fairly good correlation (i.e., a correlation of greater than 0.80) was observed for many of the cases. The factors described previously in this chapter can contribute to a lowering of the cross-correlation values for the profiler and Dipstick. The IRI cross-correlation values for a profiler and Dipstick are not necessarily a function of the roughness, but rather will depend on the type of features within the section that contribute to the roughness.

A cross-correlation analysis was performed for the North Atlantic and Western profilers to determine how well they reproduced their IRI results. The data collected during the 2003 LTPP profiler comparison were used in this analysis. One representative run for each profiler was obtained at each section to perform this analysis. Results of this analysis are shown in table 11.

 

Table 11. Cross correlation of IRI for North Atlantic and Western profilers
Wheelpath Site
1 2 3 4 5
Left 0.98 0.94 0.91 0.92 0.94
Right 0.95 0.84 0.91 0.97 0.96

The cross-correlation values shown in table 11 indicate that there is excellent reproducibility between the two profilers in their ability to obtain IRI values. All cross-correlation values were greater than 0.90, except along the right wheelpath at site 2. There were significant distresses along this wheelpath, and the lower cross correlation is attributed to lateral variations in the path followed by the two profilers. The results shown in table 11 indicate that the two profilers are picking up similar features within the test sections. The cross-correlation values obtained for the two profilers are much higher than those obtained for each of these profilers and Dipstick.

The IRI obtained from Dipstick measurements usually is considered as the reference to evaluate the accuracy of the profilers. However, because of the deficiencies in the device that were described previously in this chapter, Dipstick cannot be considered as a device for measuring the reference profile of a pavement. There is currently no other device available that can overcome the limitations of Dipstick. Some agencies use the ARRB walking profiler as a reference device for measuring reference profiles; however, this device is subject to the same limitations as Dipstick. The rod and level has similar limitations. It has been shown that measurement errors are possible with the rod and level that can cause errors in the computed roughness indices for smooth pavements.(12) Obtaining rod-and-level measurements at a closer sampling interval or obtaining measurements with Dipstick with the footpads set to a smaller sampling interval can overcome some of the errors that are introduced because of the 304.8-mm (1-ft) sampling interval of Dipstick. However, obtaining measurements at shorter sampling intervals can be very time consuming

Current procedures for LTPP profiler comparisons use the average profiler IRI obtained from five error-free runs for comparison with the Dipstick IRI. This procedure helps to smooth out some of the variability in the profiler runs. In spite of all of these limitations with Dipstick, data from past LTPP comparisons have shown that good agreement between profiler IRI and Dipstick IRI, typically within ±0.16 m/km (±10 inches/mi), can be obtained at many sections (see chapter 3). Usually, agreement between profiler IRI and Dipstick IRI that is within ±0.16 m/km (±10 inches/mi) is possible on AC and PCC pavements that do not have distress. However, significant differences in IRI between Dipstick and the profiler can occur on rough pavements that have distress. The magnitude of the difference in IRI is not necessarily a function of the IRI from the section, but rather it depends on the type of roughness features that are present in the section.

In spite of all of the limitations with Dipstick, it still can be used as a device for checking the IRI obtained from the profilers. However, it cannot be considered as a device for checking the accuracy of the profilers on pavements that have rough features or distress. The current LTPP comparison procedure uses an IRI difference of ±0.16 m/km (±10 inches/mi) between the profiler IRI and the Dipstick IRI to judge the accuracy of LTPP’s profilers. If differences outside of this limit are obtained, it does not necessarily mean that there is a problem with the profiler. Such difference can occur because of one or more of the causes that were described previously in this chapter. If such situations are encountered, a more detailed analysis of the data should be performed to identify the cause of the difference in IRI.

EXPECTED DIFFERENCES BETWEEN DIPSTICK AND PROFILER IRI

An analysis was performed to identify the range of differences in IRI that can be expected between the Dipstick IRI and the profiler IRI. This analysis was performed for the DNC 690 (using data from the 1992 LTPP profiler comparison), T-6600 (using data from the 1998 LTPP comparison), and ICC profilers (using data from the 2003 LTPP comparison).

The following procedures were used to perform the analysis for each case:

  1. For each wheelpath at each site, use the IRI values for the repeat runs of each profiler to compute the differences between the IRI from each profiler run and the IRI from the Dipstick measurements (i.e., profiler IRI — Dipstick IRI).
  2. Use the computed differences for that wheelpath from all of the runs for all of the profilers to compute the 15th percentile, 85th percentile, and median of the differences in IRI between the profiler IRI and the Dipstick IRI.

The computed values are presented for the DNC 690, T-6600, and ICC profilers in tables 12, 13, and 14, respectively. In each table, the results are grouped according to the surface type, and then under each surface type, the results are sorted according to the Dipstick IRI.

The data presented in these tables were not used to compute expected IRI differences for different roughness ranges. This is because the differences in IRI between the profiler and Dipstick are not necessarily a function of the roughness, but rather will depend to a great extent on the profile features within the section that contribute to the roughness.

 

Table 12. Differences between the K.J. Law Engineers DNC 690 profiler IRI and Dipstick IRI.
Surface type Wheelpath Test Section IRI from Dipstick(m/Km) Profiler IRI — Dipstick IRI (m/km)
15th Percentile 85th Percentile Median
Asphalt Left 6 0.66 -0.07 -0.01 -0.05
Asphalt Right 6 0.76 -0.02 -0.03 -0.01
Asphalt Right 5 0.93 -0.08 0.01 -0.04
Asphalt Left 5 1.31 -0.14 -0.07 -0.09
Asphalt Right 3 1.37 -0.04 -0.08 0.00
Asphalt Left 3 1.47 -0.09 -0.05 -0.07
Asphalt Left 7 1.85 -0.37 -0.09 -0.31
Asphalt Right 7 3.52 -0.45 -0.17 0.00
Concrete Right 8 1.10 -0.03 0.02 0.00
Concrete Left 8 1.39 -0.11 0.09 0.05
Concrete Right 2 1.69 -0.11 -0.04 -0.08
Concrete Left 1 1.81 -0.20 -0.06 -0.13
Concrete Right 1 2.13 -0.18 -0.10 -0.14
Concrete Left 2 2.32 -0.19 0.02 -0.05
Concrete Left 4 4.12 0.16 0.44 0.26
Concrete Right 4 5.63 -0.06 0.11 0.06

1 m/km = 5.28 ft/mi

 

Table 13. Differences between the K.J. Law Engineers T-6600 profiler IRI and Dipstick IRI.
Surface type Wheelpath Test Section IRI from Dipstick(m/Km) Profiler IRI — Dipstick IRI (m/km)
15th Percentile 85th Percentile Median
Asphalt Left 1 0.95 0.06 0.09 0.07
Asphalt Right 1 0.96 0.07 0.11 0.09
Asphalt Right 2 2.46 -.0.09 0.15 -0.01
Asphalt Left 2 2.52 -0.01 0.06 0.03
Concrete Right 3 1.13 0.10 0.13 0.11
Concrete Left 3 1.17 0.09 0.13 0.12
Concrete Left 4 2.87 -0.09 0.03 -0.01
Concrete Right 4 3.17 -0.13 0.02 -0.05

1 m/km = 5.28 ft/mi

 

Table 14. Differences between the ICC profiler IRI and Dipstick IRI.
Surface type Wheelpath Test Section IRI from Dipstick(m/Km) Profiler IRI — Dipstick IRI (m/km)
15th Percentile 85th Percentile Median
Asphalt Left 1 1.17 0.08 0.12 0.11
Asphalt Right 1 1.81 -0.18 -0.06 -0.09
Asphalt Right 2 2.79 -.0.25 0.23 -0.04
Asphalt Left 2 2.80 -0.07 0.01 -0.04
Concrete Left 3 0.88 0.02 0.00 0.04
Concrete Right 3 0.99 -0.03 0.13 0.00
Concrete Left 4 1.32 0.11 0.22 0.12
Concrete Right 4 1.64 0.04 0.08 0.06
Chip Seal Left 5 2.24 -0.11 0.02 -0.05
Chip Seal Right 5 2.63 -0.13 -0.06 -0.09

SUMMARY OF THE FINDINGS

Several factors that can cause IRI obtained from the Dipstick data to differ from IRI obtained from the profiler data. These factors are:

  • Sampling qualities of Dipstick: A theoretical analysis indicated that Dipstick does not measure wavelengths less than 2 m (7 ft) accurately. The sampling interval of Dipstick also may cause features with wavelengths shorter than twice the sampling interval of Dipstick to contaminate the measurement of longer wavelengths because of aliasing.
  • Variations in the path followed by the profiler: Variations in the path followed by the profiler from the path where Dipstick measurements were obtained can cause differences in IRI between the two devices. Significant differences in IRI can occur because of this factor in pavements with distress.
  • Features recorded by the profiler that are missed by Dipstick: The spacing between the two footpads of Dipstick is 0.305 m (1 ft). A profiler with a 25-mm (1-inch) recording interval obtains 12 readings within this distance. Thus, Dipstick can miss features measured by the profiler. Dipstick has footpads that are approximately 32 mm (1.25 inches) in diameter. The footpads can bridge over narrow downward features such as cracks and can cause the IRI from Dipstick to be lower than that obtained from the profiler.
  • Features recorded by the profiler that are underestimated by Dipstick: Sometimes there is some settlement close to the cracks in the pavement. Although Dipstick can record a dip at such features, the magnitude of the depth of the dip measured can be lessthan that recorded by the profilers because the profilers have a much smaller sampling interval and, thus, can capture the deepest part of the dip.
  • Averaging effects of profiler data: In the LTPP program, profile data obtained at 25 mm (1-inch) intervals are processed using ProQual, which applies a 300-mm (11.8 inch) moving average onto the 25-mm (1-inch) data and then extracts data at 150 mm (5.9-inch) intervals. The IRI is computed using this averaged profile, which is an artificial profile. When Dipstick measurements are performed, readings will be obtained on the actual pavement and not on this artificial profile.
  • Dipstick data errors: Errors in Dipstick measurements can occur because of incorrect readings being recorded, a malfunction of Dipstick, or data entry errors during data input to compute IRI. These errors can cause a bias in the IRI computed from the Dipstick data.
  • IRI computation procedures for Dipstick data: In the LTPP program, IRI values are computed using the ProQual software. When ProQual computes the IRI values from the Dipstick data, the data are manipulated and filtered before computing the IRI. Comparison of IRI values obtained from ProQual and RoadRuf showed that the IRI computed using ProQual had a slight upward bias. This is attributed to the filtering that is performed on the Dipstick data before ProQual computes the IRI.

It is possible to have good agreement between profiler IRI and Dipstick IRI because the various errors compensate for each other. The cross-correlation technique can be used to determine whether there is agreement between the Dipstick IRI and the profiler IRI in both magnitude and spatial distribution

Despite all of these limitations with Dipstick, data from past LTPP comparisons have shown that good agreement between profiler IRI and Dipstick IRI, typically within ±0.16 m/km (±10 inches/mi), can be obtained at many sections. Current procedures for LTPP profiler comparisons use the average profiler IRI obtained from five error-free runs for comparison with the Dipstick IRI. This procedure helps to smooth out some of the variability in the profiler runs. Although Dipstick can be used to check the IRI obtained from the profilers, it cannot be considered for checking the accuracy of profilers on pavements that have rough features or distress. The current LTPP comparisons use an IRI difference of ±0.16 m/km (±10 inches/mi) between the profiler IRI and the Dipstick IRI to judge the accuracy of LTPP's profilers. If differences outside of this limit are obtained on pavements having distress, it does not necessarily mean that there is a problem with the profiler

Previous | Table of Contents | Next

 

Federal Highway Administration | 1200 New Jersey Avenue, SE | Washington, DC 20590 | 202-366-4000
Turner-Fairbank Highway Research Center | 6300 Georgetown Pike | McLean, VA | 22101