4.0 Analysis and Evaluation of Alternative Condition Indicators continued

Evaluation of Highway Performance Measures for a Multi-Study Corridor - A Pilot Study

The following are some observations regarding the basic statistics:

In order to manage the size of Tables 4.4 and 4.5, some of the data from Virginia are not displayed. These values, which include lane joints, bleeding, and delaminations, are not significantly different than the values that are shown.
Counts only include segments where the measure is not missing. It is interesting to note that IRI was not provided for all Delaware segments. Delaware only recently has begun to collect IRI for all pavement segments. Previously, Delaware only collected IRI for HPMS segments.
In some cases, the standard deviation and the variance are very high. In fact, for the cracking measures in Virginia, these values were too high to be credible and were replaced with "N/A." Even without the standard deviation, a review of the minimum, maximum, average, and median values for these measures provides a sense of the variability. CS does not attach any particular significance to this except to note that variability in condition is expected in different road sections across a state the size of Virginia.
IRI is the only common measure reported by all states. Generally, these measures show a reasonable uniformity across states. The numbers also correspond well when comparing any one state with the values for all states. This can be attributed to:
- A reasonable level of uniformity in the conditions along I-95; and
- A high degree of standardization in how IRI is measured, driven by the length of time this value has been used.
It is interesting to note that Virginia does have a lower standard deviation and variance than the other states. These differences point to a higher level of uniformity in Virginia pavement condition.

Table 4.6 presents some of the same basic statistics, except that values have been weighted by segment length. Only the key measures are reported and only the statistics necessary to support calculations of probability and cumulative distributions were calculated.

Table 4.6 Basic Pavement Statistics - Weighted
State	Statistic	IRI - Average	OPC	CCI
DE	Count	91	193	Not Applicable
DE	Average	125.11	67.29	Not Applicable
DE	Std Dev	81.85	15.82	Not Applicable
DE	Variance	6699.20	250.27	Not Applicable
MD	Count	2177	Not Applicable	Not Applicable
MD	Average	82.71	Not Applicable	Not Applicable
MD	Std Dev	46.68	Not Applicable	Not Applicable
MD	Variance	2179.25	Not Applicable	Not Applicable
VA	Count	192	Not Applicable	190
VA	Average	83.95	Not Applicable	74.17
VA	Std Dev	21.57	Not Applicable	19.81
VA	Variance	465.18	Not Applicable	392.41
ALL	Count	2459	Not Applicable	Not Applicable
ALL	Average	84.69	Not Applicable	Not Applicable
ALL	Std Dev	35.46	Not Applicable	Not Applicable
ALL	Variance	1257.41	Not Applicable	Not Applicable

The following are some observations regarding the weighted statistics:

For Maryland, the count dropped by two because there are two segments with a length of zero. This appears to be a result of the fact that Maryland records values for fixed-length segments (0.1 miles) and resets the mileposts at county boundaries. There are two records where the beginning and ending mileposts have the same value and fall at the edge of a county, which leads us to believe that the ending milepost is rounding down to the same value as the beginning milepost.
The only state that shows any significant difference between weighted and nonweighted statistics is Delaware. When weighted by pavement length, the average IRI in Delaware dropped by approximately 17 percent. CS believes that this difference, which is not reflected in the other states, may be tied to the fact that Delaware did not collect IRI for all segments. Potentially, the segments that Delaware reported for the HPMS were skewed toward shorter segments with higher IRI values.
It is interesting to note that a difference similar to the one for IRI in Delaware is not observed when comparing weighted and nonweighted values for OPC in Delaware. This seems to support the idea that the Delaware HPMS segments are not indicative of the overall condition of Delaware roads.

For the second phase of the pavement analysis, CS measured the correlation between IRI and other condition measures. Table 4.7 presents correlations for Delaware. Table 4.8 shows correlations for Maryland. Table 4.9 presents correlations for Virginia. Because IRI was the only measure reported by all states, CS was unable to produce any correlations between states.

Table 4.7 Pavement Correlation Coefficients - Delaware
Statistic	IRI - Right	IRI - Average	OPC
IRI - Left	0.81	0.95	0.12
IRI - Right		0.95	0.14
IRI - Average			0.14

Table 4.8 Pavement Correlation Coefficients - Maryland
Statistic	Rut Depth	Rut Count	Friction Number	Cracking Index
IRI - Average	0.19	-0.02	-0.07	-0.19
Rut Depth		0.87	-0.14	-0.11
Rut Count			-0.03	-0.08
Friction Number				0.26

Table 4.9 Pavement Correlation Coefficients - Virginia
Statistic	IRI - Right	IRI - Average	LDR	NDR	CCI
IRI - Left	0.86	0.96	-0.40	-0.37	-0.44
IRI - Right		0.97	-0.39	-0.38	-0.42
IRI - Average			-0.41	-0.39	-0.45
LDR				0.72	0.94
NDR					0.88

The following are some observations regarding the correlation coefficients:

There are no firm rules regarding how close a coefficient must be to 1 or 1 in order to be considered significant. For purposes of this analysis, CS has chosen 0.70 as a reasonable threshold to identify values that are well correlated.
The only values in any state that are well correlated are values that are directly related (e.g., IRI - Left, - Right, and - Average; Rut Count and Rut Depth; and LDR, NDR, and CCI).
Correlations between IRI and a composite value like OPC or CCI are marginal, at best. This leads to the conclusion that IRI, by itself, is not a good measure of overall pavement condition.
Correlations between any two random distresses (e.g., Rut Depth and Cracking Index) are low. This result is not surprising since there is no particular reason to assume that having a distress of one type on a road segment inevitably would lead to distresses of other types.
Some of the correlations between IRI and other values are negative. This is not an issue but reflects that fact that a higher IRI value indicates a worse reading while higher values for some other measures may indicate better values (e.g., higher is better for LDR, NDR, and CCI).

In addition to computing the correlation coefficients, CS also prepared probability and cumulative distributions and associated graphs. A large number of graphs were produced and representative samples are shown below. Figure 4.5 shows probability and cumulative distributions for Delaware. Figure 4.6 shows distributions for Maryland. Figure 4.7 shows distributions for Virginia.

The distribution graphs for Delaware reflect the higher standard deviation and variance encountered for the Delaware IRI readings compared to the other states. For Maryland, the IRI Condition Index graphs show an index computed from the IRI data. The sharp curves reflect the fact that this index can assume values only between 1 and 5. Maryland also computes similar indexes, with similar distribution graphs, for the other condition measures. For Virginia, the similarity between the IRI and CCI graphs is driven by similarities in the averages and standard deviations for these measures. This would indicate that these values, although not well correlated, are related in some way. This would seem to be true even though IRI is not a component of CCI.

Graphs for other condition measures are similar. Most of the graphs for the cracking measures in Virginia are so spread out that they almost appear to be straight lines. This is a function of the extremely high standard deviations for these measures, which were noted above for Table 4.4.

Figure 4.5 Pavement Distributions - Delaware

Figure 4.6 Pavement Distributions - Maryland

Figure 4.7 Pavement Distributions - Virginia

Figures 4.5, 4.6, and 4.7 show values not weighted by segment length. Figure 4.8 compares weighted and not weighted measures in Delaware.

Figure 4.8 Weighted Pavement Distributions - Delaware

With the exception of Delaware, the distributions showing pavement measures that are weighted and not weighted by segment length are virtually identical. Figure 4.9 compares the weighted and not weighted values for IRI across all states.

Figure 4.9 Pavement Distributions - All States

As with the bridge analysis, pavement measures weighted by segment length generally are sharper (i.e., the probability distributions are narrower and the peaks higher). For pavement, these differences are extremely small, which reinforces conclusions associated with Table 4.6.

Comparisons across States

A key piece of this analysis was focused on comparing algorithms for composite measures across states. Both Delaware and Virginia use composite measures. However, given the complexity of the Virginia algorithm for CCI, CS opted to focus on using the Delaware algorithm for OPC. Also, CS determined that the breadth of data elements provided by Maryland was insufficient to satisfy the requirements of the OPC algorithm. Therefore, CS concentrated on calculating OPC using Virginia data and the Delaware algorithm.

For this analysis, CS performed the following steps:

Step 1 - Determine which Virginia data values match the Delaware distress measures used to calculate OPC.
CS opted for the following mapping:
- VA Transverse Cracking (sum of all severities) = DE TransCrack;
- VA Longitudinal Cracking (sum of all severities) = DE Block Cracking;
- VA Alligator Cracking (sum of all severities) = DE Fatigue Cracking;
- VA Patching Area (wheel + nonwheel) = DE Asphalt Patching; and
- VA IRI * Rut Depth = DE Surface Defects.
Other Virginia measures such as lane joints, delaminations, bleeding, potholes, etc., were not used. The most problematic of these assumptions is IRI * Rut Depth = Surface Defects. However, CS assumed that any issues would be covered when an extent and severity for each measure are estimated in Steps 2 and 3. Also, as long as the calculations were performed consistently, CS believes that a meaningful correlation can be obtained regardless of the validity of the mapping.
Step 2 - Convert the absolute quantities for each Virginia distress measure into a high, medium, or low extent required by the Delaware algorithm.
CS attempted to convert the absolute quantities into percents scaled by the segment length. For each measure, the absolute values for all severities were added and then multiplied by the segment length in feet. The maximum value of this calculation was computed for each measure. Then, the measure for each segment was divided by the maximum. Percentages <= 33.3 were assigned a LOW extent. Percentages >= 66.7 were assigned a HIGH extent. All other values were assigned a MEDIUM extent.
Step 3 - Convert the different severities for each Virginia distress measure into a high, medium, or low severity required by the Delaware algorithm.
For each segment, CS assumed that the severity with the greatest absolute quantity of distress could be used as the overall severity level. For measures like alligator cracking, where Virginia already defined three severity levels, severity 1=LOW, severity 2=MEDIUM and severity 3=HIGH. For measures like transverse cracking, where Virginia only defined two severity levels, CS "created" an intermediate severity category equal to one-third of severity 1 plus one-third of severity 2. This left severities 1 and 2 with 66 percent of their original value. Again, the overall severity level for the segment was defined as the severity category with the greatest absolute value.

For surface defects, CS observed that IRI * Rut Depth yielded values running approximately from 0 to 50. If this value was <= 16.67, then severity=LOW. If this value was >= 33.33, then the severity=HIGH. Otherwise, the severity=MEDIUM.
Step 4 - Using the HIGH, MEDIUM, LOW severity and extent calculated in Steps 2 and 3 for each Virginia pavement segment, lookup five distress index values using the Delaware tables shown in Figure 4.4.
Calculate OPC for each Virginia segment by calculating the average and standard deviation of the five distress index values and applying the formula OPC = avg - (1.25 * stdev). The process described above produced an OPC value for 190 of the Virginia I-95 pavement records. OPC was not computed for two Virginia records where CCI was equal to -1. CS computed the correlation coefficient between the OPC and CCI values for the 190 Virginia pavement records. The coefficient is 0.714, which is sufficiently high to say that these values are reasonably well correlated. Recall that the correlation coefficient does not address the question of which value is more correct. CS believes that the Virginia calculation for CCI is more correct because it includes more data elements and does not rely on estimations of severity and extent. Figure 4.10 demonstrates the correlation between OPC and CCI in Virginia.

Figure 4.10 OPC versus CCI in Virginia

This exercise demonstrates that composite measures (i.e., measures that combine multiple distresses into a single value) tend to track well regardless of the algorithm used to produce the measure. And because neither OPC nor CCI correlate well with IRI in Virginia (OPC versus IRI=-0.49 and CCI versus IRI= 0.45), CS concludes that composite values provide a superior measure of pavement condition when compared to IRI.

4.2 Data Collection, Validation, and Interpretation

In Task 1, CS contacted agencies responsible for the collection of bridge and pavement data for Delaware, Maryland, and Virginia. The data collection process is described in Section 2 and the specific individuals that provided the data are documented in Table 2.2.

As part of the data analysis task, CS again contacted these individuals to conduct follow-up interviews. The interviews focused on three main areas:

The data collection process used by each state;
How each state validates the quality of the data collected; and
Standards adopted by each state for ranking asset condition.

This section documents the information received from each state.

Data Collection

States inspect and record information on the condition of bridges and pavement annually. This process supports both in-house uses (e.g., determination of maintenance priorities) as well as Federal reporting requirements for the NBI and the HPMS. While all states maintain engineers and other in-house bridge and pavement experts, CS observed that Delaware, Maryland, and Virginia all outsource at least some of their inspection activities.

Every state operates both a bridge management system (BMS) and a pavement management system (PMS) that, at a minimum, allow for storing and reporting of information. These systems also may provide capabilities to assist with inspection, calculate condition measures, and manage assets.

Delaware

For its bridges, Delaware uses a combination of in-house inspectors and outside consultants. The consultants are used primarily for specialized structures such as moveable bridges. Most inspectors take laptops equipped with the Pontis BMS into the field to record both NBI and element-level inspections. The bridge inspection engineers are responsible for ensuring that information is moved into the production Pontis database. This process usually occurs within one week following the inspection. The bridge inspection engineer is responsible for calculating the sufficiency rating at a later time (i.e., after the inspection has been moved into the production system and reviewed).

For its roads, Delaware previously collected pavement data using a windshield survey but recently has contracted with Applied Research Associates, Inc. (ARA), which will use an instrumented van to collect pavement measurements for future inspections. ARA will be responsible for calibrating the inspection tools. As part of the transition to ARA, Delaware has stated that they will begin measuring IRI on all road segments, not just those required for HPMS reporting. ARA will be responsible for moving all inspection data into the Delaware PMS, which also is provided by ARA. Although this process is expected to be highly automated, it is not clear what sort of time lag will be encountered since there is considerably more data (e.g., images and video logs) to be transferred.

Maryland

For its bridges, Maryland also uses a combination of in-house inspectors and outside consultants. The situation is identical to Delaware in that the consultants provide expertise on specialized structures (e.g., moveable and fracture critical bridges). NBI and element-level inspection information is captured on paper and entered into a custom BMS on a monthly basis. An outside consultant is used to calculate sufficiency rating for the bridges.

For its roads, Maryland uses its own equipment to automatically inspect approximately 11,000 lane-miles per year. Data are captured every 52 feet and rolled up to one-tenth-mile and one-half-mile values. Maryland contracts with ARA to compute values based on the inspection data, including condition indexes for IRI, cracking, friction, and rutting. On a weekly basis, data are transferred automatically to a custom PMS.

Virginia

For its bridges, Virginia also uses a combination of in-house inspectors and outside consultants. Consultants are used to compensate for insufficient in-house staff. These inspectors use a combination of paper forms and tablet-style laptops to perform both NBI and element-level inspections. Inspectors are responsible for transferring or entering data into the state's Pontis system on a weekly basis. A bridge safety engineer calculates the sufficiency rating for each bridge using the Pontis functionality.

For its roads, Virginia uses a contractor, Furgo Roadware, to perform automated inspections using an instrumented van. As with ARA in the other states, Furgo Roadware is responsible for calibrating the van and transferring the inspection information weekly into Virginia's Agile Assets PMS. This process is supported by dedicated information technology (IT) personnel.

Data Validation

Quality assurance (QA) involves validating the data, the data collection process, or both to ensure that the information being recorded is accurate. QA can include many types of tests, from looking for missing values to double-checking unexpected results to direct validation of some percentage of the original data. Frequently, the level of QA involves a tradeoff between the desired level of accuracy and the cost, in both time and money, to perform the tests.

Although states usually will provide guidelines for both bridge and pavement inspections, CS believes that bridge inspections rely more on the engineering expertise of the inspector than do pavement inspections. As discussed in the previous section, a pavement inspection primarily is an automated process that directly measures certain attributes of the road. There is a visual component to count/measure defects such as cracks and potholes, but overall the pavement inspection process is fairly mechanical.

The differences in the bridge and pavement inspection processes seem to affect the degree to which the data may be validated by a state. CS observed that the states participating in this study generally invest more effort in reviewing bridge data than pavement data. CS believes that this is a function both of the type of information gathered for pavement sections and the automated processes used to capture this information.

Delaware

For bridges, the bridge inspection engineer performs the initial QA check at the time data are transferred to the production Pontis database. This generally involves a visual review of the information being transferred. At this point, a team lead may review the information before sending it to a supervisor, who checks for missing data and invalid entries. Changes to inspection data happen infrequently, but if an error is discovered, the inspection engineer reviews the issue and makes any necessary changes before sending the change to the team lead for signoff.

For roads, a supervisor reviews the inspection data both before and after they are moved into the Delaware PMS. It is rare for data to be questioned. The biggest issue that may be encountered are missing records.

Maryland

For bridges, a supervisor reviews the inspection data based on reports sent by the engineering team lead. These reports are created after the monthly transfer of information to the BMS. Changes to inspections based on this review are rare. Also, an independent internal inspection team will recheck a certain amount of work each year. According to Maryland, approximately 10 percent of the inspections may be verified by this other team.

For roads, office staff review the data received from ARA. It is rare for these data to be questioned or adjusted.

Virginia

For bridges, a bridge safety engineer reviews data at the end of the inspection process. This happens both at the district and state level. Inspection updates are infrequent but any changes are entered directly into the State's Pontis system.

For roads, the IT team responsible for loading the inspection data performs a QA process that includes comparing data summaries generated both inside and outside the PMS. In addition, a third party will review and verify approximately five percent of the records. Inspection data rarely are questioned. However, the team does look for and check large discrepancies in the year-to-year data for any given segment.

Data Interpretation

Numerical measures provide one means of quantifying the condition of a bridge or a pavement section. However, there also is a natural desire to classify asset conditions as "good" or "fair" or "poor" in order to obtain a qualitative assessment of the health of a network.

For bridges, the states participating in this study indicated that they rarely use health index to represent asset condition. Instead they rely on the traditional measures of structurally deficient and functionally obsolete. It is true that functional obsolescence is not considered in the health index calculation. When scheduling maintenance and other activities, states deal with bridges as discrete items. However, when expressing the overall condition of their bridge network, states typically would target "less than five percent of all bridges are structurally deficient," which was the goal expressed by Delaware.

Delaware

For pavement, Delaware defines characteristics for low, medium, and high severity and extent for different pavement types. Table 4.10 shows the definitions used to judge distress severity for flexible pavement. Table 4.11 shows the definitions for distress extent for flexible pavement. Severity and extent for pavement sections feed into the distress conversion tables, shown in Figure 4.4, which are used to compute OPC. OPC is a value between 0 (worse condition) and 100 (best condition). Delaware categorizes OPC as follows: <= 50 is poor, > 50 and <= 60 is fair, and > 60 is good. Because Delaware previously collected IRI only for HPMS sections, the state has not defined any qualitative categories for this measure.

Table 4.10 Delaware Pavement Distress Definitions for Severity
Distress	Low	Medium	High
Fatigue Cracking	Fine parallel hairline cracks	Alligator crack pattern clearly developed	Alligator crack pattern clearly developed with spalling and/or distortion
Transverse Cracking	Crack< 1/4 inch wide	Crack Width > 1/4 and < 3/4 inch and/or spalls less than 3 inches in width or sealed crack with sealant in good condition	Crack Width > 3/4 inch and/or spalls greater than 3 inches in width or significant loss of material
Block Cracking	Crack < 1/4 inch wide	Crack Width > 1/4 and < 3/4 inch and/or spalls less than 3 inches in width or sealed crack with sealant in good condition	Crack Width > 3/4 inch and/or spalls greater than 3 inches in width or significant loss of material
Patch Deterioration	Patches showing little or no defects with a smooth ride	Patches showing medium severity defects (e.g., cracking) and/or notable roughness	Patches showing high-severity defects and/or distinct roughness
Surface Defects	Aggregate has begun to wear away	Aggregate has worn away and surface is becoming rough and/or minor rutting occurring from horse and buggy traffic (less than 1 inch average depth)	Aggregate has worn away and surface is very rough and/or major rutting occurring from horse and buggy traffic (greater than 1 inch average depth)

Source: Delaware Department of Transportation.

Table 4.11 Delaware Pavement Distress Definitions for Extent
Distress	Low	Medium	High
Fatigue Cracking	1-9% (wheel path)	10-25%	> 25%
Transverse Cracking	> 50 foot spacing	25 foot < spacing <50 foot	< 25 foot spacing
Block Cracking	1-9%	10-25%	> 25%
Patch Deterioration	1-9%	10-25%	> 25%
Surface Defects	1-9%	10-25%	> 25%

Source: Delaware Department of Transportation.

Maryland

For pavement, Maryland collects absolute measures and then computes an index measure based on the definitions shown in Table 4.12.

Table 4.12 Maryland Pavement Distress Condition Indexes and Descriptions
Distress	Measurement	Condition Description	Condition Index
IRI (inch/mile)	> 0 and < 60	Very Good	1
	>= 60 and < 95	Good	2
	>= 95 and <= 170	Fair	3
	> 170 and <= 220	Mediocre	4
	> 220 and <= 640	Poor	5
Cracking Index	>= 90 and <= 100	Very Good	1
	>= 80 and < 90	Good	2
	>= 65 and < 80	Fair	3
	>= 50 and < 65	Mediocre	4
	> 0 and < 50	Poor	5
Friction Number	< 35	Poor	1
	>= 35 and < 40	Mediocre	2
	>= 40	Acceptable	3
Percent Rutting > one-half-inch	< 10%	Very Good	1
	>= 10% and < 20%	Fair	2
	>= 20%	Poor	3

Source: Maryland State Highway Administration.

Virginia

For pavement, Virginia provides guidelines, presented in Table 4.13, on how to measure and determine the severity of different types of cracking.

Table 4.13 Virginia Pavement Crack Severity Definitions
Distress	Severity Level	Severity Description	How to Measure
Transverse Cracking	Severity 1	A crack with the sealant in good condition such that the crack width cannot be determined or a closed, unsealed crack.	Record length of transverse cracks at each severity level. Evaluate each crack by highest severity level present on a significant portion of the crack as it is traversed.
Transverse Cracking	Severity 2	An open, unsealed crack or any crack (sealed or unsealed) with adjacent (within one foot) random cracking.	Record length of transverse cracks at each severity level. Evaluate each crack by highest severity level present on a significant portion of the crack as it is traversed.
Longitudinal Cracking	Severity 1	A crack with the sealant in good condition such that the crack width cannot be estimated or a closed, unsealed crack.	The minimum length of longitudinal cracking counted is one foot. Only longitudinal cracking outside the wheel paths is counted as longitudinal and each occurrence is counted separately. Longitudinal cracking in the wheel paths is counted as Severity 1 alligator cracking. Measure the length of each crack by severity level as it is traversed. Rate each crack at the highest severity level present on a significant portion of the crack. Report the total length of cracking by severity level for the section.
Longitudinal Cracking	Severity 2	An open, unsealed crack or any crack (sealed or unsealed) with adjacent random cracking.	The minimum length of longitudinal cracking counted is one foot. Only longitudinal cracking outside the wheel paths is counted as longitudinal and each occurrence is counted separately. Longitudinal cracking in the wheel paths is counted as Severity 1 alligator cracking. Measure the length of each crack by severity level as it is traversed. Rate each crack at the highest severity level present on a significant portion of the crack. Report the total length of cracking by severity level for the section.
Alligator Cracking	Severity 1	A single sealed or unsealed longitudinal crack in the wheel path or an area of cracks with no or few interconnecting cracks with no spalling.	Considering only the left and right wheel paths, measure the affected area (square feet) at each severity level. Consider only one severity level for a given area. If different severity levels in an area cannot be distinguished, rate the area at the highest severity level. The width of alligator cracking shall be the actual width while a minimum width for all severity levels shall be one foot. Report the square feet of alligator cracking by severity level for the section.
	Severity 2	An area of interconnecting cracks forming the characteristic alligator pattern; may have slight spalling.	Considering only the left and right wheel paths, measure the affected area (square feet) at each severity level. Consider only one severity level for a given area. If different severity levels in an area cannot be distinguished, rate the area at the highest severity level. The width of alligator cracking shall be the actual width while a minimum width for all severity levels shall be one foot. Report the square feet of alligator cracking by severity level for the section.
	Severity 3	An area of moderately or severely spalled cracks forming the characteristic alligator pattern.	Considering only the left and right wheel paths, measure the affected area (square feet) at each severity level. Consider only one severity level for a given area. If different severity levels in an area cannot be distinguished, rate the area at the highest severity level. The width of alligator cracking shall be the actual width while a minimum width for all severity levels shall be one foot. Report the square feet of alligator cracking by severity level for the section.

Source: Virginia Department of Transportation.

Virginia also categorizes both IRI and CCI, as shown in Table 4.14. Virginia refers to CCI as "pavement condition" while referring to IRI as "ride quality."

Table 4.14 Virginia Pavement Crack Severity Definitions
Category	Measurement	Pavement Condition/Ride Quality
CCI	>= 90	Excellent
	>= 70 and < 90	Good
	>= 60 and < 70	Fair
	>= 50 and < 60	Poor
	< 50	Very Poor
IRI	< 60	Excellent
	>= 60 and < 100	Good
	>= 100 and < 140	Fair
	>= 140 and < 200	Poor
	>= 200	Very Poor

Source: Virginia Department of Transportation.

4.3 Measures Available In ICAT

Following the analysis of the bridge and pavement data, CS selected certain measures to be displayed in the ICAT web application (WebCAT). Consistent with the design presented in Section 5, CS selected approximately six measures for bridges and six for pavement. These measures were incorporated as standard ArcGIS-compatible layers that are accessed using functionality already available in the WebCAT.

At the present time, the modified WebCAT system may be accessed at: http://ags.camsys.com/hpm/index.html.

The WebCAT created in conjunction with this project is a modified version of the WebCAT originally developed for the I-95 Corridor Coalition. The intent is to transition production hosting for the I-95 WebCAT at some point in the future to the University of Maryland. At the time of this transition, FHWA and the I-95 Corridor Coalition must determine the future of the modified WebCAT created for this project.

Within the WebCAT interface, bridge measures for this project are contained within the category "Bridges (Mid-Atlantic)" while the pavement measures for this project are under "Pavement (Mid-Atlantic)." The following sections describe each measure. Each description is followed by a table that shows the asset count based on the current data. Pavement tables also include total miles (rounded), which are computed by summing the difference between the beginning and ending milepoint for each segment. Note that the counts are not consistent because of minor variations in the data (e.g., based on its specific data, a bridge may or may not have a valid sufficiency rating and/or health index).

Also note that CS does not characterize any data as "good," "fair," "poor," etc. These characterizations, where they are used, either come from a state participating in this study or an external source, such as NBIAS. In some cases, data are divided into categories arbitrarily for purposes of comparing different measures for the same assets. These ranges are not intended to represent quality ratings.

Bridges (Mid-Atlantic)

SD/FO - Shows whether a bridge is structurally deficient (SD), functionally obsolete (FO), or not deficient (ND). All bridges on I-95 are included. Bridges that are SD are colored red. Bridges that are FO are colored yellow. Bridges that are not deficient are colored green. This layer provides a means to visualize the current SD/FO status of each bridge. The SD/FO status, in combination with the sufficiency rating, determines whether a bridge is eligible for Federal funding.

State	SD	FO	ND	Total
Delaware	1	9	52	62
Maryland	1	0	82	83
Virginia	24	33	247	304
Total	26	42	381	449

Sufficiency Rating (NBIAS) - Represents the sufficiency rating for the bridge using same the good/fair/poor classifications included in the latest version of the FHWA National Bridge Investment Analysis System (NBIAS). All bridges on I-95 where the value of the sufficiency rating is > 0 are included. Bridges where the sufficiency rating is <= 50 are colored red (poor). Bridges where the sufficiency rating is > 50 and <= 80 are colored yellow (fair). Bridges where the sufficiency rating is > 80 are colored green (good). This layer provides one means of classifying bridges using the same standard as NBIAS.

State	Good	Fair	Poor	Total
Delaware	34	26	0	60
Maryland	10	33	1	44
Virginia	190	102	6	298
Total	234	161	7	402

Sufficiency Rating - Represents the sufficiency rating for the bridge using a five-level classification system based roughly on the available measurements. All bridges on I-95 where the value of the sufficiency rating is > 0 are included. Bridges where the sufficiency rating is <= 60 are colored red. Bridges where the sufficiency rating is > 60 and <= 80 are colored orange. Bridges where the sufficiency rating is > 80 and <= 90 are colored yellow. Bridges where the sufficiency rating is > 90 and <= 95 are colored blue. Bridges where the sufficiency rating is > 95 are colored green. This layer provides an alternate means of classifying bridges.

State	Green	Blue	Yellow	Orange	Red	Total
Delaware	8	14	12	24	2	60
Maryland	1	0	9	32	2	44
Virginia	62	36	92	86	22	298
Total	71	50	113	142	26	402

Health Index - Overall - Represents the overall health index for the bridge using the same five-level classification system used for sufficiency rating. All bridges on I-95 where the value of the sufficiency rating is > 0 are included. Note: this bridge set was chosen deliberately to ensure that the same bridges are available in both this layer and the sufficiency rating layer. Bridges where the health index is <= 60 are colored red. Bridges where the health index is > 60 and <= 80 are colored orange. Bridges where the health index is > 80 and <= 90 are colored yellow. Bridges where the health index is > 90 and <= 95 are colored blue. Bridges where the health index is > 95 are colored green. This layer provides an alternate means of classifying bridges and a way to visually compare the classification of bridges by health index with the classification by sufficiency rating.

State	Green	Blue	Yellow	Orange	Red	Total
Delaware	26	21	10	4	0	61
Maryland	9	6	2	5	1	23
Virginia	150	38	55	55	6	304
Total	185	65	67	64	7	388

SR HIX Diff - Represents the absolute value of the difference between the Sufficiency Rating (SR) and the Health Index - Overall (HIX), both of which are values between 0 and 100. All bridges on I-95 with both a valid sufficiency rating and health index are included. Bridges where the difference is <= 20 are colored green. Bridges where the difference is > 20 and <= 50 are colored yellow. Bridges where the difference is > 50 are colored red. This layer is designed to highlight bridges where there is a substantial difference between these two composite measures.

State	Small Diff	Medium Diff	Large Diff	Total
Delaware	46	14	0	60
Maryland	13	3	0	16
Virginia	239	58	1	298
Total	298	75	1	374

Deck Area - Represents the bridge deck area (in square meters) using a five-level classification system. All bridges on I-95 where the value of the deck area is > 0 are included. Bridges with deck area <= 1,000 square meters are represented by a small circle. Bridges with deck area > 50,000 square meters are represented by a large circle. Between these extremes are three ranges where the size of the circle increases with the size of the deck. These three ranges are > 1,000 and <= 5,000 square meters, > 5,000 and <= 15,000 square meters, and > 15,000 and <= 50,000 square meters. This layer provides a means of visually separating bridges by size.

State	Small	Sm/Med	Medium	Med/Lg	Large	Total
Delaware	33	22	2	1	3	61
Maryland	20	36	1	1	1	59
Virginia	90	80	11	3	20	204
Total	143	138	14	5	24	324

Bridges (I-95) - Represents bridges on I-95. Each bridge is represented by a single point. There is no color coding. This layer provides a way to visually locate all bridges on I-95.

State	Total
Delaware	62
Maryland	83
Virginia	304
Total	449

Bridges (All) - Represents all bridges provided by Delaware, Maryland, and Virginia. Each bridge is represented by a single point. There is no color coding. This layer provides a way to visually locate all bridges in the states participating in this study.

Pavement (Mid-Atlantic)

CCI G/F/P (Virginia) - Represents critical condition index using Virginia's criteria for good, fair, and poor. All road segments on I-95 where CCI <> 1 are included. Note that CCI is available only for Virginia. Segments where CCI is < 60 are colored red (poor). Segments where CCI is >= 60 and < 70 are colored yellow (fair). Segments where CCI is >= 70 are colored green (good). This layer provides a way to visualize pavement condition using Virginia criteria.

State	Good	Fair	Poor	Total
Delaware	0 (0 mi)	0 (0 mi)	0 (0 mi)	0 (0 mi)
Maryland	0 (0 mi)	0 (0 mi)	0 (0 mi)	0 (0 mi)
Virginia	117 (228 mi)	27 (50 mi)	40 (72 mi)	184 (350 mi)
Total	117 (228 mi)	27 (50 mi)	40 (72 mi)	184 (350 mi)

OPC G/F/P (Delaware) - Represents overall pavement condition using Delaware's criteria for good, fair, and poor. All road segments on I-95 are included. Note that OPC is available only for Delaware, which provided this value, and Virginia, where this value was computed by CS. Segments where OPC is <= 50 are colored red (poor). Segments where OPC is > 50 and <= 60 are colored yellow (fair). Segments where OPC is > 60 are colored green (good). This layer provides a way to visualize pavement condition using Delaware criteria and to visually compare pavement condition in Virginia using both Virginia and Delaware criteria.

State	Good	Fair	Poor	Total
Delaware	224 (58 mi)	36 (10 mi)	10 (2 mi)	270 (70 mi)
Maryland	0 (0 mi)	0 (0 mi)	0 (0 mi)	0 (0 mi)
Virginia	171 (319 mi)	8 (13 mi)	5 (7 mi)	184 (339 mi)
Total	395 (377 mi)	44 (23 mi)	15 (9 mi)	454 (409 mi)

OPC versus CCI - Represents the absolute value of the difference between CCI, which was supplied by Virginia, and OPC, which was calculated for Virginia using Virginia data and the Delaware algorithm. Both CCI and OPC have values between 0 and 100. All road segments for I-95 are included but this measurement is available only for Virginia. Segments where the difference is <= 13 are colored green. Segments where the difference is > 13 and <= 34 are colored yellow. Segments where the difference is > 34 are colored red. This layer is designed to highlight pavement segments where there is a substantial difference between these two composite measures.

State	Small Diff	Medium Diff	Large Diff	Total
Delaware	0 (0 mi)	0 (0 mi)	0 (0 mi)	0 (0 mi)
Maryland	0 (0 mi)	0 (0 mi)	0 (0 mi)	0 (0 mi)
Virginia	127 (247 mi)	52 (94 mi)	7 (12 mi)	186 (353 mi)
Total	127 (247 mi)	52 (94 mi)	7 (12 mi)	186 (353 mi)

IRI G/F/P (Maryland) - Represents IRI using Maryland's criteria for good, fair, and poor. All road segments on I-95 are included. Segments where IRI is <= 95 are colored green (good). Segments where IRI is > 95 and <= 170 are colored yellow (fair). Segments where IRI is > 170 are colored red (poor). This layer provides a way to visualize pavement condition based on IRI using Maryland criteria.

State	Good	Fair	Poor	Total
Delaware	58 (21 mi)	30 (5 mi)	32 (5 mi)	120 (31 mi)
Maryland	1627 (163 mi)	328 (33 mi)	135 (14 mi)	2090 (210 mi)
Virginia	118 (231 mi)	66 (108 mi)	0 (0 mi)	184 (339 mi)
Total	1803 (415 mi)	424 (146 mi)	167 (19 mi)	2394 (580 mi)

IRI G/F/P (Virginia) - Represents IRI using Virginia's criteria for good, fair, and poor. All road segments on I-95 are included. Segments where IRI is < 100 are colored green (good). Segments where IRI is >= 100 and < 140 are colored yellow (fair). Segments where IRI is >= 140 are colored red (poor). This layer provides a way to visualize pavement condition based on IRI using Virginia criteria and to visually compare pavement condition based on IRI using both Maryland and Virginia criteria.

State	Good	Fair	Poor	Total
Delaware	59 (21 mi)	22 (4 mi)	39 (5 mi)	120 (30 mi)
Maryland	1662 (166 mi)	187 (19 mi)	241 (24 mi)	2090 (209 mi)
Virginia	126 (247 mi)	54 (87 mi)	4 (5 mi)	184 (339 mi)
Total	1847 (434 mi)	263 (110 mi)	284 (34 mi)	2394 (578 mi)

IRI Condition (Maryland) - Represents the IRI Condition Index provided by Maryland. All road segments on I-95 in Maryland are included where the index value is 1, 2, 3, 4, or 5. Segments where the index is 1 are colored green (very good). Segments where the index is 2 are colored blue (good). Segments where the index is 3 are colored yellow (fair). Segments where the index is 4 are colored orange (mediocre). Segments where the index is 5 are colored red (poor). This layer provides a way to visualize another distress condition using Maryland data and ratings.

State	Very Good	Good	Fair	Mediocre	Poor	Total
Delaware	0 (0 mi)	0 (0 mi)	0 (0 mi)	0 (0 mi)	0 (0 mi)	0 (0 mi)
Maryland	804 (80 mi)	820 (82 mi)	338 (34 mi)	86 (9 mi)	50 (5 mi)	2098 (210 mi)
Virginia	0 (0 mi)	0 (0 mi)	0 (0 mi)	0 (0 mi)	0 (0 mi)	0 (0 mi)
Total	804 (80 mi)	820 (82 mi)	338 (34 mi)	86 (9 mi)	50 (5 mi)	2098 (210 mi)

Crack Condition (Maryland) - Represents the Crack Condition Index provided by Maryland. All road segments on I-95 in Maryland are included where the index value is 1, 2, 3, 4, or 5. Segments where the index is 1 are colored green (very good). Segments where the index is 2 are colored blue (good). Segments where the index is 3 are colored yellow (fair). Segments where the index is 4 are colored orange (mediocre). Segments where the index is 5 are colored red (poor). This layer provides a way to visualize another distress condition using Maryland data and ratings.

State	Very Good	Good	Fair	Mediocre	Poor	Total
Delaware	0 (0 mi)	0 (0 mi)	0 (0 mi)	0 (0 mi)	0 (0 mi)	0 (0 mi)
Maryland	1642 (164 mi)	153 (15 mi)	73 (7 mi)	14 (1 mi)	0 (0 mi)	1882 (187 mi)
Virginia	0 (0 mi)	0 (0 mi)	0 (0 mi)	0 (0 mi)	0 (0 mi)	0 (0 mi)
Total	1642 (164 mi)	153 (15 mi)	73 (7 mi)	14 (1 mi)	0 (0 mi)	1882 (187 mi)

Rut Condition (Maryland) - Represents the Rut Condition Index provided by Maryland. All road segments on I-95 in Maryland are included where the index value is 1, 2, or 3. Segments where the index is 1 are colored green (very good). Segments where the index is 2 are colored yellow (fair). Segments where the index is 3 are colored red (poor). This layer provides a way to visualize another distress condition using Maryland data and ratings.

State	Very Good	Fair	Poor	Total
Delaware	0 (0 mi)	0 (0 mi)	0 (0 mi)	0 (0 mi)
Maryland	1754 (175 mi)	85 (9 mi)	258 (26 mi)	2097 (210 mi)
Virginia	0 (0 mi)	0 (0 mi)	0 (0 mi)	0 (0 mi)
Total	1754 (175 mi)	85 (9 mi)	258 (26 mi)	2097 (210 mi)

Friction Condition (Maryland) - Represents the Friction Condition Index provided by Maryland. All road segments on I-95 in Maryland are included where the index value is 1, 2, or 3. Segments where the index is 1 are colored red (poor). Segments where the index is 2 are colored yellow (mediocre). Segments where the index is 3 are colored green (acceptable). This layer provides a way to visualize another distress condition using Maryland data and ratings.

State	Poor	Mediocre	Acceptable	Total
Delaware	0 (0 mi)	0 (0 mi)	0 (0 mi)	0 (0 mi)
Maryland	13 (1 mi)	49 (5 mi)	310 (31 mi)	372 (37 mi)
Virginia	0 (0 mi)	0 (0 mi)	0 (0 mi)	0 (0 mi)
Total	13 (1 mi)	49 (5 mi)	310 (31 mi)	372 (37 mi)

IRI - Represents IRI using a four-level classification system. All road segments on I-95 are included. The classification system uses thicker lines for higher values of IRI. The four IRI ranges are <= 75 (thinnest line), > 75 and <= 150, > 150 and <= 225, and > 225 (thickest line). This layer provides a way to visualize IRI.

State	Best	Med/Best	Med/Worst	Worst	Total
Delaware	33 (13 mi)	52 (12 mi)	28 (4 mi)	7 (1 mi)	120 (30 mi)
Maryland	1322 (132 mi)	561 (56 mi)	163 (16 mi)	44 (4 mi)	2090 (208 mi)
Virginia	53 (116 mi)	129 (220 mi)	2 (3 mi)	0 (0 mi)	184 (339 mi)
Total	1408 (261 mi)	742 (288 mi)	193 (23 mi)	51 (5 mi)	2394 (577 mi)

4.4 Conclusions and Recommendations

The goal of this study was to use pavement and bridge data from a multi-state interstate corridor to analyze performance measures and evaluate how performance data can be used for corridor management. The project team was charged with acquiring data from three states, establishing candidate measures that summarize performance of pavement sections and bridges, and performing any necessary reductions/transformations of the data. In addition to this technical memorandum, the project deliverables include a revised version of ICAT showing pavement and bridge measures for use by FHWA and I-95 Corridor Coalition members.

Conclusions Regarding Current Results

Historically, sufficiency rating for bridge and IRI for pavement have been the measures that are available, either directly or via calculation by FHWA, to judge the condition of the nation's transportation assets. Based on our review of the pavement and bridge data provided by Delaware, Maryland, and Virginia, CS draws the following conclusions regarding these measures.

For pavement, IRI does not provide adequate information to judge overall condition. Composite measures that combine a range of distress types into a single index provide a more accurate condition indicator, regardless of the algorithm used to compute the measure. This conclusion is supported by:
- The extent of the engineering analysis used by Virginia to develop their Critical Condition Index;
- The relatively good correlation between CCI, which was provided by Virginia, and OPC, which was calculated using Virginia data and the Delaware algorithm; and
- The relatively poor correlation between IRI and CCI (in Virginia) and between IRI and OPC (in Delaware).
While Virginia and Delaware (and, presumably, many other states) have defined composite condition measures for pavement that are better than any individual distress rating, it is not clear if these composite measures represent more than the superficial condition of the pavement. Composite measures like CCI and OPC may not correlate well with the structural adequacy of the pavement. CS makes this assumption because the composite measures observed to date consist of weighted combinations of regular distresses (e.g., cracking, rutting, etc.). They do not include more sophisticated readings such as falling weight deflectometer results nor do they take into account the overlay history of the pavement.
Regardless of its value as an overall indicator of condition, IRI will continue to be a valuable measure of ride quality. At a sufficiently high level, any consistent measure can provide useful information about an asset network because analysts can make judgments about relative condition between states, counties, roads, etc., by comparing "apples to apples". There is an open question of whether IRI is captured consistently across states. However, this is a technical issue that can be overcome by ensuring consistent calibration of data collection instrumentation.
While states have different interpretations of what constitutes a good/fair/poor value for IRI, these differences, at least within the limited survey for this project, appear not to be significant. This is supported by a visual comparison within the WebCAT of the IRI good/fair/poor ratings provided by Virginia and Maryland. Although the Maryland standard has a significantly wider range for fair (almost twice as wide as Virginia), the differences between these two visualizations are not dramatic.
For bridges, it is difficult to directly compare sufficiency rating and health index. CS believes that the health index calculation, which is based on detailed element condition information weighted by the relative importance (i.e., failure cost) of each element, provides a superior measure of structural condition. However, health index does not include any measure of functional adequacy or asset essentiality. This is one reason that bridge engineers have not adopted health index more broadly and continue to rely on the traditional SD/FO ratings.
The differences in the purpose of sufficiency rating and health index explain the lower correlations between these two values. At this time, however, there is no justification for choosing one over the other in all circumstances. Depending on the specific area of inquiry, either sufficiency rating or health index may provide a more accurate answer (e.g., studies looking specifically at structural health would benefit more from using the health index).
While the correlations between sufficiency rating and health index did not meet the threshold adopted for this study, they generally were not as low as the correlations between IRI and CCI or OPC. This particularly was true when comparing individual NBI ratings (e.g., deck, superstructure, substructure) with equivalent health index subcomponents. This may reflect a greater degree of uniformity in bridge inspection, which likely is driven by the more comprehensive NBI standard and the near-universal adoption of the Pontis BMS. This also reinforces the previous conclusion that sufficiency rating and health index both are adequate, albeit slightly different, measures of condition.
States generally do not categorize sufficiency rating as good/fair/poor preferring instead to focus on whether or not a bridge qualifies for Federal funding. In the absence of a more compelling standard, it makes sense to coordinate with NBIAS and adopt the standards used by this program, if this type of categorization is needed. This will help ensure consistency when comparing NBI values with standard NBIAS reports.

Recommendations for Alternative Measures

Transportation asset management is a set of guiding principles and best practices for making informed resource allocation decisions and improving accountability for these decisions. Performance measures are a fundamental building block for any asset management effort. Defining these performance measures helps organizations support asset management in three basic ways:

Performance measures can be used to quantify policy goals and objectives in a practical way;
Performance measures help agencies evaluate resource allocation options and determine how to prioritize different investments and/or compare the impact of different funding levels; and
Performance measures provide a quantitative means to measure progress, determine program effectiveness, and chart trends over time.

When considering existing performance measures for bridges and pavement sections, the analysis performed during this project suggests that adequate measures exist for bridges but not for pavement. The following are the findings and recommendations for this analysis.

Finding - CS believes that sufficiency rating and health index both are useful measures for bridges. Between them, they address the key factors of physical condition, functional adequacy, and asset importance. The current NBI standard provides the information necessary for FHWA to compute sufficiency rating in a uniform manner at a national level. However, health index, which provides a better measure of physical condition than does sufficiency rating, is not available outside of an individual state running Pontis.
Recommendation - In order to provide FHWA with access to health index values for the national bridge network, CS recommends that FHWA consider modifying the NBI submission from states to include some element data. The element data could be added to the current NBI file or submitted in a second NBI Element file. Many states already have element data available and, once a new NBI structure has been defined, bridge management systems could be programmed to include this information with little or no additional effort for the states. FHWA would require a process to read/store the element data and a program to compute health index using this information. This Pontis-independent program would be similar to the one currently used to compute sufficiency rating and, like with sufficiency rating, FHWA would provide the calculated health index back to the states. FHWA also would adopt standards for element weights, which would ensure the validity of state-to-state comparisons of health index. Elements usually are weighted by failure cost, although the health index calculation also can use the cost of the most expensive action. However, these costs are not required. A health index can be produced using any weight factor that establishes the relative importance of the elements. This recommendation would be a natural follow-up to the current effort by the AASHTO Subcommittee on Bridges and Structures, Technical Committee T-18 on Bridge Management, Evaluation, and Rehabilitation to redefine bridge elements.
Finding - CS believes that composite values provide a better picture of pavement condition than individual measures like IRI. The HPMS Reassessment 2010+ will add significant new information on pavement distress and history. However, at this time, there is no reason to believe that any of these additional items, if used in isolation, will provide a significantly better measure of condition than IRI.
Recommendation - CS recommends that FHWA use the new HPMS 2010+ information as the foundation for developing one or more Federally approved composite values (e.g., "health indices" for pavement) that can serve as measures of structural adequacy, load-bearing capacity, remaining service life, etc., for pavement sections. As part of this process, CS also recommends that FHWA undertake a more thorough review of composite pavement condition algorithms used by states with an eye toward either adopting one or, more likely, developing a custom algorithm that leverages the existing work but is based on HPMS 2010+ data elements.
Finding - The focus on ride quality may have led some states to emphasize cosmetic treatments that improve the road surface but which do not address underlying structural problems. It is true that surface condition (and ride quality) are key concerns of the traveling public. However, this type of policy may be concealing significant roadway problems that must be addressed in the future.
Recommendation - CS recommends that FHWA, either directly or in partnership with one or more states, test sample road sections and solicit expert engineering advice to determine whether any relationship exists between health indexes computed using current distress values and the actual road condition. The goal will be to assess values such as carrying capacity and remaining service life (i.e., time to next major rehabilitation) and incorporate this information into pavement composite measures described previously. This will turn a composite measure of condition into a true measure of structural adequacy.
Finding - Sufficiency rating includes more than just bridge condition. It also incorporates serviceability and functional obsolescence as well as essentiality for public use. Equivalent measures do not always exist for roads. The Highway Economic Requirements System (HERS) does forecast roadway improvements to address functional inadequacies, but these results are more complex than SD/FO ratings for bridges. Also, HERS only recommends improvements that are justifiable from a benefit/cost perspective, which is not how sufficiency rating works.
Recommendation - In order to address this issue, CS recommends that FHWA define models of functional obsolescence and public importance for roads. These models could borrow from concepts already present in HERS but would be simpler to implement and interpret. CS believes that these measures may not carry the same weight for roads that they do for bridges because the road network is more extensive, has many alternate paths, and is less susceptible to catastrophic failure. However, even if these models are applied only to the interstate network, the information they provide could be used to understand the criticality of the structural adequacy measures discussed previously.
At the same time, FHWA should consider whether benefit/cost criteria similar to those used in HERS should be incorporated into the functional obsolescence component of the sufficiency rating calculation for bridges. This change could make the sufficiency rating a more effective measure.

<< Previous

Contents

Next >>

Asset Management

Evaluation of Highway Performance Measures for a Multi-Study Corridor - A Pilot Study