Using Random Samples to Assess Roadway Condition Faith Johnson GIS Specialist, NCDOT
North Carolina Highway System 79,600 miles of state maintained roads Interstate: 1,326 Primary: 13,736 Secondary: 64,522 163,500 paved lane miles 106,300 miles of public roads 18,300 structures 13,500 bridges 95.8M square feet bridge deck area 14 Highway Divisions 100 counties 2
Maintenance Condition and Assessment Program (MCAP) Program that uses roadway samples to assess the condition elements of the state system Data is collected on a quarterly cycle Reported annually Results are reported in condition rating score cards Statewide, division, county Interstate, primary, secondary Information is used by divisions to create work plans and to allocate funds 3
Use of MCAP Data Condition reports (scorecards) Infrastructure health index Division maintenance allocation Maintenance and operations planning Legislative report on maintenance needs 4
MCAP Assessment Methodology Random sampling by system Interstates sampled by division Primaries and secondaries sampled by county Sample sections are tenth-mile stretches of the roadway Mobile data collection Motion tablets ArcPad software Statewide orthoimagery Tabular data synchronized with Oracle database Staffing 12 2-person teams (mix of consultants and part-time DOT employees) 2 statewide QA teams Assessment Engineer GIS Support Staff 5
Elements Evaluated Shoulders Lateral Ditches Crossline Pipes Blocked Crossline Pipes Damaged Gutters Blocked Inlets (Blocked or Damaged) Brush & Tree Control Turf Condition Pavement Striping Words & Symbols Pavement Markers Does not include pavement 6
Motion Tablets 7
ArcPad Screen 8
ArcPad Assessment Form 9
MCAP Sampling Framework The network is subdivided into tenth mile sections Random samples selected using a performance-based sampling methodology Uses condition as an indicator of variability Sample sizes increase as score approaches 50% 213 sample pools Interstates sampled by division Primaries sampled by county Secondaries sampled by county Sampling confidence and precision Interstates and Primaries: 90% +/- 5% Secondaries: 80% +/- 5% 10
Sample Size Determination Performance-based sample formula Variables: n sample size c z-value d precision v performance N population Performance scores Per sample pool Based on performance from previous year Sample size increases as score approaches 50% 11
MCAP Sampling Process Original MCAP Sampling Tool Custom in-house tool Samples were generated every three years Involved substantial manual work to configure the samples Wish list for improvements to current sampling process Assign samples to quarters Assign samples as alternates Begin sampling every year Remove bridges Add ramps 12
Solution Develop a random sampling process Use current sampling formula and framework Incorporate improvement wish list Add flexibility for future changes to the design Remove manual involvement required by Assessment Engineer Result was a multi-step process using Python 13
Part 1 Build Tenth Mile Inventory Source data is our LRS Routes file Contains one segment/record per route Contains route information needed to subdivide each road Route Id Beginning Milepost Ending Milepost Create a Route List that simply lists all of the routes 14
Part 1 Build Tenth Mile Inventory LRS Routes The source data is also copied into the Section Events table Will contain the full inventory/population Each record becomes the first tenth mile section for that route in the Section Events table Uses FMEAS and TMEAS fields for the beginning and ending milepost for each section FMEAS (from measure) = beginning milepost of the route TMEAS (to measure) = beginning milepost of the route + 0.1 Section Events 15
Part 1 Build Tenth Mile Inventory Starting with the first route in the route list, an insert cursor creates a new record in the Section Events table Has the same Route Id and other basic information The from measure is set to the to measure of the previous record The to measure is set to the from measure + 0.1 16
Insert another record Part 1 Build Tenth Mile Inventory And another record Etc 17
Part 1 Build Tenth Mile Inventory Continue inserting records until the to measure exceeds the end of the route (ENDMP1) Time to start on the next route! Etc 18
Part 2 Clean Up Fixing the last record Most routes don t end in even tenth mile increments Setting the to measure equal to the end milepost Delete erroneous sections If the section is a common route or a gap Locate and flag sections ineligible for sampling 19
Part 2 Clean Up Sections that are not eligible for samples Bridges Unpaved roads Sections under 500 feet (end of roads) Ramp gores Sections that are not eligible for samples or inventory Non-System 20
Ramp Gores 21
Ramp Classification Ramps are sampled as part of the existing route categories: Interstate, Primary Secondary Each ramp has to be assigned to one of the route categories Classify based on the highest intersecting route category 22
Part 3 Select Random Samples Assign sequential ID numbers to sections in each pool Summarize and plug results into sampling spreadsheet Sample Pool Percentage of sections that are ramps Number of samples needed Additional 20% alternate samples Number of inventory sections Sample formula calculation Range of ID numbers 23
Randomly Selecting Samples Choosing the appropriate function from the random module in Python Random.randrange returns a true set of random numbers Random.sample returns a set of unique numbers Parameters for Random.sample: Top and bottom of range Add 1 to make top inclusive Number of samples Script reads values from the spreadsheet Samples are assigned to assessment cycle quarters 1, 2, 3, 4, 1, 2, 3, 4, etc. 24
Part 4 - Formatting Make samples spatial Add inventory and assessment criteria fields Delete extra fields Assessment Engineer runs the Import Samples tool Easy to use interface Loads the samples into a chosen geodatabase Sets the starting sample number 25
Import Samples Tool 26
Results Project Timeline Development: May, 2014 August 2014 Production: August, 2014 present Limitations Multi-step process Scripts take a long time to run (creates over 1 million records) Benefits More flexibility Ramps included Bridges removed Alternates assigned Samples assigned to quarters Ability to generate new samples annually Less manual work required for the Assessment Engineer 27
What s Next Requests from Assessment Engineer Assign alternates to quarters Increase number of alternates Sample based on least common denominator Other Improvements Increase automation Decrease processing time Further integrate processing that is done outside of Python (spreadsheet calculations) 28
Questions? 29
Faith Johnson GIS Specialist Management Systems and Assessments Unit North Carolina Department of 919 835 8451 office fsjohnson@ncdot.gov 30