Evaluation of Long-Term Pavement Performance (LTPP) Climatic Data for Use in Mechanistic-Empirical Pavement Design Guide (MEPDG) Calibration and Other Pavement Analysis

Chapter 2. LEGACY LTPP APPROACH TO CLIMATE DATA

A brief overview of the current approach for collection, processing, and storage of data to represent climate conditions at each test site is presented in this chapter. This discussion is limited to the methodology to include continuous long-term data stored in the climate module for all test sections because this was the objective of this portion of the study.

Virtual Weather Station Concept

To develop climate statistics to represent site conditions at each test site, LTPP developed the “virtual” weather station (VWS) concept. Because LTPP test sections are rarely located near an OWS, a method was needed to interpolate data from nearby sites over the required duration of the test section life. The methodology LTPP selected was to call the statistical climate data interpolated from nearby OWS a VWS.

The selection of the OWS to form the statistical basis for a VWS is a critical consideration. The objective of the OWS selection process was to identify weather measurement locations that are expected to be representative of conditions at the VWS site and that satisfy the data needs of the intended application. Some of the factors taken into consideration in the LTPP OWS selection process included the following:

• Distance between the OWS and VWS location.
• Elevation difference between the OWS and VWS location.
• Terrain features between OWS and VWS locations. Mountains and large bodies of water can influence temperature, precipitation, humidity, and wind patterns.
• Types of data available from OWS because weather stations vary in recorded data.
• OWS data reporting frequency. Some weather data measurement sources only include daily extremes while others contain hourly or more frequent data.
• Temporal coverage of data from the selected OWS.
• Data QC measures applied to OWS instrumentation.

The OWS selection method used by the LTPP program developed over time. It began with definition of a perfect weather station to describe the climate at LTPP test sections. The “perfect” weather station had to be within 5 mi and include data equivalent to a first order weather station.⁽⁷⁾ In theory, only one perfect weather station is needed to describe the weather events occurring at a pavement test section location that can account for events affecting the performance of the test pavement structure. When it became apparent that no perfect weather stations existed for the LTPP test sections at the start of the program, the following OWS selection process was used based on first-order and cooperative weather stations in the United States. First-order weather stations are those maintained by the National Weather Service (NWS) that operate 24 hours a day with frequent data collection intervals collecting variables such as temperature, precipitation, surface pressure, humidity, wind speed and direction, cloud cover, snow depth, visibility, and solar radiation; cooperative weather stations generally collect only temperature and/or precipitation at less frequent intervals, and collection is performed by volunteers. Equivalent categories of weather stations were also applied to available data from Canada. The following OWS selection process used by the LTPP program emerged over time:

1. Using automated methods, develop a list of candidate OWSs based on the following rules:
1. Identify at least one active first-order weather station with at least 50-percent coverage of the desired time length.
2. Identify the closest active cooperative weather stations satisfying the following criteria:
1. Has at least 50-percent temporal coverage over the desired coverage period.
2. Has temporal coverage greater than or equal to pavement age or 5 years.
3. Contains the following mandatory data elements: minimum and maximum daily air temperature, daily precipitation, and daily snowfall (where applicable).
3. Identify at least three other nearest active or inactive weather stations that provide data over the coverage time period.
2. Submit the list of candidate OWSs to LTPP regional contractors for further evaluation and development of recommendations on final selection of up to five weather stations. The following factors were taken into account by LTPP regional contractors in developing recommendations on the selected OWS:
1. Correction of test section location coordinates when necessary. When LTPP started, the use of Global Position Satellite Receiver (GPSR) technology was limited, and many of the initial pavement test section coordinates were not correct. A manual plotting of test section coordinates on a suitable paper map and comparison to other information on the known location of a test section was the best available method to confirm test section location coordinates.
2. Representativeness of the selected active first-order weather stations based on distance from the test section relative to elevation differences, mountains, large lakes, and other terrain features. The regional contractors were encouraged to contact State climatologists to obtain input on known micro-climate effects and other known OWS anomalies in the area. In some cases, it was not possible to associate a first-order weather station with a LTPP test section site.
3. Representativeness of the selected active cooperative weather stations meeting the criteria listed under A.2 above, using the same criteria mentioned above for first-order weather stations.
4. Representativeness of the other candidate OWSs.
3. Conduct a centralized evaluation of the recommended OWS list. Initial evaluation was based on reasonableness considerations on the basis for the recommended OWS, and secondarily on results of analysis of data obtained from the final list of selected OWS sites.
4. Repeat the OWS selection process over time as new test sections and test sites are added. In most instances, because new test sections in the SPS were added at existing LTPP test section sites, the same OWS previously selected could be used to generate the VWS statistics for the location. For the new pavement construction SPS experiments, i.e., SPS-1, SPS-2, and SPS-8, using the original concept of perfect weather stations, automated weather stations (AWS) were installed and operated by the LTPP program at or near the SPS projects.

As it turns out, based on LTPP experience, there is no such thing as a perfect weather station. As is subsequently explained, all weather measurement instrumentation requires continual monitoring, evaluation, and maintenance to provide reasonable data. This is compounded by secondary issues related to time conventions, equipment calibrations, and changes in measurement equipment over time.

Using the selected representative OWSs, LTPP used a VWS interpolation methodology based on a 1/R² “gravity” model for combining data from up to five nearby weather stations. As illustrated conceptually in figure 1, the interpolated value for data element V for day m at the VWS location is determined from the data elements V_mi recorded at each of the k OWSs on day m as seen in figure 2 in which R_i is the distance of weather station i from the VWS location at the project site. In figure 1, OWS₃ is highlighted to represent the closest first-order weather station. As subsequently explained, 95 percent of the daily VWS weather statistics for wind and humidity are based on only one OWS, which is typically a first-order weather station.

The LTPP VWS interpolation methodology does not directly account for elevation differences between the measurement locations and project site. This is the reason that the selection criteria for the OWSs included a limit to the difference in elevation to the VWS site. (Note: The VWS algorithm in the MEPDG is very similar to that used by LTPP except that it does include an elevation correction using the temperature lapse rate.)

Data Processing and Storage

The LTPP climate data obtained from external sources are stored in the CLM module in the pavement performance database (PPDB). A two-tiered data storage structure is used. The first tier contains raw and processed data from OWSs selected for use in computing the second tier VWS statistics.

Figure 1. Illustration. Illustration of VWS concept.

Figure 2. Equation. Gravity model equation.

All of the table names in the CLM module start with CLM as the first three-letter prefix. In the table relationship figures in this part of the report, the CLM prefix has been omitted for presentation convenience.

Figure 3 illustrates the organization of the CLM_OWS tables contained in the PPDB. Only the CLM_OWS_LOCATION table is distributed as a part of the LTPP Standard Data Release (SDR). All of the other CLM_OWS_* tables shown in this figure are stored centrally on the LTPP PPDB database server and are disseminated by request only. The data obtained from the United States and Canadian climate data sources, National Climate Data Center (NCDC) and Canadian Centre for Climate (CCC), respectively, are split into four data types and stored by time period, i.e., daily, monthly, and annual. This format was implemented in 2004 as part of the 2005 data release to address data discrepancies found in the raw data from NCDC.

Figure 3. Illustration. LTPP parsing, QC checks, and data flow relationships among the CLM_OWS* tables obtained from external data sources.

Only the OWS_LOCATION and OWS_data-type_DAILY tables are subjected to automated LTPP QC checks because they are used in the computation process for the CLM_VWS tables contained in the SDR. The four data types include precipitation, temperature, wind, and humidity. The OWS monthly and annual tables are not subject to LTPP QC checks. These tables have been used in the past by the LTPP engineering staff to perform reasonableness checks on the results of the VWS monthly and annual tables. In essence, the monthly computations from the OWS sources are compared with those computed from the VWS daily data when a new data upload is performed.

Figure 4 illustrates the CLM VWS computational structure relationships. OWSs are linked to pavement test sections through a process that considers their locations relative to the test section locations. The results of this selection process are contained in the VWS_OWS_LINK table, which controls which OWS data are included in each VWS statistic. The SPS_GPS_LINK and SITE_VWS_LINK tables are used to associate CLM_VWS statistics with collocated test sections at project sites with more than one test section. The dashed arrows in figure 4 show the hierarchical statistical summaries based on level E data from the underlying temporal statistics. Level E means that the data have passed all of the automated LTPP QC checks.

Figure 4. Illustration. LTPP CLM VWS computational relationship structure.

After the VWS daily tables are created, the VWS monthly tables are computed. The monthly tables are computed using daily data that have passed all of the daily data QC checks. In addition to the checks on the daily tables, the monthly data table calculations are subjected to QC checks on the number of valid days in each month’s daily data. Likewise, annual statistics are based on the monthly statistics and subjected to QC checks related to the number of valid days in the year for which data for each data type is available.

The data in CLM_VWS_* tables contained in the SDR includes all data at all levels of RECORD_STATUS. While only level E data are used to compute the higher-level temporal aggregation statistics, data failing the QC checks were retained for research purposes.

The derived monthly and annual climate statistics and indices computed from the daily data are shown in table 1. In addition to common statistical descriptive measures, the LTPP CLM module includes the following climate indices:

• Freeze index.
• Freeze–thaw cycles
• Intense precipitation days.
• Number of wet days.
• Number of days above freezing.
• Number of days below freezing.

Table 1. Derived climatic data statistics and computed indices stored in the LTPP database.

Issues with Legacy LTPP Climate Data Approach

The legacy LTPP program approach to general climate statistics for test sections dates back to technology available in 1991. Over time, LTPP has pursued an active and informed approach to provision of climate data statistics that has altered and adapted to changing technology. The following are some difficulties, limitations, and issues with this approach that the program has attempted to deal with over the last 20 years and that represent future challenges to its legacy approach:

• Limited spatial and temporal coverage of ground-based OWSs, close enough to test sections to be used as part of the VWS computation process, result in an uneven number of OWSs by data type. Figure 5 shows the number of OWS that provide data for a LTPP VWS by data type. The vertical scale is percentage of total daily OWS observations in each VWS climate data category. The horizontal scale is the number of OWSs used in each daily VWS statistic. When wind and humidity data are available, 95 percent of the VWS daily data come from only one OWS, whereas less than 10 percent of the VWS daily data for temperature and precipitation come from one OWS. At the other extreme, almost 50 percent of the VWS statistics for temperature and precipitation are based on the maximum of five OWSs. This situation creates an uneven balance in the basis of the VWS computed parameters for each study site.
• Selected OWSs contain data gaps or do not provide temporal coverage over the desired timeframe.
• Existing OWSs change location over time while keeping the same name. This requires changes to the current LTPP CLM data storage and computational structures to accommodate.
• If all of the selected OWSs do not contain a desired data element, the resulting VWS climate data statistic is left null or a record is not included in the LTPP database.
• New OWSs are introduced over time. To proceed with the legacy LTPP climate approach requires a new OWS selection process. While technological advances have been made to make this process more streamlined than in the past, the OWS selection process requires expenditures of significant program resources.

Figure 5. Graph. Percentage of OWSs represented in each daily VWS statistic by number of OWSs.

• The MEPDG software requires hourly climate inputs. The LTPP climate database contains only daily climate extremes.
• The existing onsite hourly weather measurements collected by LTPP as part of the AWS effort contain significant data gaps due to equipment failure and thus do not provide continuous coverage over the operational time period. This issue is the reason that the U.S. Climate Reference Network (USCRN) includes redundant measurement instrumentation at each site.
• No adjustments are made in the interpolation process to account for elevation differences between the OWS and VWS.

These issues are not criticisms of the LTPP program; they merely reflect the issues and challenges that all infrastructure research projects requiring sit- specific climate data have faced over the last 20 years. Even today, using ground-based weather data observations has some of these same issues.

The next chapter of this report contains a summary of climate data sources that are available to meet emerging needs of the modern generation of infrastructure design and performance prediction engineering models.