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
Back to Publication List        
Publication Number:  FHWA-HRT-15-019    Date:  May 2015
Publication Number: FHWA-HRT-15-019
Date: May 2015

 

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

Chapter 7. RECOMMENDATIONS

This study effort started out on the premise that improvements to the existing LTPP approach using ground-based weather observations to describe general climate statistics at its test sites could be used to satisfy current MEPDG climate input requirements and emerging infrastructure research needs. However, in the course of this research, the study team discovered a new source of climate data, MERRA, which provides a continuous hourly estimate of all of the climate-related data based on state-of-the-art global modeling.

The current MEPDG pavement performance models were used as an evaluation tool to compare and contrast the influence of weather histories obtained from onsite AWS, OWS, VWS, and MERRA data sources. In this study, limitations in the current climate data and modeling incorporated in the MEPDG and accompanying AASHTOWare Pavement ME Design® software were also discovered. These findings either suggest that the current MEPDG models are not appropriate for this type of climate data evaluation or that these models can be improved, especially if MERRA is the source of climate data. Regardless, the MEPDG models were chosen as the evaluation tool because they represent the most advanced models in pavement performance forecasting today.

Summary of RECOMMENDATIONS

Based on the results of this phase of the research effort, the study team recommends the following:

  1. The LTPP program should use the MERRA dataset as the basis for continuous hourly climate data histories for its test locations.
  2. Using the MERRA data set, LTPP should calculate the same derived computed climate statistics as shown in table 1.
  3. The LTPP program should not add the TMI as a new computed parameter to the climate module. There is no compelling reason to include it in the LTPP database. It is no longer needed if the MERRA dataset is adopted by LTPP, because MERRA contains the data used to determine TMI.
  4. The CLM module in the LTPP database should be modified to contain MERRA data for the cells where LTPP test sections are currently located.
  5. The existing LTPP CLM data module should be retained in the LTPP Information Management System as archived data so that the data are still available upon request.

The following portions of this chapter provide more details on the basis of these recommendations.

BENEFITS of USING MERRA DATA

MERRA is a new source of weather data for use in pavement and other transportation infrastructure modeling applications. As described in chapter 4, MERRA provides continuous hourly weather data on a relatively fine-grained uniform special grid for 1979 to the present. Most MERRA data elements are fundamental physics-based quantities, many of which are not available from any ground-based or other conventional climate data source. Only a small subset of the MERRA data elements is needed to develop weather inputs for current infrastructure applications such as the MEPDG. The full set of MERRA data elements may enable development of much more powerful infrastructure applications in the future.

MERRA data satisfy all of the major study objectives. They meet the climate data needs for current infrastructure applications such the MEPDG, LTPPBind, HIPERPAV®, and bridge management. The broad range of MERRA data means that it will likely meet the climate data needs for future applications as well. The attention to quality and continuity in MERRA data eliminates the need to deal with temporal changes in position and/or measurement details of OWS histories. The close and uniform spacing of MERRA grid points also eliminate the need for improved weather data interpolation and use of VWS. Lastly, MERRA makes moot the issue of continued location-specific solar radiation measurement, as MERRA provides this information directly at every grid point.

MERRA offers the following benefits compared with conventional ground-based OWS data:

  • Denser, more uniform, and broader spatial coverage. The ASOS network of first-order ground-based weather stations provides data at approximately 1,000 locations in the contiguous United States. These locations are not distributed uniformly across the country but rather are concentrated in areas with high population density (and an airport). Vast areas of the country have sparse or no ASOS coverage. MERRA data, by contrast, are currently available at more than 3,000 grid points in the contiguous United States and worldwide coverage at similar resolution. MERRA grid points are uniformly distributed at a horizontal spacing of approximately 31.1 mi by 37.3 mi. In other words, no point in the continental United States is more than 24.9 mi from the nearest MERRA grid centroid. This nearest grid point distance will become dramatically smaller when MERRA moves to an approximately 0.62 mi by 0.62 mi horizontal grid spacing; NASA is currently using this higher resolution data in-house, and it is expected to be made available to the public within the next few years.
  • Better temporal frequency and continuity. MERRA provides weather data at hourly time intervals as required by current state-of-the-art infrastructure modeling applications such as the MEPDG and HIPERPAV®. Daily, monthly, and/or annual statistics are also available directly from MERRA or can be aggregated from the hourly data. There are no gaps in the MERRA histories as often appear in the AWS data and other OWSs. All MERRA data are referenced to Greenwich Mean Time, eliminating issues of local time conversions and discontinuities in the data at changes to/from Daylight Savings Time.
  • Excellent data consistency and quality. NASA developed MERRA for use in its own modeling applications and satellite retrieval algorithms. To meet these in-house needs, NASA performs rigorous and sophisticated QC checks to ensure that all MERRA data are consistent and correct even as the mix of satellites and other sources of measurement data inevitably change across time and location. LTPP’s need for extensive QC checks such as those in place for the current CLM module will be greatly reduced.
  • Focus on fundamental physical quantities. MERRA data include data elements that are much more relevant to the fundamental inputs required by thermodynamics-based infrastructure modeling than are available from the first-order ASOS data. For example, MERRA directly provides the shortwave radiation fluxes at the top of atmosphere and ground. In the MEPDG, these quantities are estimated using empirical and semi-empirical relationships that are functions of location, time, and percent sunshine category. Given that net shortwave radiation flux at the surface is the primary driver for pavement temperature distributions, the MERRA data are much more suitable. The ready availability of MERRA data will likely foster improvements to current infrastructure modeling applications such as the MEPDG.
  • Richer and more versatile datasets. To meet NASA’s diverse modeling requirements, MERRA reports hundreds of data elements, although not all of these are at the highest temporal and spatial resolutions. Many of these data elements may be useful to future infrastructure and other modeling applications.
  • Potential for automated updates to LTPP database. The process of requesting MERRA data, downloading it from the server, extracting and processing the data elements relevant to LTPP needs, and importing these data into the LTPP database has the potential to be highly automated. This could enable more frequent updates to the CLM module at significantly less cost.
  • Improvement over time. NASA is currently enhancing MERRA to an approximately 1 km spatial resolution. This means that no location will be more than 2,297 ft from the nearest MERRA grid centroid. Significant improvement in conventional ground-based OWS coverage is very unlikely.
  • Reliability analysis capabilities. MERRA is only one, albeit the most comprehensive, retrospective reanalysis system available. Others have been developed in Europe, Japan, and elsewhere. These various modeling applications could be applied simultaneously to develop ensembles of weather histories. Statistical characterization of these ensembles could provide a rational basis for quantifying the uncertainty of predicted infrastructure performance due to the weather inputs.

MERRA does, of course, have some limitations, including the following:

  • Grid points are not at project location. The current version of MERRA will have a grid point centroid no further than about 40 km from any project location. This is generally sufficient for the roughly 70 percent of the United States with relatively flat terrain. This distance may not be sufficient in mountainous regions with isolated microclimates but in general, there will be no OWS that can realistically represent these microclimate weather histories either. The current version of MERRA is thus no worse than the OWSs in this regard, and MERRA will be significantly better after the 1 km horizontal resolution upgrade is released to the public. AWSs can obviously capture future (but not historical) climate conditions at a specific site, but it is unrealistic to expect deployment of these instruments at project sites other than for research purposes.
  • Spatial averaging over grid cell volume. Some MERRA data represent averages over the grid cell volume as opposed to point measures provided by OWSs. However, given the MERRA grid points are often located closer to project sites than any OWS, the question is “Which is worse—spatially averaged values around the vicinity of the site or point values at the wrong point?” This arguable limitation of the MERRA data will become less important as the MERRA grid point resolution increases and associated grid cell volume decreases in the future.
  • MERRA data begin in 1979. For LTPP test sections built before 1979, MERRA will not be able to provide a complete climate history over the entire service life of the project. However, OWS data prior to 1980 are also very sparse, especially at hourly time intervals. MERRA is thus no worse than OWS data in this regard. In addition, it is unclear how important these earlier climate data are to current infrastructure modeling applications.
  • MERRA hourly data storage will require changes to LTPP database. Hourly weather data are required for current state-of-the-art infrastructure modeling applications such as the MEPDG and HIPERPAV®. The LTPP database will therefore need to be restructured to accommodate these data regardless of whether they come from MERRA or OWSs.

In summary, MERRA offers many benefits and very few if any significant limitations for use as the source of climate data for transportation infrastructure modeling applications. It is thus clear that MERRA should be the source for climate data in LTPP.

In addition to providing the weather inputs required for applications such as the MEPDG, the uniformly gridded MERRA data will also make the development of more precise climate zone maps easier. For example, figure 98 shows a LTPP climate zone classification map that was published during the test section recruitment phase of the program. In some places, this map was altered to adhere to State boundaries to make it easier for participating highway agencies to complete the test section nomination form. Climate zones do not follow State boundaries. Unfortunately, this stylized climate zone map has been inappropriately used for other applications; for example this figure appears in the early MEPDG literature. The uniformly gridded MERRA data will make it much easier to develop an updated climate zone map in the style of figure 99 that much more precisely demarcates the zone boundaries.

Figure 98. Map. LTPP climate zone map. This figure is a map of the United States. Each State is outlined but not labeled. This map is divided into four climate zones. The southeastern States and the northern far west States are highlighted in black and labeled as wet-nonfreeze areas. The northeastern States are highlighted in gray and labeled as wet-freeze areas. The north and midwestern States are transparent and labeled as dry-freeze areas. The southwestern States are highlighted in white and labeled as dry-nonfreeze areas.

Figure 98. Map. LTPP climate zone map.

Figure 99. Map. Example of more realistic climate zone map. This figure is a map of the United States. Each State is outlined but not labeled. This map is divided into three temperature zones and three soil conditions. The three temperature zones are A with low temperature, B with freeze–thaw, and C with high temperature. Three soil conditions are I where soil is wet all year around, II where soil is wet for only a part of the year, and III where soil is dry all year around.

Source: M.I. Darter

Figure 99. Map. Example of more realistic climate zone map.(70)

Another potential LTPP product made possible by MERRA is a “weather anywhere” interactive application. Using a graphical map, a user can click a location and the program returns summary climate statistics. The output could be further refined to include inputs tailored to the AASHTOWare Pavement ME Design® software, HIPERPAV®, and other infrastructure modeling software. This tool is currently under production by the project team.

CLIMATE INDICES

Based on the project team’s review of the existing LTPP CLM module, there is no compelling need to eliminate any of the climate indices currently in the database. These indices all have valid use over a range of applications, and their implementation in the LTPP database involves minimal storage and processing.

One specific objective of this project was to examine the calculation and storage of the TMI in the LTPP database. The TMI is a semi-empirical method for classifying the climate of a given location that quantifies the aridity or humidity of a soil-climate system by summing the effects of annual precipitation, potential evapotranspiration, storage deficit, and runoff.(71,64) Inclusion of TMI in the LTPP database was considered early in the program because some existing work on climate zones (e.g., climate zone maps) used it in its formulations. However, TMI has the disadvantage of being a largely empirical index formulated for continental United States conditions and therefore cannot be easily extended to other locations. Zapata et al. suggest a method for extending TMI beyond the continental United States, but the robustness of this method is unclear and its general validity needs further evaluation.(64) The TMI has seen a resurgence of interest in recent years because the MEPDG uses it to determine the equilibrium moisture contents of the unbound pavement materials far above the groundwater table.

The project team has found no compelling reason to incorporate TMI into the LTPP database. The following are the three principal reasons for this conclusion:

  1. Although used currently in the MEPDG, TMI is a calculated internal quantity, not a direct model input. TMI values stored in the LTPP database could not be entered as input to the AASHTOWare Pavement ME Design® software and thus would have little direct usefulness to the MEPDG.
  2. Because TMI is a semi-empirical index with limitations (e.g., its inability to be calculated outside the continental United States), it is unlikely to be used as a direct model input in any future infrastructure modeling application.
  3. In the rare event that TMI values may be desired and assuming that MERRA data are incorporated into the LTPP database, TMI can be calculated on the fly from the appropriate MERRA data elements.

INCORPORATING MERRA DATA INTO THE LTPP DATABASE

If the recommendation to use MERRA data as an additional source of climate data for LTPP test sections is approved, the next consideration is how to add these data to the LTPP database. The MERRA dataset requires adaptation to specific user needs. The following activities and data storage constructs are recommended for adaptation of MERRA data to LTPP needs:

  • Some MERRA data units need to be converted to conventional weather units common to infrastructure uses. For example, precipitation in MERRA is expressed as a flux rate that needs to be translated into more traditional units of “depth” over a prescribed period of time. For example, rain is traditionally measured by inches per hour or day whereas MERRA units of precipitation are in units of mass/(area * time). Wind data are presented in eastward and northward vector components to avoid time-based averaging issues related to wind direction.
  • The MERRA grid data represent the weather data for all test sections included in the grid boundaries. Interpolation of weather data is no longer needed. MERRA in essence performs a physics-based interpolation at each grid cell location.
  • A new table needs to be developed to associate LTPP test site locations to the appropriate MERRA grid cell.
  • Based on previous LTPP experience with weather data, it is recommended that the hourly MERRA data should be parsed into base line datasets containing distinct types of data. While this results in more datasets/tables than originally contained in the MERRA data files, it also provides a rational basis for the higher order climate statistics using existing LTPP database software code.
  • A new LTPP database nomenclature needs to be developed to distinguish between MERRA data and the older OWS/VWS ground-based observations. The nomenclature used in this report is CLM_MERRA_datataype_time, where datatype follows the current climate categories of PRECIP, TEMP, WIND, and HUMIDITY, plus the new SOLAR category for solar radiation related data. Time is HOUR, DAY, MONTH, and ANNUAL.
  • New data can be added to tables. Examples of potential new data additions include the following:
    • Evaporation to the PRECIP tables.
    • Snow to the PRECIP tables.
    • Average soil moisture at various depths to the PRECIP tables.
    • Albedo, surface emissivity, shortwave radiation, longwave radiation, and cloud cover to the new solar radiation tables.

  • MERRA data are delivered in a flat file format. One of the first steps will be to translate the file format into a relational database format. The raw MERRA data will be stored in tables in native units for QC purposes and to make automating updates, extraction, and population of the new hourly climate data tables easier. Three MERRA analysis products are recommended for LTPP, and some of the data of interest in the files include the following:
    • tavg1_2d_slv-Nx—This includes hourly specific humidity and air temperature data.
    • tavg1_2d_flx_Nx—This includes surface evaporation, snowfall, precipitation, total precipitation, and total reevaporation of precipitation.
    • tavg1_2d_rad_Nx—This includes data associated with solar radiation, including albedo, shortwave radiation, longwave radiation, and cloud fraction.
  • The raw MERRA database tables are not envisaged as being included in the LTPP SDR because they are intended to mirror the raw data structure, including field names, obtained from the data source. This will decrease data user support requirements by LTPP because the raw data formats are not in customary units employed for civil engineering infrastructure modeling purposes and use nonintuitive variable names.

Figure 100 shows the conceptual computational and structure for climate data for LTPP based on MERRA data. The process starts with obtaining the MERRA data analysis products, which are delivered in a data file format. In the step that transforms the data file formats to database tables, the MERRA raw hourly data are parsed down to only the grid cells where LTPP test sections are located to save data storage space on the LTPP server because the total size of the three primary files are more than 17 terabytes. The MERRA raw hourly data are then both transformed into customary civil engineering units and split into five climate data categories to populate the CLM_MERRA_datataype_HOUR tables. These tables are recommended for inclusion in the LTPP SDR as are the remaining DAY, MONTH, and ANNUAL tables. After computation of the day tables from the hourly tables, the remaining month and annual tables will contain the climate indices included in the current LTPP CLM tables. Not shown in this figure is the table that will link test sections to MERRA grid cells, because the key fields in the CLM_MERRA_datataype_time tables will be based on MERRA grid cells to save data storage space.

Figure 100. Flowchart. Conceptual computational and data storage structure for MERRA-based LTPP climate data. This flowchart shows the conceptual computational and structure for climate data for Long-Term Pavement Performance based on Modern-Era Retrospective Analysis for Research and Application (MERRA) data. The process starts with obtaining the MERRA data analysis products, which are named tavg1_2d_flx_Nx, tavg1_2d_slv_Nx, and tavg1_2d_rad_Nx. The next step is transforming the data file formats to database tables of CLM_MERRA_FLX, CLM_MERRA_SLV, and CLM_MERRA_RAD. The MERRA raw hourly data are then both transformed into customary civil engineering units and split into five climate data categories to populate the CLM_MERRA_datataype_HOUR tables. The datatypes from left to right are PRECIP, HUMIDITY, TEMP, WIND, and SOLAR. After computation of the DAY tables from the HOUR tables, the MONTH and YEAR tables are computed from DAY tables.

Figure 100. Flowchart. Conceptual computational and data storage structure for MERRA-based LTPP climate data.

 

 

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