Verification of LTPP Virtual Weather Stations Phase I Report: Accuracy and Reliability of Virtual Weather Stations

Task 1: Obtain and Summarize Climatic Data

Verification of LTPP Virtual Weather Stations Phase I Report: Accuracy and Reliability of Virtual Weather Stations

TASK 1: OBTAIN AND SUMMARIZE CLIMATIC DATA

Data Sources

Climatic data needed for the conduct of this project were obtained from various sources and collected into several working databases. Five sources of information were used in this study as follows.

SMP Daily Temperature and Precipitation

The daily temperature and precipitation data for seasonal monitoring sections were taken from the DataPave 2.0 software on CD-ROM. The daily temperature and precipitation data were stored in the SMP_ATEMP_RAIN_DAY table. There were data for 63 SMP sites in this table.

IMS Virtual Weather Estimates in Climatic Module

The monthly and yearly climatic summaries for VWS in the LTPP database were also extracted from DataPave 2.0. The monthly temperature data were stored in table CLM_VWS_TEMP_ MONTH, which contained data for 880 SPS and GPS sites. The daily VWS data, which are kept offline, were obtained from LTPP program customer service (CLM_VWS_DATA_DAILY table).

Automated Weather Stations

Onsite weather stations are installed at several SPS–1, –2, and –8 sites. Data from these weather stations were available in the AWS module of the database included in DataPave 2.0. The daily temperature and precipitation data were extracted from the AWS_DAILY_DATA table, which contained data for 35 SPS sites.

Operating Weather Stations (OWS)

OWS data, stored offline, were obtained from LTPP program customer service and were used in this project (owsdaily.txt file). This file contained daily temperature, precipitation, humidity, and wind data for up to five OWS near each GPS section or SPS site.

National Climatic Data Center Cooperative Data

The weather station data for 1994 through 1996 from the NCDC cooperative program were used to verify daily temperature and precipitation estimates. Data for more than 8000 weather stations existed in this database.

Summary Climatic Databases

After obtaining the needed data files, data were extracted for the desired time period and locations, and were combined into several user-friendly databases for further analysis. Four summary databases were initially developed as explained below.

VWS Daily Database

A database was developed that contained VWS daily data for SMP and AWS sites. The database contained the climatic parameters listed in table 1 (except for the mean daily temperature and snowfall). This database, which contained data for 58 SMP sections (table 2) and 24 AWS sites (table 3) was used for verification of the precision and bias of the VWS data in the LTPP database.

mm – millimeters
m/s – meters/second

NCDC Daily Database

A database was developed from the NCDC cooperative weather station data for daily maximum and minimum temperature and precipitation. This database contained data for more than 8000 weather stations and was used for verification of the accuracy of estimated climatic data and for investigating factors affecting VWS estimates.

SMP Daily Database

This database contained daily SMP climatic data, and VWS estimates and OWS data for up to five nearby weather stations. It included maximum and minimum temperatures and precipitation for 58 SMP sections (table 2). The database was used to verify VWS calculations for SMP sites and to study factors affecting climatic estimates for a site.

AWS Daily Database

A database similar to the SMP daily database was developed for the AWS data. This database had data for 24 sites (table 3) and was used to verify VWS data for sections equipped with AWS.

Data Completeness

The available climatic data in DataPave 2.0 were reviewed to assess the adequacy and completeness of the data, and to identify the areas where the quality and quantity of LTPP climatic data can be improved. This review resulted in identification of missing or incomplete climatic data in the LTPP database.

Seasonal Monitoring Program Climatic Data

Some climatic data are collected as part of the SMP. Air temperature is collected on an hourly basis and is summarized for the day. The daily summary includes mean air temperature and minimum and maximum air temperature. Total daily precipitation is also recorded at the end of the day.

Table 4 presents the number of days of available data for minimum and maximum air temperature and the total precipitation for 68 SMP installations. Some LTPP sections have more than one SMP installation since sensors are reinstalled after each overlay. The installation date for each construction event and the first and last date of reported data (SMP_DATE) are also included in table 4. The following are some general observations regarding SMP data completeness:

• Most SMP installations provided sufficient climatic data for this analysis.
• There typically is more precipitation than air temperature data.
• Some SMP sites were missing significant temperature data (e.g., sections 131031-2, 133019-1, and 390204-1).
• Data for sections 370201-1, 370205-1, 370208-1, and 370212-1 were suspect.
• Data feedback reports outlining the problems in detail were submitted to the LTPP program.

Automated Weather Station Data

Table 5 presents counts of data for 8 different climatic parameters presently collected from 35 AWS installations. The following are some observations on data completeness:

• With few exceptions, the AWS installations provided sufficient data for analysis.
• The precipitation data for section 190100 are completely missing. The windspeed data are also deficient.
• Precipitation data are deficient for section 260100.
• Some of the precipitation and almost half of the wind and solar radiation data are missing for section 260200.
• Precipitation data from 8/14/96 to 8/26/97 are missing for section 530200.

Data marked with an asterisk (*) appears to be unusable.

*Identifies deficient data.

LTPP VWS Estimates in Climatic Module

Monthly data for selected VWS parameters were reviewed for completeness. Different temperature, precipitation, humidity, and windspeed data provided with DataPave 2.0 were considered. Table 6 includes a list of VWS parameters that were reviewed.

GPS Data Counts

The Administration module of the LTPP database contains an Experiment_Section table that is considered the master control section for the database. All test sections are supposed to have entries in the table. The table thus is both a source of data and provides a cross check on data from other areas of the database. For GPS sections, the link between the VWS climatic tables and the Experiment_Section table is fairly complete. Table 7 shows total data counts (in months) for selected parameters listed in table 6. Out of the 790 GPS sections currently in the Experiment_Section table, 788 sections had VWS data; however, data for 2 GPS sections could not be located. These sections (both GPS–1) are listed in table 8. One GPS section (section 481094), also listed in table 9, had VWS estimates, but was not listed in the Experiment_Section table.

Table 7 shows that 788 GPS sections had adequate temperature and precipitation data. There are slightly more data for precipitation than for temperature (211,523 and 211,458 months, respectively). Humidity and windspeed data, on the other hand, were missing for more than 40 percent of the sections. Minimum and maximum humidity were missing for 328 sections, average windspeed was missing for 310 sections, and maximum windspeed was missing for 348 sections. For the sections that had humidity and windspeed data, the average number of records per section was almost half of the temperature and precipitation data. This indicates that these data are incomplete during the data collection period.

SPS Data Counts

For SPS sites, the link between the VWS climatic data and the Experiment_Section table is not direct. The CLM_SITE_VWS_LINK table is designed to provide the link between the VWS data and other tables. However, this table currently only includes data for GPS sections and SPS control sections. Expanding this table to other SPS sections (noncontrol sections and SPS sections that are linked to GPS) can significantly improve the VWS data accessibility for SPS.

Out of the 200 SPS sites that are represented in the Experiment_Section table, 120 sites are linked to GPS through the SPS_GPS_Link table. Data for these sections are the same as the GPS data. Data for seven SPS sites were not located. These sections are listed in table 8. The data counts for the other 74 SPS sites are listed in table 10. There is also one site that is both linked to GPS and has VWS data (VWS_ID 340900 linked to GPS section 341011). SPS sections listed in table 9 had VWS data, but were not listed in the Experiment_Section table. These data were not included in this review since pertinent data could not be located.

As with the GPS data, almost half of these SPS sites were missing humidity and windspeed data. The average months of data per section for these data were about 40 percent less than the temperature data, indicating incomplete coverage.

Data Coverage and Adequacy

Climatic data coverage was calculated from the first and last month of climatic data for a section and shows the maximum number of years of climatic data available for every section. Table 11 shows the distribution of years of climatic data for 862 GPS sections and SPS sites. This table shows that about half of the sections (408 sections) have between 16 and 20 years of climatic data, while the other half of the sections have more than 20 years of data. The most extensive data are for the SPS site (ID 810500), with more than 56 years of data. Two GPS sections (ID 541640 and ID 134111), each with about 13 years of data, have the least data.

Data adequacy was only evaluated to identify the quantity of the missing data in the LTPP database and is not relevant to the analysis of the data. Data adequacy for each parameter is assessed by the percentage of the available data for the period covered. This value is derived by dividing the number of months of available data by the number of months of data coverage. Table 12 includes the distribution of data adequacy for GPS and SPS sections for different climatic parameters. The precipitation data for all 862 sections cover the entire data period (100 percent adequacy) for all parameters. For temperature parameters, 859 sections were 100 percent complete, while 3 sections had data for 95 percent of the covered period. Humidity data for 363 sections and mean windspeed for 342 sections were missing (not available), and the remaining data had less than 80 percent adequacy (no section had data for the entire data coverage). The maximum windspeed was missing for 388 sections; however, about 15 percent of the sections were 100 percent complete, while another 15 percent were 95 percent complete.

Data Quality

The number of weather stations on which VWS estimates are based is related to the quality of the estimated data since better estimates are made from five weather stations rather than a single weather station. This fact was demonstrated by comparing different estimates (refer to the discussions for table 27).

Table 13 shows the average number of contributing weather stations for each VWS table. Data in table 13 show that the temperature and precipitation estimates are generally made based on data from more than the four closest weather stations. The humidity and windspeed estimates, however, are based on data from a single weather station.