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

Conclusions

During this project, the LTPP virtual weather data for SMP- and AWS-instrumented sites were compared with the measured data. Different methods of developing climatic estimates for the SMP, AWS, and NCDC weather stations were considered and evaluated. Factors affecting the precision and bias of the climatic estimates were investigated. The following are the major findings of this project:

• Daily, monthly, and yearly climatic estimates for anywhere in North America can be developed with reasonable accuracy.
• Using several nearby weather stations for estimating the climatic parameters for a project site provides more accurate estimates than using the closest weather station.
• A simple average of the data (without a weight) for the five closest weather stations provides the best overall estimate for temperature. Weighting the average by the inverse distance provides closer estimates than weighting by the inverse square distance.
• Elevation differences (between the project site and the nearby weather station) of more than 250 m significantly affect the climatic estimates. In this case, temperatures could be corrected to reduce the bias of the estimate. A relationship was proposed for correcting the maximum temperature by the elevation difference.
• The estimated climatic parameters are not severely affected by either the distance or the latitude (north-south distance) of the contributing weather stations within the 60-km range.
• The variation of the measured climatic data (calculated from two or more instruments in one location) was around 40 percent of the variation of the estimated data from the five nearby weather stations.
• Significant year-to-year variation in climatic data was observed. The year-to-year variation (COV) of the yearly precipitation was 21 percent and the FI was 34 percent. On average, the year-to-year variation of the monthly temperature data was 6 percent.
• Review of the VWS data for completeness showed that the quality and quantity of the windspeed and humidity data are significantly lower than the temperature and precipitation data.
• Several data anomalies in VWS, SMP, and AWS data were found and reported.
• LTPP VWS estimates were found to be within a reasonable range, considering the within-site variation of the measured data and also the year-to-year variation of the data.