1	Mobile Vehicle Road and Weather Observation Quality Check Methods September 7, 2011 Dan Koller daniel.koller@und.edu Surface Transportation Weather Research Center University of North Dakota
2	Outline Motivation for developing quality checks for maintenance trucks Development of the quality check tests Case Studies Results Summary Next Steps
3	Motivation Current road weather observations are in static locations leaving data gaps in between RWIS. Many maintenance trucks have been equipped with Mobile Data Collection and Automatic Vehicle Location (MDC/AVL) units that collect data. The shortcoming of these data is the unverified accuracy of the received data.
4	Data Collected from MDC/AVL Vehicles Bolded items are used in the quality checking algorithm.
5	Quality Check Comparison
6	Quality Check Sequence The quality check algorithm begins with primary tests. If they pass then secondary tests are performed.
7	Primary Tests Speed Test
8	Primary Tests (Cont.) Gross QC Test
9	Secondary Tests
10	Truck-to-Truck Spatial Test
11	IQR Test For RWIS
12	Barnes Spatial Test for RWIS
13	Barnes Spatial Analysis The Barnes spatial test uses neighboring observations and weights them based on their distance from the target sensor. The weights from the neighboring observations drop exponentially as the distance from the target increases. Observations outside of the radius of influence receive a weight of zero.
14	Persistence Test
15	Test Cases Black Hills, SD Cases Dec 30, 2010 - Jan 1, 2011 January 15, 2011 February 24, 2011 March 8, 2011 March 26, 2011 Sisseton Moraine, SD Dec 30, 2010 - Jan 2, 2011 February 2-3, 2011 February 8-9, 2011 February 13-14, 2011 February 17-18, 2011 Eastern ND Cases: November 29-30, 2010 Dec 30, 2010 - Jan 1, 2011 March 11-12, 2011 March 22-23, 2011 April 15-16, 2011 St. Cloud, MN Cases November 22, 2010 December 11, 2010 February 20-22, 2011 March 22-23, 2011 April 20, 2011
16	Cases Location
17	Case March 22-23, 2011 Focuses on Eastern ND and St. Cloud, MN area Trucks that were processed include: MN-AT-205569, MN-AT-206572, MN-AT-208503, MN-AT-208562, MN-AT-208563, MN-AT-209507 ND-9303, ND-9311, ND-9372, ND-9519, ND-9644, ND-9757, ND-9784 Trucks ND-9372 and MN-AT-208562 show a sample of some results.
18	ND-9372 Results
19	ND-9372 Error Counts
20	Notes for ND-9372 ND-9372 experienced an issued during the snow event. At 12:20UTC (6:20 am CST)on March 23 the sensors “got stuck” at a 32.2 F for Air temperature and 52.8 F for Pavement Temperature for 40 minutes. At 13 UTC (7am CST) on March 23 those values switched over to 0 F for both of the sensors until the end of the run at 19UTC (1pm CST).
21	MN-AT-208562 Results
22	MN-AT-208562 Error Counts
23	Notes for MN-AT-208562 MN-AT-208562 shows that the tests were able to complete without any errors. The pavement sensor compared well against surrounding stations and trucks. The air temperature senor on board did not fair as well. Reported temperatures were typically 5-20 F degrees warmer than surrounding observations.
24	Summary Amount of included data Frequency of GPS Data VS. Observation data Timing of data Data collection from third party data is delayed Data “Getting Stuck” at 0^oF Significant figures in data (xxx.xxx F or xxx F) Missing observations for comparison and/or differentiation between surface observation types No pavement/surface temperature sensors installed Missing “reference locations” i.e. bridge or roadway Limitations Post or Real time analysis.
25	Next Steps Study the quality check algorithm against other trucks and other wintertime events. Determine alternative way to run the quality checks to improve algorithm performance for high volumes of mobile observations.
26	Acknowledgments UND collaborators Matt Clegg Jennifer Hershey Damon Grabow Thesis Advisor: Prof. Leon Osborne Data contributions for the study provided by: North Dakota DOT Minnesota DOT South Dakota DOT
27	Questions?