1

Site-specific Wind Data Acquisition and Analysis with Geospatial Mobile TestingTechnology

Feng Chen1, Suren Chen2, Juhua Liu3 and Jun Wu4

1 Graduate Research Assistant of Colorado State University, Fort Collins, Colorado, USA, feng.chen@colostate.edu2 Assistant Professor of Colorado State University, Fort Collins, Colorado, USA, suren.chen@colostate.edu

3 Research Scientist of Colorado State University, Fort Collins, Colorado, USA, juhua.liu@colostate.edu4 Graduate Research Assistant of Colorado State University, Fort Collins, Colorado, USA,jun.wu@colostate.edu

ABSTRACT

Traffic safety of high-sided vehicles under strong wind gust, other adverse environmental andtopographical conditions has been a pressing issue for modern highway transportation and economy. Arational traffic safety assessment relies on accurate and specific data of wind, terrain and otherenvironmental characteristics. It is known the actual wind loading at the typical height of vehicles variessignificantly from one location to another even on the same highway at the same time simply because ofthe site-specific terrain and surroundings. As a result, the actual measurements by nearby weatherstations, historical records or simulations from existing wind spectrums can only provide generic,approximated and scattered wind data in the whole area. Therefore, accurate crosswind velocity data inboth time and spatial domains are needed, for a rational assessment of traffic safety risks for variousmoving vehicles on highways in windy conditions. Challenges remain on how to get more accurate, site-specific and continuous wind data along the highway for traffic safety assessment as well as highwayenvironmental evaluations. In addition to site-specific wind data which can be used for most vehicles,vehicle-specific crosswind velocity is often required for an accurate safety assessment of high-sidedvehicles with unique shapes. A mobile measurement strategy is developed to collect geospatial wind andother environmental data on a typical highway by integrating advanced wind measurement and globalpositioning system (GPS) techniques. Such technology integrates a 3-D sonic anemometer and geospatialvideo mapping system, mounted on a vehicle driven along highways at a normal (cruising) speed. As aresult, both vehicle-specific and general site-specific crosswind velocity can be directly “sensed” andcollected by using a high-sided vehicle or a streamlined car as the test vehicle. A field test of thedeveloped technology with a high-sided truck driven on mountainous sections of the interstate I-70 (inColorado) was conducted. The crosswind data at six selected feature locations along I-70, representingdifferent roadside environments, was analyzed. Wind-tunnel investigations employing the scaled modelsof the truck used in the field test as well as a common streamlined sedan car were conducted to evaluatethe accuracy and the feasibility of the developed technology.

1. Introduction

Each year, adverse weather alone is associated with more than 1.5 million vehicular crashes, whichresult in 800,000 injuries and 7,000 fatalities nationwide (The National Academies, 2006). Crashes due toadverse natural environments usually cause serious traffic congestion and further deteriorate drivingconditions on highways. For example, vehicle accidents by strong crosswind gust have been frequentlyreported around the country (Brassfield and Allison, 2001; U.S. Department of Transportation, 2003;Willett, 2005). Over the past decade, a number of researchers have been working on safety assessmentunder crosswind for high-sided commercial trucks (Baker, 1991; Baker, 1999; Chen and Cai, 2004;Snaebjornsson et al., 2007) and fire trucks (Pinelli et al., 2004). Recently, Chen et al. (2009) studied thesingle-vehicle crash risk assessment under adverse environmental conditions, including crosswind,inclement weather, complex terrain, and adverse driving manners. Most of the existing studies wereanalytical works. In these studies, the wind velocity data was typically obtained from simulations based

Correspondence author.

2

on empirical spectra formulated using the wind data measured at fixed locations (e.g. Simiu and Scanlan,1998). It is well known that the wind velocity at the typical height of vehicles varies significantly fromone highway to another, due to the differences in roadside environments and surrounding terrain. Even fordifferent segments on the same highway, the actual wind environments at the same time can beconsiderably different due to topographic effects, for example, highway turns, nearby mountains andtrees. Therefore, site-specific wind velocity data along any particular highway is desired for realistictraffic safety assessments of various vehicles passing through everyday. In addition to general site-specific wind velocity data which can be applied to most vehicles with common and streamlined shapes,vehicle-specific wind velocity data is often required in order to conduct a reasonable assessment of thosehigh-sided vehicles with unique shapes.

In principle, site-specific wind velocity data can be obtained from weather stations located close tothe highway. However, it is well known that wind data between two weather stations is usually notavailable. With a typically scattered distribution of weather stations in proximity to the highway, onlygeneric, approximated and scattered wind data at several fixed points along a highway are available.Thus obtaining accurate, site-specific and continuous wind data is challenging, as appropriate accuracy isessential for a reliable safety assessment of various vehicles moving along a specific highway. Therefore,to collect both site-specific and vehicle-specific wind velocity data, the best way is probably to conductfield testing using the actual vehicle as a full-size moving “sensor”.

A limited number of studies on field wind data collection at the typical height of vehicles have beenreported in literature. Pinelli et al. (2004) measured static wind pressure on a parking fire truck. Schmidlin(1998) investigated in full scale the impacts from strong wind, caused by a tornado. Snaebjornsson et al.(2007) conducted the wind velocity measurement using an anemometer attached to a minivan, butdetailed results and further applications were not reported. In the present study, a mobile wind velocitymapping technique is developed to collect site-specific as well as vehicle-specific wind velocity data inboth time and spatial domains. A test vehicle equipped with the advanced 3-D sonic anemometer and thegeospatial video mapping system is employed. The developed mobile testing technique can be used to: 1)generate site-specific wind velocity data along any highway from the measurements for vehicles withcommon and streamlined shapes; and 2) directly measure vehicle-specific wind velocity for thosevehicles with unique and high-sided shapes. A field test was conducted on the interstate I-70 corridor inColorado to prove the idea and demonstrate the technology. In order to evaluate the feasibility and theaccuracy of the introduced technology, a wind tunnel investigation was conducted and the scaled modelof the test vehicle used in the field test, as well as a common streamlined sedan car, was employed in thelaboratory testing.

2. Geospatial Wind Mobile Field Testing

2.1. Mobile Testing Equipment

(a) 3-D anemometer (b) VMS system (c) AccelerometerFig. 1. Geospatial field testing equipment installed on testing vehicle

2.1.1. Sonic anemometer

3

To accurately measure time-dependent wind velocity, including wind speed and direction, anultrasonic 3-D anemometer manufactured by R. M. Young’s Inc. was adopted (Fig. 1(a)). Theanemometer can measure wind speed from 0 through 90 mph (40 m/s), with a resolution of 0.022 mph(0.01 m/s). Wind direction ranging from 0 through 360 degrees can be covered. The anemometer hasbeen calibrated in the wind tunnel by the manufacturer, before its use in the field test. Digital output ofthe measurements is acquired by a laptop computer, via a serial RS-232 connection. The data samplingfrequency is 15 Hz. The acquired data includes turbulent wind speed components in u (longitudinal), v(lateral) and w (vertical) directions. The anemometer was installed with the “north direction” (ucomponent) aligned with the longitudinal axis of the vehicle.

2.1.2. Geospatial video mapping systemThe geospatial video mapping system consists of two major components: a navigation system and

mapping sensors (Fig. 1(b)). The navigation system, typically a GPS receiver combined with an inertialnavigation system or laser range finder, is capable of continuously determining the position informationof the system. Mapping sensors include digital cameras, video cameras, and audio devices. The videomapping system (VMS 300) developed by Red Hen System Inc. (Red Hen System Inc., 2005) is used tocollect the geospatial multimedia information on what a driver can actually see in the front (Fig. 1(b)).The GPS coordinates as well as the time stamp are recorded by the digital camcorder continuously on onechannel of the audio track of the videotape. By using the accompanying software (VMS MediaMapper) aswell as the VMS 300 unit, the captured video can be played back for indexing during the transfer of theGPS data from the videotape to a computer, during the data processing.

During the test, the SONY digital video camera and VMS 300 were mounted behind the wind shieldon the front passenger side. The zooming of the camcorder was adjusted to ensure the almost samescenery that the driver sees is captured, such as the view of the highway and roadside features (e.g. speedlimit and other road signs, roadside trees and bushes). The GPS receiver antenna was mounted on the topof the driving cab outside of the vehicle in order to receive signals from up to eight satellites,simultaneously. The sampling frequency of the GPS receiver was 1 Hz. During the test, the wind data wasrecorded on the laptop. The geospatial video information was recorded on a video tape (of the VMS) andwas subsequently processed and transferred to the computer. Time stamps were utilized to synchronizethe data originating from various measurement equipments.

2.2. Test vehicle and site

The test truck was a GMC SAVANA G3500 16' Truck. The primary parameters of this truck include:117 square feet of floor space, 800 cubic feet of loading space, 2,700 lb. load capacity and the interiordimensions of 15'3"L x 7'8"W x 6'2"H. Figure 2 shows the truck with the anemometer installed above thetop outside surface of the truck. The adoption of the high-sided vulnerable truck as the test vehicle servestwo purposes: (1) to collect general site-specific crosswind data along the highway; and (2) to acquire thevehicle-specific crosswind velocity associated with this particular high-sided truck at the same time.

The Interstate I-70 Mountain Corridor, from Denver to Grand Junction, is an important interstate highway inColorado and it is well known for complicated terrain and severe snowstorms during winter seasons. ForColorado residents, visitors and businesses, I-70 is a gateway to recreation, commerce and everydaynecessities. At some locations along the corridor, steep grades and curves, coupled with extreme weatherconditions, pose serious safety threats on passing vehicles, especially high-sided trucks. Congestion caused byhigh volume of traffic and number of accidents has caused significant economic and societal impacts in the pastdecades. In the present study, I-70 is selected to demonstrate the proposed measurement technology for assessmentof wind and topographic effects on traffic safety. Two routes were selected during the field test on theinterstate I-70 between Exits 252 and 266: one on I-70 W (7.3 miles) and another on I-70 E (15 miles),which are marked on the GIS highway map in Fig. 4.

4

Fig. 2. The test truck with equipments

Fig. 3. Wind velocity interpretation of the introduced technology

x

Yy

X

v

u

VD

North (driving direction)

X

Y

v

u

(a) Wind speed components in moving coordinate system

(b) Wind speed componentsin fixed coordinate system

Z(z)

Z

w w

5

2.3. Vehicle-specific and site-specific wind velocities

As shown in Fig. 3, the fixed Cartesian coordinate system XYZ was used to define the general windenvironment on the highway and this reference frame is not specific to the test vehicle. The wind velocity in thefixed coordinate system (on the ground) can be separated into components v, u and w, in the X, Y and Zdirections, respectively. The coordinate system denoted as xyz in Fig. 3 is the moving coordinate system attachedto the test vehicle, and the wind velocity components in this system are v, u and w, respectively in the x, y and zdirections. As discussed earlier, the anemometer installed on the top of the test vehicle is positioned by aligningthe “north direction” of the anemometer with the longitudinal axis of the vehicle (Fig. 3(a)). The wind velocitymeasurements by the anemometer can be separated as vm, um and wm in its “east”, “north” and “vertical”directions, respectively.

Most highways have slopes and camber angles, and these angles usually do not have considerable impact oncrosswind velocity measurements for assessments of traffic safety due to their relative small values. If it isapproximated that the X-Y plane is always horizontal and the Z axis is parallel to z axis, the w components of thewind velocities in both the coordinate systems are approximately the same, namely mw w w . Themeasured wind velocity component along the “north” direction um is actually equals to u+VD and the measuredcomponent in the “east” direction of the anemometer is vm= v. The moving coordinate system xyz is ideal fortraffic safety study, as it directly gives vehicle-specific wind velocity components (u, v and w) applying on amoving vehicle which is used to quantify wind loading on the vehicle.

Obviously the vehicle-specific wind velocity defined based on the moving coordinate system is dependentnot only on the environment of the highway, but also the specific shapes, and steering angles of the test vehicle.In order to provide spatially continuous general wind velocity data which can be used for various vehicles driventhrough the same highway, site-specific wind velocity components (u, v, w), which are dependent on theenvironment of the specific highway but with little or no dependence on the specific test vehicle being used, areoften needed,. The wind velocity measurements by the anemometer on the moving coordinate system can beeasily converted to the general site-specific wind velocity data in the fixed coordinate system through followingformulas (Fig. 3(b)):

m D mu u V cos v sin (1)

m D mv u V sin v cos (2)

mw w (3)where VD is the vehicle velocity; is the angle between the driving direction and the absolute Y direction.Both the driving speed VD and the can be obtained from the GPS data provided by the VMS installedon the top of the vehicle.

3. Wind data analysis

The geospatial wind data and multimedia information were collected for the total of 22 miles, duringthe field test carried out on I-70. Among all the data, six feature points (FP) along the tested routes wereselected as representatives of typical traffic scenarios: a sharp turn, driving on a ramp in the presence ofrelatively strong wind, passing a large truck on a straight road as well as on a curve and under a bridge.The detailed information on these feature points, including their coordinates and features, are presented inTable 1. These feature points (FP) are labeled 1 through 6 on the GIS-based map in Fig. 4. With thegeospatial information of the points, the geo-referenced data can be easily integrated into a geographicinformation system (GIS) database using ArcMap or Google Earth. Since crosswind is of the primaryconcern for traffic safety studies under windy conditions, only the crosswind speed component in themoving coordinate v direction is discussed hereafter.

6

Table 1. GPS data for specified feature points

FeaturePoint LON LAT ALT Speed

(m/s)Course(degree) Description

No.1 -105.20767472 39.69651722 1980.590 19.190 220.600 Turning right

No.2 -105.27344278 39.70567333 2312.070 23.740 280.400 Driving by a large truckNo.3 -105.29392167 39.71006000 2381.040 26.120 286.200 Under a bridgeNo.4 -105.32594833 39.70508444 2386.330 19.490 235.900 On rampNo.5 -105.28405139 39.70888139 2334.380 27.860 97.000 Turning right

No.6 -105.25537472 39.70442472 2209.310 25.860 82.600 Driving by a large truckand curving

Fig. 4. Testing site on I-70 and selected feature points

3.1. Time Histories of Wind Data

Time histories of the crosswind speed measured at the six feature points are shown in Figs. 5-6. In Fig.4, the time histories of wind speed measured at FP-1, 2 and 3 are displayed from the top to the bottom ofthe figure, while the time histories for FP-4, 5 and 6 are shown in Fig. 6. For each feature point,corresponding time duration has been identified to describe each event according to the geo-referencedvideo clips, for example, a curving sequence. Respective time period for each feature point is marked bytwo vertical black lines as shown in each figure, representing the starting and the finishing time of theevent, respectively. Two still pictures are extracted from the video clips, visualizing driving conditions forthe starting and the finishing time instant. The GPS coordinates and the time stamps are shown in eachpicture. As a result, the actual surrounding information and the corresponding spatial position on thehighway can be linked with each time history of wind speed measurements.

7

Fig. 5. Geospatial crosswind speed time histories for FP-1, 2 & 3

8

Fig. 6. Geospatial crosswind speed time histories for FP-4, 5 & 6

9

3.2. Spectral characteristics of measured wind velocity

Fig. 7. HHT Analysis of crosswind speed for FP-1

Hilbert-Huang Transform (HHT) analysis is often used to disclose time-variant spectralcharacteristics of time histories of signal (Huang et al. 1998). Fig. 7 shows the HHT analysis results forfeature point 1: FP-1. In Fig. 6, IMF 1-6 illustrate the turbulent components of wind speed at differentfrequencies. For FP-1 (turning right), the raw time history of wind speed has the overall mean andstandard deviation of -1.7 m/s and 1.2 m/s, respectively. The HHT results in Fig. 7 suggest that theturbulent components of wind velocity with high frequencies are relatively weak (with a magnitude lessthan 1m/s) compared to the low frequency components. The residual plot in Fig. 7, which is essentiallythe mean wind speed, varies significantly from 0 to -4m/s within 22 seconds, the duration of the wholecurving process of the vehicle. The considerable time-dependent mean wind speed at FP-1 suggests thatthe crosswind speed measured at FP-1 is a nonstationary random process. It can be found that thenonstationarity of the crosswind speed at FP-1 is more significant than that exhibited by other featurepoints. The different degrees of nonstationarity among various feature points may originate from the site-specific terrain, environment and/or curving maneuvers. It is known that the stationary assumption is

10

typically adopted in the wind load simulations carried out for traffic safety studies. The results presentedherein, however, suggest the significantly nonstationary character of wind at some situations, e.g. curving.Thus the typical assumption of stationary process imbedded in traditional analytical studies of trafficsafety may not always be realistic and this assumption may lead to erroneous results. Quantification ofthe effects of the stationary assumption on the traffic safety risk evaluations requires furtherinvestigations.

4. Wind tunnel evaluation

For the mobile testing technique demonstrated on I-70, the anemometer was installed on the outside topsurface of the high-sided truck to collect wind velocity data, primarily for the two purposes: (1) to collectgeneral site-specific wind velocity data on I-70, and (2) to collect vehicle-specific wind velocity data forthe particular test truck. For both the cases, it is important to evaluate the impact of the experimentalsetting on the accuracy of the measurements. In order to address this issue, a series of wind tunnelexperiments were conducted with the scaled vehicle models. The wind tunnel study was conducted in theEnvironmental Wind Tunnel (EWT) at the Wind Engineering and Fluids Laboratory (WEFL), atColorado State University. A 1:10 geometrical scale model of the truck used in the field testing wasfabricated of foam. The wind tunnel testing showed that for the vehicle-specific wind velocity data onhigh-sided vehicles, the measurements with the introduced technology can give accurate results with noor limited adjustment (about 5% to 10%) when the windward wind velocity is measured. The adoption ofthe dual-anemometer setup is suggested - one anemometers on the windward and another on thetransversely leeward side of the vehicle. Such an arrangement would significantly improve the accuracyof crosswind velocity measurements acquired for high-sided vehicles. Due to the limit of space, thedetails of the experiment are not reported in this paper.

5. Conclusions

The present study introduced a new mobile testing technology developed to collect crosswindvelocity data in both time and spatial domains along any highway, for traffic safety studies. Thedeveloped technology can be used for two primary purposes: (1) acquisition of general site-specific windvelocity along any highway, independent of the choice of the test vehicle; and (2) direct measurement ofvehicle-specific wind velocity, at the roof height of a specific vulnerable vehicle driven along a highway.A field test was carried out on I-70 corridor to evaluate the performance of the developed technology.Subsequently, wind tunnel investigation was performed, using the scaled model of the test truck and thesedan car, to investigate the effects of proximity of the car cabin surface on the anemometer (crosswindvelocity) readouts.

The following conclusions are drawn from the present study:

1) The developed technology was proven to be feasible to collect crosswind speed data in both time andspatial domains. This technology can be used to collect both site-specific wind data (which is independentof the selection of the test vehicle) and vehicle-specific data associated with the specific shape of the testvehicle. The field testing of this system was carried out using a high-sided truck on I-70 corridor, west ofDenver, Colorado;

2) The crosswind velocity data collected for the six selected feature points (highway locations ortraffic/car/trucks configurations) showed unique crosswind characteristics which indicated the need forcollecting site-specific as well as vehicle-specific wind data. The field testing confirmed that thedeveloped mobile testing technology is suitable for acquisition of such data;

11

Acknowledgement

This study was partially sponsored by the United States Department of Transportation (through theMountain Plains Consortium), Colorado Department of Transportation, and Grant Number1T42OH009229-01 from CDC NIOSH Mountain and Plains Education and Research Center. The contentof this paper reflects the views of the authors, who are responsible for the facts and the accuracy of theinformation presented.

References

Baker, C. J., 1991. Ground vehicles in high cross winds. 3. The interaction of aerodynamic forces and thevehicle system. Journal of Fluids and Structures, 5, 221-241.

Baker, C.J., 1994. Quantification of accident risk for road vehicles in cross winds, Journal of WindEngineering and Industrial Aerodynamics, v 52, 1-3, 93-107.

Brassfield, M., Allison, W., 2001. Skyway sees wind-blown crash, fatal leap, St. Petersburg Times,http://www.sptimes.com/News/032101/TampaBay/Skyway_sees_wind_blow.shtmlChen, S.R., Cai, C.S., 2004. Accident assessment of vehicles on long-span bridges in windy

environments. Journal of Wind Engineering and Industrial Aerodynamics, 92, 12, 991-1024.Chen, S. R., Cai, C. S., Wolshon, B. 2009. From normal operation to evacuation: single-vehicle safety

under adverse weather, topographic and operational conditions, Natural Hazards Review, ASCE (inpress).

Huang, M. L. Wu, Long, S. R., Shen, S. S., Qu, W. D., Gloersen, P., Fan, K.L., 1998. The empirical modedecomposition and the Hilbert spectrum for nonlinear and non-stationary time series analysis. Proc.Roy. Soc. Lond., 454A, 903-993.

Pinelli, J.P., Subranmanian, C., Plamondon, M., 2004. Wind effects on emergency vehicles, Journal ofWind Engineering and Industrial Aerodynamics, 92, 7-8, 663-685.

Redhen Inc., 2004. VMS 200-Video Mapping System Hardware User’s Guide.Schmidlin, T.W., King, P.s., Hammer, B.O., Ono, Y., 1998. Behavior of vehicles during tornado winds.

Journal of Safety Research, 29 (3), 181-186.Simiu, E., Scanlan, R. H., 1996. Wind Effects on Structures - Fundamentals and Applications to Design,

3rd Edition, Wiley, New York.Sigbjornsson, R., Snabjornsson, J. T., 1998. Probabilistic assessment of wind related accidents of road

vehicles: A reliability approach. Journal of Wind Engineering and Industrial Aerodynamics, 74-76,1097-1090.

The National Academies, 2006. Where the weather meets the road- a research agenda for improving roadweather system, Report in Brief, The National Academies.

U. S. Department of Transportation (USDOT)-Intelligent Transportation Systems, 2003. I-90 truck/windwarning system, Columbia River, http://www.its.dot.gov/resources/IntegrationProject/projects /CVO-141.htm

Willett, A., 2005. Wind wreak havoc, Gazette, The (Colorado Springs), November 4, 2005,http://www.findarticles.com/p/articles/mi_qn4191/is_20051104/ai_n15844795