{"Bibliographic":{"Title":"A field comparison of in situ meteorological sensors","Authors":"","Publication date":"1985","Publisher":""},"Administrative":{"Date created":"08-17-2023","Language":"English","Rights":"CC 0","Size":"0000064141"},"Pages":["NOAA/ERL Wave Propagation Laboratory\nA FIELD COMPARISON OF IN SITU\nMETEOROLOGICAL SENSORS\nA","90\n851\nB6\nA FIELD COMPARISON OF IN SITU\nno.\n6\nMETEOROLOGICAL SENSORS\nJ.C. Kaimal\nJ.E. Gaynor\nP.L. Finkelstein\nM.E. Graves\nT.J. Lockhart\nReport Number Six\nDecember 1984\nLIBRARY\nNOAA\nBoulder Atmospheric Observatory\nJUN 1 2 1985\nN.O.A.A.\nU.\nS.\nDept.\nof\nCommer\nAND\nATMOSPHERIC\nNOAA\nU.S. Department of Commerce\nAMOUNT\nNational Oceanic and Atmospheric Administration\nEnvironmental Research Laboratories\nCommunity\nA NOAA publication available from NOAA/ERL, Boulder, CO 80303.","","A FIELD COMPARISON OF IN SITU METEOROLOGICAL SENSORS\nby\nJ. C. Kaimal and J. E. Gaynor\nNOAA/ERL/Wave Propagation Laboratory\nBoulder, Colorado 80303\nP. L. Finkelstein*\nEnvironmental Protection Agency\nResearch Triangle Park, North Carolina 27711\nM. E. Graves\nNorthrop Services, Inc.\nResearch Triangle Park, North Carolina 27709\nT. J. Lockhart**\nMeteorology Research, Inc.\nAltadena, California 91001\nThis study was conducted for\nU.S. Environmental Protection Agency\nunder Interagency Agreement No. DW1-3F2A059\nWave Propagation Laboratory\nEnvironmental Research Laboratory\nU.S. Department of Commerce\nBoulder, Colorado 80303\n*\nOn assignment from National Oceanic and Atmospheric Administration\nPresent Affiliation: Meteorological Standards Institute\nFox Island, Washington 98333","NOTICE\nAcquisition of information provided in this document was funded\nin part by the United States Environmental Protection Agency\nunder Interagency Agreement No. DW1-3F2A059. This study was con- -\nducted jointly by NOAA/Environmental Research Laboratory and the\nEnvironmental Protection Agency.\nMention of a commercial company or product does not constitute\nan endorsement by NOAA/Environmental Research Laboratories or\nthe Environmental Protection Agency.\niv","CONTENTS\nPage\nAbstract\nvi\nFigures\nvii\nTables\nX\nAcknowledgments\nxi\n1.\nIntroduction\n1\n2.\nDescription of the Field Experiment\n3\n3.\nData Acquisition, Processing, and Analysis\n17\n4.\nOperational Maintenance and Quality Control\n21\n5.\nComparison of Wind Speed Measurements\n25\n6.\nComparison of Wind Direction Measurements\n27\n7.\nComparison of ow O Values\n51\n8.\nComparison of o Values\n63\n9.\nComparison of Sigma Meters\n75\n10.\nSensor Response to Wind Fluctuations\n79\n11.\nConclusions\n91\nReferences\n92\nAppendix\n93\nV","ABSTRACT\nMeasurements of wind speed, wind direction, and the vertical\nwind component from five conventional in situ meteorological systems\nwere compared with similar measurements from a fast-response sonic\nanemometer. The systems tested were an orthogonal three-axis pro-\npeller anemometer, a light bivane and cup anemometer, a bivane pro-\npeller anemometer, a light cup and vane with a vertical propeller,\nand a vane-mounted propeller anemometer with a vertical propeller.\nComputed accuracy and field precision variables measured by each\nsystem are presented. The response characteristics of the sensors\ntested are discussed.\nvi","FIGURES\nNumber\nPage\n1\nLocation of the 10-m towers at the BAO site and the in situ\nwind sensors they support\n4\n? Sensors on the 10 m tower, viewed from the east\n5\n3\nGill UVW anemometer on Tower 1\n11\n4\nCup anemometer and bivane on Tower 2\n12\n5\nPropeller bivane on Tower 3\n13\n6 RAO sonic anemometer on Tower 4\n14\n7\nCup and vane with vertical propeller on Tower 5\n15\n8 Propeller-vane and vertical propeller on Tower 6\n16\n9 Bias in o from U-V-W shown as a function of wind\ndirection\n31\n10 Bias in o from C-BIV shown as a function of wind\ndirection\n32\n11\nBias in o from P-BIV shown as a function of wind\ndirection\n33\n12 Bias in o from C-V-W shown as a function of wind\ndirection\n34\n13 Bias in o from P-V-W shown as a function of wind\ndirection\n35\n14\nComparability in o from U-V-W shown as a function of\n0\nwind speed\n36\n15 Comparability in o 0 from C-BIV shown as a function of\nwind speed\n37\n16\nComparability in o 0 from P-BIV shown as a function of\nwind speed\n38\n17\nComparability in o\nfrom\nC-V-W\nshown\nfunction\nof\nas\na\n0\nwind speed\n39\n18\nComparability in O 0 from P-V-W shown as a function of\nwind speed\n40\n19\nBias in o 0 from U-V-W shown as a function of o\n41\n0\n20 Bias in ° 0 from C-BIV shown as a function of o\n42\n0\nvii","43\n21 Bias in ° O from P-BIV shown as a function of of\n44\n22 Bias in of from C-V-W shown as a function of o 0\n45\nBias in of from P-V-W shown as a function of o 0\n23\n46\n24 Comparability in of from U-V-W shown as a function of o 0\n47\nComparability in 00 from C-BIV shown as a function of 0\n25\n48\nComparability in °0 from P-BIV shown as a function of\n26\n0\nComparability in °0 from C-V-W shown as a function of 0\n49\n27\n28\nComparability in of from P-BIV shown as a function of o\n50\n0\n29 Comparability in from U-V-W shown as a function of\n53\nwind direction\n30 Comparability in o w from C-BIV shown as a function of\n54\nwind direction\n31\nComparability in from P-BIV shown as a function of\nW\n55\nwind direction\n32 Comparability in o W from C-V-W shown as a function of\n56\nwind direction\n33 Comparability in from P-V-W shown as a function of\n57\nwind direction\n34 Comparability in O W from U-V-W shown as a function of\n58\nwind speed\n35 Comparability in °W from C-BIV shown as a function of\n59\nwind speed\n36 Comparability in O W from P-BIV shown as a function of\n60\nwind speed\nComparability in ow from C-V-W shown as a function of\n37\n61\nwind speed\nComparability in O W from P-V-W shown as a function of\n38\n62\nwind speed\n39 Bias in 0 $ from U-V-W shown as a function of wind speed\n65\n66\n40 Bias in o from C-BIV shown as a function of wind speed\n67\n41 Bias in o from P-BIV shown as a function of wind speed\n42 Comparability in o from U-V-W shown as a function of\n68\nwind speed\nviii","43\nComparability in O from C-BIV shown as a function of\nwind speed\n69\n44\nComparability in o from P-BIV shown as a function of\nwind speed\n70\n45\nComparability in o from U-V-W shown as a function of\n71\no\n46\nComparability in o from C-BIV shown as a function of\n72\no\n47\nComparability in o from P-BIV shown as a function of\n73\no\n48 Comparability in o from the analog and digital sigma\n0\nmeters shown as functions of o\n77\n0\n49\nComparability in o from the analog and digital sigma\nmeters shown as functions of\n78\no\n50 Spectra of W from SONIC and C-BIV for three stability\nconditions\n81\n51 Spectra of W from SONIC and P-BIV for three stability\nconditions\n82\n52 Spectra of W from SONIC and C-V-W for three stability\nconditions\n83\n53 Transfer functions for (a) W response in C-BIV, P-BIV,\nC-V-W and P-V-W and (b) three-axis response in U-V-W\n85\n54 Transfer functions for (a) wind speed response and (b)\nwind direction response in the in situ sensors\n86\nix","TABLES\nPage\nNumber\n3\nInstrument selection summary\n1\n25\nBias and comparability for wind speed\n2\n27\n3 Bias and comparability for wind direction\n27\nBias and comparability for\n4\no\n0\n5 Bias and comparability for so during the day (0600-1800 MST)\n29\nBias and comparability for 00 at night (1800-0600 MST)\n6\n29\n7\nBias and comparability for o\n51\nW\n8 Bias and comparability for ow during the day\n52\nBias and comparability for ow at night\n52\n9\n63\nBias and comparability for o\n10\n0\nBias and comparability for o during\n11\nthe\nday\n64\n64\n12\nBias and comparability for o\nat night\n13 Bias and comparability of standard deviations computed by\nthe sigma meters and the wind sensor\n76\n14 Data summary for periods represented in Figs. 50-52\n79\nAverage of W variances (m 2/s2) for unstable and stable\n15\n88\nconditions\nX","ACKNOWLEDGMENTS\nThe authors acknowledge the valuable contributions made by David McSorley,\nJanet Lockhart, Joan Hart, Jim Newman, Daniel Wolfe, Norbert Szczepczynski,\nLeslie Mosley, and Mary Sue Phillips to the field experiment and to the analy-\nsis of the data. The experiment and the data analysis were sponsored by the\nU.S. Environmental Protection Agency under Interagency Agreement No. DW1-\n3F2A059.\nxi","1. INTRODUCTION\nIt has recently become clear through advances in both theoretical and\nexperimental meteorology, that improvements in modeling the transport and\ndispersion of pollutants will require on-site measurements of the atmosphere.\nThis requirement has in turn generated questions about our ability to make\nsuch measurements. To help answer these questions the Environmental\nProtection Agency sponsored at NOAA's Boulder Atmospheric Observatory (BAO) an\nexperiment designed to assess the ability of in situ and remote sensors to\nmeasure the mean and turbulent properties of the lower atmosphere. The tests\nwere carried out over a 3 week period in September 1982. They were designed\nand conducted with the goal of gaining a knowledge of the accuracy, precision,\nand general performance characteristics of a variety of meteorological sensors\nthat are commonly used in environmental studies. The results should prove\nvaluable in designing experiments, understanding data from field studies, and\ninterpreting the inherent limits of accuracy and precision possible in\ntransport and diffusion models.\nThe BAO was chosen as the site for the experiment because of the availabi- -\nlity of precise profile and turbulence data from accurate fast-response sen- -\nsors on a 300 m tower, as well as comprehensive data-logging facilities\n(Kaimal and Gaynor, 1983). Two categories of sensors were tested. One con -\nsisted of lightweight in situ sensors of types that have been frequently used\nin the recent past for boundary layer studies. The other category consisted","of four commercially available Doppler sodars, with the capability to measure\nwind speed, wind direction, and vertical component of turbulence, all at\nvarious heights above the ground. The sodar comparison has been described by\nKaimal et al. (1984). This report deals only with the in situ instrument\ncomparison.\n2","2. DESCRIPTION OF THE FIELD EXPERIMENT\nThe sensors were selected to provide a measure of turbulence in addition\nto mean speed and direction at 10 m above the ground. The systems compared in\nthis study (see Table 1) were selected because they represent types commonly\nused in meteorological monitoring programs related to air pollution. The\nselection was also made to include all usual configurations of instruments\ncapable of describing three-dimensional flow. The reference standard for com-\nparison was the BAO three-axis sonic anemometer, noted as SONIC throughout\nthis report and placed on Tower 4.\nTable 1. Instrument selection summary\nTower no.\nDesignation\nSensor type\nManufacturer\n1\nU-V-W\nGill UVW propeller\nR.M. Young Co.\n2\nC-BIV\nBivane and cup\nMeteorology Research,\nInc.\n3\nP-BIV\nPropeller bivane\nR.M. Young Co.\n4\nSONIC\nSonic anemometer\nApplied Technology,\nInc.\n5\nC-V-W\nCup and vane\nClimatronics, Inc.\nVertical propeller\nR.M. Young Co.\n6\nP-V-W\nPropeller vane\nR.M. Young Co.\nVertical propeller\nR.M. Young Co.\nThe instruments were mounted on 10 m towers, erected in a line to the west\nof the 300 m BAO tower (see Figs. 1 and 2). Each tower held a different\n3","Figure 1. Location of the 10-m towers at the BAO site and the in situ wind sensors they support.\n94°\n154°\nN\n26°\nTower\nBAO\n10\nScale (m)\n0\nTower 1\nU-V-W\n1\nTower 2\nC-BIV\n2\nTower 3\nP-BIV\nA\n10 m towers\n3\nTower 4\nSONIC\nA\n4\nTower 5\nC-V-W\n5\nTower 6\nP-V-W\nA\n6","5","instrument set. The installation of the instruments was intended to simulate\nthe best practice found in operational field programs and not the best\npossible practice if money were of no concern. Various levels of quality\ncontrol checks were used to achieve the best practical results, including\ndaily review of instrument performance. One benefit of the program was the\nevaluation of the quality control procedures, calibration checks, and opera-\ntional methods.\nThe sensors on the towers, going from east to west, are described in\ndetail below.\n2.1\nTower 1\nThe sensor mounted on this tower was a Gill UVW (U-V-W) anemometer (Fig.\n3) manufactured by the R. M. Young Company, Traverse City, Michigan. It has\nthree helicoid propeller anemometers oriented north-south, east-west, and up -\ndown. The implied assumption when fixed propellers are used is that they have\na nearly cosine response and will respond linearly to the component of the\nwind vector parallel to their axis of rotation. How well they perform is one\nof the questions to be answered by this study.\nThe propeller selected was made of polystyrene with 19 cm diameter and\n0.30 m pitch. This propeller offers a good compromise between response and\ndurability. The distance constant is listed by the manufacturer as 0.8 m (63%\nrecovery) and the starting threshold as 0.3 m/s. It should be remembered that\nthe starting threshold relates to the component of wind parallel to the\nturning axis of the propeller and not to the wind speed. Each propeller is\nflexibly coupled to a small d.c. generator. The output voltage was carried to\n6","the BAO instrument trailer (about 42 m away) through signal cables, which were\nconnected to a standard R. M. Young translator. Here the sensor signal was\nnoise filtered and amplified to provide a linear output of 0.100 volts per\nm/s. The W propeller was scaled to a different pitch than the U or V pro-\npeller, even though the propellers are identical, in order to compensate for\nthe non-cosine response when the wind is nearly perpendicular to the axis of\nrotation. This practice is recommended by the manufacturer.\n2.2 Tower 2\nThis tower held two instruments (C-BIV) manufactured by Meteorology\nResearch, Inc. of Altadena, California (see Fig. 4). The wind vane was a\nModel 1041 bivane with freedom to turn 360 degrees in azimuth and +60 degrees\nin elevation. The bivane has a front damping vane to improve performance.\nThe manufacturer lists a damping ratio of 0.6 for this instrument and a delay\ndistance (50% recovery) of 0.3 m. The sensor output is a resistance propor-\ntional to the vane shaft position.\nThe anemometer was a Model 1074 speed and direction sensor from which the\ndirection vane had been removed. This cup anemometer provided the mean hori- -\nzontal wind speed measurement with which the azimuth and elevation wind direc-\ntion measurements could be transformed to vector components. It has rather\nlarge cups, which have a response distance (63% recovery) of about 5 m. The\nsensor output is a pulse train from a 132 slot light-chopper disc. Both sen-\nsors were connected by a 47 m cable to standard signal-conditioning circuits\nin the BAO instrument trailer.\n7","2.3 Tower 3\nThis tower supported a single instrument, the Gill Anemometer Bivane\n(P-BIV), manufactured by the R. M. Young Company (see Fig. 5). It provides\nthe two directions and speed necessary to define the three-dimensional flow.\nThe helicoid propeller mounted at the front of the -directional wind vane\nprovides some damping of the vane. The manufacturer lists a damping ratio of\n0.63 and a delay distance (50% recovery) of 1.0 m. The 23 cm diameter x 0.30\nm pitch propeller has a response distance (63% recovery) of 1.0 m.\nThe vane position is measured by two potentiometers. The propeller is\nflexibly coupled to a d.c. generator for the speed measurement. The sensor\nis\nconnected by a 52 m cable to a standard translator unit in the trailer. The\nspeed measurement is noise filtered and scaled to 0.100 volts per m/s. The\nazimuth range is 1 to 351 degrees and the elevation range is +45 degrees.\n2.4 Tower 4\nThe sonic anemometer (SONIC) installed on this tower (Fig. 6) was one of\nthe three-axis anemometers used for routine wind measurement at each of the\neight levels on the BAO 300 m tower. It measures the wind component along\neach of its three orthogonal acoustic paths (vertical and two horizontal).\nFor convenience of data handling and processing, the acoustic array was\noriented in the same direction (approximately SSE) as the anemometers on the\nBAO tower. The high-frequency response of the sonic anemometer is limited by\nthe attenuation introduced by line averaging along the acoustic paths. The\nresponse function for line averaging has its half-power at a wavelength\nroughly the length of the path which, for these arrays, is 25 cm. In a linear\n8","first-order system the equivalent distance constant would be smaller by a fac-\ntor of 2TT, or roughly 4 cm.\nAn important source of error in a sonic anemometer is the systematic\nunderestimation in the measured horizontal velocity components from partial\nshadowing of the acoustic path by the transducers as the wind direction\napproaches the array axes (Kaimal and Gaynor, 1983). A simple algorithm in\nthe data acquisition software corrects this error in real time.\n2.5 Tower 5\nThree instruments (C-V-W) were mounted on this tower (Fig. 7) . Wind speed\nand direction were measured by a Model F460 sensitive cup and vane set manu- -\nfactured by Climatronics, Inc., of Bohemia, New York. The vertical wind com-\nponent was measured by an R. M. Young propeller anemometer, similar to that on\nTower 1. The vane position is measured by a potentiometer. The manufacturer\nlists the damping ratio of the vane to be 0.4 at a 10 degree initial angle of\nattack, and the distance constant to be 1.1 m. The rate of rotation of the\ncup wheel is measured by a pulse train from a 32 slot light chopper. The sen-\nsor distance constant is listed as 2.4 m for the stainless steel cup wheel.\nA\n62 m cable carried the instrument outputs to standard signal conditioners\n(Climatronics and Young) in the BAO trailer. The signal conditioner for the\nF460 normally has a built-in 10 S averaging circuit, but for this experiment\nit was removed by the manufacturer at our request so that we could observe the\nfull response of the instrument.\n9","2.6 Tower 6\nThe sensors (P-V-W) on this tower (Fig. 8) were a Gill Propeller Vane for\nthe horizontal wind speed and direction measurement, and a vertical propeller\nidentical to those on Towers 1 and 5, both made by R.M. Young Co. The pro- -\npeller and vane were changed slightly after the tower was knocked down by high\nwinds on 9/13/82. Before the accident the Model 35005 with aluminum vane and\npolypropylene propeller was used. The damping ratio of this vane is listed by\nthe manufacturer as 0.45 with a delay distance (50% recovery) of 1.3 m. The\ndistance constant (63% recovery) for the propeller is 3.3 m. The tower was\nreinstalled on 9/16/82 with a polystyrene tail and propeller, making the sen-\nsor a Model 35003. The damping ratio of this vane is 0.54 with a delay\ndistance of 1.2 m. The distance constant of the propeller is 1.0 m. The sen- -\nsor outputs were connected through a 67 m cable to standard R.M. Young signal -\nconditioning circuits.\n10","Figure 3. Gill UVW anemometer on Tower 1.\n11","Figure 4. Cup anemometer and bivane on Tower 2.\n12","Figure 5. Propeller bivane on Tower 3.\n13","Figure 6. BAO sonic anemometer on Tower\n4.\n14","Figure 7. Cup and vane with vertical propeller on Tower 5.\n15","Figure 8. Propeller-vane and vertical propeller on Tower 6.\n16","3. DATA ACQUISITION, PROCESSING, AND ANALYSIS\nThe output signals from the anemometers were connected through shielded\ncables to their signal conditioners located in an instrument trailer at the\nbase of the tower. The outputs of the signal conditioners were sampled, digi- -\ntized, and processed by the BAO data acquisition system, along with all the\nsignals from the standard sensors on the 300 m tower. The sonic anemometer\nsignals were processed in exactly the same manner as all the other sonic ane-\nmometer signals from the tower. All the channels were sampled ten times per\nsecond. See Kaimal and Gaynor (1983) for details of the BAO operation. The\ndata acquisition computer calculates in real time the means and variances of\nsamples of the outputs, converts them into meteorological units, and prints\nout the results at the end of every 20 min non-overlapping averaging period.\nFor selected periods the complete time series (10 samples/s) can be recorded\non magnetic tape. The rest of the time only 10 S non-overlapping averages are\nsaved. From each of the 10 S periods a grab sample (one of these instan-\ntaneous values) is also saved. These grab samples are needed especially for\ncomputing standard deviations since they retain the high-frequency information\ninherent in the data sample. Thus, we have 120 grab samples for calculating\neach 20-min standard deviation.\nScalar averages and standard deviations were computed for all variables.\nFor U-V-W and SONIC, the instantaneous direction was computed from the\nmeasured horizontal wind components. For the 20 min periods in which there\nwere major shifts in wind direction, and for which the scalar average direc-\n17","tion would be meaningless, the wind direction values were edited out of the\ndata set. Data were excluded when the wind was parallel to the line of the\ntowers to avoid any shadowing effect. Care was also taken to ensure that\nmisleading values were not generated when the wind was from the north.\nNortherly winds cast a shadow on the SONIC W axis and also caused errors in\nsome vane direction readings from their discontinuities at 0 deg. The entire\ndata set was also carefully edited to remove spurious values caused by instru-\nment malfunction, line noise, birds, rain, and the like.\nThe intercomparisons of first and second moments (means and standard\ndeviations) are made against the SONIC measurements, which are considered the\nreference or true values. The statistics of comparison are the bias, b, and\ncomparability, C, as defined by Hoehne (1971) and as used by Kaimal et al.\n(1984). They are defined as\n(1)\n(2)\nN = number of 20 min observations\nwhere\nY i = ith observation of the test instrument\nS j = ith observation of the reference instrument.\nThe field observations for the experiment extended from 1 to 22 September.\nTwo periods were selected for recording data at the 10 samples/s rate. The\nfirst period, 0800-1540 MST recorded on 9 September, represents typical con- -\nvective conditions encountered at the BAO, while the second period, 1600-2320\nMST recorded on 18 September, represents neutral and stable conditions. These\nrapidly sampled data are used in subsequent sections to explore details of the\n18","sensor response to the turbulence in the flow. In the sections to follow, we\nwill examine first comparisons of the means and standard deviations of wind\nspeed and wind direction (derived from the 20 min observations) and then exa-\nmine how the response characteristics of the various sensors contribute to the\nobserved performance.\n19","","4. OPERATIONAL MAINTENANCE AND QUALITY CONTROL\nA field monitoring program from which data of known quality are required\nshould be planned with a quality assurance effort aimed toward that goal.\nThe quality assurance plan for this experiment required that the data\ngathering period be bracketed by calibrations and that an independent audit be\nconducted during the field program. The ideal time for an audit is at the\nbeginning of the field measurement program so that errors can be corrected\nbefore too many data have been collected. An independent audit serves two\npurposes. (1) The expectation that someone else will check to be sure things\nhave been done right inspires the person responsible for the operation of the\nsystem to be sure everything is done and documented within the planned metho-\ndology. (2) The auditor gives another layer of authority to the claim that\ndata are valid and representative. His report and the records kept for\ncalibration are an important part of the data history.\nSince this program was short (3 weeks), it was important to monitor the\noperations daily. When problems arose, corrective action was taken at once.\nWhen, for example, a strong wind blew down Tower 4 (because of rain-softened\nsoil and inadequate anchors for such conditions) and Tower 4 fell across the\nguy wires of Tower 5, causing Tower 5 to fall across the guy wires of Tower 6,\nour daily presence allowed immediate action. Towers were repaired and re-\nerected with stronger anchors. Sensors were inspected and the necessary parts\nfor replacement ordered. Towers down on noon of the 13th were back on line on\n21","the 16th with new parts from Traverse City, Mich., and Bohemia, N.Y. Such\nfast action could happen only with maximum cooperation from both vendors and\nfield personnel. A normal monitoring program with weekly visits would have\nlost 2 or 3 weeks of data.\nThe initial calibration documented the response of the system elements to\nartificial known conditions, such as rate of rotation for speed shafts and\nposition change for wind vane shafts. Because we monitored the 20 min average\ndata on hard copy continuously, we were able to detect consistent differences\nin wind direction that led us to correct the orientation before the measure-\nment program began officially. Normally a monitoring program does not have\nredundant sensors, so differences are not there to see. It would be prudent\nto have the orientation checked by a different person than the one who did the\norientation initially. This may be the most important field task and perhaps\nthe most difficult.\nThe final calibration marks the termination of the data gathering period.\nIt remains for the data themselves to support the claim that the instruments\nperformed the same between calibrations as they did during calibrations. When\ncollocated instruments are used, the data may show that the instruments are\nperforming more accurately than the calibration suggests. Calibration methods\nhave their own assumptions and uncertainties and it is difficult to know some-\ntimes whether the error reported in a calibration comes from the measurement\ninstruments or from the calibrator. In field calibrations of meteorological\ninstruments, it is unusual to have the luxury of test equipment ten times as\naccurate as the instrument, as is usually required in routine quality control\nprograms. The best quality control in a field monitoring program is to have\nan experienced meteorologist looking at the data streams in as near real time\n22","as the budget will allow.\nAs technology advances, new problems advance in parallel. For instance,\nin the decades of analog strip chart recording it was not necessary to filter\nhigh-frequency noise spikes from the output signals since the recorders could\nnot respond to them. Now that high speed digital data loggers are becoming\nless expensive and more common, it is necessary to look at the output signals\nwith an oscilloscope to be sure that the value being recorded represents the\nmeasurement output and not some combination of measurement and high frequency\nnoise. Since the BAO system can detect frequencies as high as 5 Hz, it was\nnecessary to look at all outputs (from the signal conditioner--input to the\ndata logger) to be sure they were free of noise. Where noise was present,\nfiltering was used but at frequencies higher than the sensors could produce so\nthat no data were lost or distorted.\nDuring this experiment initial and final calibrations were conducted by\none of the authors (Lockhart). The independent audit was conducted by Dr.\nFred V. Brock of the National Center for Atmospheric Research (see Appendix).\nWe believe the quality of data collected for this study to be as good as\nanything one can achieve under field conditions. Accuracies and precisions\nreported are illustrative of what one might expect with careful attention\ngiven to calibration and installation of the sensors.\n23","","5. COMPARISON OF WIND SPEED MEASUREMENTS\nSpeed readings from U-V-W and SONIC are scalar averages of the instan-\ntaneous resultant speeds from two horizontal velocity measurements. All the\nother anemometers measure scalar speed directly. Bias and comparability of\nwind speed for the five anemometer systems compared are given in Table 2. The\nmost striking fact here is the substantial negative bias in U-V-W. The other\nsensors show a smaller bias. The non-cosine response in the propellers is\nthe most likely reason for this underestimation. No corrections were made at\nany stage in the acquisition and processing of the data to correct for this\neffect. Another point of interest is the large positive bias in P-BIV. Upon\nreviewing the calibration data, we found that this instrument had a small,\nuncorrected error of +0.3 m/s, which accounts for this positive bias. P-V-W,\nwhich has a propeller similar to P-BIV, but attached to a vane fixed in the\nhorizontal plane, had a rather small bias.\nTable 2. Bias and comparability for wind speed\nInstrument\nb (m/s)\nC (m/s)\nN\nU-V-W\n-0.43\n0.53\n1279\nC-BIV\n-0.13\n0.35\n760\nP-BIV\n0.33\n0.48\n760\nC-V-W\n-0.13\n0.36\n760\nP-V-W\n-0.16\n0.34\n760\nb = bias\nC = comparability\nN = number of observations\n25","The two cup anemometers compared in this experiment are very different in\nsize. C-BIV - cups were rather large, while the C-V-W cups were small and\nlight. However, they showed almost identical performance in measuring wind\nspeed, both in terms of bias and comparability. The latter figure is a\nmeasure of the degree of scatter in the data. The results indicate that the\noverspeeding problem attributed to cups (Izumi and Barad, 1970; Busch and\nKristensen, 1976) is not a problem here.\n26","6. COMPARISON OF WIND DIRECTION MEASUREMENTS\nBias in wind direction measurements is usually an indication of our inabi-\nlity to align the sensor properly, since we have no reason to suspect that the\nvanes do not line up with the wind. In the experiment, great care was taken\nto line up the instruments, and they were checked several times by independent\nobservers under field conditions. Therefore these bias data provide a measure\nof the expected absolute accuracy possible from any wind direction observation\nunder normal conditions. The bias, b, and comparability, C, for the measure-\nments of wind direction, 0, and standard deviation of the wind direction, '0''\nare given in Tables 3 and 4.\nTable 3. Bias and comparability for 0\nb (deg)\nC (deg)\nInstrument\nN\n1035\nU-V-W\n-1.47\n5.40\n1057\nC-BIV\n-2.69\n5.54\n1055\nP-BIV\n-0.16\n4.61\nC-V-W\n1.44\n4.48\n819\nP-V-W\n-0.31\n4.03\n897\nTable 4. Bias and comparability for °0\nb (deg)\nC (deg)\nN\nInstrument\nU-V-W\n-0.18\n3.29\n1024\n1045\nC-BIV\n-0.86\n2.63\nP-BIV\n0.26\n3.46\n1041\n810\nC-V-W\n-0.28\n2.25\n879\nP-V-W\n0.11\n2.92\n27","With the exception of P-BIV and P-V-W, the observed biases in Table 3 are\nlarger than the sensor resolution. The comparability figures in the same\ntable are somewhat disturbing. They indicate that the scatter in the obser-\nvations is about 5°. This scatter is a measure of the confidence one can\nplace in any one or small group of wind direction observations. These results\nshould be of interest, and concern, to those involved in such fields as dif-\nfusion model verification, where a 5° difference in wind direction can cause a\nconsiderable difference in the prediction of ground level concentrations.\nThe values of 0 0 bias in Table 4 show, on average, good agreement of °6\nwith SONIC. The scatter in the measurements, represented by C in Table 4, is\nsmall, about 3°, and there is not much difference between instruments. The\nlightest vanes (in C-BIV and C-V-W) did have the least scatter with respect to\nthe reference. It is to be expected that all these instruments will measure a\nslightly smaller ° 0 than the reference instrument because of their insensitiv-\nity to very small turbulent eddies. This is indeed the case, except for the\nvanes that have a propeller mounted on the front (in P-BIV and P-V-W), which\nshow slightly higher values of 00 It is possible that the propeller con-\nfiguration is causing some underdamping. This matter is discussed in Sec. 10.\nTo see if stability had an effect on the measurement of '0'' the data were\ndivided into day and night categories. (Other stability-related categories,\nsuch as lapse rate, were tried and gave similar results, so this classifica-\ntion scheme was used for simplicity.) The results are given in Tables 5 and\n6. In all but one case the bias and comparability are smaller at night than\nduring the day. During the day the propeller vane (P-V-W) and propeller\nbivane (P-BIV) again seem to be underdamped, measuring a ° O that is higher\nthan the sonic anemometer, but at night they show a slight negative bias.\n28","Table 5. Bias and comparability for o during the day\n0\n(0600 - 1800 MST)\nInstrument\nb (deg)\nC (deg)\nN\nU-V-W\n-0.23\n3.54\n469\nC-BIV\n-1.07\n3.07\n481\nP-BIV\n0.60\n4.20\n475\nC-V-W\n-0.27\n2.71\n383\nP-V-W\n0.32\n3.59\n405\nTable 6. Bias and comparability for so at night\n1800 - 0600 MST)\nInstrument\nb (deg)\nC (deg)\nN\nIJ-V-W\n-0.14\n3.06\n555\nC-BIV\n-0.69\n2.20\n564\nP-BIV\n-0.02\n2.70\n566\nC-V-W\n-0.30\n1.74\n427\nP-V-W\n-0.08\n2.19\n474\nFigures 9-13 contain plots of o 0 bias for the various instruments as a\nfunction of wind direction. The pattern in Fig. 9 for the U-V-W is most\nstriking. It is obviously a sinusoidal wave with minima at 0° , 100°, 180°, ,\n270°, , and 360°, , and with maxima in between. This pattern may be caused by two\nfactors, the slight departure from cosine response of the propellers, and the\nfact that when the wind is not perpendicular to the propellers, the threshold\nresponse speed is higher. Both factors cause the bias to be more negative for\nwind directions along either axis. Interference from the sensor-supporting\narms is not the cause since that would tend to produce maxima along those\ndirections. For P-BIV and P-V-W, Figs. 11 and 13 show a rapid increase in b\nnear 360° when the readings are left in the data set. This is caused by the\n29","small dead band in the direction potentiometer at this position.\nFigures 10-13 show what might be a pattern of noise, which persists in\nalmost the same form in all the plots. There is no symmetry about 180° but a\nfaint one around 154°, the direction in which the sonic anemometer is pointed.\nAlthough the pattern is not the same on either side of 154°, the minima at\n134° and 185° point to possible residual errors in the SONIC correction for\ntransducer shadowing as the likely cause of the deviations.\nFigures 14-18 show plots of ° 0 comparability as a function of wind speed.\nAs expected, all the instruments show more scatter at lower wind speeds. The\nsecondary peak in Fig. 16 (P-BIV) is not understood.\nFigures 19-28 are plots of 00 bias and comparability as a function of\n° 0 itself (as measured by SONIC). This is roughly equivalent to plotting it\nas a function of stability. The bias stays close to zero for o 0 up to 30°,\nbut falls off rapidly above that, approaching -4° at 00 = 50°. The com-\nparability becomes larger with increasing of (or with increasing instability),\nimplying that the sensors are less well able to follow the large-amplitude\nfluctuations.\n30","360\n330\nx\nx\n120\n60\n5\n10","360\n330\nFigure Bias in °0 from C-BIV shown as a function of wind direction.\n300\n270\n240\nWind Direction (deg)\nTower 2: C-BIV\n210\n180\n150\n+\n120\n90\n10.\n60\n+\n+\n30\n+\n0\n5\n0\n-5\n10\n-10","360\n+\n330\nFigure 11. Bias in 00 from P-BIV shown as a function of wind direction.\n300\n270\n240\nWind Direction (deg)\nTower 3: P-BIV\n210\n180\n150\n120\n+\n90\n60\n+\n30\n+\n+\n0\n5\n0\n10\n-5\n-10","360\n+\n330\nFigure 12. Bias in °0 from C-V-W shown as a function of wind direction.\n300\n270\n+\n240\nWind Direction (deg)\n210\nTower 5: C-V-W\n180\n+\n+\n150\n+\n120\n+\n+\n90\n+\n60\n+\n30\n+\n0\n0\n5\n-5\n10\n-10","360\n+\n+\nFigure 13. Bias in °0 from P-V-W shown as a function of wind direction.\n330\n300\n270\n240\nWind Direction (deg)\nTower 6: P-V-W\n210\n180\n+\n150\n120\n+\n90\n60\n+\n30\n+\n0\n5\n0\n10\n-5\n-10\n35","2\n(6ep)\n36","12","12","12\n10\n8\nTower T 5: C-V-W\nWind\n+\n2\n17.\nFigure\n10","2","90\n80\n70\n60\nTower 1: U-V-W\n+\nof (deg)\n50\n40\n30\n19.\nFigure\n20\n+\n19\n6","90\n80\n08°\nFigure 20. Bias in 00 from C-BIV shown as a function of\n70\n60\nTower 2: C-BIV\n+\nO (deg)\n50\n40\n30\n20\n+\n10\n+\n0\n6\n2\n-2\n-6","90\n68\nTower 3: P-BIV\nx\n% (deg)\n50\n30\n20\n10\n6","90\n80\nFigure 22. Bias in 00 from C-V-W shown as a function of\n70\n60\nTower 5: C-V-W\n+\nof (deg)\n50\n+\n40\n+\n30\n20\n10\n0\n6\n4\n-2","6","","90\n80\nTower 2: C-BIV\n60 (deg)\n50\n30\n20\n10\n8\n2","90\n80\n70\n6e\nTower 3: P-BIV\n08 (deg)\n50\n30\n26.\n2e\nFigure\n10\n0\n2\n10","90\n80\nFigure 27. in °0 from C-V-W shown as a function of\n70\n60\nTower 5: C-V-W\n+\n0g (deg)\n50\n40\n30\n20\n10\n8\n2\n6\n10","+\n+\n+\n+\n+\n+\na\n8\n(6ep)\n50","7. COMPARISON OF ow VALUES\nThe vertical wind velocity, W, was measured with vertically oriented pro-\npellers, bivanes and the sonic anemometer. In the case of the bivanes, hori -\nzontal speeds measured by the cups (C-BIV) and the propeller (P-BIV) were used\nin combination with the inclination angles measured by the bivanes to compute\nthe vertical velocity. On three towers (1, 5, and 6) identical propeller ane-\nmometers were used to measure vertical velocity. The reference anemometer\n(SONIC) measures W directly. Standard deviations, 'w' were computed from the\ntime series of w provided by each of the sensor systems.\nValues for bias and comparability for o ow are given in Table 7. The slight\nnegative bias of the vertical propellers is to be expected; however, the posi- -\ntive bias of the bivanes indicates an overshoot, or underdamping. The close\nagreement in the magnitude of b and C for the propellers in U-V-W, C-V-W, and\nP-V-W is reassuring.\nTable 7. Bias and comparability for o W\nInstrument\nb (m/s)\nC (m/s)\nN\nU-V-W\n-0.08\n0.10\n840\nC-BIV\n0.06\n0.11\n809\nP-BIV\n0.01\n0.06\n803\nC-V-W\n-0.07\n0.08\n782\nP-V-W\n-0.07\n0.09\n780\n51","Daytime and nighttime values of b and C can be compared by use of Tables 8\nand 9. Again it is clear that all instruments do slightly better under more\nstable conditions and that the bivanes show remarkably little bias at night.\nTable 8. Bias and comparability for ow during the day\nb (m/s)\nC (m/s)\nN\nInstrument\n-0.10\n0.11\n460\nU-V-W\n0.10\n0.15\n429\nC-BIV\n0.03\n0.07\n440\nP-BIV\n0.09\n428\nC-V-W\n-0.07\n426\n-0.08\n0.10\nP-V-W\nTable 9. Bias and comparability for ow at night\nb (m/s)\nC (m/s)\nN\nInstrument\n0.08\n380\nU-V-W\n-0.07\n0.01\n0.05\n380\nC-BIV\n-0.01\n0.04\n363\nP-BIV\n-0.06\n0.07\n354\nC-V-W\n0.07\n354\nP-V-W\n-0.06\nFigures 29-33 show plots of ow comparability as a function of wind direc-\ntion. A maximum near 94° , which is particularly pronounced in C-BIV, is\ncaused by interference from the neighboring towers and possibly the structures\naround the BAO 300 m tower. Why this maximum is not seen in P-BIV is not\nclear.\nIncreases in wind speed increase the scatter in the measurements, but\nonly at wind speeds above 5 m/s as seen in the comparability plots of Figs.\n34-38. The largest scatter at the higher speeds is displayed by C-BIV. The\nscatter is significantly lower at all wind speeds for P-BIV.\n52","360\n330\n270\nTower 1: U-V-W\n210\n150\n+\n120\n90\n60\n30\n+","360\n60\nx","+\n+\n+\n+\n(s/w)\n55","360\n+\ndirection.\n330\n+\nwind\n306\n270\n240\nWind Direction (deg)\nTower 5: C-V-W\n210\n180\n150\n120\n90\n60\n32.\nFigure\n30\n+\n0.10\n0.30\n0.25\n0.20\n0.15\n56","+\n(s/w)\n57","+\n+\n+\n+\n+\n+\n(s/w)\n58","12\n10\n8\n+\n2","12\nspeed.\nwind\n10\nof\n+\n8\nWind Speed (m/s)\nTower 3: P-BIV\n6\n4\n2\n36.\nFigure\n0\n0.5\n0.3\n0.2\n0.0\n0.1\n60","12\nspeed.\nin from C-V-W shown as of wind\n10\n8\n+\nWind Speed (m/s)\nTower 5: C-V-W\n6\n2\n37.\n+\nFigure\n0\n0.5\n0.3\n0.2\n0.1\n0.0","+\n+\n+\n+\n+\n(s/w)\n62","8.\nCOMPARISON OF o VALUES\nThe wind elevation angle ( ) ) was measured directly by the two bivanes and\nobtained indirectly from the wind components measured by the U-V-W system.\nThese measurements and their standard deviations, o '''' were compared with the\ncomputed from SONIC. Agreement is good according to the values in Table\no\n10.\nThe slight positive bias of the lighter bivane is attributed to under-\ndamping. That the U-V-W instrument agrees so closely with the bivanes is\nreassuring in view of the fact that it is used so frequently for turbulence\nmeasurements.\nTable 10. Bias and comparability for o\n0\nInstrument\nb (deg)\nC (deg)\nN\nU-V-W\n-0.30\n2.07\n890\nC-BIV\n0.65\n1.80\n804\nP-BIV\n-0.81\n1.85\n806\nComparisons of o measurements during the day and night are given in\nTables 11 and 12. The differences shown here are interesting. C-BIV has a\nhigh positive bias during the day and a slightly negative one at night. The\nscatter as indicated by comparability is uniformly larger during the day than\nat night.\n63","Table 11. Bias and comparability for o 0 $ during the day\nb (deg)\nC (deg)\nN\nInstrument\n-0.27\n2.49\n437\nU-V-W\n1.25\n2.26\n433\nC-BIV\n-0.83\n2.14\n435\nP-BIV\nTable 12. Bias and comparability for o at night\no\nb (deg)\nC (deg)\nN\nInstrument\n1.43\n372\nU-V-W\n-0.96\nC-BIV\n-0.05\n1.02\n371\n371\nP-BIV\n-0.79\n1.43\nFigures 39-44, which plot o bias and comparability as functions of wind\nspeed, bring a different perspective to the measurement. Although the U-V-W\nand C-BIV systems show only a small bias dependent on wind speed, the P-BIV\nhas a definite negative bias at low wind speeds. This is probably due to the\nlarger mass of P-BIV compared with the other two. All instruments show signi\n-\nficant scatter at low wind speeds, but only C-BIV shows the scatter increasing\nagain for speeds higher than 4 m/s (Fig. 43). The measurements are least\ndependable when the winds are light and variable; then the effects of bivane\nmass become more pronounced.\nThe o comparability is shown as a function of o itself in Figs. 45-47.\nAll three instruments show an increase of C with ' o' Since large o values\nare associated with low wind speeds, the trends in these figures are not\nsurprising.\n64","12\n6","12\n10\n8\nTower 2: C-BIV\nWind\n+\nFigure\n2\n+\n0\n6\n2","12\nspeed.\n10\nFigure 41. Bias in from of wind\n8\nWind Speed (m/s)\nTower 3: P-BIV\n6\n4\n2\n0\n6\n4\n2\n-6\n-2\n6)","","12\n5","12\n+\n+\n+\n+\nS\nr\n(6ep)\n70","30\n20\nTower 1: U-V-W\n09 (deg)\n15\n10\n5\n8\n6\n10\nN","+\n+\n+\n+\n+\n(6ep)\n72","30\n2","","9. COMPARISON OF SIGMA METERS\nThere are various devices on the market that purport to compute an effec- -\ntive standard deviation from input analog signals. These have frequently been\nused with meteorological equipment, usually wind vanes, to produce a ° 0 value\nfor preselected averaging times. Early sigma meters processed analog signals,\nusing variations on an R-C circuit, to estimate the standard deviation of the\nsignal. With the availability of microprocessor chips, digital computation is\nnow widely used. In this experiment two sigma meters, denoted A (analog) and\nD (digital), were tested. Several different input signals were used to see if\nany significant differences could be detected; in each case the two sigma\nmeters saw the same input at the same time. Standard deviations estimated by\nthe meters were compared with o values computed by the BAO data logging system\nfrom the same input signals. The same data logging and averaging procedures\nused for the rest of this study were followed. No comparison is made with the\nsonic anemometer in this evaluation.\nThe results of this comparison (Table 13) indicate that for 0, 0, and W\ninputs the analog sigma meter significantly underestimates the standard\ndeviation. Both systems show considerably more scatter for ° 0 than for o or\n°W' but the analog system shows scatter almost twice as large. The scatter is\napproximately equivalent for o ; for ow the analog system performs slightly\nbetter than the digital in terms of both bias and scatter. Not surprisingly\nthe performance of both systems deteriorates with increasing levels of tur- -\nbulence, as is shown in Figs. 48 and 49. The reasons for this seem clear for\n75","the analog system, but are less so for the digital. In any event, the trend\ntoward more digital electronics and on-site digital data processing and\nlogging should produce improvements in the digital meters and in the develop- -\nment of new algorithms for real-time analyses of meteorological data.\nTable 13. Bias and comparability of standard deviations\ncomputed by the sigma meters and the wind sensors\nInput\nType of\nb (deg)\nC (deg)\nN\nSignal\no meter\n0 (P-BIV)\nA\n-3.6\n10.3\n595\nD\n-1.5\n5.1\n653\n0 (C-BIV)\nA\n-3.8\n10.8\n354\nD\n-1.2\n5.5\n354\n(P-BIV)\nA\n-0.8\n1.3\n480\nD\n0.7\n1.3\n479\n(C-BIV)\nA\n-1.3\n2.1\n88\nD\n0.6\n1.9\n88\nW (U-V-W)\nA\n-0.05\n0.06\n157\nD\n0.12\n0.14\n157\n76","(6ep)\n77","(6ep)\n78","10. SENSOR RESPONSE TO WIND FLUCTUATIONS\nAn important objective of this study was to determine how well our sensors\nrespond to turbulent fluctuations in the flow. Published data on response\nlengths and distance constants enable us to derive response functions\nindirectly. The bias and comparability statistics presented in the foregoing\npages offer additional clues. A direct approach is to compare spectra for\ntypical flows from the candidate sensor and a reference sensor, such as the\nSONIC. Toward this end, we had recorded data from all our sensors at the full\n10 samples/s rate on two days, 9 and 18 September. Time series from these\nrecords were subjected to the spectrum analysis procedures outlined by Kaimal\nand Gaynor (1983) and are used in the discussion that follows. Typical plots\nof frequency-weighted spectral intensities as functions of frequency, n, are\npresented in Figs. 50-52. Mean wind speeds, wind directions, and stability\nconditions for these periods presented are given in Table 14.\nTable 14. Data summary for periods represented in Figs. 50-52\nDate\nTime (MST)\nz/L\nU(m/s)\n0 (deg)\n9 Sept. 82\n1040-1200\n-1.43\n3.4\n348\n18 Sept. 82\n1720-1840\n0.05\n6.5\n192\n18 Sept. 82\n2000-2120\n0.37\n2.1\n252\nU: mean wind speed at 10 m\n0: mean wind direction at 10 m\nz/L: stability parameter (height / Obukhov length) at 10 m\n79","There is, in general, good agreement between the sensor and the SONIC\nspectra at middle and low frequencies. At the high-frequency end, the SONIC\nspectra fall off at a rate consistent with predictions for the inertial\nsubrange, at a rate less steep than the sensor spectra (see Figs. 52-54). The\nsensor response starts to separate from the SONIC spectrum at approximately\nthe same wavelength in all three stabilities. (We assume wavelength l = U/n,\nfollowing Taylor's hypothesis.) The vertical arrows represent our best esti- -\nmate of this separation point, which corresponds to wavelengths of 4.4 m for\nC-BIV, 7.0 m for P-BIV, and 32 m for C-V-W.\nOn closer examination, one finds a tendency in both bivanes to overesti- -\nmate spectral contributions in the middle- to low-frequency range in unstable\nand neutral air, and to underestimate contributions in the middle- to high-\nfrequency range in very stable air. For the vertical propeller, the\nunderestimation in stable air is more extensive (Fig. 52), almost a factor of\n2 across the entire spectral bandwidth. Intermittent stoppage of the pro- -\npeller, when the wind drops below its response threshold, can produce such a\ndepression in spectral levels. (The effect would be comparable to the effect\nof adding zeros to a time series. When time series are thus expanded, a\ncorrection factor is usually applied to restore the spectrum to its proper\nlevel. )\nTransfer functions derived from composite plots of spectra from each sen. -\nsor normalized by the SONIC spectra, are presented in Figs. 53 and 54. For\nvariable X, the transfer function Tx(2) is defined as\nSensor SONIC\n(3)\n,\nwhere Sx(x) is the spectral estimate at wavelength 1. The light wind stable\ncases are not included in the composites, although they too would have fitted\n80","10\n0\n10°\nStable (U=2.1 m/s)\n-1\n10\nFigure 50. Spectra of W from SONIC and C-BIV for three stability conditions.\n10-2\n10-3\n10\n10°\nNeutral (U=6.5 m/s)\n1\n10-superscript(1)\nn (Hz)\n10-2\n10-3\n10\n10°\nUnstable (U = 3.4 m/s)\n-1\n10\nSONIC\nC-BIV\nn-2\n10-\n10-3\n10-4\n0\n-1\n10-Superscript(1)\n10-2\n10-3\n10-4\n10\n10","101\n10°\nFigure 51. Spectra of W from SONIC and P-BIV for three stability conditions.\nStable (U=2.1 m/s)\n10-1\n10-2\n10-3\n10\n10°\nNeutral (U = 6.5 m/s)\n10-1\nn (Hz)\n10-2\n10-3\n10\n0\n10°\nUnstable (U = 3.4 m/s)\n10\nSONIC\nP-BIV\n10-2\n10-3\n10-4\n10-2\n10-3\n10-4\n10\n10°\n10","101\n10°\nSpectra of W from SONIC and C-V-W for three stability conditions.\nStable (U=2.1 m/s)\n10-1\n10-2\n10-3\n10\n10°\nNeutral (U=6.5 m/s)\n10-1\nn (Hz)\n10-2\n10-3\n10\n10°\nUnstable (U=3.4 = m/s)\n10-1\nFigure 52.\nC-V-W\nSONIC\n10-2\n10-3\n10-4\n10-2\n10-3\n10\n10°\n10-4\n10","with only a small upward adjustment along the ordinate.\nThe overestimation in bivanes shows up very clearly in Fig. 53 (a).\nUnderdamping explains the increase near 'c' the cut-off frequency, but the\ncontinued overestimation at lower frequencies remains a puzzle. Low-frequency\nenergy in the horizontal fluctuations appears to be finding its way into the\nmeasured W component. Two possible paths come to mind: one through the gyro\neffect from the rotation of the propeller, and the other through 'cross-talk'\nfrom misalignment of the fins in the bivanes.\nLeif Kristensen1 (private communication) has investigated both those\npossibilities. His calculations show the gyro effect to be no more than 1%\nfor the lightweight propeller in P-BIV. The misalignment effect, on the\nother hand, does not lend itself to a simple theoretical treatment.\nKristensen's linear model, with two coupled second-order differential\nequations, could not account for the cross-talk produced by non-restoring\nforces on the fins. For now, we are left without a definitive explanation of\nthe bivane's behavior.\nOf the two bivanes, C-BIV has the better wavelength response, presumably\nbecause of its lighter construction. Dependence on the slower cup anemometer\nfor its wind speed information has had little effect on C-BIV's W response.\n(This conclusion is supported by the fact that the transfer function for W\nfollows and not the speed.) These favorable results notwithstanding, we had\nfound earlier that C-BIV's bias for °W was the largest and P-BIV's the\nsmallest (see Table 7 and Figs. 30 and 31), which shows that an enhanced high-\nfrequency response does not necessarily imply better accuracy.\n1Risoe National Laboratories, Roskilde, Denmark.\n84","(a)\nW component\n2.0\nP-BIV\nC-BIV\n1.0\n0.5\n10\n0.2\nP-V-W\nEq. (4) for d = 2 m\nC-V-W\n0.1\n10\n1.0\n1000\n100\nl (m)\n(b)\nWind components from U-V-W\n(for winds along N-S axis)\n2.0\n1.0\n0.5\nN-S\nE-W\n0.2\nVertical\n0.1\n1000\n100\n10\n1.0\nl (m)\nFigure 53. Transfer functions for (a) W response in C-BIV, P-BIV, C-V-W and\nP-V-W and (b) three-axis response in U-V-W.\n85","(a)\nWind speed\n2.0\n1.0\nP-BIV\nP-V-W\n0.5\n0.2\n0.05\n0.1\n0.2\nn (Hz)\n0.2\nC-BIV\nC-V-W\n0.1\n1000\n100\n10\n1.0\nl (m)\n(b)\nWind direction\nP-BIV\n2.0\nP-V-W\n1.0\nC-V-W\n0.5\n0.05\n0.1\n0.2\n0.5\nn (Hz)\n0.2\nC-BIV\n0.1\n1000\n100\n10\n1.0\nl (m)\nFigure 54. Transfer functions for (a) wind speed response and (b) wind direc-\ntion response in the in situ sensors. Frequency scales apply only where sen-\nsor response is controlled by filtering in the translator circuit (see text).\nFrequency and wavelength scales match at U = 6.5 m/s.\n86","The poor wavelength response of the W propeller (Fig. 53) is, at first\nglance, surprising, especially in view of the small distance constants (~1 m)\nreported from wind tunnel tests. But the discrepancy may not seem so serious\nif we recognize that (1) advertised distance constant is for flow in the axial\ndirection, (2) distance constant increases from 1 to 2 m as the flow deviates\nfrom axial to 80° off axis (approaching infinity at 90°), and (3) factors as\nlarge as 4 TT2 can exist between the existing distance constant and the wave-\nlength at which the sensor response begins to degrade.\nThe propeller response is often represented by a first order differential\nequation with a response function of the form\n(4)\nwhere T(x) is the power transfer function and d is the distance constant or\nthe length of travel of the wind before the sensor attains 63% of a step wind\nchange. This transfer function has a half-power wavelength, 10' at l = 2nd\nand a 98% power wavelength, ^c' at 2 Thus,\n(5)\n.\nThe dashed curve in Fig. 53(a) represents Eq. (4) for a sensor with d = 2 m.\nThis curve differs somewhat from the actual response curve for the propeller,\nbut their respective 'o's coincide. The predicted 1c is larger by a factor of\n2.5 than the observed ^c (~32 m). The attenuation in the observed W variance\ncan still be significant. Daytime variances are smaller by a factor of 20%\nand nighttime variances by 50%, measured at 10 m height (see Table 15). These\npercentages are quite similar to the 30% and 60% attenuations, respectively,\npredicted by Horst (1973).\n87","Table 15. Average of W variances (m2/s2) for unstable and stable conditions\nPeriod (MST)\nSONIC\nC-V-W\nP-V-W\nN\nU-V-W\nC-BIV\nP-BIV\n0800-1540\n0.208\n0.326\n0.332\n0.286\n0.226\n0.225\n24\n0.053\n0.063\n0.031\n0.032\n22\n1600-2300\n0.024\n0.065\nN = number of 20 min variances averaged\nThe W propeller in U-V-W responded somewhat differently from the pro-\npellers in C-V-W and P-V-W. In Fig. 53(b), the W curve is slightly depressed,\npossibly due to a higher response threshold from bearing friction. For the\nE-W propeller, the observed depression can be attributed to a degraded\nresponse for winds that were predominantly N-S. The curve for the N-S sensor\nthus represents the true axial response of the propeller.\nFigures 54 (a) and (b) show the response functions for wind speed and wind\ndirection. In both plots a frequency scale is inserted (positioned to match\nthe l scale at U = 6.5 m) to accommodate sensors that have roll-offs\ncontrolled by the time constants of the filters in their translator circuits.\nThe response of the sensors themselves would have followed the l scaling, but\nthe filter time constant predominates in all cases. The reason for the nearly\nidentical bias and comparability figures for C-BIV and C-V-W (Table 2) is now\nclear: the two systems have identical filters on them. Transfer functions\nfor P-BIV and P-V-W are also identical, which is not surprising in view of\ntheir similar construction. Both systems show a 1c of approximately 20 m,\nthe same as for the N-S propeller in U-V-W.\nAs for wind direction, only C-BIV follows frequency scaling, because of\nthe time constant of the filter that was inadvertently left in the 0 transla- -\ntor. The filter in C-V-W's 0 translator, on the other hand, was removed prior\n88","to the experiment; C-V-W's response is therefore comparable with that of P-BIV\nand P-V-W. The effect of the filter on °0 measurement, as reflected in\nbias values of Table 4, is not significant.\n89","","11. CONCLUSIONS\nThe results of the experiment demonstrate the best one can expect from\nconventional wind-monitoring systems, when careful attention is given to\ncalibration and installation. Our findings can be summarized as follows:\n1. Mean wind speed measurement seem the most reliable. They are subject to\nlittle, if any, overspeeding from the cup anemometers. However, mean wind\ndirection measurements show scatter of about 5° , larger than expected.\n2. Standard deviations 00 and o are measured with reasonable accuracy\n(scatter of +3° in '0' +2° in of's The scatter increases linearly with\nmagnitude up to a point (50° in '0' 15° in o of)\n3. When transfer functions for W are compared, a clear difference emerges\nbetween the bivanes and the propellers. The bivanes tended to overesti-\nmate W but also responded to wavelengths as short as 4.4 m. The pro -\npellers did not overestimate W, but neither did they respond well to\nwavelengths shorter than 32 m. Intermittent stoppage of the propeller was\nprobably responsible for the drop in spectral levels observed in the light\nwind stable case. The response to is the same as for W.\n4. Sigma meter performance degrades with increasing turbulence. The digital\nmeter shows smaller bias and less scatter than the analog meter in most\ncases.\n91","REFERENCES\nBusch, N. E. , and L. Kristensen, 1976: Cup anemometer overspeeding. J. Appl.\nMeteor., 15, 1328-1332.\nHoehne, W. E., 1971: Standardized functional tests. NOAA Tech. Memo.\nNWST&EL-12, Sterling, VA. U.S. Department of Commerce. 23 pp.\nHorst, T. W., , 1973: Corrections for response errors in a three-component pro-\npeller anemometer. J. Appl. Meteor., 12, 1072-1075.\nIzumi, Y., and M. L. Barad, 1970: Wind speeds measured by cup anemometers and\ninfluenced by tower structure. J. Appl. Meteor., 9, 851-856.\nKaimal, J. C., J. E. Gaynor, P. L. Finkelstein, M. E. Graves and T. J.\nLockhart, 1984: An Evaluation of Wind Measurements by Four Doppler\nSodars. BAO Report No. 5, Wave Propagation Laboratory, NOAA/ERL, Boulder,\nCO 80303. 110 pp.\nKaimal, J. C., and J. E. Gaynor, 1983: The Boulder Atmospheric Observatory.\nJ. Appl. Meteor., 22, 863-880.\n92","APPENDIX\nAudit of Sensor Calibration\nAn audit was conducted in the vicinity of the Boulder Atmospheric Observa-\ntory (BAO) operated by NOAA on 3 and 4 September 1982 for Thomas J. Lockhart.\n3 September 1982 Calibration of a Gill U-V-W propeller anemometer Model\n27005, Serial 1005 equipped with 19 cm diameter polystyrene propellers.\nThe procedure used was to remove the propeller and rotate the shaft with a\nbattery powered motor at approximately constant speed, monitored with a\ncounter. The counter output was recorded and compared with the computer indi-\ncated speed. It was possible to monitor the voltage signal output of the\ninterface which was sent to the analog-to-digital converter, and the raw digi-\ntal count output by the ADC. It is a good practice to check calibration in as\ncomplete a manner as possible as was done here. In this way, one does not\nrely upon the calibration of the ADC or upon having the correct calibration in\nthe computer. These happened to be correct because the BAO facility is very\nwell run but it is not a good idea to rely upon these things. The raw count\nwas converted by the computer to physical units by application of the\nfollowing polynomial (used for both U and V):\ny = 0.0061035 X - 30\nwhere X = raw count from the ADC and y = scaled data. Table 1 lists the\ncounter values which were taken to be the calibration reference values and the\nscaled computer output (y values). The counter data were converted to\nmeters/sec using the following equation:\nS = 0.30 k/60\nwhere k is the counter output in revolutions/minute (RPM) and S is the speed\nin meters/sec. The constant 0.30 is the pitch of the propeller (0.30\nmeters/revolution) and the 60 converts the time base from minutes to seconds.\nThe BAO convention is that South and West components are taken to be\npositive.\nThe errors listed are the observed values minus the true values.\nNote the W-component was not checked.\n93","Table 1. Calibration of Gill U and V propellers. All speeds are in units of\nmeters/sec. Observed data are as indicated by the computer. Times are MST.\nThe direction of rotation is indicated as clockwise (CW) or counterc lockwise\n(CCW).\nRotation\nComp.\nError\nTime\nObserved\nCalib.\n1342\n1.51\n1.50\nCCW\nU\n0.01\n0.00\n1343\n1.50\n1.50\nCCW\nU\n1358\n1.50\n1.51\nCCW\nU\n-0.01\n1359\n1.51\n1.51\nCCW\nU\n0.00\n1400\n1.51\n1.52\nCCW\nU\n-0.01\n1401\n1.51\n1.52\nCCW\nU\n0.01\n1402\n1.50\n1.51\nCCW\nU\n-0.01\n1403\n1.49\n1.51\nCCW\nU\n-0.02\n1406\n-1.52\n1.52\nCW\nU\n0.00\n1407\n-1.52\n1.52\nCW\nU\n0.00\n1408\n-1.53\n1.50\nCW\nU\n-0.03\n1409\n-1.51\n1.50\nCW\nU\n-0.01\n1410\n-1.52\n1.50\nCW\nU\n-0.02\n1411\n-1.53\n1.50\nCW\nU\n-0.03\n1414\n-2.87\n2.87\nCW\nU\n0.00\n1415\n-2.88\n2.87\nCW\nU\n-0.01\n1420\n-1.45\n1.52\nCW\nV\n0.07\n1421\n-1.44\n1.52\nCW\nV\n0.08\n94","4 September 1982 Calibration of the elevation angle of the MRI bivane and\nthe R. M. Young bivane.\nThe bivanes were removed from the mast and set on their mounting plates.\nThe elevation angle was set using jigs provided by the manufacturer. As\nabove, the observed data presented in Table 2 are the computer output.\nTable 2. Calibration of MRI and R. M. Young bivane elevation angles. Times\nare recorded to the nearest minute, all angles are recorded to the nearest\ndegree.\nMRI\nR. M. Young\nTime\nObs. True Error\nObs. True Error\nAdjusted R. M. Young bivane\n0803\n0\n0\n0\n0\n0\n0\n0805\n15\n15\n0\n17\n15\n2\n0808\n30\n30\n0\n32\n30\n2\n0810\n45\n45\n0\n51\n45\n6\n0812\n0\n0\n0\n1\n0\n1\n0814\n-15\n-15\n0\n-15\n-15\n0\n0816\n-30\n-30\n0\n-30\n-30\n0\n0818\n-45\n-45\n0\n-39\n-40\n1\nAdjusted R. M. Young bivane\n0829\n0\n0\n0\n0830\n15\n15\n0\n0831\n31\n30\n1\n0832\n50\n45\n5\n0834\n41\n40\n1\n0835\n1\n0\n1\n0837\n-15\n-15\n0\n0838\n-30\n-30\n0\n0839\n-39\n-40\n1\nNote that the R. M. Young bivane was aligned before the calibration proce-\ndure was started and again before repeating the procedure. The adjustment of\nthis bivane was difficult and appeared to shift. Also the Young calibration\nfixture was not as good as the MR I fixture. It was more difficult to use and\n95","Table 3. Calibration of the R. M. Young bivane azimuth\nError\nTime\nObserved\nTrue\n0844\n181\n180\n1\n0845\n210\n210\n0\n0846\n240\n240\n0\n270\n-1\n0847\n269\n-1\n0848\n299\n300\n0849\n329\n330\n-1\n0850\n350\n350\n0\n0853\n35\n30\n5\n0854\n64\n60\n4\n0855\n94\n90\n4\n0856\n123\n120\n3\n0857\n152\n150\n2\nThe two temperature sensors mounted on the mast assembly to measure tem-\nperature difference between the top and the bottom were removed and inserted\ninto a thermal mass at about 0950. The thermal mass gradually warmed during\nthe day and the results were monitored by the computer. The average dif-\nference was 0.09°C with the top sensor being the warmer. The sensors were\nwell matched, with almost identical time constants.\nIn general, the calibration check went smoothly. It was evident that the\nsensors had all been carefully aligned and calibrated before the audit com-\nmenced. As noted above, the BAO facility is maintained very well and the com-\nputer system used to log the data was working well. This was an easy audit.\nHowever, it should be noted that there was considerable difficulty with the R.\nM. Young bivane. It was difficult to align and the calibration indicated con-\nsiderable error. It is beyond the strict scope of this audit to speculate\nabout probable long term calibration shifts but it would not be surprising to\nfind that this bivane had shifted during the project.\nThis audit did not treat one aspect of the propeller anemometer perfor-\nmance, the cosine correction. R. M. Young data supplied with the propeller\nanemometers indicates that these propellers deviate from the desired cosine\nresponse. The worst case is when the wind vector is aligned about 45 degrees\nfrom the propeller axis. In that case the propeller underestimates the wind\ncomponent by up to 20%. There was no indication that any cosine correction\nwas being applied during the time of the audit. It would be possible, of\ncourse, to apply this correction later.\nAudited by Fred V. Brock Zed VBrock\n*U.S. GOVERNMENT PRINTING DFFICE:1985-576-000 / 20015\n96","AOBA"]}