• Albers, A., , and Janssen A. W. , 2008: Evaluation of Windcube. Deutsche WindGuard Consulting GmbH, Internal Project VC 08007, Rep. PP 08007, 29 pp. [Available online at http://www.leosphere.com/file/deusche_windguard_report_windcube_evaluation.pdf.]

  • Andreas, E. L, 1988: Estimating over snow and sea ice from meteorological data. J. Opt. Soc. Amer., 5A, 481495.

  • Charles, L. A., , Chaw S. , , Vladutescu V. , , Wu Y. , , Moshary F. , , Gross B. , , Gedzelman S. D. , , and Ahmed S. , 2007: Application of CCNY lidar and ceilometers to the study of aerosol transport and PM2.5 monitoring. Extended Abstracts, Third Symp. on Lidar Atmospheric Applications, San Antonio, TX, Amer. Meteor. Soc., P1.14. [Available online at http://ams.confex.com/ams/87ANNUAL/techprogram/paper_119919.htm.]

  • Foote, G. B., , and du Toit P. S. , 1969: Terminal velocity of raindrops aloft. J. Appl. Meteor., 8, 249253.

  • Frehlich, R., 1996: Simulation of coherent Doppler lidar performance in the weak-signal regime. J. Atmos. Oceanic Technol., 13, 646658.

    • Search Google Scholar
    • Export Citation
  • Fujii, T., , and Fukuchi T. , 2005: Laser Remote Sensing. Taylor & Francis Group, 912 pp.

  • Gomes, L., , Mallet M. , , Roger J. C. , , and Dubuisson P. , 2008: Effects of the physical and optical properties of urban aerosols measured during the CAPITOUL summer campaign on the local direct radiative forcing. Meteor. Atmos. Phys., 102 (3–4), 289306.

    • Search Google Scholar
    • Export Citation
  • Gottschall, J., , and Courtney M. , 2010: Verification test for three WindCubeTM WLS7 LiDARs at the Høvsøre test site. Risø-R-Rep. Risø-R-1732(EN), 43 pp. [Available online at http://130.226.56.153/rispubl/reports/ris-r-1732.pdf.]

  • Im, J.-S., , Saxena V. K. , , and Wenny B. N. , 2001: An assessment of hygroscopic growth factors for aerosols in the surface boundary layer for computing direct radiative forcing. J. Geophys. Res., 106 (D17), 20 21320 224.

    • Search Google Scholar
    • Export Citation
  • Lawson, J. K., , and Carrano C. J. , 2006: Using historic models of to predict r0 and regimes affected by atmospheric turbulence for horizontal, slant, and topological paths. Atmospheric Optical Modeling, Measurement, and Simulation II, S. M. Hammel and A. Kohnle, Eds., International Society for Optical Engineering (SPIE Proceedings, Vol. 6303), 630304-1 to 630304-12. [Available online at https://e-reports-ext.llnl.gov/pdf/335213.pdf.]

  • Lindelöw, P., 2007: Fiber based coherent lidars for remote wind sensing. Ph.D. dissertation, Technical University of Denmark, 186 pp.

  • Sharma, R., , Pervez Y. , , and Pervez S. , 2005: Seasonal evaluation and spatial variability of suspended particulate matter in the vicinity of a large coal-fired power station in India—A case study. Environ. Monit. Assess., 102, 113.

    • Search Google Scholar
    • Export Citation
  • Sonnenschein, C. M., , and Horrigan F. A. , 1971: Signal-to-noise relationships for coaxial systems that heterodyne backscatter from the atmosphere. Appl. Opt., 10, 16001604.

    • Search Google Scholar
    • Export Citation
  • Tunick, A., 2003: CN2 model to calculate the micrometeorological influences on the refractive index structure parameter. Environ. Modell. Software, 18, 165171.

    • Search Google Scholar
    • Export Citation
  • U.S. Environmental Protection Agency, 2010: Our nation’s air: Status and trends through 2008. Office of Air Quality Planning and Standards EPA Publ. EPA-454/R-09-002, 54 pp.

  • Wagner, R., , Antoniou I. , , Pedersen S. M. , , Courtney M. S. , , and Jørgensen H. E. , 2009: The influence of the wind speed profile on wind turbine performance measurements. Wind Energy, 12, 348362.

    • Search Google Scholar
    • Export Citation
  • Wiser, R., , and Bolinger M. , 2010: 2009 wind technologies market report. National Renewable Energy Laboratory Rep. NREL/TP-6A2-48666, 88 pp.

  • Wulfmeyer, V., , and Feingold G. , 2000: On the relationship between relative humidity and particle backscattering coefficient in the marine boundary layer determined with differential absorption lidar. J. Geophys. Res., 105 (D4), 47294741.

    • Search Google Scholar
    • Export Citation
  • Yura, H. T., 1979: Signal-to-noise ratio of heterodyne lidar systems in the presence of atmospheric turbulence. J. Mod. Opt., 26, 627644.

    • Search Google Scholar
    • Export Citation
  • View in gallery

    Theoretical estimate of CNR reduction factor vs for the Windcube at 40 m AGL; see Eqs. (2) and (3).

  • View in gallery

    Measurement heights for the experimental setup in Boulder. Dashed lines indicate laser beam directions, while solid lines indicate altitudes at which measurements were collected.

  • View in gallery

    Flux station measurements in Iowa.

  • View in gallery

    Scatterplot of CNR vs backscatter at 100 m AGL. Data are for 22–29 Aug 2010 in Boulder, Colorado.

  • View in gallery

    Average level of backscatter corresponding to each MAA for August 2010 in Boulder, Colorado. Error bars indicate the standard deviation of each set.

  • View in gallery

    Modeled Windcube collection efficiency vs altitude.

  • View in gallery

    Average ceilometer backscatter vs average Windcube CNR and modeled collection efficiency for each maximum available altitude. The best-fit line is also shown.

  • View in gallery

    Average CNR at 40 m AGL corresponding to each level of PM2.5 for the first two weeks of August 2010 in Boulder, Colorado. Error bars indicate the standard deviation of each set.

  • View in gallery

    Average MAA corresponding to each level of PM2.5 for the first two weeks of August 2010 in Boulder, Colorado. Error bars indicate the standard deviation of each set.

  • View in gallery

    Average CNR diurnal cycle at 40 m AGL for August 2010 in Boulder, Colorado.

  • View in gallery

    Average diurnal cycle of MAA for August 2010 in Boulder, Colorado.

  • View in gallery

    Scatterplot of CNR at 40 m AGL vs at 9 m from 1 to 4 Jul 2010 at a wind farm near Ames, Iowa.

  • View in gallery

    Scatterplot of CNR at 40 m AGL vs RH at 9 m AGL. Data are from 28 Jun to 9 Jul 2010, excluding periods of measurable precipitation, at a wind farm near Ames, Iowa. The bold black line indicates the best-fit curve.

  • View in gallery

    Comparison of (top) Windcube-measured vertical wind speed at 40 m AGL and (bottom) precipitation during the Iowa field deployment. Positive vertical winds are toward the ground. Each tick mark on the horizontal axis corresponds to midnight of the respective date.

  • View in gallery

    Histogram comparing data availability at different altitudes for the Boulder and Iowa field deployments.

  • View in gallery

    Histogram comparing maximum available Windcube altitude for the Boulder and Iowa field deployments.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 145 145 48
PDF Downloads 116 116 34

Performance of a Wind-Profiling Lidar in the Region of Wind Turbine Rotor Disks

View More View Less
  • 1 University of Colorado, Boulder, Colorado
© Get Permissions
Full access

Abstract

As the wind energy sector continues to grow, so does the need for reliable vertical wind profiles in the assessment of wind resources and turbine performance. In situ instrumentation mounted on meteorological towers can rarely probe the atmosphere across the full span of modern turbine rotor disks, which typically extend from 40 to 120 m above the surface. However, by measuring the Doppler shift of laser light backscattered by particles in the atmosphere, remote sensing lidar is capable of estimating wind speeds and turbulence at several altitudes in this range and above. Consequently, lidar has proven a promising technology for both wind resource assessment and turbine response characterization. The aim of this study is to quantify data availability for a coherent detection wind-profiling lidar—namely, the Leosphere Windcube.

To determine situations of suitable data return rates, a Windcube, collocated with a Vaisala CL31 ceilometer, was deployed as part of the Skywatch Observatory at the University of Colorado at Boulder. Aerosol backscatter, as measured by the ceilometer, and lidar carrier-to-noise ratio (CNR) are strongly correlated. Additionally, lidar CNR was found to depend on atmospheric turbulence characteristics and relative humidity in another deployment at a location in the United States Great Plains. These relationships suggest an ability to predict lidar performance based on widely available air quality assessments (such as PM2.5 concentration) and other climatic conditions, thus providing guidance for determining the utility of lidar deployments at wind farms to characterize turbine performance.

Corresponding author address: Matthew Aitken, Department of Physics, 390 UCB, University of Colorado, Boulder, CO 80309-0390. E-mail: matthew.aitken@colorado.edu

Abstract

As the wind energy sector continues to grow, so does the need for reliable vertical wind profiles in the assessment of wind resources and turbine performance. In situ instrumentation mounted on meteorological towers can rarely probe the atmosphere across the full span of modern turbine rotor disks, which typically extend from 40 to 120 m above the surface. However, by measuring the Doppler shift of laser light backscattered by particles in the atmosphere, remote sensing lidar is capable of estimating wind speeds and turbulence at several altitudes in this range and above. Consequently, lidar has proven a promising technology for both wind resource assessment and turbine response characterization. The aim of this study is to quantify data availability for a coherent detection wind-profiling lidar—namely, the Leosphere Windcube.

To determine situations of suitable data return rates, a Windcube, collocated with a Vaisala CL31 ceilometer, was deployed as part of the Skywatch Observatory at the University of Colorado at Boulder. Aerosol backscatter, as measured by the ceilometer, and lidar carrier-to-noise ratio (CNR) are strongly correlated. Additionally, lidar CNR was found to depend on atmospheric turbulence characteristics and relative humidity in another deployment at a location in the United States Great Plains. These relationships suggest an ability to predict lidar performance based on widely available air quality assessments (such as PM2.5 concentration) and other climatic conditions, thus providing guidance for determining the utility of lidar deployments at wind farms to characterize turbine performance.

Corresponding author address: Matthew Aitken, Department of Physics, 390 UCB, University of Colorado, Boulder, CO 80309-0390. E-mail: matthew.aitken@colorado.edu

1. Introduction

The importance of reliable vertical wind profiles for both resource assessment and evaluation of turbine performance continues to rise with the rapidly escalating use of wind power in both domestic and worldwide energy production. Meteorological towers used to collect wind data, however, are usually constructed no higher than 60 m in the United States for reasons concerning structural stability, cost, and zoning regulations. With hub heights of 80–100 m and rotor diameters of 80 m or more, modern wind turbines are so tall that in situ instrumentation mounted on meteorological towers can rarely probe the atmosphere across even the lower half of the rotor disk. And while turbine nacelles may be equipped with cup and sonic anemometry, hub-height point measurements are not representative of the wind speeds over the entire swept area of the rotor (Wagner et al. 2009). As a result, remote sensing techniques, such as lidar and sodar, will play an increasingly prominent role in the measurement of atmospheric conditions at wind farms. In particular, the Leosphere Windcube, a pulsed coherent Doppler wind lidar, is capable of measuring wind speeds within a claimed uncertainty of 0.1 m s−1 at 10 different altitudes greater than 40 m above ground level (AGL). Yet the availability of wind speed measurements, quantified by the carrier-to-noise ratio (CNR) of the laser signal, is dependent upon weather conditions such as aerosol backscatter, turbulence, humidity, and precipitation. By examining the impact of these four parameters on CNR, this study aims to determine the most appropriate circumstances for implementing wind lidar in the assessment of turbine performance. Despite particular reference to the Windcube, our results and experience may serve to guide characterization of other lidar systems, such as Natural Power’s ZephIR and SgurrEnergy’s Galion as well.

2. Background and previous work

In optical remote sensing, it is possible to image features on the order of the signal wavelength or larger. Typically operating in the visible or near-infrared range, lidar is highly sensitive to aerosols and has many applications in atmospheric research and meteorology. Wind lidar, in particular, relies on deflection by small airborne objects, which is a process known as backscattering. In coherent detection lidar (e.g., the Windcube), a portion of the scattered radiation is collected by the lidar receiver, and the returned signal is mixed with a local oscillator beam and then detected with a photodetector. The resulting beat signal is a radio frequency (RF) photocurrent whose frequency is a direct measure of the Doppler shift, which in turn is used to calculate wind speed. Lidar data availability is specified by CNR, which is defined as the ratio of the mean RF signal power to the mean noise power (Fujii and Fukuchi 2005). In the case of very weak signal strength, velocity estimates are dominated by random outliers and, consequently, subject to significant estimation error. If CNR falls below a predetermined threshold, the uncertainty in the velocity estimate is deemed too large for the data to be useful and no measurement is recorded (Frehlich 1996). For example, the Windcube automatically disregards all measurements for which CNR is less than −22 dB, although this default setting may be adjusted in the expert version of the controlling software. Henceforth, Windcube data are considered to be “available” if CNR is greater than −22 dB. Schematics and details concerning Windcube operation can be found in brochures published by the manufacturer, which are available at http://www.leosphere.com/file/leosphere_windcube.pdf and http://www.lidarwindtechnologies.com/pdf/windcube.pdf.

Neglecting the loss due to atmospheric refractive turbulence, the parameter dependence of coherent lidar CNR is derived from the lidar equation in Fujii and Fukuchi (2005):
e1
where η represents an overall efficiency factor, T is atmospheric extinction, Ex is laser pulse energy, h is Planck’s constant, ν is the transmitted laser beam frequency, B is the receiver noise-equivalent bandwidth, c is the speed of light, β is backscatter coefficient as measured by the lidar, Ar is aperture area, and R is target range. The collection efficiency of a coherent focused lidar, such as the Windcube, is maximized for scattering at the focus point and falls off with distance as the target is moved away from this location. Extinction is weak at the Windcube wavelength of 1.54 μm, which avoids molecular absorption features. In fact, the one-way near-infrared extinction is on the order of 0.2 dB km−1 in the planetary boundary layer (Fujii and Fukuchi 2005). At a given height, then, Windcube CNR is expected to be linearly proportional to aerosol backscatter since all other parameters are approximately constant.
While extinction due to molecular absorption is negligible, the additional loss due to refractive turbulence can be quite severe. The reduction factor ψ is defined as the ratio of CNR in the presence of turbulence to CNR in the absence of turbulence. Yura (1979) found that, for a focused coaxial system, the reduction factor is given by
e2
where a2 is the 1/e2 intensity radius (the distance at which the beam intensity has dropped by a factor of 1/e2 from its center value) and ρR is the round-trip coherence length, defined as
e3
for uniform turbulence conditions. Here, k is the wavenumber of the signal, is the refractive index structure parameter, and z is propagation distance. High values of on the order of 10−13 to 10−12 m−2/3 are indicative of a strongly turbulent atmosphere; for lower values on the order of 10−16 to 10−15 m−2/3, atmospheric optical turbulence is considered negligible over optical paths of less than 2 km (Tunick 2003). Figure 1 shows the theoretical reduction factor versus for the Windcube at 40 m AGL. Although nearly unaffected for weak and moderate turbulence levels, CNR can be reduced by 20%–80% in the strong turbulence regime.
Fig. 1.
Fig. 1.

Theoretical estimate of CNR reduction factor vs for the Windcube at 40 m AGL; see Eqs. (2) and (3).

Citation: Journal of Atmospheric and Oceanic Technology 29, 3; 10.1175/JTECH-D-11-00033.1

Aerosol scattering is a complex function of refractive index and particle size, both of which depend on relative humidity (RH). Although the refractive index tends to decrease with increasing humidity, this effect is small and is dominated by the swelling of hygroscopic particles near the saturation point. Wulfmeyer and Feingold (2000) found that scattering remains nearly constant for low-to-moderate levels of humidity and increases rapidly for RH > 0.8. It is also possible for the lidar signal to scatter from water droplets, enhancing CNR during precipitation events but also causing a false indication of vertical wind speed.

In summary, the primary atmospheric conditions expected to influence CNR, and thus lidar performance, are aerosol backscatter, atmospheric refractive turbulence, relative humidity, and precipitation. The effects of each are examined using the two datasets described in the following section; observations are presented in section 4.

3. Data and methods

A Windcube lidar was deployed in late summer 2010 as part of the Skywatch Observatory—a set of meteorological instruments on the roof (approximately 15 m AGL) of the Duane Physics building at the University of Colorado at Boulder, elevation 1663 m. Backscatter was measured at a vertical resolution of 10 m from ground level to 7690 m AGL using a Vaisala CL31 ceilometer, which has an operational wavelength of 910 nm and was also located on the roof, approximately 5 m east of the Windcube. Similar to the Windcube, the ceilometer consists of a vertically pointing laser and a collocated receiver. Short laser pulses are sent through the atmosphere, and a small component of the light is scattered by aerosols, water droplets, and low-level clouds and then returned to the receiver. The strength of the returned signal is recorded and its timing is transformed into a spatial range using the speed of light, thus producing a vertical profile of backscatter within the atmosphere. Measurements of lidar CNR and ceilometer backscatter were collected at heights of 40, 50, 60, 80, 100, 120, 140, 160, 180, and 200 m throughout August 2010; see Fig. 2. (Please note that all further references to backscatter measurements should be understood to come from the ceilometer.) Lidar and ceilometer data were taken at rates of 1 and 0.0625 Hz, respectively, and then averaged over 2-min intervals for comparison.

Fig. 2.
Fig. 2.

Measurement heights for the experimental setup in Boulder. Dashed lines indicate laser beam directions, while solid lines indicate altitudes at which measurements were collected.

Citation: Journal of Atmospheric and Oceanic Technology 29, 3; 10.1175/JTECH-D-11-00033.1

In addition, a second dataset was collected with the Windcube at the same altitudes during a deployment at a wind farm in central Iowa in late June and early July 2010. A flux station 142 m due west of the Windcube collected measurements of wind speed, temperature, and absolute humidity at 20 Hz. Wind speed and temperature were measured at 6.45 m AGL with a Campbell Scientific sonic anemometer (CSAT3), while absolute humidity measurements were collected by a Li-Cor LI-7500 gas analyzer at this same height. Precipitation and relative humidity were also measured at this same flux station with a Campbell Scientific tipping-bucket rain gauge (TE525) at 3 m AGL and a Vaisala probe (HMP45) at 9 m AGL, respectively. A sketch of the flux station measurements is provided in Fig. 3. These data were used to determine momentum flux, heat flux, and latent heat flux over 30-min averaging periods; in turn, the refractive index structure parameter was then calculated and compared to the lowest available CNR measurement at 40 m AGL. Following Andreas (1988), is given by
e4
where z is altitude and ζ = z/L is the stability parameter. Here, the length scale is the Obukhov length
e5
where γ is the acceleration of gravity, κ = 0.4 is von Kármán’s constant, and are representative values of temperature and absolute humidity in the surface layer, and ρ is the density of moist air. The temperature and humidity scales are defined as
e6
e7
respectively, where is the friction velocity, w is vertical wind velocity, and primes indicate turbulent fluctuations from the mean, while the similarity function is
e8
For visible and near-infrared wavelengths (0.36–3 μm), the coefficients A and B are
e9
where P is atmospheric pressure measured in hectopascals, T is temperature measured in kelvins, and
e10
are functions of wavelength λ in micrometers. For the Windcube (λ = 1.54 μm), then, the two coefficients are A = −7.77 × 10−5(P/T2) and B = −5.80 × 10−5.
Fig. 3.
Fig. 3.

Flux station measurements in Iowa.

Citation: Journal of Atmospheric and Oceanic Technology 29, 3; 10.1175/JTECH-D-11-00033.1

4. Results

Lidar performance is expected to be influenced by aerosol backscatter, atmospheric refractive turbulence, humidity, and precipitation. The effects of these parameters are examined in the following subsections.

a. Aerosol backscatter

From section 2, CNR is expected to be linearly proportional to the backscatter coefficient at fixed altitude in the absence of turbulence. In Fig. 4, CNR and β at 100 m AGL are compared for the week 22–29 August 2010 in Boulder; note the logarithmic scale on the horizontal axis, as CNR is measured in decibels. The correlation coefficient is 0.7, indicating a fairly high degree of linear dependence between the variables. Outliers are likely due to variations in atmospheric conditions, such as turbulence and humidity.

Fig. 4.
Fig. 4.

Scatterplot of CNR vs backscatter at 100 m AGL. Data are for 22–29 Aug 2010 in Boulder, Colorado.

Citation: Journal of Atmospheric and Oceanic Technology 29, 3; 10.1175/JTECH-D-11-00033.1

Accordingly, the maximum available altitude (MAA) of the Windcube—defined as the highest altitude for which CNR > −22 dB at a given measurement time—should be influenced by the amount of aerosol backscatter. For each Windcube measurement during the month of August 2010, the MAA was determined along with the corresponding backscatter coefficient at that height. The points in Fig. 5 represent the average level of backscatter corresponding to each MAA, and the error bars indicate the standard deviation of each set; the rightmost point actually represents all altitudes greater than or equal to 200 m, thus the larger standard deviation for this point than the rest. To verify the trend in Fig. 5, one must consider both the average CNR at each MAA and the range-dependent collection efficiency of the Windcube (see Fig. 6)—which Lindelöw (2007) modeled using the results of Sonnenschein and Horrigan (1971)—since from Eq. (1) backscatter varies as β ~ (CNR/η)R2. Indeed, Fig. 7 shows that the left- and right-hand sides of this proportionality are linearly related for the 10 distinct MAAs, with a correlation coefficient of 0.932.

Fig. 5.
Fig. 5.

Average level of backscatter corresponding to each MAA for August 2010 in Boulder, Colorado. Error bars indicate the standard deviation of each set.

Citation: Journal of Atmospheric and Oceanic Technology 29, 3; 10.1175/JTECH-D-11-00033.1

Fig. 6.
Fig. 6.

Modeled Windcube collection efficiency vs altitude.

Citation: Journal of Atmospheric and Oceanic Technology 29, 3; 10.1175/JTECH-D-11-00033.1

Fig. 7.
Fig. 7.

Average ceilometer backscatter vs average Windcube CNR and modeled collection efficiency for each maximum available altitude. The best-fit line is also shown.

Citation: Journal of Atmospheric and Oceanic Technology 29, 3; 10.1175/JTECH-D-11-00033.1

To relate lidar performance to air quality assessments, we note that optical backscatter and PM2.5—a measurement of the concentration of atmospheric particles less than 2.5 μm in diameter—have been found to be highly correlated in the lowest 200 m of the boundary layer (Charles et al. 2007). A similar relationship is thus expected between CNR and PM2.5. Hourly measurements of PM2.5 at a resolution of 1 μg m−3 were available for the first two weeks of August in Boulder from an air monitoring station located approximately 500 m due north of the Windcube and operated by the Colorado Department of Public Health and Environment. Indeed, as indicated in Fig. 8, greater concentrations of particulate matter correspond to higher levels of CNR in general. Notably, Fig. 9 shows that, on average, a concentration of just 2–3 μg m−3 is required for a maximum range of 120 m—a height corresponding to the top of the rotor disk of a typical modern wind turbine. By comparison, the average annual PM2.5 concentration for 766 monitoring stations across the United States has hovered between 11 and 13 μg m−3 over the last decade, and just 10% of sites measured average annual concentrations of less than 8 μg m−3 in that time span (U.S. Environmental Protection Agency 2010). While particulate matter concentrations obviously vary with location and time, these results suggest that the Windcube is generally well suited for wind energy applications throughout much of the United States.

Fig. 8.
Fig. 8.

Average CNR at 40 m AGL corresponding to each level of PM2.5 for the first two weeks of August 2010 in Boulder, Colorado. Error bars indicate the standard deviation of each set.

Citation: Journal of Atmospheric and Oceanic Technology 29, 3; 10.1175/JTECH-D-11-00033.1

Fig. 9.
Fig. 9.

Average MAA corresponding to each level of PM2.5 for the first two weeks of August 2010 in Boulder, Colorado. Error bars indicate the standard deviation of each set.

Citation: Journal of Atmospheric and Oceanic Technology 29, 3; 10.1175/JTECH-D-11-00033.1

Also noteworthy is that the Boulder dataset exhibits a regular diurnal cycle in CNR, inducing a diurnal pattern in the MAA of the Windcube. The diurnal variation of CNR at 40 m AGL was averaged over the days of August 2010 in Boulder and can be seen in Fig. 10. Typically reaching a local minimum in early morning around 0600 local time (LT), CNR increases as aerosols are lifted with the development of convective conditions after sunrise. Countering this effect is the increase of the boundary layer height, which tends to lower aerosol concentration with the entrainment of cleaner air from above, and thus CNR reaches a peak in mid- to late morning. The local maximum around 1500 LT is due to rain, which is a common occurrence in midafternoon for the local climate. The last local maximum near 1730 LT is likely the result of increased vehicular aerosol emissions during rush hour. Although the height of the boundary layer diminishes after sunset, anthropogenic aerosol production also decreases and aerosols close to the surface are lost by sedimentation, leading to lower CNR. This diurnal cycle—due to variations in local anthropogenic activity, meteorological conditions, boundary layer height, and removal mechanisms—is typical of urban areas (Gomes et al. 2008). A similar pattern is apparent for the average diurnal variation of MAA in Fig. 11, which indicates how high the Windcube can be expected to operate throughout the course of the day. On average, the MAA in summertime Boulder ranges from about 140 m AGL in early morning to about 180 m AGL just before noon—well above the reach of modern turbine blades.

Fig. 10.
Fig. 10.

Average CNR diurnal cycle at 40 m AGL for August 2010 in Boulder, Colorado.

Citation: Journal of Atmospheric and Oceanic Technology 29, 3; 10.1175/JTECH-D-11-00033.1

Fig. 11.
Fig. 11.

Average diurnal cycle of MAA for August 2010 in Boulder, Colorado.

Citation: Journal of Atmospheric and Oceanic Technology 29, 3; 10.1175/JTECH-D-11-00033.1

b. Atmospheric refractive turbulence

While aerosol backscatter exerts the primary influence on CNR, strong levels of atmospheric refractive turbulence can cause signal degradation. Unfortunately, independent flux measurements for quantifying the effect of atmospheric refractive turbulence on lidar performance were only available during the field campaign in Iowa. From Fig. 1, CNR should only noticeably decrease in strongly turbulent conditions—say > 5 × 10−14 m−2/3⅔—a situation that occurred less than 1% of the time in Iowa. Because refractive turbulence levels were most often weak or moderate, there is no discernible relationship between the two variables, as indicated in Fig. 12. Despite the fact that the lowest available CNR measurement at 40 m AGL is compared to at 9 m, refractive turbulence is expected to be even weaker at 40 m based on several boundary layer models (Lawson and Carrano 2006).

Fig. 12.
Fig. 12.

Scatterplot of CNR at 40 m AGL vs at 9 m from 1 to 4 Jul 2010 at a wind farm near Ames, Iowa.

Citation: Journal of Atmospheric and Oceanic Technology 29, 3; 10.1175/JTECH-D-11-00033.1

c. Humidity and precipitation

Figure 13 shows the effect of RH on CNR during the Iowa field deployment. The bold black line is a nonlinear regression of the form CNR = a(1 − RH)b, following the regression for backscatter versus RH used in Im et al. (2001). Here, a = −16.3 and b = 0.0734. There is almost no correlation for RH < 0.8, but CNR does increase above this level of humidity, thus confirming the results in Wulfmeyer and Feingold (2000); the increase is especially sharp near saturation. During August 2010 in Boulder, which has a high desert climate, RH was never greater than 0.9 even during precipitation events: CNR and RH were almost completely uncorrelated for this time period.

Fig. 13.
Fig. 13.

Scatterplot of CNR at 40 m AGL vs RH at 9 m AGL. Data are from 28 Jun to 9 Jul 2010, excluding periods of measurable precipitation, at a wind farm near Ames, Iowa. The bold black line indicates the best-fit curve.

Citation: Journal of Atmospheric and Oceanic Technology 29, 3; 10.1175/JTECH-D-11-00033.1

One possible advantage of lidar over sodar technologies is that sodar is known to collect erroneous measurements during precipitation events. Although the manufacturer suggests that the Windcube is capable of taking measurements during periods of rain, Fig. 14 indicates that measurements of vertical velocity are likely invalid; horizontal velocity measurements may be uncontaminated (Albers and Janssen 2008). During precipitation events, w ranges between 1 and 5 m s−1, which is of the same order as the terminal velocity of a raindrop (Foote and du Toit 1969) and almost certainly too large to be actual vertical wind speed, as no microbursts or similar type of meteorological phenomena were reported to be in the area at the time according to forecast discussions released by the National Weather Service. Similarly, previous studies, such as Gottschall and Courtney (2010), have found it necessary to filter out Windcube data in the presence of rain. Thus, rain may interfere with operation of the Windcube, and measurements of vertical velocity during precipitation events must be evaluated with caution.

Fig. 14.
Fig. 14.

Comparison of (top) Windcube-measured vertical wind speed at 40 m AGL and (bottom) precipitation during the Iowa field deployment. Positive vertical winds are toward the ground. Each tick mark on the horizontal axis corresponds to midnight of the respective date.

Citation: Journal of Atmospheric and Oceanic Technology 29, 3; 10.1175/JTECH-D-11-00033.1

d. Summary of data availability

A comparison of data availability at different heights for the Boulder and Iowa field deployments can be seen in Fig. 15; note that the data availability in Iowa is higher at all altitudes. Additionally, the fractional occurrence of a particular MAA was calculated by dividing the number of measurements for which that height was the maximum available by the total number of Windcube measurements in the given dataset. Figure 16 compares the occurrence of each MAA in Boulder and Iowa. The Windcube could “see” at or above 200 m only about 20% of the time in Boulder, whereas measurements were available at or above this height over half the time in Iowa.

Fig. 15.
Fig. 15.

Histogram comparing data availability at different altitudes for the Boulder and Iowa field deployments.

Citation: Journal of Atmospheric and Oceanic Technology 29, 3; 10.1175/JTECH-D-11-00033.1

Fig. 16.
Fig. 16.

Histogram comparing maximum available Windcube altitude for the Boulder and Iowa field deployments.

Citation: Journal of Atmospheric and Oceanic Technology 29, 3; 10.1175/JTECH-D-11-00033.1

The mean PM2.5 concentrations during the month of August 2010 in Boulder and during the Iowa field deployment were 5.7 and 8.2 μg m−3, respectively, suggesting that the improved Windcube performance in Iowa can probably be explained by higher concentrations of particulate matter. While humidity was higher in Iowa than in Boulder, PM2.5 is expected to be the dominant influence on optical backscatter and, hence, on CNR. We hesitate to attribute the greater amount of data availability in Iowa to the effects of humidity with such a limited dataset, especially since PM2.5 and humidity tend to be negatively correlated (Sharma et al. 2005).

5. Conclusions

To determine situations of acceptable data availability, the theoretical parameter dependence of lidar CNR has been investigated in the field using a Windcube. The four most significant factors found to influence lidar performance are aerosol backscatter, atmospheric refractive turbulence, humidity, and precipitation. In summary, Windcube CNR tends to be higher during the day than at night and is linearly proportional to aerosol backscatter, which is highly correlated to PM2.5 in the lowest part of the boundary layer. Because CNR is proportional to the inverse of the square of propagation distance, more backscatter is needed at higher altitudes for suitable data return. Accordingly, it should be possible to monitor Windcube performance during future field deployments by using a ceilometer to measure vertical aerosol concentrations and the height of the boundary layer. Although the loss due to atmospheric refractive turbulence can be quite severe in theory, such strongly turbulent conditions did not occur in the present datasets even though data were taken in midsummer daytime, and therefore this effect could not be discerned. While practically uninfluenced by low levels of humidity, CNR increases sharply near the saturation point. Windcube performance is adversely affected by precipitation, as rainfall is measured instead of vertical wind speed.

Overall, our results show that lidar can be expected to reliably provide profiles of wind speed, direction, and turbulence intensity at altitudes within a typical wind turbine rotor disk. The average hub height and rotor diameter of wind turbines installed in 2009 in the United States was 78.8 and 81.6 m, respectively (Wiser and Bolinger 2010), meaning that measurements between about 40 and 120 m above the surface will be required for accurate resource assessment and evaluation of turbine performance at modern wind farms. In particular, for the Windcube considered in this study, data were available up to 120 m AGL more than 90% of the time both in Boulder and at a wind farm in the Great Plains. Given that the mean PM2.5 concentrations during both experiments were below the national annual average, this type of lidar seems a promising candidate for widespread use in the wind energy industry, especially at humid sites characterized by relatively high concentrations of particulate matter.

Acknowledgments

We would like to recognize NRG Systems, Inc. for their support during the Windcube experiments, particularly the efforts of Mr. Evan Osler. Special thanks also to the research group led by Dr. Gene Takle at Iowa State University (ISU) for helping to deploy the Windcube in Iowa, and to both ISU and Dr. Jerry Hatfield of the National Laboratory for Agriculture and the Environment for supplying the flux station measurements. This paper benefited tremendously from the many valuable suggestions provided by Dr. Andrew Clifton of the National Renewable Energy Laboratory and three anonymous reviewers. Support for this effort was provided by the Department of Energy Wind and Hydropower Program’s “20% by 2030” program via a project managed by Dr. Jeff Mirocha at Lawrence Livermore National Laboratory. Skywatch Observatory is supported by the National Science Foundation under Grant DUE-0837388. The Windcube considered here was purchased by the University of Colorado at Boulder from NRG Systems, Inc.

REFERENCES

  • Albers, A., , and Janssen A. W. , 2008: Evaluation of Windcube. Deutsche WindGuard Consulting GmbH, Internal Project VC 08007, Rep. PP 08007, 29 pp. [Available online at http://www.leosphere.com/file/deusche_windguard_report_windcube_evaluation.pdf.]

  • Andreas, E. L, 1988: Estimating over snow and sea ice from meteorological data. J. Opt. Soc. Amer., 5A, 481495.

  • Charles, L. A., , Chaw S. , , Vladutescu V. , , Wu Y. , , Moshary F. , , Gross B. , , Gedzelman S. D. , , and Ahmed S. , 2007: Application of CCNY lidar and ceilometers to the study of aerosol transport and PM2.5 monitoring. Extended Abstracts, Third Symp. on Lidar Atmospheric Applications, San Antonio, TX, Amer. Meteor. Soc., P1.14. [Available online at http://ams.confex.com/ams/87ANNUAL/techprogram/paper_119919.htm.]

  • Foote, G. B., , and du Toit P. S. , 1969: Terminal velocity of raindrops aloft. J. Appl. Meteor., 8, 249253.

  • Frehlich, R., 1996: Simulation of coherent Doppler lidar performance in the weak-signal regime. J. Atmos. Oceanic Technol., 13, 646658.

    • Search Google Scholar
    • Export Citation
  • Fujii, T., , and Fukuchi T. , 2005: Laser Remote Sensing. Taylor & Francis Group, 912 pp.

  • Gomes, L., , Mallet M. , , Roger J. C. , , and Dubuisson P. , 2008: Effects of the physical and optical properties of urban aerosols measured during the CAPITOUL summer campaign on the local direct radiative forcing. Meteor. Atmos. Phys., 102 (3–4), 289306.

    • Search Google Scholar
    • Export Citation
  • Gottschall, J., , and Courtney M. , 2010: Verification test for three WindCubeTM WLS7 LiDARs at the Høvsøre test site. Risø-R-Rep. Risø-R-1732(EN), 43 pp. [Available online at http://130.226.56.153/rispubl/reports/ris-r-1732.pdf.]

  • Im, J.-S., , Saxena V. K. , , and Wenny B. N. , 2001: An assessment of hygroscopic growth factors for aerosols in the surface boundary layer for computing direct radiative forcing. J. Geophys. Res., 106 (D17), 20 21320 224.

    • Search Google Scholar
    • Export Citation
  • Lawson, J. K., , and Carrano C. J. , 2006: Using historic models of to predict r0 and regimes affected by atmospheric turbulence for horizontal, slant, and topological paths. Atmospheric Optical Modeling, Measurement, and Simulation II, S. M. Hammel and A. Kohnle, Eds., International Society for Optical Engineering (SPIE Proceedings, Vol. 6303), 630304-1 to 630304-12. [Available online at https://e-reports-ext.llnl.gov/pdf/335213.pdf.]

  • Lindelöw, P., 2007: Fiber based coherent lidars for remote wind sensing. Ph.D. dissertation, Technical University of Denmark, 186 pp.

  • Sharma, R., , Pervez Y. , , and Pervez S. , 2005: Seasonal evaluation and spatial variability of suspended particulate matter in the vicinity of a large coal-fired power station in India—A case study. Environ. Monit. Assess., 102, 113.

    • Search Google Scholar
    • Export Citation
  • Sonnenschein, C. M., , and Horrigan F. A. , 1971: Signal-to-noise relationships for coaxial systems that heterodyne backscatter from the atmosphere. Appl. Opt., 10, 16001604.

    • Search Google Scholar
    • Export Citation
  • Tunick, A., 2003: CN2 model to calculate the micrometeorological influences on the refractive index structure parameter. Environ. Modell. Software, 18, 165171.

    • Search Google Scholar
    • Export Citation
  • U.S. Environmental Protection Agency, 2010: Our nation’s air: Status and trends through 2008. Office of Air Quality Planning and Standards EPA Publ. EPA-454/R-09-002, 54 pp.

  • Wagner, R., , Antoniou I. , , Pedersen S. M. , , Courtney M. S. , , and Jørgensen H. E. , 2009: The influence of the wind speed profile on wind turbine performance measurements. Wind Energy, 12, 348362.

    • Search Google Scholar
    • Export Citation
  • Wiser, R., , and Bolinger M. , 2010: 2009 wind technologies market report. National Renewable Energy Laboratory Rep. NREL/TP-6A2-48666, 88 pp.

  • Wulfmeyer, V., , and Feingold G. , 2000: On the relationship between relative humidity and particle backscattering coefficient in the marine boundary layer determined with differential absorption lidar. J. Geophys. Res., 105 (D4), 47294741.

    • Search Google Scholar
    • Export Citation
  • Yura, H. T., 1979: Signal-to-noise ratio of heterodyne lidar systems in the presence of atmospheric turbulence. J. Mod. Opt., 26, 627644.

    • Search Google Scholar
    • Export Citation
Save