Estimation of Wind Vector Profile Using a Hexarotor Unmanned Aerial Vehicle and Its Application to Meteorological Observation up to 1000 m above Surface
1. Introduction
Unmanned aerial vehicles (UAVs), also known as drones, have recently become promising tools for meteorological measurements. In particular, small and lightweight (<20 kg; Anderson and Gaston 2013) UAVs have become easy to operate and are cost efficient, owing to technological advances in general, and the reduction in the cost of inertial measurement units, GPS, and onboard computers, in particular. Small, lightweight UAVs are potentially applicable to a wide range of research purposes. These include air quality studies (reviewed by Villa et al. 2016), ecological studies (reviewed by Anderson and Gaston 2013), postdisaster surveys (e.g., Nishijima et al. 2015), and meteorological measurements, such as wind vector profiling, which is the main topic of this study.
Meteorological vertical profile measurements have traditionally been conducted using towers and radiosondes. These have the advantages of long-term stable measurements and long-range vertical profiling. On the other hand, their disadvantages are spatial inflexibility for towers, and high dependence on wind condition and no reusability for radiosondes. Meteorological vertical profile measurements have also been obtained by Doppler effect instruments, such as the lidar and the sodar (e.g., Lang and McKeogh 2011), with some specific instruments having been tested in harsh and hostile environments (Argentini et al. 2013; Casasanta et al. 2014). The Doppler effect instruments can estimate the meteorological profile continuously with higher vertical resolution. However, the instruments are expensive and limited in spatial location, as they require electric power to function. Therefore, using UAVs could compensate for the abovementioned disadvantages, owing to their flexibility, controllability, and repeatability. It is important to note that manned operations are necessary for continuous monitoring of UAV measurements.
The idea of meteorological vertical profile measurements by UAV is not new. It was proposed in the 1970s (e.g., Konrad et al. 1970). More recently, Egger et al. (2002) observed vertical profiles of temperature and humidity up to 2000 m in a Himalayan valley using a fixed-wing UAV. In addition to temperature and humidity, successive experiments to measure vertical profiles of wind vectors by fixed-wing UAVs [more specifically the Small Unmanned Meteorological Observer (SUMO) and the Meteorological Mini Aerial Vehicle(M2AV)] have been conducted in Europe (Reuder et al. 2009, 2012; Mayer et al. 2012; Van den Kroonenberg et al. 2008, 2012; Martin et al. 2011). In addition, Niedzielski et al. (2017) showed the feasibility of wind measurement using consumer-grade fixed-wing UAVs. Since fixed-wing UAVs cannot measure wind vectors directly, they are indirectly derived from airspeed vectors measured by pressure sensors and ground speed vectors measured by the GPS.
Rotor-based UAVs (RB-UAVs) differ from fixed-wing ones with respect to functionality. RB-UAVs can hover at designated locations, fly at lower altitude, offer excellent spatial flexibility, and take off/land in significantly smaller spaces. Therefore, RB-UAVs are more suitable for vertical profiling experiments (Anderson and Gaston 2013). This combination of features enables meteorological measurements over complicated topography (e.g., mountainous areas). So far, there are fewer studies on wind profile measurement using RB-UAVs than those using fixed-wing ones. A wind estimation method based on the relationship between wind vectors and UAV inclination angle was proposed by Neumann and Bartholmai (2015). This estimation method is explained in section 4. Neumann and Bartholmai (2015) showed that the root-mean-square error (RMSE) of wind speed estimation was 0.6 m s−1. Brosy et al. (2017) measured wind profiles, temperature, humidity, and methane concentration up to 150 m, based on the method proposed by Neumann and Bartholmai (2015). Brosy et al. (2017) concluded that the measured wind profile was in accordance with results from other kinds of observations, such as those by lidar and sodar. Palomaki et al. (2017) compared wind measurements of an onboard ultrasonic anemometer and the inclination angle–based wind estimation (Neumann and Bartholmai 2015). Their conclusion was that the RMSE of wind speed and wind direction were typically 0.5 m s−1 and 30°, respectively, using both approaches. The inclination angle–based method requires wind tunnel experiments or several field experiments for the calibration of the drag coefficient. Furthermore, the accuracy of estimations of higher wind speeds (outside the range of calibration experiments) is undetermined. Brosy et al. (2017) and Palomaki et al. (2017) conducted their field experiments only under low to moderate wind conditions (5 m s−1). Therefore, an in-depth investigation of such wind profile measurements by RB-UAV is required.
In this study we investigated the feasibility of wind vector profile measurement using an ultrasonic anemometer installed on a 1-m-wide hexarotor UAV. Comparisons with tower measurements under low to relatively higher wind speed conditions (11 m s−1) were carried out for validation. Then, its application for meteorological observations around a volcano was conducted for up to 1000 m.
2. Methodology
a. UAV
The UAV used in this study was a hexarotor-based UAV (SPIDER CS6; Fig. 1) developed by Luce Search Co., Ltd. (Japan; http://luce-s.net/services/spider/overview). It had horizontal dimensions of 95 cm × 95 cm, a height of 40 cm, and a weight of 3.8 kg. The additional available payload weight was 4 kg, and the available flight duration was 20–25 min with lithium polymer batteries.
Hexarotor UAV installed with an ultrasonic anemometer and a thermohygrometer.
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0186.1
Hexarotor UAV installed with an ultrasonic anemometer and a thermohygrometer.
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0186.1
Hexarotor UAV installed with an ultrasonic anemometer and a thermohygrometer.
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0186.1
Installed on the UAV was a 2D ultrasonic anemometer (FT702; FT Technologies). The accuracy is ±2° within ±10° datum and ±4° beyond ±10° datum for direction, ±0.5 m s−1 for wind speed lower than 15 m s−1 and ±4% for wind speed higher than 15 m s−1 (details described at http://www.fttechnologies.com/Wind-Sensors/Products/FT702). The ultrasonic anemometer can measure wind at 1-s intervals. It was installed on a 47-cm aluminum pole so that the downwash generated by the rotors would not disturb the wind measurements by the anemometer. Three wire stays prevented the pole from vibrating excessively. A datalogger and a battery for the ultrasonic anemometer were located on the deck at the center of the UAV. Another device (FUJITSU Arrows M02 FARM06006; http://www.fmworld.net/product/phone/m02/) was placed on the deck to record UAV vehicle position data (e.g., altitude, roll, and pitch). A thermohygrometer [HYT939; Innovative Sensor Technology (IST AG)] was also mounted on the deck. The accuracy of temperature is ±0.2 K (0°–60°C), while that of relative humidity (RH) is ±1.8% at 23°C (0% to 90% RH). Details are provided online (in https://www.ist-ag.com/sites/default/files/DHHYT939_E.pdf).
Although the objective of this study was wind vector profiling by UAV, other meteorological measurements could easily be captured by the UAVs. Temperature and mixing ratio profiles are discussed in section 3b, in addition to wind vector profiles.
b. Comparison with tower observations
For purposes of validation of wind measurements by the UAV, we compared the wind speed and direction measurements by the ultrasonic anemometer on the UAV with those measured by vane anemometers mounted on a meteorological observation tower (55 m high), located in the Uji-Gawa Open Laboratory of the Disaster Prevention Research Institute (DPRI), Kyoto University (Kyoto, Japan). Wind speed and direction were measured by vane anemometers (WS-B16; Ogasawara Keiki Seisakusho) at 40- and 55-m height on the tower. The accuracy of the vane anemometer was ±3° for direction, ±0.3 m s−1 for wind speed lower than 10 m s−1, and ±3% for wind speed higher than 10 m s−1. The starting wind speed was 0.3 m s−1. The wind speed and direction were recorded as 1-min averaged values and instantaneous values per minute, respectively.
Five UAV flights were operated on 26 October and 9 November 2016. The wind speed was low (~2 m s−1) on 26 October and high (~10 m s−1) on 9 November. The UAV was controlled to hover at 40 or 55 m above the ground level, beside the tower, for approximately 15 min, and the wind speed and direction were measured during the hovering. Information about the five flights is summarized in Table 1. The flights were denoted as 1026No1, 1026No2, 1109No1, 1109No2, and 1109No3. Wind vectors measured during flights 1026No1 and 1026No2 were compared with those measured by a vane anemometer at 40 m, and measurements during 1109No1, 1109No2, and 1109No3 were compared with the data from the anemometer at the 55-m height.
Overview of UAV flights for tower comparison.
c. Application for meteorological field observations
To demonstrate the potential of this UAV, meteorological observations were carried out up to 1000 m AGL. The observation site was Sakurajima, Japan (Fig. 2), at the center of which is an active volcano that erupts frequently (e.g., Poulidis et al. 2018). Spatially dense observations for wind vector profiles are required for accurate forecasting of volcanic ash transport and deposition (Poulidis et al. 2017). Observations were conducted on 19 April 2017. The observations were collected during 10 flights: 5 morning flights within 1 h, 17 min [1004–1121 Japan standard time (JST)] and 5 more in the afternoon within 1 h, 11 min (1353–1504 JST). The weather was fair during the observation. The wind speed and direction at 0900 JST was 17 m s−1 and 317° at 850 hPa, which was observed by a radiosonde at the Kagoshima station, approximately 10 km west of the UAV flight point (http://www.jma.go.jp/jma/en/Activities/upper/upper.html). The UAV was controlled to fly vertically while remaining stationary horizontally, and hovering for 3 min at 50, 100, 150, 200, 300, 500, 750, or 1000 m above the ground level, depending on the flight. The flight location is shown in Fig. 2 and the ground-level elevation was 85 m. Information about the 10 flights (denoted as 0419No1–No10) is summarized in Table 2. Flights 0419No1–No5 were also grouped as 0419 a.m. (morning observations), while flights 0419No6–No10 were grouped as 0419 p.m. (afternoon observations). Temperature, humidity, and UAV vehicle positions (altitude, roll, and pitch) were measured using onboard sensors. However, the UAV vehicle positions were not recorded during flight 0419No1 (hovering at 500 m) because of human error.
Field observation site at Sakurajima. The red round marker indicates the flight point of UAV. The two green markers indicate Doppler lidar locations. Color scale indicates elevation (m). Elevation data are based on data from the Geospatial Information Authority of Japan (http://www.gsi.go.jp/download).
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0186.1
Field observation site at Sakurajima. The red round marker indicates the flight point of UAV. The two green markers indicate Doppler lidar locations. Color scale indicates elevation (m). Elevation data are based on data from the Geospatial Information Authority of Japan (http://www.gsi.go.jp/download).
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0186.1
Field observation site at Sakurajima. The red round marker indicates the flight point of UAV. The two green markers indicate Doppler lidar locations. Color scale indicates elevation (m). Elevation data are based on data from the Geospatial Information Authority of Japan (http://www.gsi.go.jp/download).
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0186.1
Overview of flights for field meteorological observation at Sakurajima
The wind vector profiles were also measured by the optical-fiber Doppler lidar (LR-07F type III S; Mitsubishi Electric Corporation). Its wavelength was 1.5–1.6 μm; the resolution was 30 m; the available measurement range was 30–600 m, depending on the weather condition; the laser pulse width was 200 ns; and pulse repetition rate was 4 kHz. The Doppler lidar measurements were conducted at two locations in Sakurajima (shown in Fig. 2). The ground-level elevation was 408 and 85 m for the Haruta and Kurokami locations, respectively. The wind profile generated using Doppler lidar data is discussed and compared with that measured by UAV.
Japanese aviation law requires official permission from the Ministry of Land, Infrastructure, Transport and Tourism (MLIT) of Japan for UAV flights above 150 m. Our flights were carried out with permission from MLIT.
3. Results
a. Comparison with tower observations
The validation of the wind vectors measured by the ultrasonic anemometer on the UAV is described here. First, the time series of wind speed and direction are compared between the UAV and tower observations. Figure 3 shows the time series during flight 1109No3 as an example. The wind speed during flight 1109No3 (11 m s−1) was the highest among all five flights (Table 1). The ultrasonic anemometer on the UAV can measure wind at 1-s intervals, while the vane anemometers measure winds averaging over 1 min. For this reason the wind speed measured by the UAV was also averaged for 1 min. In contrast, the wind direction is an instantaneous value recorded each minute. The time series of 1-min average wind speed and instantaneous direction measured by the UAV were in good agreement with the vane anemometer on the tower. The bias was 0.1 m s−1 for wind speed and −9° for wind direction. The RMSE was 0.4 m s−1 and 10°, respectively. The RB-UAV measurements of wind in previous studies (Neumann and Bartholmai 2015; Palomaki et al. 2017; Brosy et al. 2017) were not validated at such high wind speeds as in this study.
Time series of measured (a) wind speed and (b) wind direction during flight 1109No3. The red lines with circle markers are 1-min average values for wind speed in (a) and 1-min instantaneous values for wind direction in (b).
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0186.1
Time series of measured (a) wind speed and (b) wind direction during flight 1109No3. The red lines with circle markers are 1-min average values for wind speed in (a) and 1-min instantaneous values for wind direction in (b).
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0186.1
Time series of measured (a) wind speed and (b) wind direction during flight 1109No3. The red lines with circle markers are 1-min average values for wind speed in (a) and 1-min instantaneous values for wind direction in (b).
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0186.1
The results for all five flights are summarized below. Figure 4 shows the overall comparison of the results for the five UAV flights with those for the tower. Wind speed and direction measured by the UAV and tower are in good agreement from low to higher wind speeds. The overall bias is 0.5 m s−1 for wind speed and −9° for wind direction. The RMSE is 0.6 m s−1 and 12°, respectively. Subtracting the bias from the UAV values, the RMSE values become 0.4 m s−1 and 9°, respectively. The correlation coefficient is 0.993 and 0.997, respectively. The wind speed measured by the UAV is higher by 0.5–1 m s−1 than the tower observations, except for 1109No3. We have confirmed that the instrumental difference (bias) between the ultrasonic anemometer and the vane anemometer is insignificant. Compared with the wind speed measured by the UAV while remaining on the ground and hovering near the ground, we found that wind speeds measured during hovering were higher by approximately 0.5 m s−1 than while remaining on the ground. Therefore, we conclude that the positive bias of wind speed (+0.5 m s−1) was due to the rotors. Palomaki et al. (2017) also estimated the average positive bias of an ultrasonic anemometer from the rotors on the UAV to be approximately 0.5 m s−1, although our ultrasonic anemometer was vertically separated from the rotors by 47 cm, which was more than the 30-cm separation of Palomaki et al. (2017). The bias (+0.5 m s−1) is corrected in the following section. It is possible that tilting of the UAV affects the measured horizontal wind speed. The tilt of the UAV tends to increase with increasing wind speeds (details in section 4). The slope of the regression line of the measured wind speed between the vane anemometer on the tower and the ultrasonic anemometer on the hovering UAV is 0.95 (Fig. 4a), while the instrumental difference between the two anemometers is 0.94. Since the difference between the results is nearly equal, the effects of the tilt of the UAV can be ignored.
Overall comparison of (a) 1-min average wind speed and (b) 1-min instantaneous wind direction between UAV and tower. Each marker indicates a different flight. Bias, RMSE, RMSE subtracted by bias (RMSEsb), and correlation coefficient (R) are also mentioned. The least squares regression line is plotted (broken line).
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0186.1
Overall comparison of (a) 1-min average wind speed and (b) 1-min instantaneous wind direction between UAV and tower. Each marker indicates a different flight. Bias, RMSE, RMSE subtracted by bias (RMSEsb), and correlation coefficient (R) are also mentioned. The least squares regression line is plotted (broken line).
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0186.1
Overall comparison of (a) 1-min average wind speed and (b) 1-min instantaneous wind direction between UAV and tower. Each marker indicates a different flight. Bias, RMSE, RMSE subtracted by bias (RMSEsb), and correlation coefficient (R) are also mentioned. The least squares regression line is plotted (broken line).
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0186.1
Brosy et al. (2017) concluded that the RMSEs of the measured wind speed and direction were 0.7 m s−1 and 14.5°, respectively, for 10-s average wind, built on the UAV inclination–based estimation. Palomaki et al. (2017) concluded that the RMSEs of the measured wind speed and direction were typically 0.5 m s−1 and 30°, respectively, for 1-s wind using both UAV inclination–based estimation and an onboard anemometer estimation. Our results show that the accuracy of wind estimation by onboard ultrasonic anemometers is comparable to the previous studies, even for higher wind speeds (11 m s−1). Therefore, we conclude that wind estimation using ultrasonic anemometers on RB-UAVs is sufficiently accurate and can be applied over a wide range of wind speed conditions.
b. Field meteorological observations at Sakurajima
The conclusion from the previous section is that ultrasonic anemometers installed on RB-UAVs can measure wind vectors accurately. In this section the method is applied to meteorological field observations at Sakurajima. Figures 5 and 6 show the wind speed and direction profiles measured by the UAV up to 1000-m altitude. Figures 5 and 6 represent 0419 a.m. and 0419 p.m., respectively. Wind values measured during the same flight were plotted using the same marker. Therefore, five types of markers were plotted for the five flights. Wind profiles measured by Doppler lidar are shown by black lines, while sidebars indicate standard deviations of 3-min averaged values. The values are averaged from 1000 to 1130 JST for 0419 a.m. and 1400–1500 JST for 0419 p.m., corresponding to the flight time (Table 2). There are two Doppler lidar observation site locations (Haruta and Kurokami) as shown in Fig. 2. Haruta, Kurokami, and the UAV flight point are located to the west, east, and south of the mountain in Sakurajima, respectively (Fig. 2). The horizontal separation between Haruta and the UAV flight point is 5 km, while that between Kurokami and the UAV flight point is 4 km. The Doppler lidar measurements at Haruta are plotted above 323 m (408–85 m) because the flight start point and Kurokami were at the same ground elevation of 85 m.
(a) Wind speed and (b) wind direction profiles measured by the UAV during flight group 0419 a.m. The wind profiles measured by the two Doppler lidars are also plotted (black lines). The sidebar indicates the standard deviation of the 3-min averaged value of Doppler lidar.
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0186.1
(a) Wind speed and (b) wind direction profiles measured by the UAV during flight group 0419 a.m. The wind profiles measured by the two Doppler lidars are also plotted (black lines). The sidebar indicates the standard deviation of the 3-min averaged value of Doppler lidar.
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0186.1
(a) Wind speed and (b) wind direction profiles measured by the UAV during flight group 0419 a.m. The wind profiles measured by the two Doppler lidars are also plotted (black lines). The sidebar indicates the standard deviation of the 3-min averaged value of Doppler lidar.
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0186.1
As in Fig. 5, but for flight group 0419 p.m.
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0186.1
As in Fig. 5, but for flight group 0419 p.m.
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0186.1
As in Fig. 5, but for flight group 0419 p.m.
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0186.1
In Fig. 5, for the flight group 0419 a.m., the range of wind speeds measured by the UAV are from 6.5 to 13.4 m s−1, depending on the altitudes of the measurements. The wind directions range from 287° to 335°. For the Doppler lidar at the Kurokami location, the minimum wind speed was 7.0 m s−1 at 59-m altitude, while the maximum wind speed was 9.8 m s−1 at 353 m. For the Doppler lidar at the Haruta location, the minimum wind speed was 7.6 m s−1 at 648 m, while the maximum wind speed was 11.2 m s−1 at 707 m. The wind speed profiles measured by the UAV tend to increase from the surface to 300 m, being constant between 300 and 750 m, and subsequently increasing again between 750 and 1000 m. This tendency can be considered to represent the features of the surface, convective boundary, and entrainment layers. The wind direction profiles measured by the UAV tend to be constant below 750 m. Although the values measured at the locations of the UAV and Doppler lidar are supposed to vary quantitatively because the locations were several kilometers apart, the wind vector profiles of the UAV were qualitatively consistent with those from the Doppler lidar.
For the flight group 0419 p.m. (Fig. 6), the wind speeds measured by the UAV range from 6.6 to 9.7 m s−1, depending on the altitudes of measurement, while the wind directions range from 242° to 336°. For the Doppler lidar at the Kurokami location, the minimum wind speed was 7.0 m s−1 at 59 m, while the maximum wind speed was 9.8 m s−1 at 353 m. For the Doppler lidar at the Haruta location, the minimum wind speed was 7.6 m s−1 at 648 m, while the maximum wind speed was 11.2 m s−1 at 707 m. The wind speed profile measured by the UAV tends to increase up to 300 m, becoming constant between 300 and 1000 m. The wind direction profile measured by the UAV tends to rotate clockwise from the surface to 1000 m. This tendency is consistent with other Doppler lidar results. The differences between the UAV and Doppler lidar results can be attributed to the mountain effects. Because of this, a UAV can estimate the wind vector profile like a Doppler lidar does, at least qualitatively. Therefore, UAV can spatially compensate for wind profile measurements made using Doppler lidar. Doppler lidars are expensive and limited in spatial location because they require a power source.
A thermohygrometer was also installed on the RB-UAV. Since the temperature and humidity were measured in addition to the wind vector profile, meteorological conditions can be discussed in detail. Figure 7 shows the temperature and mixing ratio profile measured by the UAV, in addition to the wind vector profile. In this figure, the average instantaneous values measured at a certain height during the ascent and descent of the UAV (i.e., not hovering) are shown by thin lines. The wind direction, temperature, and mixing ratio during the vertical moving did not vary significantly among the flights, as compared to the wind speed. Up to 750 m, the temperature decreased by 1.02°C per 100 m as a result of the nearly dry adiabatic lapse rate. This decrease ceased between 750 and 1000 m. The mixing ratio sharply decreased between 750 and 1000 m, while the wind speed increased in this altitude range. Therefore, it was found that the top of the mixed layer was located at approximately 850 m for the duration of the measurement. Figure 8 is the same as Fig. 7, but it shows the flight group 0419 p.m. As described above, the wind speed profiles tend to be constant above 300 m. Moreover, the temperature and mixing ratio show no drastic change, with the temperature decreasing by 0.99°C per 100 m, up to 1000 m. Therefore, it was found that the mixed layer was above 1000 m at the time of measurement.
(a) Wind speed, (b) wind direction, (c) temperature, and (d) mixing ratio profile measured by the UAV during 0419 a.m. Markers indicate averaged values during hovering, and thin lines indicate the averaged value measured during the ascent and descent of the UAV (not hovering). The same flights are indicated by lines and markers of the same color. Markers are the same as in Fig. 5. Gray horizontal bars indicate standard deviation during hovering.
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0186.1
(a) Wind speed, (b) wind direction, (c) temperature, and (d) mixing ratio profile measured by the UAV during 0419 a.m. Markers indicate averaged values during hovering, and thin lines indicate the averaged value measured during the ascent and descent of the UAV (not hovering). The same flights are indicated by lines and markers of the same color. Markers are the same as in Fig. 5. Gray horizontal bars indicate standard deviation during hovering.
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0186.1
(a) Wind speed, (b) wind direction, (c) temperature, and (d) mixing ratio profile measured by the UAV during 0419 a.m. Markers indicate averaged values during hovering, and thin lines indicate the averaged value measured during the ascent and descent of the UAV (not hovering). The same flights are indicated by lines and markers of the same color. Markers are the same as in Fig. 5. Gray horizontal bars indicate standard deviation during hovering.
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0186.1
As in Fig. 7, but for flight group 0419 p.m.
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0186.1
As in Fig. 7, but for flight group 0419 p.m.
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0186.1
As in Fig. 7, but for flight group 0419 p.m.
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0186.1
To summarize this section, we found that the quantitative estimates of wind vector profiles up to 1000 m by the UAV and the Doppler lidar were similar. The UAVs can easily carry this type of thermohygrometer. The vertical meteorological conditions (e.g., the mixed layer altitude) are well estimated when combined with the wind vector profile.
4. Discussion: Inclination angle–based wind speed estimation
In this study we deployed an ultrasonic anemometer installed on an RB-UAV for wind estimation. Brosy et al. (2017) and Palomaki et al. (2017) used UAV inclination angle–based wind speed estimation, as explained by Eqs. (1) and (2) below. For the meteorological observations obtained in Sakurajima (described in section 3b), the UAV flight positions (pitch and roll) were also recorded by an onboard sensor. The relationship between the UAV inclination angle and wind speed, as measured by an onboard ultrasonic anemometer, is explored as follows.
Relationship between UAV inclination angle and squared wind speed measured during flights 0419No2–No10 while hovering (Table 2). The 15 filled markers and small dots indicate 3-min and 1-s averaged values, respectively. Black bars are standard deviations of 1-s values. Color coordinates show the wind direction measured by the ultrasonic anemometer.
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0186.1
Relationship between UAV inclination angle and squared wind speed measured during flights 0419No2–No10 while hovering (Table 2). The 15 filled markers and small dots indicate 3-min and 1-s averaged values, respectively. Black bars are standard deviations of 1-s values. Color coordinates show the wind direction measured by the ultrasonic anemometer.
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0186.1
Relationship between UAV inclination angle and squared wind speed measured during flights 0419No2–No10 while hovering (Table 2). The 15 filled markers and small dots indicate 3-min and 1-s averaged values, respectively. Black bars are standard deviations of 1-s values. Color coordinates show the wind direction measured by the ultrasonic anemometer.
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0186.1
With regard to the wind direction dependence, a measurement with the same wind direction as the exception (330°) can be seen to have a squared wind speed of 90 m2 s−2 and a 
On the other hand, if stronger downward flow existed, then the inclination angle could be considered to decrease because additional uplift force is required. As described in section 3b, the wind speed (for 0419No5) can be measured just above the mixed layer (in other words, the entrainment layer). It is possible that a stronger downward flow exists as the entrainment flow, or a downward flow by the mountain wave because the flight was near the mountain (volcano).
The standard deviations of 
5. Conclusions
Small unmanned aerial vehicles (UAVs), also known as drones, have been adapted for use in various fields recently and appear to be promising tools for meteorological measurements. We investigated the feasibility of wind vector profile measurement using an ultrasonic anemometer installed on a small (1 m wide) hexarotor UAV. First, the wind vectors measured by the UAV were compared with observations from a 55-m-high meteorological tower. The range of wind speed investigated in this study was up to 11 m s−1. To demonstrate the method, it was applied to field meteorological observations near a volcano, and the wind vector profiles (as well as temperature and humidity) were measured by the UAV up to 1000 m.
Upon comparison with the meteorological tower observations, the wind speed and direction measured by the UAV and the tower were in good agreement for wind speeds up to 11 m s−1. The overall bias was 0.5 m s−1 for 1-min averaged wind speed and −9° for wind direction. The RMSE was 0.6 m s−1 and 12°, respectively. The positive bias of wind speed (+0.5 m s−1) was due to the rotors. Subtracting the bias from the UAV values, the RMSE values become 0.4 m s−1. Therefore, we conclude that an ultrasonic anemometer installed on a RB-UAV can measure wind vectors appropriately. This further demonstrates that the UAV is useful for meteorological field observation. The UAV can also capture wind vector profiles up to 1000 m, with its quality being consistent with the Doppler lidar estimation at two locations several kilometers away from the UAV location. The measured wind vector profile can represent the features of the surface layer, the mixed layer, and the entrainment layer. UAVs can easily carry a thermohygrometer, allowing vertical meteorological conditions to be well measured by coordinating with the wind vector profile data.
We demonstrated the feasibility of wind vector profiling using an RB-UAV for a wider range of wind speeds (up to 11 m s−1) and higher altitudes (up to 1000 m) than those of previous studies (Palomaki et al. 2017; Brosy et al. 2017). This study focused on two-dimensional wind speed measurements, leaving the three-dimensional one for future work. Other sensors such as gas sensors and particle counters could be easily carried and could fly horizontally, although the effect of RB-UAV horizontal motion on wind measurements needs to be estimated. Therefore, small RB-UAVs are very effective for meteorological and related measurements.
Acknowledgments
We thank Dr. Mitsuaki Horiguchi for providing the wind data measured by the meteorological observation tower at the Uji-Gawa Open Laboratory of DPRI of Kyoto University, and Mr. Tomoki Kobayashi and the team ReNA of the Japan Weather Association for supporting our field observations. The UAV flights above 150-m altitude at Sakurajima were conducted with permission from the Ministry of Land, Infrastructure, Transport and Tourism (MLIT) of Japan and from air traffic control at the Kagoshima airport. This research was partially supported by the Integrated Program for Next Generation Volcano Research and Human Resource Development of the Ministry of Education, Culture, Sports, Science, and Technology (MEXT), Japan.
REFERENCES
Anderson, K., and K. J. Gaston, 2013: Lightweight unmanned aerial vehicles will revolutionize spatial ecology. Front. Ecol. Environ., 11, 138–146, https://doi.org/10.1890/120150.
Argentini, S., I. Petenko, A. Viola, G. Mastrantonio, I. Pietroni, G. Casasanta, E. Aristidi, and C. Ghenton, 2013: The surface layer observed by a high-resolution sodar at DOME C, Antarctica. Ann. Geophys., 56, F0557, https://doi.org/10.4401/ag-6347.
Brosy, C., K. Krampf, M. Zeeman, B. Wolf, W. Junkerman, K. Schäfer, S. Emels, and H. Kunstmann, 2017: Simultaneous multicopter-based air sampling and sensing of meteorological variables. Atmos. Meas. Tech., 10, 2773–2784, https://doi.org/10.5194/amt-10-2773-2017.
Casasanta, G., I. Pietroni, I. Petenko, and S. Argentini, 2014: Observed and modelled convective mixing-layer height at DOME C, Antarctica. Bound.-Layer Meteor., 151, 597–608, https://doi.org/10.1007/s10546-014-9907-5.
Egger, J., and Coauthors, 2002: Diurnal winds in the Himalayan Kali Gandaki valley. Part III: Remotely piloted aircraft soundings. Mon. Wea. Rev., 130, 2042–2058, https://doi.org/10.1175/1520-0493(2002)130<2042:DWITHK>2.0.CO;2.
Konrad, T., M. Hill, J. R. Rowland, and J. Meyer, 1970: A small, radio-controlled aircraft as a platform for meteorological sensors. APL Tech. Dig., 10, 11–19.
Lang, S., and E. McKeogh, 2011: LIDAR and SODAR measurements of wind speed and direction in upland terrain for wind energy purposes. Remote Sens., 3, 1871–1901, https://doi.org/10.3390/rs3091871.
Martin, S., J. Bange, and F. Beyrich, 2011: Meteorological profiling of the lower troposphere using the research UAV “M2AV Carolo.” Atmos. Meas. Tech., 4, 705–716, https://doi.org/10.5194/amt-4-705-2011.
Mayer, S., G. Hattenberger, P. Brisset, M. O. Jonassen, and J. Reuder, 2012: A ‘no-flow-sensor’ wind estimation algorithm for unmanned aerial systems. Int. J. Micro Air Veh., 4, 15–29, https://doi.org/10.1260/1756-8293.4.1.15.
Neumann, P. P., and M. Bartholmai, 2015: Real-time wind estimation on a micro unmanned aerial vehicle using its inertial measurement unit. Sens. Actuators, 235A, 300–310, https://doi.org/10.1016/j.sna.2015.09.036.
Niedzielski, T., and Coauthors, 2017: Are estimates of wind characteristics based on measurements with Pitot tubes and GNSS receivers mounted on consumer-grade unmanned aerial vehicles applicable in meteorological studies? Environ. Monit. Assess., 189, 189–431, https://doi.org/10.1007/s10661-017-6141-x.
Nishijima, K., N. Mori, T. Yasuda, T. Shimura, J. T. Gogon, D. Gibson, and F. Jockley, 2015: DPRI-VMGD joint survey for Cyclone Pam damages. DPRI and VMGD Doc., 17 pp., http://www.taifu.dpri.kyoto-u.ac.jp/wp-content/uploads/2015/05/DPRI-VMGD-survey-first-report-Final.pdf.
Palomaki, R. T., N. T. Rose, M. van den Bossche, T. J. Sherman, and S. F. De Wekker, 2017: Wind estimation in the lower atmosphere using multirotor aircraft. J. Atmos. Oceanic Technol., 34, 1183–1191, https://doi.org/10.1175/JTECH-D-16-0177.1.
Poulidis, A. P., T. Takemi, M. Iguchi, and I. A. Renfrew, 2017: Orographic effects on the transport and deposition of volcanic ash: A case study of Mt. Sakurajima, Japan. J. Geophys. Res. Atmos., 122, 9332–9350, https://doi.org/10.1002/2017JD026595.
Poulidis, A. P., T. Takemi, A. Shimizu, M. Iguchi, and S. F. Jenkins, 2018: Statistical analysis of dispersal and deposition patterns of volcanic emissions from Mt. Sakurajima, Japan. Atmos. Environ., 179, 305–320, https://doi.org/10.1016/j.atmosenv.2018.02.021.
Reuder, J., P. Brisset, M. Jonassen, M. Müller, and S. Mayer, 2009: The Small Unmanned Meteorological Observer SUMO: A new tool for atmospheric boundary layer research. Meteor. Z., 18, 141–147, https://doi.org/10.1127/0941-2948/2009/0363.
Reuder, J., M. Jonassen, and H. Ólafsson, 2012: The Small Unmanned Meteorological Observer SUMO: Recent developments and applications of a micro-UAS for atmospheric boundary layer research. Acta Geophys., 60, 1454–1473, https://doi.org/10.2478/s11600-012-0042-8.
Van den Kroonenberg, A. C., T. Martin, M. Buschmann, J. Bange, and P. Vörsmann, 2008: Measuring the wind vector using the autonomous mini aerial vehicle M2AV. J. Atmos. Oceanic Technol., 25, 1969–1982, https://doi.org/10.1175/2008JTECHA1114.1.
Van den Kroonenberg, A. C., S. Martin, F. Beyrich, and J. Bange, 2012: Spatially-averaged temperature structure parameter over a heterogeneous surface measured by an unmanned aerial vehicle. Bound.-Layer Meteor., 142, 55–77, https://doi.org/10.1007/s10546-011-9662-9.
Villa, T. F., F. Gonzalez, B. Miljievic, Z. D. Ristovski, and L. Morawska, 2016: An overview of small unmanned aerial vehicles for air quality measurements: Present applications and future prospectives. Sensors, 16, 1072, https://doi.org/10.3390/s16071072.
