A Tethered And Navigated Air Blimp ( TANAB ) for observing the microclimate over a complex terrain

This study presents the first environmental monitoring field campaign of a newly developed Tethered And Navigated Air Blimp (TANAB) system to investigate the microclimate over a complex terrain. The use of a tethered balloon in complex terrains such as mines and tailings ponds is novel and the focus of the present study. The TANAB system was fully developed and launched at a mine facility in northern Canada in May 2018. This study describes the key design features, the sensor payload onboard, and the observations made by the TANAB system. The system measured meteorological conditions including wind 5 speed in three directions, temperature, relative humidity, and pressure over the first few tens of meters of the atmospheric boundary layer. The system also performed earth surface thermal imaging, or temperature mapping, of the underlying surface. The measurements were made at two primary locations in the facility: i) near a tailings pond and ii) in a mine pit. TANAB measured the dynamics of the atmosphere at different diurnal times (e.g. day versus night) and locations (near tailings pond versus inside the mine). Such dynamics include mean and turbulence statistics pertaining to flow momentum and energy, and 10 they are crucial in the understanding of emission fluxes from the facility in future studies. In addition, TANAB can provide boundary conditions and validation datasets to support mesoscale dispersion modelling or Computational Fluid Dynamics (CFD) simulations for various transport models.

Response: We appreciate your comments. We have included a detailed literature survey and references in the revised version of the manuscript as the following in Sect. 1.1.
Airborne systems are increasingly being used for atmospheric measurements (Martin et al., 2011;Palomaki et al., 2017) although recently their use is being regulated more restrictively. For instance, rotary or fixed-wing drones are not permitted to fly in complex environments such as busy urban areas and airports. On the other hand, tethered balloon-based atmospheric measurement techniques have been used widely for obtaining the turbulence structure as well as the mean vertical profiles of the ABL in complex environments (Thompson, 1980). One of the main advantages of a tethered balloon system is its ability to profile a significant portion of the planetary boundary layer starting from the surface, which is not possible or economical by ground-based or aircraft-based atmospheric measurement techniques (Egerer et al., 2019). The use of ultrasonic anemometers in tethered balloons have been reported in many studies (Stevens et al., 2013;. In comparison, one of the disadvantage of Pitot tubes is their inability to measure the low wind speeds. So they require a fast flying probe which cannot fly in a complex environment for safety and logistic reasons. Ultrasonic anemometers, on the other hand, are popular because of their continuous measurement characteristics, high accuracy, and their ability to be levitated to measure low velocities (Martikainen et al., 2010;. In addition, studies have shown that tethered balloon-based probes can be used for continuous measurement of the important parameters in the ABL without influencing the flow compared to tower-mounted instruments, where the tower structure can perturb the flow measurement, or presence of complex topography or nearby buildings that can influence the measurement (Haugen et al., 1975;Stevens et al., 2013).
Tethered balloon-borne atmospheric turbulence measurements have a long history of observations over the land (Smith, 1961) and sea (Thompson, 1972) to measure fluxes of heat and moisture at heights up to a few hundred metres. The most notable tethered balloon systems deployed collected data in campaigns in the late 1960s and 1970s including the Barbados Oceanographic and Meteorological Experiment (BOMEX) (Davidson, 1968;Garstang and La Seur, 1968;Friedman and Callahan, 1970), the Joint Air-Sea Interaction (JASIN) Experiment (Pollard, 1978), and the Global Atmospheric Research Programme (GARP) Atlantic Tropical Experiment (GATE) (Berman, 1976). In BOMEX a tethered balloon system was operated from the deck, which measured temperature, wind, and humidity continuously, at different levels in the range of 0 to 600 m in the ocean area north and east of the Island of Barbados. In JASIN tethered balloons were used to measure the structure of ABL to understand the air-sea interaction in the North Atlantic. In the recent past, tethered balloon systems have been used in Boundary-Layer Late Afternoon and Sunset Turbulence (BLLAST) field campaign that was conducted in southern France (Lothon et al., 2014).  used an ultrasonic anemometer mounted on a tethered balloon system for turbulent flux and variance measurements. Egerer et al. (2019) used the BELUGA (Balloon-bornE moduLarUtility) tethered balloon system for profiling the lower Atmosphere by turbulence and radiation measurements in the Arctic.
3. More recently, ultrasonic anemometers (rather than say Gill anemometers) on tethered systems were first made in the 1990s, and in the last 5 years remotely pilot aircraft systems (RPAS) have become the standard platforms for ABL turbulence, and this has been reported in dozens of publications.
Response: Thank you for your comments. We have incorporated more information about the use of ultrasonic anemometers on airborne measurements and the associated advantages. Please refer to the previous comment.
4. Lack of instrumental rigour. Both the ultrasonic anemometer data and the remote sensed surface temperatures are not suitably calibrated. I could not find any mention of the means to convert the anemometer wind measurements of [u,v,w] into the reference of the earth: this requires a very significant effort in both analysis of the platform movement (via Inertial Navigation Unit (IMU), not solely from three-axes tilt) and calibration of the resulting variances and co-variances. The wind tunnel data are not suitable for this analysis as both units (I assume) were static.
Response: Thank you for pointing out the fact that the ultrasonic anemometer data calibration section were missing important details. Now we have added the gondola motion as well as the calibration experimental set-up in detail in Sect. 2.1 and Sect. 3.1 of the revised manuscript, respectively. We have considered separate calibration equations for both mean and turbulence statistics of the flow. We did not require to convert TriSonica TM Mini's measurement of flow statistic to a fixed reference frame of the earth. We only required to understand the flow statistics in the along-wind, cross-wind, and vertical-wind directions. As shown below, the gondola turns against the mean wind most of the time and it is level within ±5 • in both the pitch and roll angles more than 90% of the time. The gondola only rotates in the yaw direction when the main wind direction changes. This enables us to interpret the flow statistics in the three directions mentioned. Sect. 2.1 of the manuscript now contains the following text.
The gondola is a part of the tethered balloon system controlled by multiple ropes on the ground. The acting forces on the system are 1) the lift force due to the helium-filled balloon, 2) the force of gravity, 3) the tension forces due to the ropes, and 4) the drag forces due to the wind. In the absence of the drag force the remaining forces are acting in the vertical direction and the system ascends up or down, but the presence of the drag force displaces the system in the horizontal direction. At all times these forces are balanced so that the TANAB is in a quasi-stationary position.
The balloon is equipped with a net that helps facing the balloon against the main wind direction at any moment. The net guides the air on one side, and the pressure force stops the balloon from rotating. In case of wind direction changing, the pressure force builds up on the net creating a torque around the centre of rotation that repositions the balloon facing the main wind. Up to three ropes are used to tether the balloon to help with stabilizing the system, especially during high winds associated with convective boundary layers. The ropes, however, impose extra weight on the system so that the vertical range of the system reduces when three ropes are used as opposed to one.
The evidence that the net mechanism succeeds in aligning the balloon against the main wind direction can be found in Fig. 1. The wind-rose shows the wind direction records by TriSonica TM Mini (at TriSonica TM Mini's coordinate) for over 56 hr of flight. According to the wind-rose the gondola mostly faces against the main wind direction because wind direction is recorded from the local north direction most of the time. Note that TriSonica TM Mini's north axis is the sensor local coordinate. If desired, this reading can be converted into the fixed inertial body of reference. For this to occur, the sensor yaw, pitch, and roll angles need to be used. This conversion, however, was not desired in the present analysis.
Figure 2 (left) shows the balloon operation in an unstable atmosphere with high winds when three ropes are used to stabilize it. A sudden drag force on the gondola may drive the system out of its stable position momentarily. Hence it may affect the quality of measurements by creating instabilities. Such phenomenon can be prevented by deploying extra ropes connecting the gondola directly to ground operators. The tension in these ropes balances the sudden drag force exerted on the system. This arrangement places the gondola in a quasi-stationary position in the air that indeed helps the stability of measurement in gusty conditions. Figure 2 (right) shows the balloon operation in a stable atmosphere with low winds when only two ropes are used to stabilize it.
A T-connector connects the balloon to the gondola using ropes allowing the gondola to hang freely while minimizing the pitch and roll angles to result in better measurements from a levelled gondola. Figure 3 shows the distributions for pitch and roll angles recorded by TriSonica TM Mini's compass over 56 flight hours. According to the bar charts, the gondola is level within ±5 • of pitch and roll angles more than 90% of the time. The gondola only rotates in the yaw direction when the main wind direction changes. The gondola is positioned 2 m below the balloon, so the effects of the balloon motion on the gondola are reduced. Figure 4 shows the wind tunnel used for ultrasonic anemometer calibration onboard of TANAB, which is an open circuit tunnel designed for turbulent boundary layer research. The reference sensor for calibration was a pitot tube which was used to calibrate an R.M. YOUNG 81000 ultrasonic anemometer and the ultrasonic anemometer onboard of TANAB while all three components of wind were considered from the R.M. YOUNG 81000 ultrasonic anemometer. The azimuth angle, elevation angle, and wind levels were changed, independently, to derive calibration coefficients for both mean and turbulence statistics as measured by the TriSonica TM Mini and calibrated against the R.M. YOUNG 81000. Figures 5 and 6 show the set-up for R.M. YOUNG 81000 and gondola calibration, respectively. Note that the gondola was hung from the tunnel with rope attachments to simulate its motion in the real atmosphere. Section 3.1 of the manuscript now contains the following text.
All the experiments were conducted in the University of Guelph's wind tunnel, which is an open circuit tunnel designed for turbulent boundary layer research. The cross sectional area is 1.2 m × 1.2 m. The tunnel is 10 m long. The tunnel's air speed is controlled by a gauge that sets the fan speed. The tunnel achieves wind speeds up to 10 m s −1 . The turbulence intensity is typically    less than 2 % if no roughness blocks are placed upstream of the flow. The Reynolds number characterizes the turbulence level of the fluid flow and is defined as the ratio of the inertial to viscous forces given by Re = ρ×U ×L µ , where U is the flow velocity, L is the characteristic length scale of the system (commonly, the hydraulic diameter of the wind tunnel), and µ and ρ are the dynamic viscosity and density of the fluid, respectively. In the present study, the wind tunnel's Re number varied between 150,000 and 1,100,000. Considering the size of the wind tunnel, it is capable to generate eddies as large as its physical dimensions.
The performance of the gondola (or the effects of the frame on TriSonica TM Mini measurements) in reading the mean and turbulence statistics of the flow field is studied with respect to the R.M. YOUNG 81000 anemometer, which is already calibrated, and used for cross comparison to derive the calibration coefficients for the TriSonica TM Mini using line fits. By adding multiple degrees of freedom, the set-up for this test was designed to further simulate the gondola's movements in real cases. The gondola is attached to the ceiling with tow ropes (featuring the ropes to the balloon) and a single rope to the bottom floor (resembling the ground controller). Now, the gondola faces the main flow (as it does in the real atmosphere), but it has some degrees of freedom to slightly wobble. The azimuth angle, elevation angle, and wind levels were changed, independently, to derive calibration coefficients for both mean and turbulence statistics as measured by the TriSonica TM Mini and calibrated against the R.M. YOUNG 81000. Both sensors were set up at similar airflow condition while wind speed was varied at few wind levels in the range 2-10 m s −1 . At each wind speed level, data recording continued for 5 min. Each recording was time averaged to calculate mean and turbulence statistics.

5.
Similarly, the over-complex calibration of the FLIR camera's land Surface Temperature (LST) using Plank's Law is inappropriate. These cameras are bolometers, and give an output with a response related to the overall incoming thermal radiation according to Stefan-Boltzmann law, which is far simpler and with fewer coefficients. It is then a matter of calibrating the camera against a known temperature with a known emissivity: using a certificated thermometer in a large lake for instance.
Response: Thank you for your comments regarding the operation and use of the thermal camera onboard the TANAB. The DJI Zenmuse XT was included in the TANAB platform to record Earth surface temperature variation within the mining facility perimeter. Using the default camera constants does not provide absolute surface temperature measurements. Therefore, if one does not calibrate the sensor with known surface temperatures, the relative surface temperatures of the mine, tailings pond, the facility to the East of the mine, and the forest beyond the facility, will be reported. To be able to measure absolute temperatures, we have calibrated the camera constants against known surface temperatures. This method has improved the accuracy of our method. In addition, we added more literature review to address the use of uncooled thermal cameras for numerous similar applications. The following text has been added in the revised manuscript in Sect. 1.1.
From literature, the use of uncooled thermal cameras for quantification of relative surface temperatures have been documented many times for a variety of applications. Improvements in Unmanned Aerial System (UAS) and thermal imaging technology have allowed waterbodies to be mapped with a high spatiotemporal resolution. Specifically, understanding the spatial and temporal vari-ation of thermal plumes from sources, such as influent stormwater or geothermal activity, on waterbodies is of significant importance for many hydrological applications . Water temperature is also a parameter used to quantify water quality as it is known to impact aquatic organisms . Localized areas of higher surface water temperature were effectively identified by using a DJI Zenmuse XTR uncooled thermal camera without additional radiometric calibration . Mallast and Siebert (2019) utilized a UAS-based FLIR Tau2 thermal camera to spatially map submarine groundwater discharge in the Dead Sea, Israel. The intrusion of cooler groundwater rising to the surface of the Dead Sea was captured by the thermal camera, and the variation of spatial perturbations with respect to time were represented. Similarly, uncooled microbolometer thermal imaging systems have been employed to quantify the spatial distribution of surface temperatures of glaciers . This data can be used to increase the accuracy of energy budget models developed for glaciers . Luo et al. (2018) used a Zenmuse XT uncooled thermal camera to spatially map surface temperature variations of permafrost slopes. Understanding the variation of surface temperature of permafrost slopes is very important for infrastructure management and planning of railways, pipelines, and roads constructed on permafrost (Luo et al., 2018).
Uncooled microbolometer thermal cameras have also been used extensively in the agricultural industry, especially in the emerging precision agriculture sector. Poblete et al. (2018) used a FLIR Tau 2 640 uncooled thermal camera to quantify plant water stress which is a measure of water availability for crops . A FLIR Tau 2 640 uncooled microbolometer thermal camera was used by  to image plant leaf temperatures to identify Sclerotinia sclerotiorum on the leaves of oilseed rape plants.
UAS-based microbolometer thermal imaging cameras have also been noted to have some niche applications in literature. For example, Murray et al. (2018) successfully conducted a study using the FLIR Systems Vue Pro to identify relative temperature variations between human remains and the surrounding environment. Gallardo-Saavedra et al. (2018) detailed numerous studies where uncooled microbolometer thermal cameras were used to inspect photovoltaic panels and quantify performance of photovoltaic plants. Zhong et al. (2019) completed a study using a FLIR Tau 2 640 to assist in identifying pipeline leakages within a district heating system.
It is well reported in literature that uncooled microbolometer temperature measurements are impacted by variation of sensor and camera/ambient temperatures (FLIR-Systems, 2012;Lin et al., 2018). Considering these three sources of measurement error, it is not acceptable to consider calculated surface temperatures from uncooled microbolometers to be absolute without a calibration. Conversely, cooled thermal cameras are designed to measure temperatures with a high degree of accuracy such that absolute temperatures could be determined (Ribeiro-Gomes et al., 2017). Although cooled thermal cameras are very accurate, they are significantly heavier, more expensive, and require more power to operate (Sheng et al., 2010;Ribeiro-Gomes et al., 2017). As a result, cooled thermal cameras are not commonly used for small UAS (such as drones and blimps) (Sheng et al., 2010).
The following information has been added in Sect. 2.3.1 of the revised manuscript.
Although the camera utilized in this paper used an uncooled microbolometer to record thermal energy, it was calibrated against known surface temperatures. The use of the Stefan-Boltzmann Law is not applicable because the DJI Zenmuse XT is based on FLIR radiometric thermal imaging technology where the recorded microbolometer value is represented as a signal value comprising energy recorded from the surface, reflected energy from the surface, and atmospheric radiation energy (Zeise et al., 2015). Furthermore, it should be noted that many FLIR radiometric thermal imaging cameras (including the DJI Zenmuse XT) have a 14-bit radiometric resolution capable of recording pixel signal values derived from the camera's A/D converter between 0 and 16383 (FLIR-Systems, 2012;Sagan et al., 2019). 1 The radiometric image pixel signal values can be converted to temperature in Kelvin by performing a radiometric calibration between the recorded pixel signal values and corresponding object surface temperatures . The relationship between the radiometric signal value and the object temperature can be approximated with a Planck curve, as noted by Horny (2003) and similar to Martiny et al. (1996), in Eq. 1 where U Obj represents the radiometric pixel signal value, T Obj represents the surface temperature of the object, R represents the uncooled camera system response, B is a constant derived from Planck's Radiation Law, F accounts for the non-linear nature of the thermal camera system, and O represents an offset . Eq. 1 is rearranged to solve for T Obj as per Eq. 2 Tempelhahn et al., 2016) These four values, R, B, F and O were calculated by the camera manufacturer through the completion of a non-linear regression from the radiometric calibration data (the blackbody surface temperature and the corresponding radiometric pixel signal value) . However, the authors of this study performed a non-linear regression to fit the R = R 1 /R 2 , B, O, and F constants based on land use type and known surface temperatures in an off-site calibration activity. Note that the complex mining environment did not allow an on-site calibration activity due to access restrictions and safety measures. The raw signal value (U Tot ) recorded by the camera is governed by Eq. 3 as described by Usamentiaga et al. (2014) where U Obj is the raw output voltage of a blackbody recorded in a laboratory calibration experiment in the absence of reflection and atmospheric influence in the measured signal. To back calculate U Obj , U Refl and U Atm must be determined. U Refl is the theoretical camera output voltage for a blackbody of temperature T Refl according to the calibration. T Refl is the effective temperature of the object surroundings or the reflected ambient temperature. U Atm is the theoretical raw output voltage of a blackbody based on the assumed atmospheric temperature. is the emissivity of the object and τ is the atmospheric transmissivity. The transmissivity of the atmosphere is generally close to 1.0 (Usamentiaga et al., 2014) under clear weather conditions, so U Atm does not have to be calculated. From the camera metadata, the assumed reflective temperature (T Refl ) was 22 • C. This value was extracted using ExifTool. The U Refl value was calculated using the same Eq. 1 (Zeise and Wagner, 2016) where R = R 1 /R 2 , B, F and O are Planck constants of the camera that could be extracted through ExifTool (default constants) or set using fitted constants separately as detailed in Sect. 3.3.
The emissivity of the land surface was determined to be a function of geographic position. The Moderate Resolution Imaging Spectroradiometer (MODIS) instrument MOD11B3 data product was used to derive Land Surface Emissivity (LSE) (Wan et al., 2015). The monthly data product with a resolution of 6 km was used to derive LSE for data collected during the observation campaign in May 2018. The emissivity values were calculated using bands 29, 31, and 32. Since the thermal camera used a Longwave Infrared Radiation (LWIR) detector, radiation within the 7.5 µm to 13.5 µm spectral range was included in the camera voltage output. 2 MODIS band 29 records radiation within the 8.4 µm to 8.7 µm spectral range, band 31 records radiation within the 10.78 µm to 11.28 µm spectral range, and band 32 records radiation within the 11.77 µm to 12.27 µm spectral range. These three bands are used in conjunction with the BroadBand Emissivity (BBE) derivation to calculate emissivity as a function of geographical area (Wang et al., 2005). The BBE formula used is where a, b, and c are constants that vary based on the land surface material. Wang et al. (2005) determined that the constants a, b, and c do not vary significantly between soil, vegetation, or anthropogenic materials. However, water, ice, and snow resulted in noticeably different BBE coefficients. Based on Wang et al. (2005), the BBE coefficients for a, b, and c were chosen to be 0.2122, 0.3859, and 0.4029, respectively. After quantifying U Refl , it is possible to back calculate U Obj by rearranging Eq. 3 and finally calculate T Obj via Eq. 2.
The following paragraphs about the calibration are added in Sect. 3.3 of the revised manuscript.
In an attempt to quantify surface temperature measurement inaccuracies of DJI Zenmuse XT, an experiment was conducted to fit the R, B, O and F parameters as per Eq. 2. Three radiometric images were recorded approximately thirty seconds apart every hour between 0600 Local Daylight Time (LDT) and 2300 LDT over a two day period for four distinct surfaces including water, soil, developed land (urban surfaces), and grass. Each radiometric image captured included a certified thermometer, which measured a corresponding temperature for each land surface. The time delay of approximately thirty seconds between each consecutive image was chosen as Olbrycht and Więcek (2015) noted that uncooled thermal cameras without recent calibration experienced temperature drift as much as 1 • C per minute.
FLIR Tools was used to calculate surface temperatures from the radiometric images recorded by the DJI Zenmuse XT on top of the certified thermometer. For each hourly interval, the average of the surface temperatures from the thermal camera were calculated and were used in the following calculations and figures. The pixel value (U Obj ) was calculated using Eq. 1, where constants R, B, O and F were defined during camera factory calibration and stored in the metadata of each image. As per Table 1 these values are referred to as the default camera constants. The temperatures recorded by the calibrated thermometer were scaled appropriately as the outdoor field test occurred in Guelph, Ontario, Canada which is 334 m above sea level.
The empirical line method as described by Smith and Milton (1999) was used to relate A/D Counts to the corresponding certified temperatures to calibrate (fit) the R, B, O and F constants as in Fig. 7. Using the Non-Linear Least-Squares Minimization and Curve-Fitting for Python (LMFIT) library, 3 the constants were fitted and residuals were minimized for each specific surface material imaged during the calibration experiment using Eq. 2. The default and calibrated camera constants are detailed in Table 1. Fig. 7 displays the experimental, default, and calibrated temperatures as a function of camera pixel signal value for water, soil, developed land, and grass, respectively. The non-linear curve fitting library used the certified temperature obtained during the experiment, the corresponding pixel signal value, and Eq. 2 to derive the calibrated camera parameters. Fitting of the camera constants for each material resulted in reduced bias and Root Mean Square Error (RMSE) values for water, soil, developed land, and grass, respectively, as compared to the bias and RMSE values considering default camera constants, as shown in Table 2. Gallardo-Saavedra et al. (2018) reported that the manufacturer stated accuracy of the FLIR Vue Pro 640 thermal camera was ±5 K. Kelly et al. (2019) also used the empirical line calibration method for a FLIR Vue Pro 640 thermal camera and determined that the accuracy of the thermal camera was ±5 K, which is in agreement with our findings. The fitted R, B, O and F constants for water, soil, developed land, and grass were applied to surfaces in the actual mining facility with coordinates corresponding to the closest land use categories. Furthermore, the emissivity of the surface was considered by applying the BBE Eq. 5  derived by Wang et al. (2005). The boundaries for each land use type were determined by visually inspecting the Landsat 8 Operational Land Imager image recorded on May 17, 2018 with a pixel resolution of 30 m. In QGIS, land use type corresponding to geographic coordinates with a spatial resolution of 1 km were applied and the surface temperatures were calculated according to Fig. 8.
The updated land surface temperature maps for each four-hour time period, the corresponding box plots for each four-hour time period, and the comparison of median surface temperatures between the data collected via the TANAB and MODIS on May 24, 2018 are included in Figs. 9, 10, and 11, respectively. The updated figures, Fig. 9, 10, and 11 are added in the revised manuscript in Sect. 4.7. In Fig. 11, the resulting maximum relative percentage error in temperature for the corrected pond and mine boundaries and using fitted camera constants was calculated as 3.9%, while the median percentage error was calculated as 0.9%. In comparison to the previous calculation not using correct boundaries and fitted camera constants, these errors were 4.8% and 1.0%, respectively. Clearly, the modified method reduces errors.
6. Minor but significant points: when regressing it is best to pivot the data to reduce error.
Response: We addressed this comment earlier. Please see our response to comment 5 of the same reviewer. For non-linear fitting of the camera constants, the residuals were minimized.
7. With such fundamental concerns over the methodology, I am unable to comment on the results.
Response: Thank you very much for all your comments. We have worked rigorously to incorporate all your suggestions in the revised manuscript. We hope this new version meets the standards of  2 Reviewer 2 1. This paper needs some updates/inclusions/improvements in order to be suitable for publication detailed below.
Response: Thank you very much for your valuable comments that will help make the content and presentation of the manuscript better. We have performed our best to improve the manuscript based on the comments.
2. Use of sensors (lack of rigorous or appropriate calibration, and combining of sensor data to allow for appropriate analysis and conclusions to be drawn).
Response: We have performed rigorous calibration for the TriSonica TM Mini and the DJI camera. We have addressed this comment. Please see our response to reviewer 1 comment 4 (calibration for the TriSonica TM Mini) and comment 5 (calibration of DJI camera). We have also updated the manuscript rigorously.
3. GPS, there is no mention of ground control, differential or precise point processing (PPP), whichever is the chosen method of processing GPS and thus geolocating the collected data accurately for analysis and interpretation. The author may have carried this out but without a short in calibration of sensors, I cannot comment on the results.
Response: Establishment of physical Ground Control Points (GCP) at the remote mining site was not possible due to the complex nature of the environment. The remote site was actively being excavated during TANAB flights and physical access to desired GCP locations was prohibited.
The following three paragraphs has been added to the Sect. 2.3.1 in the revised manuscript.
The N3 flight controller is equipped with two Inertial Measurement Units (IMUs), a Global Navigation Satellite System (GNSS), and Compass system. 4 This flight controller is not capable of Real-Time Kinematic (RTK) positioning.
Post processing of GPS data with Precise Point Positioning (PPP) or RTK methods have been noted to reduce geographical positioning error down to a few centimetres (Satirapod et al., 2003;Remondino et al., 2011;Padró et al., 2019). On the other hand, georeferencing capabilities of GNSS and IMU systems are known to have horizontal geographical positional errors of a few meters Chiabrando et al., 2013;Zhuo et al., 2017). Direct georeferencing without the use of differential corrections provide GPS coordinates of a land point within 2 to 5 m of accuracy (Turner et al., 2014;Whitehead et al., 2014). For the purpose of the TANAB system, direct georeferencing with no corrections were used. Vertical positioning of GNSS/IMU systems have been noted in the literature to be prone to significant errors of up to 50 m especially when compared to other RTK and PPP methods (Eynard et al., 2012;Padró et al., 2019). As a result, vertical positioning of the TANAB gondola was completed by considering the hypsometric equation as per Eq. 6 using the airborne barometric pressure and temperature recorded by the ultrasonic anemometer with respect to ground level barometric pressure at the beginning of each TANAB launch.
Even without the use of RTK or PPP, the resulting temperature maps for the surface of the mining facility have a higher spatio-temporal resolution as compared to conventional satellite based sensors. MODIS 5 located on both the Terra and Aqua satellites records two distinct thermal images daily, approximately three hours apart at a 1 km × 1 km spatial resolution (Crosson et al., 2012;Kumar, 2014;Liu et al., 2017). The Advanced Baseline Imager located on the Geostationary Operational Environmental Satellite (GOES) satellites 6 , are capable of capturing thermal images every 5 min with a spatial resolution of 2 km × 2 km Schmit et al., 2017). Furthermore, Landsat satellites are capable of recording Thermal InfRared (TIR) images. The Landsat 7 Enhanced Thematic Mapper Plus 7 can capture TIR images at a spatial resolution of 60 m × 60 m and the Landsat 8 Thermal Infrared Sensor can record TIR images at a spatial resolution of 100 m × 100 m. 8 Both Landsat satellites have a time resolution of 16 days . Although errors are inherently introduced into the directly georeferenced surface temperatures, the advantages of collecting both high spatial and temporal resolution land surface temperature data from the TANAB platform outweigh geographical positioning error from the GNSS/IMU system. 4. IMU, there is no detail on the use of the IMU to determine the attitude and motion of the platform this is key to the later calibration and processing of anemometer u,v,w data into the reference of the earth.
Response: We have addressed this comment earlier as a response to reviewer 1 comment 4 (please see the section about gondola motion). We have added more detail in Sect. 2.1 in the revised manuscript. It was not desired to convert the wind velocity measurements to an earth-fixed reference frame. The system is intended to measure along-wind, cross-wind, and vertical components of wind velocity vector. Further, it was found that the gondola was mostly levelled within ±5 • 90% of the time and only rotated in the yaw direction. We used IMU roll, pitch, and yaw at a frequency of 200 Hz to quantify the camera line of sight and ultimately performed pixel geographical positioning accordingly. For altitude, we used barometric pressure and temperature measured by the ultrasonic anemometer to calculate the altitude using the standard hypsometric equation Stull, 2015).
where P 1 and P 2 are the pressure measurement at two altitudes z 1 and z 2 . The system measured pressure in mBar although the equation is insensitive to the units of pressure. The unit of altitude is m. T v is the average virtual temperature between altitudes z 1 and z 2 . The constant a = R d /g is equal to 29.3 m K −1 (Stull, 2015). Given the uncertainty of temperature and pressure measurement, the uncertainty of altitude measurement is estimated as 1.2 m.
The uncertainty in altitude calculation is quantified using the theory of error propagation analy- sis. For the ultrasonic anemometer, the uncertainty of temperature measurement is 2 K and the uncertainty of pressure measurement is 0.01 kPa. A sample calculation for the uncertainty of error on the hypsometric equation is detailed below in Eqs. 7, 8 and 9 assuming P 2 = 100 kPa, P 1 = of 101.3 kPa, and T v = 300 K (Ku, 1966). The resulting uncertainty was calculated to be 1.2 m.
5. Anemometer, though wind tunnel calibration had been carried out, it did not detail the means of use of processed IMU/GPS data to derive the u,v,w reference to the earth.
Response: We have already addressed this comment. Please see our response to reviewer 1 comment 4.
6. FLIR Zenmuse camera is an uncooled VOx microbolometer so reported measurement related to the total incoming IR radiation and requires calibration by means of accurately measuring a ground surface target. Any calibrated ground based temperature sensor measuring one of the ground targets temperature, probably easiest is to measure the lake with a thermometer.
Response: Thank you for your comments. We have addressed this comment in detail. Please see our response to reviewer 1 comment 5.
7. Balloons and blimps are not a novel platform and as platforms, there is over a centuries worth of documented use. There is no detail on why the Aero Drum balloon was chosen as opposed to a Unmanned Aerial Vehicle (UAV) platform that many of the other researchers in this field are choosing to use. I would expect a more in depth review and detail on why the balloon was the best platform, with references to other papers on the use of balloon, kites, or UAV as platforms and their use of for observing microclimates.
Response: Thank you for your comments. We have included a detailed literature review and references in the revised version of the manuscript in Sect. 1.1 (also please see our response to reviewer 1 comment 2.). We reiterate that we are proposing TANAB as a practical, simple, and low cost system for atmospheric turbulence measurements at lower elevation. Aero Drum balloon was chosen because of its low volume round-oval shape that is aerodynamic and requires less helium compared to larger common Zeppelins. We have explained the advantages of tethered balloon over UAVs as response to reviewer 1 comment 3.
8. On the choice of field site there needs to be considered and justified arguments why the characteristics of the mine are suitable for study of other microclimate environments.
Response: Open-pit mining involves altering surface topography and land use at large scales beyond 10 to 20 km. Unlike urban neighbourhoods at micro scale, open-pit mining has a large footprint beyond the meso scale so meteorological observations require tools that are able to map such scales. e.g. land fills, industrial ponds, farms, dams, reservoirs, and airport. One cannot easily fly a UAV and map such areas. Also there are safety concerns (please see our response to reviewer 1 comment 3) with flying UAVs in such heavily industrialized and complex facilities. This can provide justification why the tethered balloon approach is useful.
Wang, K., Wan, Z., Wang, P., Sparrow, M., Liu, J., Zhou, X., and Haginoya, S. (2005). Estimation of surface long wave radiation and broadband emissivity using moderate resolution imaging spectroradiometer (modis) land surface temperature/emissivity products. J. Geophys. Res.-Atmos., 110(D11).  Abstract. This study presents the first environmental monitoring field campaign of a newly developed Tethered And Navigated Air Blimp (TANAB) system to investigate the microclimate over a complex terrain. The use of a tethered balloon in complex terrains such as mines and tailings ponds is novel and the focus of the present study. The TANAB system was fully developed and launched at a mine facility in northern Canada in May 2018. This study describes the key design features, the sensor payload onboard, ::::::::: calibration, : and the observations made by the TANAB system. The system measured meteorological conditions 5 including wind speed in three directions, temperature, relative humidity, and pressure over the first few tens of meters of the atmospheric boundary layer. The system also performed earth surface thermal imaging, or temperature mapping, of the underlying surface. The measurements were made at two primary locations in the facility: i) near a tailings pond and ii) in a mine pit. TANAB measured the dynamics of the atmosphere at different diurnal times (e.g. day versus night) and locations (near tailings pond versus inside the mine). Such dynamics include mean and turbulence statistics pertaining to flow momentum 10 and energy, and they are crucial in the understanding of emission fluxes from the facility in future studies. In addition, TANAB can provide boundary conditions and validation datasets to support mesoscale dispersion modelling or Computational Fluid Dynamics (CFD) :::::::::::: computational :::: fluid :::::::: dynamics : simulations for various transport models.

Introduction
The Atmospheric Boundary Layer (ABL) is the lowest portion of the air near the earth surface that responds to surface pro-15 cesses in one hour or less (Stull, 1988;Aliabadi, 2018). The understanding of the atmospheric turbulent processes governing the transfer of heat, moisture, and momentum between the surface and the free troposphere : in ::: the ::::::: surface ::::: layer is of practical importance for many applications such as weather and climate prediction, pollution dispersionmodelling, and urban air quality studies (Zilitinkevich and Baklanov, 2002;Pichugina et al., 2008;Aliabadi et al., 2016b, c). Most of the research in atmospheric turbulence have been focused on relatively smooth terrain and horizontally homogeneous environments mainly 20 due to the limitation in the availability of adequate observation platforms and difficulty in acquiring data from the complex environments. However, the study of the ABL and surface-atmosphere interaction over complex terrain is very important for many applications. Surface heterogeneity can cause horizontal gradients of momentum or temperature, and it can influence or complicate the horizontal and vertical transport mechanisms, for instance driven by slopes or thermals (Mahrt and Vickers, (Roth, 2000).
As far as earth surface properties are concerned, Land Surface Temperature (LST) is a crucial meteorological parameter to be measured that influences the energy budget and dynamics of the atmosphere. It ::: LST : has historically been de- Furthermore, satellite-based sensors are known to have missing and skewed data due to a variety of environmental factors including atmospheric effects, land surface emissivity : , and sensor failure (Li et al., 2013;Malamiri et al., 2018).
investigation of surface layer within ABL overcoming some of the above limitations :::::::: limitations :::::::::: mentioned :: in ::: the :::::::: literature ::::: review. TANAB is lifted by the buoyancy force and requires no propulsion power for navigation compared to drones. This 5 allows turbulence measurements over long periods without disturbing the surrounding air. It provides accurate in-situ measurements unlike remote sensing technologies such as LIDARs, SODARs, and satellites. It is safer to operate at low altitudes compared to manned aircraft. It can be navigated and redeployed using a tether, unlike radiosondes, and it is very cost effective.
TANAB can collect high time-resolution observations of the weather to characterize the turbulence properties in low altitudes in almost all weather conditions. In case of extreme wind, TANAB can still be used with the help of additional stabilizing 10 tethers (usually :: up :: to : three). The variables it measures include wind speed in three directions, temperature, relative humidity, and pressure. It also performs earth surface thermal imaging, or temperature mapping, of the underlying terrain.
TANAB observations can be utilized in numerous ways. Atmospheric dynamics as measured by TANAB can determine transport mechanisms that drive emission fluxes (Steudler et al., 1991). Factors such as mean wind speed, atmospheric diffusion coefficient, and thermal stability, greatly influence emission fluxes (Bowden et al., 1993), all of which are measured by There are only a few comprehensive field studies that focus on the ABL over a mine environment while the structure of ABL 20 in an orographically complex terrain such as a mine can be complicated (Rotach and Zardi, 2007;Fitzjarrald, 2014, 2015).
In this study, the TANAB system is demonstrated while being tested to reveal the surface layer meteorology of a complex terrain in a mining facility. In the surface layer, flows are highly influenced by the terrain geometry, while Coriolis effects have still negligible influences (Arroyo et al., 2014). TANAB measurements are reported at different diurnal times (e.g. day versus night) and locations (near tailings pond versus inside the mine).

Structure of the Paper
The structure of the paper is organized as follows. Section 2 briefly describes the TANAB system and the sensors payload.
Calibration experiments are explained in Sect. 3. Section 4 presents field experiments and results of the environmental monitoring campaign where TANAB was used in a complex mining facility. Conclusions and future recommendations are provided in Sect. 5.

Tethered And Navigated Air Blimp (TANAB) Specification
TANAB consists of fixed and variable payloads. The fixed payload is comprised of a helium balloon, the navigation tether, a tether reel, and a gondola platform housing the sensors. The variable payload is comprised of microclimate sensors, such as a variable payload can be altered to use different sensors suitable for a particular application.

15
The TriSonica TM Mini weather station is an ultrasonic anemometer manufactured by Anemoment TM and is mounted onto the gondola of TANAB 5 .. 5 This mini weather station is ideal for applications that require a miniature, lightweight, and low velocity anemometer, and are suitable particularly for airborne systems. It has a measurement path length of 35 mm and a weight of 50 gr. The light weight makes it an ideal candidate to use with the TANAB system. It can measure the 3D wind speed, air temperature, relative humidity, and the barometric pressure at a sampling rate up to 10 Hz. The open path provides the least 20 possible distortion of the wind field. Its design with four measurement pathways provides a redundant measurement and the path with the most distortion is removed from the calculations to provide accurate wind measurements. It is also equipped with a compass and a tilt sensor. Because of its low power consumption (only 30 mA at 12 V), it is highly power efficient and can record data for hours.  Mini's ::::::: compass.
A data logger by Applied Technologies Inc. 6 is used as the data synchronization and data collection device for the TriSonica TM

25
Mini weather station. This data logger records the measurements on an SD card onboard that can be retrieved after every flight.
In addition, the TriSonica TM Mini can be monitored or programmed using serial communication via the data logger. The

Thermal Camera and Flight Controller
The uncooled thermal camera used is the Zenmuse XT, 19-mm lens, manufactured by FLIR Systems 7 . ::: DJI. 7 This radiometric camera has a resolution of 640 × 512, and it is capable of capturing thermal images at 30 Hz. This camera is powered by DJI and mounted underneath the gondola of the TANAB using a gimbal kit. The camera is operated by the DJI N3 flight controller 8 with associated hardware securely attached to the gondola of the TANAB. This aerial imaging device is designed to be mounted and operated using drone aircraft systemswhile it is the first time such a camera has been considered for a balloon system. The camera gimbal system compensates for gondola movement due to wind to keep the camera orientation. The camera system roll angle is generally very close to zero due to the self-stabilization of the device. Nevertheless, the gondola itself is connected to the balloon using ropes and a swivel mechanism to hang freely without being influenced by aerodynamic vibrations of 15 the balloon envelop. The gondola is designed to include a safety feature to protect the camera from impact damage. The DJI Lightbridge 2 (LB2) controller is used with either an iOS or Android device to communicate with the N3 and Zenmuse XT during flight. The LB2 has a maximum communicable range for image transmission of 5 km. Pictures and videos can be recorded with precision while the gondola is in flight as the camera can move independently with respect to the gondola.
During flight, the N3 and DJI Zenmuse XT function in parallel such that the GPS location and altitude of the gondola as well 20 as tilt and heading angles of the camera are recorded by the N3 and are included in the meta-data ::::::: metadata : of each image.
Using the recorded information for each image, the GPS coordinates (latitude and longitude) for individual pixels within each image can be derived. Through using additional software, surface temperature from individual pixels are calculated.

Image Processing Methodology
The images obtained from field observations were processed utilizing Python (version 3.6), ExifTool ( Mini : was not operational when the image was captured), the raw signal value recorded by the thermal camera, and the reflected apparent temperature. When airborne, the thermal camera is stabilized, as a result, the camera system is independent of the gondola up to the camera's mechanical range. As a result, images beyond the mechanical extent of the camera were omitted ( :::::: gondola : roll greater than 45 • or less than −45 • and , ::::::: gondola pitch greater than 45 • and less than −135 • : , ::: and :::::: gimbal ::::: pitch :::::: greater :::: than :: or 5 :::: equal :: to :: 0 • : , where the recorded pitch angle is located at the center ::::: centre of the image, as measured from the horizontal plane).
The Emissivity :::::::: emissivity : of the land surface was determined to be a function of geographic position. The Moderate Resolution Imaging Spectroradiometer (MODIS) instrument MOD11B3 data product ::::::: acquired ::::: from ::::::: MODIS : was used to derive Land Surface Emissivity (LSE) (Wan et al., 2015). The monthly data product with a resolution of 6 km was used to 5 derive LSE for data collected during the observation campaign in May 2018. The emissivity values were calculated using bands 29, 31, and 32 values. :: 32. : Since the thermal camera used a Longwave Infrared Radiation (LWIR) detector, radiation within the 7.5 µm to 13.5 µm spectral range was included in the camera voltage output 16 . : . 16 MODIS band 29 records radiation within the 8.4 µm to 8.7 µm spectral range, band 31 records radiation within the 10.78 µm to 11.28 µm spectral range, and band 32 records radiation within the 11.77 µm to 12.27 µm spectral range. These three bands are used in conjunction with the Broad 13 Band ::::::::: BroadBand : Emissivity (BBE) derivation to calculate emissivity as a function of geographical area (Wang et al., 2005).
The BBE formula used is where a, b, and c are constants that vary based on the land surface material. Wang et al. (2005) determined that the constants a, b, and c do not vary significantly between soil, vegetation, or anthropogenic materials. However, waterice , ::: ice, : and snow 15 resulted in noticeably different BBE coefficients. Based on Wang et al. (2005), the BBE coefficients for a, b, and c were chosen to be 0.2122, 0.3859, and 0.4029, respectively.

20
In order to check the validity of the high frequency data from the anemometer onboard of TANAB, the anemometer performance was validated against calibrated sensors prior to the field campaign in a series of wind tunnel and outdoor calibration experiments.
3.1 Wind Velocity CalibrationThe mounted anemometer performance was characterized in a highly turbulent flow generated by a wind tunnel at the University of Guelph.

Temperature Calibration
Temperature ::::::::::: Temperatures measured by TriSonica TM is :::: Mini ::: are calibrated with respect to the Campbell Scientific HMP60 sensor 17 . : . 17 The latter collected minute-averaged temperatures, to which the TriSonica TM :::: Mini temperatures were also averaged and compared. The experiment was carried out under a set of climate ::::::: weather conditions to cover a wide range of temperatures outdoors.

10
The TANAB system was launched at a mine facility in northern Canada (above 56 • N) for an environmental monitoring field campaign in May 2018. A schematic of the mine facility can be seen in Fig. 6. The depth of the mine is approximately 100 m.

Sampling Time
TANAB measured wind speed, the turbulence kinetic energy, variances, and fluxes for both momentum and heat rigorously.
Each balloon launch lasted approximately 15−30 min while the tether was carefully controlled to obtain a profile with constant ascent and descent rates. The sampling time for calculating the mean and turbulence quantities were 3 min while typically it is 10 to 30 min for flux tower measurements (Aliabadi et al., 2018a). For turbulence statistics, the time series is first detrended to ensure that background weather variations that have inherently very large time and length scales are filtered out without influencing the turbulence statistics calculations. It is known that finite time sampling, instead of ensemble averaging, will introduce random and systematic errors in the prediction of turbulence statistics such as variances and fluxes (Lenschow et al., always underpredict the magnitude of the turbulence statistic. These errors have been reported in an aircraft campaign to be anywhere in the range 10 to 90 % % of the measured value (Aliabadi et al., 2016b). While repeated measurements, increasing the averaging time, and detailed error analysis are possible to eliminate such errors from predictions, the focus of this study was not to investigate errors associated with finite sampling time. Nevertheless, operating the TANAB involves a delicate choice of the sampling time. On one hand, short sampling times have inherent large errors but provide profiles at high vertical resolutions.
On the other hand, long sampling times have inherent small errors but provide profiles at low vertical resolutions.

5
It was observed that both wind speed (Fig. 7) and turbulence kinetic energy (Fig. 8) exhibited a significant diurnal variation, indicating calm conditions at nights and early mornings, when atmospheric diffusion coefficient is low, and gusty conditions in the mid-afternoons when the atmospheric diffusion coefficient is high. When calculating statistical percentiles, the data is combined over all altitudes and aggregated for both the mine and tailings pond locations.
It is noteworthy that maximum flight altitude is usually less under windy conditions, therefore most profiles obtained under windy conditions in mid day are shorter than those obtained under calm conditions at night and early mornings. It is observed that in the surface layer within ABL the highest gradients of turbulence properties occur at the lowest 100 m. This statement must be considered with caution. Certainly it is only valid for the short length and time scales considered as a result of the 3-25 min time averaging. The diurnal variation of turbulence statistics are clear from the plots. The magnitude for most statistics are greatest during mid day time interval 1200−1600 LDT due to gusty conditions, while the magnitudes are smallest during nighttime and early-morning time interval 0400 − 0800 LDT associated with calm conditions. The :::::::::: along-wind vertical momentum flux sw ::: uw : is always negative, confirming boundary layer physics that imply momentum should sink to the surface due to skin drag with turbulent processes. This also implies confidence in the measurement of velocity fluctuation covariance. The vertical 30 sensible kinematic heat flux wθ is negative at night time but positive during day time. This confirms that the earth surface acts as a heat sink at night time due to heat loss by radiation to the sky, and that the earth surface acts as a heat source during day time due to heat gain by radiation from the sky. The variance of potential temperature θ 2 is not strictly diminished under calm night time conditions. This can be interpreted as the presence of differential near surface horizontal gradients of temperature due to thermal structures :::::: caused. It is expected that such gradients must exist because of the heterogeneity of land surface and anthropogenic activities in such a complex terrain of a mining facility.

5
The atmospheric dynamical condition can be described using two parameters: 1) wind speed and 2) thermal stability. The wind speed determines mechanical advection and usually the higher the wind speed the greater :: is, ::: the :::::: greater :::: will ::: be the  atmospheric transport and diffusion coefficient. Thermal stability determines the buoyant transport in the vertical direction in the atmosphere. Thermal stability is reported using various methods such as i) the vertical gradients of the potential temperature (Liu and Liang, 2010), ii) the bulk Richardson number (Mahrt, 1981;Aliabadi et al., 2016a), or iii) Monin-Obukhov length 5 (Obukhov, 1971;Wilson, 2008).
If vertical gradients of the potential temperature or the bulk Richardson number are positive, the atmosphere is stable and the buoyant transport is suppressed. This occurs during the nights and early mornings. If the vertical gradients of the potential temperature and bulk Richardson number are negative, the atmosphere is unstable and buoyant transport is enhanced. This occurs during the mid afternoons. If the vertical gradients of potential temperature and bulk Richardson number are close to 10 zero, the atmosphere is neutral, in which case buoyant transport is still present but weak. Figures 11 and 12 show evidence for the variation of vertical profiles of thermal stability ::::::: potential ::::::::::: temperature and wind speed as a function of diurnal time, respectively. Here the data is statistically processed in 3-min intervals, such that a median is calculated for each 3-min interval. Since observations for these plots are not aggregated over many days, the height interval is not binned. The thermal stability and magnitude of wind speed differences between day and night are clearly evident. However, the profiles of wind speed do not exhibit the expected power law or logarithmic law as a function of height. This is likely due 5 to limited time of sampling.

Atmospheric Dynamical Condition
As shown in Fig. 13, all observations were used to determine the atmospheric dynamical condition on a two-dimensional map consisting of thermal stability, i.e. bulk Richardson number Ri b , and mean wind speed S. Here the bulk Richardson number is defined as where g is gravitational acceleration, H is the maximum altitude for each launch, S H is mean horizontal wind speed at maximum altitude, S S is mean horizontal wind speed near the surface, Θ H is mean potential temperature at maximum altitude, Θ S is mean potential temperature near the surface, and Θ A is mean potential temperature for the entire launch.  the atmosphere in the surface layer has preferred states. For instance, it was never observed to be very stable and gusty at the 10 same time.

Comparison Between the Mine and the Tailings Pond
Almost the same number of hours were spend :::: spent : measuring the surface layer at the mine and near the tailings pond using TANAB. The measurements attempted to cover a 24-hr time period in three to four days so that it ::::::: TANAB would capture the diurnal variations completely. Note that due to logistical difficulties, it was impossible to measure the surface layer in either 15 location for 24 hr continuously. The base location for the launches at the mine was approximately 100 m below grade and at the center ::::: centre : of the mine, while the base location near the pond was on the east side. We have compared the diurnal variations in various surface layer properties between the mine and near the pond observations. These include the mean horizontal wind speed, turbulence kinetic energy, vertical momentum flux, vertical sensible kinematic heat flux, variance of vertical velocity vector, and variance of potential temperature. We have also compared the atmospheric dynamical condition in terms of the bulk Richardson number and mean wind speed. The idea behind this comparison was to quantify any differences in surface layer properties as a result of terrain complexity and land use. calm conditions at nights and early mornings, when atmospheric diffusion coefficient is low, and gusty conditions in the mid 10 afternoons when the atmospheric diffusion coefficient is high. The data is aggregated for all altitudes.
There are subtle differences between observations at the mine and near the tailings pond. The trends in mean wind speed S and turbulence kinetic energy k suggest that the mine experiences calm conditions at early morning hours, i.e. from 0500 to 0700 LDT, when the atmospheric stability is at its maximum, while at the same time near the pond : , higher mean wind speeds ::::: speed and turbulence kinetic energy are observed. During early afternoons and under unstable conditions, i.e. from 1500 to 1700 LDT, near the pondhigher mean wind speeds : , ::::: lower ::::: mean ::::: wind ::::: speed : and turbulence kinetic energy are observed in comparison to the mine. Even though day to day variations of meteorological conditions may contribute to this, there is evidence that such features can also result from terrain complexity and land surface heterogeneity. As far as the vertical 5 sensible kinematic heat flux wθ and variance of potential temperature θ 2 are concerned, the mine shows significantly higher activities compared to the location near the pond. In fact, the mine shows significantly positive vertical heat flux and substantial potential temperature variance from 0200 to 0300 LDT. It is known that excavation activities using heavy machinery is very active in the mine over night, possibly explaining the significance of heat flux and potential temperature variance.

Comparison of Atmospheric Dynamical Condition
10 Figure 16 shows the frequency plots of atmospheric dynamical condition as a function of the bulk Richardson number Ri b and mean wind speed S in the mine and near the tailings pond separately. It was found that in the case of tailings pond, the atmosphere spends a considerable amount of time under near-neutral and unstable conditions in comparison to the mine. Again : , the frequency plots should be read with care at Ri b ∼ 10 due to the lack of a reliable TANAB predictions for vertical mean wind speed gradients under calm conditions. 15

Earth Surface Thermal Imaging
An important measurement of the TANAB system was the earth surface thermal imaging. The surface temperature values and thermal gradients at the mining site are important physical parameters that can potentially drive unique airflow patterns. For instance, hot spots can create rising thermal structures in the boundary layer that subsequently carry surface-level atmospheric constituents, such as GHGs, upward. On the other hand, cold spots can create stable boundary layers and initiate subsiding 20 flow that prevent mixing of gaseous compounds within the boundary layer. Inevitably, the thermal condition places a significant impact on the gaseous emission fluxes. TANAB took more than 10,000 thermal images during the May 2018 campaign, all of which are processed in detail to determine the surface temperature characteristics of the site during different diurnal times and at different locations. An analysis of selected thermal images is included below. Figure 17 shows the median temperature maps for the site for different diurnal periods (4-hr intervals) of the day. Data is 25 aggregated over all observations measured by the experiments as mentioned in Table 3. The diurnal variation of the temperature is evident. The surface temperature is the greatest during mid day time interval 1200 − 1600 LDT, while the magnitudes are smallest during night time and early morning time interval 0400 − 0800 LDT. In order to quantify the difference in temperatures as a consequence of terrain heterogeneity and other complexities between the mine and tailings pond, we have compared the surface temperatures using box plots in Fig. 18 at different diurnal periods 30 (4-hr intervals) aggregated over the entire observation period as detailed in Table 3. The thermal imaging data is in agreement with the TriSonica TM :::: Mini : data as plotted in Fig. 15. For many periods the mine shows higher activity in heat related turbulence statistics such as the turbulent kinematic vertical heat flux during 0800−2000 :::: LDT : and variance of potential temperature during 0000 − 0400 :::: LDT. These time intervals correspond to periods when the mine shows a statistically significant higher surface temperature than the pond using the box plots.
The median temperatures measured by TANAB are compared with satellite measurements of MODIS on 24 May 2018. For this comparison, the selected images for TANAB were captured during 1211 − 1401 LDT, and for MODIS the selected data source corresponded to 1230 LDT. The horizontal resolution for this comparison was 1 km × 1 km. The comparison and the 5 calculated percentage relative error are shown in Fig. 19. The percentage relative error is less than 6 :: 3.9 : % everywhere within the perimeter of the mining facility and the median relative error within the mining facility is 1.1 ::: 0.9%. However, the ::: The LST errors are higher in :::: areas ::::::: between ::: the ::::: mine ::: and : the northwest direction and lower in the southeast direction. These directions are along the topographical height variation, which is highest in the northwest and lowest in the southeast :::: pond. It appears that TANAB temperature predictions in the northwest direction :: this :::::: region result from more oblique-angle observations, which are
As shown by this surface temperature analysis, TANAB offers many advantages over existing satellite systems that attempt to observe the earth surface temperatures. On the one hand, geostationary satellites, such as the GOES-R, exhibit very low horizontal resolutions (∼ 1 − 2 km) and high temporal resolutions (15 minutes) at such high altitudes 15 Schmit et al., 2017). On the other hand, orbiting satellites capable of thermal imaging at high horizontal resolutions (∼ 10−100 m), such as the LANDSAT group of satellites, pass over northern sites very intermittently once every sixteen days :: 16 : days (Weng, 2012;. The TANAB overcomes these difficulties by allowing frequent observations at high horizontal resolutions.

20
The vertical structure of the Atmospheric Boundary Layer (ABL) in an orographically complex terrain, such as a mine, can be complicated. The Tethered And Navigated Air blimp (TANAB) has provided an acceptable platform for the meteorological measurements inside the surface layer within ABL and atmospheric dynamical condition over a complex terrain of a mining facility. The field campaign occurred in May 2018 in northern Canada. In the surface layer, most atmospheric transport mechanisms are highly influenced by the terrain geometry ::::::::: complexity. TANAB measured the microclimate in the complex terrain by quantifying mean and turbulence statistics of the atmospheric meteorology. This was achieved by sensing wind speed, wind direction, temperature, relative humidity, and pressure. The calculated parameters included mean horizontal wind speed, turbulence kinetic energy, vertical momentum flux, vertical sensible kinematic heat flux, variance of potential temperature, and variance of vertical wind velocity. In addition, TANAB performed earth surface thermal imaging (temperature mapping) to TANAB observed that both wind speed and turbulence kinetic energy exhibit a significant diurnal variation, indicating calm conditions at nights and early mornings, when atmospheric diffusion coefficient is low, and gusty conditions in the mid 10 afternoons, when the atmospheric diffusion coefficient is high. TANAB also revealed the vertical structure of the atmosphere near the surface for most meteorological parameters. The highest turbulence kinetic energies occurred in the lowest 100 m above the surface, albeit for small fluctuation time and length scales probed. Experiments provided evidence for the variation of thermal stability and wind speed as a function of diurnal time. The atmosphere spent a considerable amount of time under near-neutral and stable conditions, with implications on atmospheric diffusion coefficient and emission fluxes of atmospheric 15 constituents released near the surface. The experiments specifically observed differences in the microclimate in the mine pit in comparison to near the tailings pond. The overall pattern of diurnal variation was found to be similar for both the mine and the tailings pond, but subtle meteorological differences were observed. The mean wind speed and turbulence kinetic energy were comparatively lower in the mine than near the tailings pond in early morning hours, suggesting that the mine boundary layer may have been isolated from the boundary layer above grade. In addition, significantly positive potential temperature variance 20 and turbulent heat fluxes were observed in the mine in various diurnal time windows in comparison to the area near the pond.
This was likely due to terrain complexity and anthropogenic activities in the mine. TANAB detected a significant horizontal temperature difference between the mine and surrounding ::::::::::: surroundings in comparison to the tailings pond. The mine exhibited greater surface temperatures in many diurnal time periods, likely due to land surface modifications and anthropogenic activities.
Higher surface temperatures at the mine could explain the meteorological effects observed in the surface layer of the mine.

25
A particular challenge in operating the TANAB system is the proper choice of sampling time. On the one hand, short sampling times impose inherent errors in mean and turbulence statistics predictions while enable high resolution vertical measurements. On the other hand, long sampling times impose less inherent errors in mean and turbulence statistics predictions but they only enable low vertical resolution measurements. Another drawback of TANAB system is that it is not autonomous, so it requires intensive operator effort to fly it. Future development requires advanced techniques for autonomous control of TANAB.
Overall, TANAB offers a simple and cost-effective platform for microclimate measurements within the atmospheric surface layer. The light high-frequency weather sensor onboard enables measurement of mean and turbulence statistics of the