Errors of naturally ventilated air temperature

0 downloads 0 Views 2MB Size Report
2 May 2007 - Weidinger, 2007: The Energy Balance Experiment EBEX-2000. Part I: Overview and. 352 energy balance. Bound.-Layer Meteor., 123, 1-28.
AMERICAN METEOROLOGICAL SOCIETY Journal of Atmospheric and Oceanic Technology

EARLY ONLINE RELEASE This is a preliminary PDF of the author-produced manuscript that has been peer-reviewed and accepted for publication. Since it is being posted so soon after acceptance, it has not yet been copyedited, formatted, or processed by AMS Publications. This preliminary version of the manuscript may be downloaded, distributed, and cited, but please be aware that there will be visual differences and possibly some content differences between this version and the final published version. The DOI for this manuscript is doi: 10.1175/2008JTECHA1046.1 The final published version of this manuscript will replace the preliminary version at the above DOI once it is available.

© 2008 American Meteorological Society

1 2

Errors of naturally ventilated air temperature measurements

3

in a spatial observation network

4 5

MATTHIAS MAUDER, R. L. DESJARDINS, ZHILING GAO, AND RONALD VAN HAARLEM

6

Agriculture and Agri-Food Canada, Research Branch, Ottawa, Ontario, Canada

7 8 9

Technical Note

10 11 12

Submitted to Journal of Atmospheric and Oceanic Technology

13

2nd Revision

14 15 16 17 18 19 20 21

Corresponding author address: Matthias Mauder, Agriculture and Agri-Food Canada, Research Branch,

22

960 Carling Ave., Ottawa, ON, Canada, K1A 0C6.

23

Email: [email protected]

1

24 25

ABSTRACT

26 27

A spatial network of 25 air temperature sensors was deployed over an area of

28

3.5 km x 3.5 km of agricultural land, aiming to calculate the sensible heat flux by spatial

29

averaging instead of temporal averaging. Since temperature sensors in naturally

30

ventilated solar radiation shields were used for these measurements, a correction for

31

radiative heating had to be applied. In this study, the approach of Anderson and

32

Baumgartner (1998) was adapted to the cuboidal shaped HOBO Solar Radiation Shields.

33

This semi-empirical correction depends on the shield’s area normal to the sun in addition

34

to solar radiation and wind speed. The required correction coefficients, which can be

35

universally applied for this type of shield, were obtained through comparison with fan-

36

aspirated temperature measurements at one site. The root-mean-square error of the

37

HOBO temperature measurements was reduced from 0.49 °C to 0.15 °C after applying

38

this radiation correction.

39 40 41 42

1. Introduction

43 44

Horizontal homogeneity of the air temperature field is considered a prerequisite

45

for tower-based micrometeorological techniques to accurately measure the turbulent heat

46

exchange between the surface and the atmosphere (Kaimal and Finnigan, 1994). In reality,

2

47

such conditions are probably rarely met. Measuring the spatial distribution of air

48

temperature allows checking whether spatial differences are small enough that this basic

49

requirement can be assumed. If not, it allows quantifying additional fluxes, such as

50

horizontal advection for example. Flux contributions that originate from spatially

51

stationary structures also cannot be captured with single-tower measurements and can

52

therefore only be measured through spatial Reynolds averaging (Mahrt, 1998; Steinfeld et

53

al., 2007). The resulting underestimation of turbulent fluxes is often referred to as the

54

energy balance closure problem (e.g. Culf et al., 2004; Mauder and Foken, 2006; Foken,

55

2007; Oncley et al., 2007).

56

Since evidence of large scale circulations was found through aircraft

57

measurements by Mauder et al. (2007), an experiment was designed to measure the

58

spatial distribution of air temperature from ground based measurements. A network of

59

naturally ventilated air temperature sensors was deployed to obtain the spatial average

60

that is required for a spatial eddy covariance calculation. Turbulent fluctuations were

61

calculated by subtracting the spatial instead of the temporal average of a scalar.

H=

1 N

30 min

∑ ( w − [ w]) (T − [T ]) , t =0

(1)

62

where H is the sensible heat flux, N is the number of samples, w is the vertical wind

63

velocity, T is air temperature, and [] denote a spatial average.

64

Temperature fluctuations were measured by a sonic anemometer/thermometer.

65

Since this instrument is based on a completely different measurement principle, great

66

care had to be taken in correcting systematic errors of the spatially averaged temperature

67

measurement. The dominant error source of a temperature sensor unit in a multi-plate

68

radiation shield is radiative solar heating.

3

69

According to Steinfeld et al. (2007), a number of about 25 measurement sites is

70

necessary to obtain a good estimate of the spatially averaged air temperature if they are

71

equally distributed over a 10 km x 10 km area at a measurement height of 20 m.

72

Although the size of the observation area and the measurement height were different for

73

our set-up, we decided to follow this recommendation and deployed 25 HOBO 12-Bit

74

Smart Sensors (Onset Computer Corporation, Bourne, Massachusetts, part # S-TMB-

75

M002) in combination with HOBO Solar Radiation Shields (Onset, part # M-RSA) as a

76

practical and economical solution. These HOBO systems are an improved version of the

77

temperature datalogger HOBO H8 Pro, which has been evaluated by Whiteman et al.

78

(2000). It has been shown that these sensors are durable, compact, easy to maintain, and

79

very low in power consumption. Nakamura and Mahrt (2005, NM05 in the following)

80

have developed a correction for the radiative error of the HOBO M-RSA radiation shield.

81

However, this correction is not sufficient for the purpose of this experiment, i.e.

82

determining the sensible heat flux by replacing the temporal average with a spatial

83

average when computing turbulent air temperature fluctuations. Turbulent temperature

84

fluctuations are usually on the order of 1.0 °C (standard deviation of sonic temperature),

85

and the measurement error of the HOBO sensors should preferably be one order of

86

magnitude less.

87

Therefore, our goal is to reduce the measurement error of the HOBO systems as

88

much as possible. In this study, an improved correction for the radiative heating error of

89

naturally ventilated HOBO radiation shields is presented. This correction is able to

90

provide better results than the approach of NM05 because it considers the shield’s

91

geometry. The approach of NM05 uses wind speed and the shortwave radiation normal to

4

92

the earth’s surface only. It does not consider the shield’s area normal to the sun and is

93

therefore not as effective. The method presented in this paper uses an adaptation of the

94

correction developed by Anderson and Baumgartner (1998, AB98 in the following) for

95

cylindrical shaped shields to the cuboidal shaped HOBO shields.

96 97 98

2. Model for the radiative heating error

99 100

Since our correction for radiative heating is based on the work by AB98, we will

101

briefly summarize this method and point out necessary modifications for its application to

102

the HOBO shield. Sensor and shield are considered as one unit. The three major effects

103

that influence the temperature of this system are

104



105 106

radiative heating, meaning primarily the incoming shortwave solar radiation that hits the shield’s surface,



natural convection, meaning the heat transfer by fluid circulation or

107

movement solely due to the natural forces of buoyancy at zero wind speed;

108

this process is called “conductive cooling” by AB98, and

109



forced convection, meaning the heat transfer due to movement of the fluid

110

with the mean wind, which is called “convective cooling” by AB98.

111

According to AB98, the incoming shortwave radiation is the dominant term in the

112

radiation budget, at least during daytime. In comparison the longwave radiation balance

113

is small and the reflected shortwave radiation is linked to the incoming shortwave

114

radiation through a relatively constant factor, the albedo. This means that only the

5

115

measurement of the incoming shortwave radiation is required for a radiation correction of

116

temperature measurements. The error due to longwave radiation was determined by

117

NM05 during nocturnal conditions and these measurements showed a nighttime error of

118

less than 0.1 °C on average. This effect was neglected for the correction presented in this

119

paper. As a result of the three processes listed above, the measured temperature in the

120

shield is usually greater than or equal to the actual air temperature during daytime. The

121

heat budget of this system can be written as

α s Rs As = L + S ,

(2)

122

where radiative heating term is the product of an absorption coefficient αs, the shortwave

123

flux density Rs, and the area normal to the incident solar radiation As. L and S are the heat

124

transfer terms for forced and natural convection. In accordance with AB98, Rs can be

125

described as Rs =

SW ↓ , rd + (1 − rd ) sin θ

(3)

126

where θ is the sun’s elevation angle, and SW↓ is the flux density of shortwave radiation

127

normal to the surface. Here, a constant ratio rd of diffuse radiation to total downward

128

radiation of 0.1 is assumed. The justification is that the largest radiative heating occurs

129

under clear skies rather than under cloudy conditions (AB98).

130

The calculation of the area normal to the sun As has to be adapted for the HOBO

131

shield. Its geometry can be described with a cuboid of 0.213 m length l, 0.188 m depth d,

132

and 0.152 m height h. The contribution of the top area Atop of the cuboid to As is a

133

function of θ.

6

Atop = l ⋅ d ⋅ sin θ .

(4)

134

The contribution of the sides of the cuboid Asides to the total area As is a function of θ, the

135

sun’s azimuth α, and the orientation of the HOBO shield ϕ. It can be written as the sum

136

of the contribution from the longer side Al and the contribution from the shorter side Ad.

Asides = Al + Ad = h ⋅ cos θ ⋅ l ⋅ sin (α − ϕ ) + h ⋅ cos θ ⋅ d ⋅ cos (α − ϕ ) .

(5)

137

The sun’s angles θ and α were computed according to a parameterization

138

(http://www.jgiesen.de/SME/tk/index.htm). The orientation angle ϕ is defined as the

139

direction, where the mounting plate on the longer side of the cuboid is facing.

140 141

The sum of the cooling terms L and S is a function of the shield surface temperature T and the ambient air temperature Ta. It is written as L + S = hu (T − Ta ) Ac + ho (T − Ta ) Ac ,

(6)

142

where Ac is the surface area of the shield, which is the actual heat exchange area; hu and

143

ho are the heat transfer coefficients for forced convective and natural convective cooling.

144

The coefficient ho is a constant, because natural convection is defined for zero wind and it

145

is solely based on buoyancy. The unit of ho is W m−2 °C−1. The heat transfer coefficient

146

for forced convection hu is usually described by an empirical model (Incropera and

147

DeWitt, 1985). In accordance with AB98, hu can be expressed as a function of wind

148

speed V and two empirical parameters C and m. hu = CV m ,

(7)

149

where m is a non-dimensional constant and the coefficient C incorporates another non-

150

dimensional constant C ′ , the thermal conductivity k, the viscosity of the air ν, and the

151

Prandtl number Pr.

7

C=

C ′ Pr k . ν

(8)

152

It is assumed that all the parameters in Eq. (8) are constant over the observed air

153

temperature range. Thus, the unit of C is J m−3 °C-1. Combining Eqs. (1), (5), and (6) and

154

dividing by αs, the sum of both heat transfer coefficients, δ, is

δ = CV m + ho =

Rs As . (T − Ta ) Ac

(9)

155

Ta can be substituted by the reference (aspirated) temperature measurements, while T is

156

the shielded HOBO temperature. Here, we assume that the shield surface temperature can

157

be substituted by the temperature measured inside the shield by the HOBO sensor. All the

158

variables on the right-hand-side of Eq. (7) and wind speed V are available, which allows

159

determining the three empirical constants C, m, and ho from a regression analysis as

160

described in AB98. The corrected ambient air temperature measurement in the naturally

161

ventilated shield then becomes

Tc = T −

Rs As . δ Ac

(10)

162 163 164

3. Experimental Set-up

165 166

The HOBO 12-Bit Smart Sensor is a thermistor with a built-in A/D converter. Its

167

digital output was recorded on HOBO Micro Station dataloggers (Onset, part # H21-002).

168

The sensor was placed inside the HOBO Solar Radiation Shield, which was mounted on a

169

3-Meter Tripod (Onset, part # M-TPB). The measurement height of the temperature

8

170

sensor was 2.60 m above ground level. The accuracy of the sensor itself is ±0.2 °C

171

according to the manufacturer (Table 1). The response time specification of < 2 min is

172

approximately confirmed by the findings of Whiteman et al. (2000) and NM05. This is

173

adequate for measuring horizontal temperature gradients to calculate advective fluxes.

174

We also do not expect the spatial mean to change at a faster rate over our study area of

175

3.5 km x 3.5 km, because the spatial averaging acts a low-pass filter.

176

The measurements were carried out over farmland in southwest Ottawa, Ontario,

177

Canada (45°18’13” N, 75°46’12” W, 88 m a.s.l.). The observation period was from 17

178

May 2007 to 20 June 2007. Wheat, corn, soybean, grass and alfalfa were cultivated on

179

this land. 25 HOBO sensors were distributed over the 3.5 km x 3.5 km area in a regular 5

180

x 5 grid. The orientation of the HOBO shields ϕ was 54° for all the 25 sensors of our set-

181

up. The central site of that grid was located on a 700 m x 300 m large grassland area. In

182

addition to the HOBO sensor datalogger combination, this site was equipped with a high-

183

precision temperature sensor of type 063 (MetOne Instruments Inc., Grants Pass, Oregon)

184

in a radiation shield of type 076B, also produced by MetOne. This system served as

185

reference for determining the correction coefficients ho, C, and m, since its fan-aspirated

186

shield reduces the radiative error to < 0.03 °C according to the manufacturer (MetOne

187

Instruments, 1997). This extremely low error is safeguarded though the shield’s

188

construction including a large umbrella-shaped cover plate of 51 cm in diameter, triple-

189

sided walls, double-cross mask bottoms towards the ground, and a relatively large flow

190

rate of 37.7 l s-1.

191

Further, a CSAT3 sonic anemometer (Campbell Scientific Inc., Logan, Utah) and

192

a CMA6 first-class albedometer (Kipp&Zonen, Delft, The Netherlands) were deployed at

9

193

this site. All these sensors were collocated in an area of 20 m x 20 m. The sonic data were

194

recorded on a Campbell CR23X datalogger at a sampling rate of 20 Hz. All slow-

195

response sensors were recorded at 30 s intervals.

196 197 198

4. Field intercomparison

199 200

In order to estimate the precision of the HOBO systems under field conditions, a

201

side-by-side intercomparison was conducted before the actual deployment in the main

202

experiment. All 25 HOBO sensors were set-up together with the aspirated MetOne sensor

203

over grassland on the Central Experimental Farm of Agriculture and Agri-Food Canada

204

in Ottawa, Ontario. The 25 HOBO sensors were set-up in a 5 x 5 grid in an area of 10 m

205

x 10 m. The MetOne sensor was located 2 m beside this grid (Fig. 1). To ensure

206

comparable conditions to the later deployment in the main experiment, all sensors were

207

mounted at a height of 2.60 m above ground level facing the same direction, and the

208

sampling interval was set to 30 s.

209

The systems recorded air temperature on 2 May and 3 May 2007. Temperatures

210

ranged from 9.5 °C to 16 °C during this period. Root-mean-square errors (RMSE) were

211

calculated for all HOBOs sensors choosing one of them randomly as a reference. The

212

RMSE values ranged between 0.05 °C and 0.14 °C with a median of 0.09 °C. However,

213

the maximum deviation of the HOBO measurements from the fan-aspirated MetOne

214

temperature was +1.54 °C. This clearly shows the need for a radiation correction.

215

Fortunately, it is possible to use the same radiation correction for all 25 sensors because

10

216

of the good precision of the HOBO units and because all shields were oriented in the

217

same direction. However, it has to be assumed that the incoming shortwave radiation and

218

the wind speed were similar over the entire study area, and that possible differences

219

cancel out when calculating the spatial average from the 25 sensors. The topography of

220

the measurement domain is generally flat with elevation changes of only a few meters

221

over the entire area. The assumption of a constant V is therefore reasonable.

222 223 224

5. Results and discussion

225 226

Wind speeds V and δ values were calculated as 30-min averages in order to

227

determine the empirical coefficients for the radiation correction. A dependency of δ on

228

either wind direction or the sun’s azimuth, as found by AB98, could not be detected for

229

our measurements. Therefore, it was not necessary to apply wind blocking or shading

230

adjustments. From an analysis of nighttime data, an offset of +0.32 °C was found for the

231

HOBO sensor compared to the MetOne system. This was probably related to the

232

characteristics of the sensor/data acquisition system and was therefore generally

233

subtracted from the HOBO readings. After discarding data from periods with rain, 593

234

pairs of V and δ values remained from the six-week observation period. These samples

235

were classified into 30 equal sample size bins of ascending wind speed. The wind speed

236

range was 0.04 to 6.36 m s−1. A nonlinear regression analysis was conducted, which

237

resulted in the coefficients ho = 242.24 W m−2 °C−1, C = 44.87 J m−3 °C-1, and m = 2.05

238

(nondimensional).

11

239

Fig. 2 shows examples of magnitude of the correction term (Tc − T) as a function

240

of wind speed V for various combinations of Rs and As. For calm wind, the correction

241

term reaches up to −1.2 °C for Rs = 1000 W m−2 and As = 0.06 m2, which are both at the

242

upper limit of the range of values that occurred during this experiment. The correction

243

term decreases rapidly for wind speeds larger than 1 m s−1. At 3 m s−1 the correction term

244

is 0.5 °C for the highest product of Rs and As values. For wind speeds larger than 5 m s−1

245

the correction term is generally less than 0.2 °C. For radiative heating term of 500 W m−2,

246

which corresponds to the midday conditions on an overcast day in summer, the maximum

247

correction term is 0.6 °C for calm winds.

248

In theory, the parameters ho, C, and m can be universally applied for this type of

249

HOBO radiation shields over similar surfaces. However, the exposure of the sensors

250

during this experiment in May and June of 2007 was limited and only a selection of all

251

possible meteorological conditions was covered. Thus, these parameters are only valid

252

for daytime conditions as long as wind speed, solar radiation, and solar angles are within

253

the range of the observations during this experiment (V < 6.36 m s−1; SW↓ < 1008 W m−2 ;

254

θ < 68.1°). This correction cannot be applied for periods of precipitation and when the

255

shields surface characteristics are altered, e.g. through adherent water or snow. A site

256

with a different albedo than grass (≈ 0.2), which formed the reflecting surface during this

257

experiment, would also require a recalibration of the correction parameters. Regular

258

weather stations mostly use measurement heights of 1.5 m or 2.0 m. The HOBO sensors

259

in our set-up were located at a height of 2.6 m. We do not expect that the radiative error

260

of the HOBO shields would be significantly different between these heights.

12

261

The effect of the consideration of the cuboidal HOBO geometry on the solar

262

heating term is shown in Fig. 3 for three very different consecutive days. The first day, 19

263

May 2007, was cloudless with a maximum incoming solar radiation of 970 W m−2 and

264

low wind speeds between 0.1 m s−1 and 2.5 m s−1. The second day was overcast, and the

265

maximum incoming solar radiation was 444 W m−2. Wind speeds were much higher

266

ranging between 2.3 m s-1 and 6.0 m s-1. The third day was cloudless in the morning

267

before some cumulus clouds developed in the afternoon. Incoming shortwave radiation

268

was undisturbed from clouds at solar noon and reached 984 W m−2. Wind speeds were

269

between 3.1 m s−1 and 6.0 m s−1.

270

The shield’s area normal to the sun As had a trimodal diurnal course. On the clear

271

sky day, 19 May 2007, the highest maximum was at 0842 local solar time (LST), a

272

secondary maximum at 1615 LST, and a tertiary maximum at 1220 LST. In comparison,

273

the diurnal course of RsAs for a cylindrical shaped shield is bimodal with a distinct

274

minimum at solar noon (AB98). The geometry factor As modulated the incoming

275

shortwave radiation in such a way that the largest radiative heating of the HOBO shield

276

RsAs occurred at 0855 LST. The maximum radiation correction on 19 May 2007 was

277

1.17 °C. This maximum was reached twice on that day, firstly between 0830 LST and

278

0900 LST when the RsAs term had its absolute maximum, and secondly between

279

1130 LST and 1200 LST when wind speeds were lowest. On 20 May 2007, the maximum

280

temperature correction was only 0.30 °C at maximum. The radiative heating was lower

281

and wind speeds were higher compared to the previous day. On 21 May 2007, the

282

radiation conditions were similar to 19 May 2007 in the morning when the radiative

13

283

heating is usually largest. However due to higher wind speeds, the maximum temperature

284

correction was only 0.37 °C.

285

The overall improvement of the HOBO temperature measurements through the

286

radiation correction is shown in Fig. 4. Without this correction, the HOBO temperature

287

measurements were sometimes up to 1.2 °C higher. The differences between the

288

uncorrected HOBO temperature T and the aspirated temperature Ta scattered around an

289

average of +0.42 °C with a standard deviation of 0.27 °C. The median was +0.38 °C.

290

After applying the radiation correction, the HOBO temperatures Tc and Ta were almost

291

identical on average with significantly less scatter. The mean difference between Tc and

292

Ta was +0.01 °C with a standard deviation of 0.15 °C and a median of +0.02 °C. The

293

root-mean-square difference between the HOBO temperature measurements and the fan-

294

aspirated reference measurements was reduced from 0.49 °C to 0.15 °C. The remaining

295

error is only slightly larger than the precision of the HOBO sensor-shield combinations

296

themselves.

297 298 299

6. Conclusions

300 301

The principle of AB98’s correction for radiative heating of naturally ventilated

302

shields was successfully adapted to the cuboidal geometry of the HOBO Solar Radiation

303

Shields. This modified correction method makes it possible to deploy a network of many

304

relatively low-cost and low-maintenance temperature sensors in naturally ventilated

305

shields with a data quality close to aspirated systems, provided wind and radiation

14

306

measurements are available for at least one of sites over the entire measurement period.

307

The study area also must be small enough that similar conditions for incoming shortwave

308

radiation and wind can be assumed. At many weather stations, shortwave radiation and

309

wind speed belong to the standard set of measured variables, so that this radiation

310

correction can be applied for the temperature measurements in HOBO Solar Radiation

311

Shields. An aspirated temperature sensor was deployed over a 35-day period in May and

312

June 2007, in order to obtain estimates for the three empirical correction coefficients

313

required. These parameters may be generally applied for clean and dry HOBO Solar

314

Radiation Shields of type M-RSA during daytime periods over surfaces with similar

315

radiative properties. Further measurements are needed to validate these parameters for

316

meteorological conditions that are not covered by the field observations presented here.

317 318 319

Acknowledgements. We are grateful to Doug Glowenlock, manager of the Canadian Food

320

Inspection Agency research farm in Ottawa, and Bert Moore, manager of the Greenbelt

321

research farm in Ottawa, for their cooperation during the realization of the presented

322

measurements.

323 324

15

324 325

REFERENCES

326 327

Anderson, S. P. and M. F. Baumgartner, 1998: Radiative heating errors in naturally

328

ventilated air temperature measurements made from buoys. J. Atmos. Oceanic Technol.,

329

15, 157-173.

330

Culf, A. D., T. Foken, and J. H. C. Gash, 2004: The energy balance closure problem.

331

Vegetation, Water, Humans and the Climate. A New Perspective on an Interactive System,

332

P. Kabat and M. Claussen, Eds., Springer, 159-166.

333

Foken, T., 2007: The energy balance closure problem - an overview. Ecological

334

Applications, accepted.

335

Incropera, F. P. and D. P. DeWitt, 1985: Fundamentals of Heat Transfer. 2nd revised

336

edition ed. John Wiley and Sons, 802 pp.

337

Kaimal, J. C. and J. J. Finnigan, 1994: Atmospheric Boundary Layer Flows: Their

338

Structure and Measurement. Oxford University Press, 289 pp.

339

Mahrt, L., 1998: Flux sampling errors for aircraft and towers. J. Atmos. Oceanic Technol.,

340

15, 416-429.

341

Mauder, M. and T. Foken, 2006: Impact of post-field data processing on eddy covariance

342

flux estimates and energy balance closure. Meteor. Z., 15, 597-609.

343

Mauder, M., R. L. Desjardins, and J. I. MacPherson, 2007: Scale analysis of airborne flux

344

measurements over heterogeneous terrain in a boreal ecosystem. J. Geophys. Res., 112,

345

D13112, doi: 10.1029/2006JD008133.

16

346

MetOne Instruments, Inc., 1997: Fan Aspirated Radiation Shield. [Available online from

347

http://www.metone.com/documents/076BShieldDSTemp.pdf.]

348

Nakamura, R. and L. Mahrt, 2005: Air temperature measurement errors in naturally

349

ventilated radiation shields. J. Atmos. Oceanic Technol., 22, 1046-1058.

350

Oncley, S. P., T. Foken, R. Vogt, W. Kohsiek, H. de Bruin, C. Bernhofer, A. Christen, D.

351

Grantz, E. Lehner, C. Liebethal, H. Liu, M. Mauder, A. Pitacco, L. Ribeiro, and T.

352

Weidinger, 2007: The Energy Balance Experiment EBEX-2000. Part I: Overview and

353

energy balance. Bound.-Layer Meteor., 123, 1-28.

354

Steinfeld, G., M. O. Letzel, S. Raasch, M. Kanda, and A. Inagaki, 2007: Spatial

355

representativeness of single tower measurements on the imbalance problem with eddy-

356

covariance fluxes: results of a large-eddy simulation study. Bound.-Layer Meteor., 123,

357

77-98.

358

Whiteman, C. D., J. M. Hubbe, and W. J. Shaw, 2000: Evaluation of an inexpensive

359

temperature datalogger for meteorological applications. J. Atmos. Oceanic Technol., 17,

360

77-81.

361 362 363

17

363

FIGURE CAPTIONS

364 365 366 367 368 369 370 371 372

Fig. 1: Intercomparison pre-experiment of the 25 naturally ventilated HOBO sensors and one aspirated MetOne reference sensor (on the right side). Fig. 2: Difference between corrected and measured HOBO temperature (Tc − T) as a function of wind speed V for various combinations Rs and As values. Fig. 3: Shield area normal to the sun (As) and radiative heating term (RsAs) for three selected days in May 2007. Fig. 4: Histograms of the difference between the HOBO temperature and the aspirated temperature before and after applying the radiation correction.

373 374

18

374

Table 1: Selected specifications of the HOBO 12-Bit Smart Sensor

Specifications Measurement range

−40 °C to +100 °C

Accuracy

< ±0.2 °C from 0 °C to 50 °C

Resolution

< 0.03 °C from 0 °C to 50 °C

Drift

< 0.1 per year

Response time

< 2 min (in 2 m s−1 moving air flow)

375 376

19

376 377 378

Fig. 1: Intercomparison pre-experiment of the 25 naturally ventilated HOBO sensors and one aspirated MetOne reference sensor (on the right side).

379

20

379 380 381

Fig. 2: Difference between corrected and measured HOBO temperature (Tc − T) as a function of wind speed V for various combinations Rs and As values.

382

21

383 384 385

Fig. 3: Shield area normal to the sun (As) and radiative heating term (RsAs) for three selected days in May 2007.

386

22

387 388 389

Fig. 4: Histograms of the difference between the HOBO temperature and the aspirated temperature before and after applying the radiation correction.

390

23