You are on page 1of 57

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/353299572

Quantifying the Impacts of Land Surface Modeling on Hub-Height Wind Speed


under Different Soil Conditions

Article  in  Monthly Weather Review · July 2021


DOI: 10.1175/MWR-D-20-0363.1

CITATIONS READS

0 23

4 authors, including:

Geng Xia Caroline Draxl


National Renewable Energy Laboratory National Renewable Energy Laboratory
20 PUBLICATIONS   163 CITATIONS    42 PUBLICATIONS   1,383 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

IEA Wind Task 36 Wind Power Forecasting View project

Soil moisture linkages to the Great Plains low-level jet View project

All content following this page was uploaded by Geng Xia on 19 July 2021.

The user has requested enhancement of the downloaded file.


1 Quantifying the Impacts of Land Surface Modeling on Hub-Height Wind Speed under

2 Different Soil Conditions

4 Geng Xia1, Caroline Draxl1,2, Larry K. Berg3, David Cook4


1National
5 Wind Technology Center/ National Renewable Energy Laboratory, Golden, CO 80401

6 USA
2Renewable
7 and Sustainable Energy Institute, Boulder, CO 80309, USA
3
8 Pacific Northwest National Laboratory, Richland, WA 99352 USA
4333
9 Waxwing Avenue, Naperville, IL 60565-1243

10

11 Correspondence to Geng Xia (Geng.Xia@nrel.gov)

12

13 Resubmitted to Monthly Weather Review

14

15

16

17

1
18 Abstract

19 We investigate the impact of three land surface models (LSMs) on simulating hub-height wind

20 speed under three different soil regimes (dry, wet, and frozen) to improve understanding of the

21 physics of wind energy forecasts using the Weather Research and Forecasting (WRF) model. A

22 six-day representative period is selected for each soil condition. The simulated wind speed, surface

23 energy budget and soil properties are compared with the observations collected from the second

24 Wind Forecast Improvement Project (WFIP2). For the selected cases, our simulation results

25 suggest that the impact of LSMs on hub-height wind speed are sensitive to the soil states but not

26 so much to the choice of LSM. The simulated hub-height wind speed is in much better agreement

27 with the observations for the dry soil case than the wet and frozen soil cases. Over the dry soil,

28 there is a strong physical connection between the land surface and hub-height wind speed through

29 near-surface turbulent mixing. Over the wet soil, the simulated hub-height wind speed is less

30 impacted by the land surface due to weaker surface fluxes and large-scale synoptic disturbances.

31 Over the frozen soil, the LSM seems to have limited impact on hub-height wind speed variability

32 due to the decoupling of the land surface with the overlying atmosphere. Two main sources of

33 modeling uncertainties are proposed. The first is the insufficient model physics representing the

34 surface energy budget, especially the ground heat flux, and the second is the inaccurate initial soil

35 states such as soil temperature and soil moisture.

36

2
37 Significance Statement

38 To facilitate better wind and wind energy forecasts, the Second Wind Forecasting Improvement

39 Project aims to improve the representation of boundary layer physics in numerical models. Even

40 though significant modeling advancements have been achieved, additional sources of uncertainties,

41 such as those arising from the land surface model, have not been addressed. In this study, we

42 examine the impact of land surface model on hub-height wind speed under different soil conditions.

43 Our results indicate a better agreement of the modeled hub-height wind speed with observations

44 for the dry soil case than the wet and frozen soil cases, and that is largely insensitive to the choice

45 of the land surface model. Two main sources of modeling uncertainties have been proposed. These

46 results are useful for providing potential guidance for future land surface model development.

3
47 1.Introduction

48

49 Wind energy has experienced a drastic growth during the most recent decade, exceeding

50 hydropower as the nation’s largest renewable energy source (U.S. Department of Energy (DOE);

51 DOE 2018). Accompanying such rapid development is a growing demand for more accurate wind

52 forecasts for the wind energy sectors (Veers et al. 2019). However, uncertainties in weather

53 forecasts continue to pose a challenge to the wind industry (Lundquist et al. 2019, Veers et al.

54 2019). This could be partially attributed to the fact that less emphasis has been devoted to

55 improving the forecasts of wind at heights of 50-200m above ground level (AGL), compared to

56 the traditional high-impact weather events (Olson et al. 2019). Therefore, understanding and

57 improving the numerical forecasts of rotor-layer (40m ~ 120m AGL) wind will be critical for a

58 sustainable growth of wind energy.

59

60 In 2015, the DOE and the National Oceanic and Atmospheric Administration (NOAA) launched

61 the second Wind Forecast Improvement Project (WFIP2) with the specific goal of improving the

62 representation of boundary layer physics and mesoscale processes in numerical models for better

63 wind and wind power forecasts in complex terrain (Shaw et al. 2019, Wilczak et al. 2019, Olson

64 et al. 2019, Bianco et al. 2019, Draxl et al. 2021, Xia et al. 2020, Irina et al. 2020, Pichugina et al.

65 2020). The Columbia Basin of eastern Oregon and Washington was selected as the targeted region

66 and an extensive field campaign was conducted to collect observations in support of numerical

67 weather prediction (NWP) model development. Through WFIP2, significant forecast

68 improvements were achieved by improving the treatment of complex terrain, vertical mixing

69 between the surface and the upper atmosphere, and the treatment of turbulent mixing in the

4
70 horizontal as well as vertical (Olson et al. 2019). Despite these successes, there are additional

71 known significant sources of rotor-layer forecast errors that have not yet been investigated, such

72 as the errors arising from land surface model (LSM).

73

74 The LSMs are coupled to various schemes, such as the surface-layer, radiation, microphysics and

75 convective schemes, together with the land’s state variables and land surface properties, to provide

76 heat and moisture fluxes over land and sea-ice points. These fluxes provide a lower boundary

77 condition for the vertical transport computed in the planetary boundary layer (PBL) schemes.

78 Multiple modeling studies have indicated that land surface process plays a key role in regulating

79 regional weather and climate (e.g. Sobel et al. 2008, Chen et al. 2014, Sun et al. 2017, 2020, Ma

80 et al. 2017, Lee et al. 2019, Zhuo et al. 2019, Grachev et al. 2020, Zhang et al. 2020). However,

81 not many studies focused on the rotor-layer wind, which is of significant importance for wind

82 renewable energy. Wharton et al. (2016) examined the role of surface energy exchange for

83 simulating wind turbine inflow. They found that there is a relationship between surface energy

84 partitioning and near-surface wind shear and the relationship is stronger during the summer than

85 during the autumn. However, their research focused on wind shear rather than wind speed and

86 their model simulation was performed as a single continuous run without frequent reinitialization,

87 which is not a common practice for wind energy forecasts.

88

89 By leveraging the multi-scale dataset from the WFIP2, the goal of this paper is to evaluate and

90 compare the skills of three LSMs applied in WRF: Noah, Noah with multi-physics (NoahMP) and

91 Rapid Update Cycle (RUC), in simulating the hub-height wind speed (100m AGL). The results

92 will provide useful information about the relative performance of these LSMs for wind energy

5
93 forecasts and suggest physical processes that may require further evaluations. In this study, three

94 cases of distinctly different soil regimes are selected to conduct model simulations, which will

95 further address the impact of the land surface on hub-height wind under different soil conditions.

96

97 The paper is organized as follows. Section 2 introduces the case selections, validation data, model

98 configuration and land-surface schemes tested. Section 3 discusses the model results and their

99 evaluations with observations. Section 4 examines the factors responsible for the differences

100 between model simulations and observation, followed by the conclusions in Section 5.

101

102 2. Data and Methodology

103

104 a. Case selection

105 Three cases with distinctively different soil regimes are selected for this study. The soil regime is

106 first determined by Gravchev et al. (2020) using the soil moisture (SM) and soil temperature (ST)

107 measurements from the WFIP2 field campaign. The first case, which represents the dry soil

108 condition, ranges from 13 to 19 August 2016. During this period, an upper level ridge persisted

109 over the Cascades and eastern Columbia River Gorge and the atmosphere was relatively stable.

110 The second case represents wet soil conditions from 4 to 9 November 2016. During this period, a

111 Pacific storm was approaching the northwestern coast, causing falling pressure west of the

112 Cascades. This created an offshore (~4mb) pressure gradient across the Cascades which resulted

113 in easterly gap flows. Cold pools were developed during the overnight hours due to radiational

114 cooling. The third case represents the frozen soil condition, lasting from 16 to 21 February 2017.

115 In this case, a low pressure dominated the flow and moved slowly eastward over the campaign

6
116 region. Even though upper-level westerlies persisted, the near surface was weakly decoupled with

117 the overlying atmosphere due to a persistent cold pool. All the weather patterns described here are

118 documented in the WFIP2 event log (Atmosphere to Electrons (A2e); A2e 2015). Interested

119 readers can refer to that document for more details.

120

121 To further confirm that the surface conditions during these three periods are indeed representative

122 of dry, wet and frozen soil conditions at our study region, we examine the timeseries of observed

123 ST and SM measurements at the Physics Site 3 (PS03; Fig. 1). Note that the observed ST and SM

124 are measured at the depth of 5 cm beneath the soil surface. For the dry soil case, the ST is around

125 30℃ while the SM exhibits a constant line at 0.12 m3m-3. During this period, the SM measurements

126 have reached the low limitation value of the measurement technique whereas in reality, the actual

127 SM content should be less than shown in Figure 1. For the wet soil case, the ST is around 10℃

128 while the SM is nearly constant at around 0.43 m3m-3, indicating that the soil is very wet during

129 this period. For the frozen soil case, the ST is about 0℃ while the SM increases from 0.33 m3m-

3to
130 0.50 m3m-3. This suggests that the top soil layer is thawing and the thawed water drained down

131 the sandy silty soil where the sensor was situated, allowing the measured SM to elevate over the

132 next six days.

133

134 Overall, the observations clearly confirm that our selected case days are representative of dry,

135 wet and frozen soil conditions.

136

137 b. WFIP2 observations

138

7
139 1) Observation sites

140 Two sites from the WFIP2 campaign are selected to provide observational data for this work. The

141 first site is PS03 providing measurements for the surface energy flux. Even though other stations

142 also have surface flux measurements, PS03 is the only measurement site that has the complete

143 measurements for the surface energy budget terms (Eq. 1),

144 NR + SH + LH + GH =0 (1)
145 where NR is the net radiation, SH is the sensible heat flux, LH is the latent heat flux and GH

146 represents the ground heat flux. However, there are no wind speed measurements at PS03 (45.638 o

147 N, 120.671o W). The wind speed observations used in this study are obtained from the SoDAR

148 profiler at the Physics Site 01 (PS01; 45.637o N, 120.679o W). The distance between these two

149 sites is about 0.65 km. Therefore, the surface flux and wind speed measurements are close together

150 that they are represented by local impacts. In fact, these two sites are so close together that they

151 are represented as a single grid point in the domain setup. The quality-controlled data are archived

152 at the Data Archive and Portal (DAP; https://a2e.energy.gov/about/dap).

153

154 At both sites, the soil type is sandy silt loam which can be very loose when dry. The vegetation

155 type is the soft white winter wheat which is normally planted in mid-October. In August (dry soil

156 case), there is no green vegetation covering the surface. In November, the crop would probably be

157 up and green, but it would have been only about 5 to 10 cm tall and thus not cover much of the

158 soil surface. The wheat is pretty much dormant throughout the winter and will be green again after

159 fertilization by early March. Therefore, the measured LH flux is negligible during the dry soil case

160 whereas during the wet and frozen soil cases virtually all LH was from evaporation from the soil

161 rather than transpiration from the plants (Grachev et al. 2020).

8
162

163 2) Surface flux data

164 The Energy Balance Bowen Ratio (EBBR) system from PS03 produces 30-minute estimates of

165 the vertical fluxes of sensible and latent heat at the local surface (ARM 2015). Key instrumental

166 components include Radiation and Energy Balance Systems (REBS) net radiometer, REBS soil

167 moisture probes, REBS soil heat flow plates and REBS soil temperature probes. Surface energy

168 fluxes are calculated from the collected observations, and are quality checked to remove suspicious

169 spikes that are greater than 1000 Wm-2 or smaller than -200 Wm-2. The data are later compiled into

170 1-hour averages in order to compare with the model output. The uncertainty of the measurement

171 is on the order of 10% for the surface fluxes and that is determined based on differences in

172 measurements using the same technique by different investigators at the same location (Twine et

173 al. 2000, Jiang et al. 2004, Liu et al. 2013). Note that the observed GH flux is defined as the soil

174 heat flow plate measurement at 5cm under the ground, adjusted for the change in energy storage

175 in the soil above the heat flow plate. More details about instrumentation, data quality and data

176 uncertainty can be found from the user handbook (Cook 2018).

177

178 3) SoDAR data

179 The Vaisala Triton SoDAR wind profiler (Vaisala 2015) at PS01 measures wind speed, direction,

180 and turbulence intensity at heights from 30 m to 200 m above ground every 10 minutes. Two

181 automated procedures have been applied to remove erroneous data due to precipitation and

182 measurement error. Similar to the surface flux data, the observed wind speeds are also compiled

183 into hourly averages. Because this study focuses on renewable wind energy, wind speeds at

184 common hub-heights (100 m AGL) are used in this analysis.

9
185

186 c. Model simulation

187

188 1). Simulation design

189 The simulation design is similar as employed in Xia et al. (2020), and the following text is derived

190 from there with minor modifications.

191

192 The Weather Research and Forecasting model (WRF) version 4.1.2 is used to conduct the model

193 simulations in this study (Skamarock and Klemp 2008, Powers et al. 2017). The boundary and

194 initial conditions are derived from the National Centers for Environmental Prediction/North

195 American Mesoscale Forecast System 12-km analysis (NAM; https://doi.org/10.5065/G4RC-

196 1N91).

197

198 The simulations are performed with three nested domains centered on the WFIP2 campaign region

199 (Fig. 2). The first domain consists of 95 x 80 grid points with a horizontal grid spacing of 9 km,

200 the second domain consists of 88 x 85 grid points with a grid spacing of 3km while the innermost

201 domain consists of 91 x 88 grid points with a grid spacing of 1 km. Corresponding topography,

202 soil characteristics and the MODIS-based land cover dataset are used to match domain grid spacing

203 respectively. Physical packages that are applied include Rapid Radiative Transfer Model for

204 shortwave and longwave radiation (Iacono et al. 2008), Thompson aerosol-awareness

205 microphysics scheme (Thompson et al. 2008, Thompson and Eidhammer 2014), Mellor-Yamada

206 Nakanishi and Niino (MYNN) Level 2.5 planetary boundary layer scheme and MYNN surface

207 layer scheme (Mellor and Yamada 1982, Nakanishi and Niino 2009). Note that cumulus

10
208 convection is treated explicitly for second and third domains while it is parameterized for the

209 outermost domain using the Kain-Fritsch scheme (Kain and Fritsch 1990, 1993, Kain 2004). A

210 total number of 52 vertical levels is employed with finer resolution at lower levels (16 within the

211 lowest 200m) and coarser resolution at higher levels. The lowest model level height is 5 m AGL.

212 Note that the reference grid point for PS01 and PS03 is the same, represented by the black point

213 in the center of the innermost domain.

214

215 Three LSMs, the Noah, NoahMP and RUC, are examined in this study. The differences between

216 these LSMs are described in detail in the next section. For each LSM, simulations were conducted

217 for all three case periods using the 3-day reinitialization method, in which the first day of each 3-

218 day run is discarded as spinup and the next two days are retained for further analysis (Xia et al.

219 2017, 2019).

220

221 2) Land surface models

222 Several important features of these LSMs, including vegetation types, soil levels, snow layers and

223 canopy separation, are compared in Table 1. The Noah LSM has four soil layers with a total soil

224 depth of 2 m and a surface layer of vegetation and soil surface to consider biophysical and carbon

225 cycling processes (Mahrt and Ek 1984, Mahrt and Pan 1984, Pan and Mahrt 1987, Ek and Mahrt

226 1991, Chen et al. 1996, Koren et al. 1999, Chen and Dudhia 2001a, 2001b, Ek et al. 2003). This

227 scheme relies on soil and vegetation processes to simulate SM and ST in the soil layer, water and

228 snow stored on the canopy and surface flux exchange between the land and the atmosphere.

229

11
230 The NoahMP LSM is an improved version of Noah LSM in terms of better representation of

231 biophysical and hydrological processes (Niu et al. 2011, Yang et al. 2011). The default options of

232 each physical process parameterization sub-scheme were adopted in this experiment

233 (https://www.jsg.utexas.edu/noah-mp/scheme-options/). Major improvements include but are not

234 limited to 1) a vegetation canopy layer separated from the original surface layer; 2) a TOPMPDEL

235 (TOPography based hydrological MODEL) based runoff scheme (Niu et al. 2005) and a simple

236 groundwater model (Niu et al. 2007) for soil hydrology; and 3) an introduction of a more

237 permeable frozen soil (Cai 2015).

238

239 The RUC LSM was originally developed for NOAA weather prediction (Benjamin et al. 2004a,

240 2004b) but has now been incorporated into the WRF and High-Resolution Rapid Refresh (HRRR;

241 Smith et al. 2008) models. It has nine soil levels with a default soil depth of 3 m. The vegetation

242 processes as well as surface flux calculation are treated similarly to Noah LSM following the

243 concept developed by Pan and Mahrt (1987). The frozen soil processes are included to improve

244 snow treatment and phase change in soil (Smirnova et al. 1997, 2000, 2016, Benjamin et al. 2016).

245

246 d. Quantifying the impact of LSMs on hub-height wind speed

247 In WRF, the simulated wind speed is determined by solving a prognostic set of nonlinear equations

248 which includes multiple physical processes. Even though it is impossible to assess the contribution

249 of a single process (e.g., LSM) on hub-height wind speed, we can still assess its potential influence

250 on hub-height wind by establishing an idealized physical framework based on our current

251 understanding of the process.

252

12
253 As previously stated, the surface fluxes from LSMs serve as the lower boundary for turbulent

254 mixing computed by the PBL scheme. In the surface layer (approximately the bottom 10% of the

255 boundary layer), stronger surface mixing leads to weaker wind shear and wind speed (Stull 1998).

256 Therefore, we hypothesize that if the LSM has a potential impact on wind at hub-height level, we

257 would expect that stronger surface energy flux (e.g., SH) corresponds to weaker near-surface wind

258 shear and reduced hub-height wind speed. Such physical connection will be examined across all

259 soil cases from both the observations and model simulations. Following Wharton et al. (2016), we

260 used near-surface wind shear as an estimated measure for near-surface turbulence mixing. Wind

261 shear is estimated using Eq. 2,

262 (2)

263 where U is the mean horizontal wind speed (m/s) at height z (m), 𝑈𝑅 is the mean horizontal wind

264 speed (m/s) at a reference height 𝑍𝑅 (m) and α is a wind shear exponent used to describe the

265 variations in wind speed with height. In this analysis, the wind shear is calculated between 40 m

266 (𝑍𝑅 ) and 120 m (z). We use 120 m as the upper height limit because the available wind speed

267 measurements from the SoDAR decrease significantly above this height, especially for the wet

268 and frozen soil cases.

269

270 At nighttime, the land surface is mostly decoupled from the atmosphere. Therefore, only

271 measurements from daytime hours are used in this analysis. In addition, the daytime hours are

272 separated into three periods, morning (0600-0900 local time), noon (1000-1300 local time) and

273 afternoon (1400-1700 local time), to further demonstrate the temporal variability of such

274 connection.

13
275

276 3. Results

277

278 a. Comparing simulated hub-height wind speeds with observations

279 Figure 3 shows the timeseries of the observed and simulated hub-height wind speed from the three

280 soil cases. Note that about 20% and 40% of the observations are missing for the wet and frozen

281 soil cases respectively due to measurement errors and the WRF wind speed is vertically

282 interpolated to the 100 m AGL in order to compare with the observations. The simulated wind

283 speeds between the three LSMs are in good agreement with one another. However, there are

284 certainly some discrepancies between the simulated wind speed and observations, especially for

285 the wet and frozen soil cases. Figure 4 shows the Taylor diagram to further quantify how well the

286 simulated wind speed matches with the observations for these three cases, in terms of their root-

287 mean-square difference (RMSD), temporal correlation as well as variances (Taylor 2001). The

288 simulated wind speed during the dry soil case has a much higher correlation with the observations

289 than those from the wet and frozen soil cases. With respect to standard deviation, the closer the

290 dot is to the reference line (red), the better the agreement with the observation in terms of wind

291 speed variability. For the dry soil case, the three LSMs closely cluster around the reference line,

292 indicating similar variability between the observed and simulated wind speed. For the wet and

293 frozen soil cases, the simulated wind speed from the three LSMs exhibit larger variability than the

294 observations. As for RMSD, the values from the dry soil case are significantly smaller than those

295 from the wet and frozen soil cases, indicating the difference between the observed and simulated

296 wind speed is smallest under the dry soil condition. Overall, the simulated hub-height wind speeds

14
297 from all three LSMs are in much better agreement with observations during the dry soil case than

298 the wet and frozen soil cases.

299

300 b. Comparing the simulated surface energy budget with the observations

301 The measured and simulated NR, SH, LH and GH fluxes are shown in Figures 5–7 for each soil

302 case respectively to illustrate how well each LSM simulated the full surface energy budget. Table

303 2 shows both the observed and simulated mean midday (1100-1400 local time) Bowen ratio, which

304 is the ratio between SH and LH flux, to further describe surface energy partitioning.

305

306 For the dry soil case, the SH flux is the most dominant surface forcing from both the observations

307 and model simulation (Fig 5a). Since the magnitude of the SH flux is significantly larger than the

308 LH flux, this results in a large Bowen ratio, indicating a very dry surface condition. The RUC has

309 the largest Bowen ratio because it has the weakest LH flux (Fig. 5c). All the LSMs are able to

310 reproduce the temporal variability of the observed SH fluxes. However, they tend to underestimate

311 the peak daytime value with Noah having the most significant underestimation by about 50 Wm-
2.
312 The simulated daytime LH fluxes are generally weak with near zero value from RUC and around

313 20 Wm-2 from Noah and NoahMP, but the observations suggest high variability. This is because

314 the EBBR system, by definition, forces the energy balance to equal to the net radiation. Therefore,

315 in the case of dominant SH fluxes, the observed LH flux could very likely be overestimated and

316 appear with the wrong sign. In reality, however, the site area is quite dry in August so the LH flux

317 may actually be near zero, which is very similar to what the LSMs predicted. The simulated GH

318 flux, on the other hand, differs significantly from the observations, in terms of both magnitude and

319 pattern. All the LSMs tend to simulate stronger GH flux at daytime but weaker GH flux at

15
320 nighttime, as compared with observations. In addition, the daytime peak of the simulated GH flux

321 seems to occur too early compared with the observations.

322

323 For the wet soil case, the magnitude of the SH flux decreases during the daytime while the LH flux

324 increases. As a result, both the observed and simulated Bowen ratios drop significantly, suggesting

325 a semi-arid surface condition. In this particular case, the Noah and NoahMP produce a larger

326 midday Bowen ratio than the observation while the value from RUC is smaller. All the LSMs

327 manage to reproduce the variability of the observed SH flux. However, only the RUC manages to

328 capture the observed LH flux. Note that a precipitation event occurred on 06 November 2016 and

329 the RUC is capable of reproducing the observed LH flux fairly well. This could be attributed to

330 the fact that the soil layers and drivers of water flux exchanges are based on atmospheric

331 temperature and humidity instead of physiologically driven controls (Wharton et al. 2016). The

332 simulated precipitation from simulations using all three LSMs are very similar to each other in

333 terms of both the spatial pattern and intensity. Supporting Figure 1 (SFigure 1) shows the observed

334 and simulated hourly precipitation at Goldendale, which is located near the center of domain 3,

335 and all three simulations manage to qualitatively capture the precipitation event. The problem

336 associated with reproducing the observed GH flux is still evident for all three LSMs. Compared

337 with the observations, the simulated GH fluxes are too large during daytime and too small during

338 nighttime.

339

340 The frozen soil case exhibits the largest difference in terms of the total energy budget between the

341 observations and model simulations (Fig. 7a). This discrepancy is mainly contributed by the LH

342 and GH fluxes during the daytime. Compared with the observations, all the LSMs tend to

16
343 overestimate these two fluxes. As a result, the simulated Bowen ratios are larger than the observed

344 value. Note that even though the RUC predicts the most accurate LH fluxes during the wet soil

345 case, it also overestimates the LH flux by far for the frozen soil case. The observed GH flux is

346 essentially zero throughout the case period whereas the simulated values from all three LSMs

347 exhibit large variability. Over frozen soil, the difference in GH flux between observations and

348 model simulations is greater than that of the wet and dry soil, regardless of the LSM used. A more

349 detailed discussion about this issue is provided in section 4.

350

351 c. Examining the impacts of LSMs on hub-height wind speed

352 Following section 2d, the impact of LSMs on hub-height wind speed is examined by illustrating

353 the relationship between surface flux, hub-height wind speed and wind shear during the daytime.

354 If the land surface plays an important role in determining the hub-height wind speed, we will

355 expect a significant physical relationship associated with surface flux as well as wind speed and

356 shear.

357

358 Figures 8 and 9 show the scatter plots of SH flux versus hub-height wind speed and wind shear,

359 respectively, from both the observations and LSMs for the dry soil case. From the observations,

360 there is a strong physical connection, suggesting that stronger SH fluxes correspond with weaker

361 wind shear and lower wind speeds. This is consistent with our hypothesis that the land surface has

362 a significant impact on hub-height level over the dry soil. During the morning hours, the SH flux

363 is weak but wind shear and wind speed at the hub-height are strong. However, the strong wind

364 shear and wind speed should rather be more associated with the decay of the nocturnal boundary

365 layer. As time progresses toward noon, the SH flux drastically increases, reaching the maximum

17
366 value of 450 Wm-2 around solar noon time (Fig. 5b). During this period, the corresponding wind

367 shear and wind speed are smallest because of vigorous near-surface turbulent mixing in the

368 boundary layer. As the afternoon progresses, the surface fluxes decrease, indicating a weaker

369 coupling between the land surface and the hub-height level. This is accompanied by a slight

370 recovery of stronger wind shear and hub-height wind speed. The results from the various LSMs

371 general depict a similar picture, except that the simulated wind shear shows no difference or

372 slightly decreases in the afternoon as compared to noon. In addition, the variability of simulated

373 wind shear during the daytime is smaller than that from the observations.

374

375 Figures 10 and 11 show the similar plots but for the wet soil case. In this case, the observations

376 suggest that stronger SH flux corresponds with weaker hub-height wind but stronger wind shear.

377 This differs from our original hypothesis that wind shear should get weaker as surface fluxes get

378 stronger. Certainly, the impact of land surface on hub-height wind for the wet soil case is smaller

379 than that from the dry soil case. Similar to the dry soil case, the weak surface flux corresponds

380 with strong wind speed and wind shear during morning hours. However, that is more associated

381 with the decay of the nocturnal boundary layer as previously mentioned. Between noon and

382 afternoon, the Noah and RUC LSM manage to mostly reproduce the temporal variability of the

383 observed wind shear and wind speed. However, changes in surface fluxes from the NoahMP do

384 not seem to capture such impact.

385

386 For the wet soil case, the LH flux becomes a much more important surface forcing. To examine

387 the impact from LH flux, SFigures 2 and 3 show similar plots as Figure 10 and 11 but for LH flux.

388 Evidently, most of the identified relationships are determined by the SH flux rather than LH flux.

18
389 This is not a surprise because the SH flux is the buoyancy flux which mainly drives the turbulence

390 mixing at the near surface. Even though the magnitude of the LH flux is comparable to the SH

391 flux, it has a negligible impact on determining the magnitude and variability of hub-height wind

392 speed and wind shear.

393

394 The weaker coupling between the surface and hub-height winds for the wet soil case can be

395 attributed to two main factors. The first is the drastically reduced magnitude of the SH flux. Note

396 that the magnitude of the SH flux from the wet soil case is only about one third of that from the

397 dry soil case. This significantly weakens the near-surface turbulence mixing, thus reducing the

398 coupling between the land surface and hub-height level. The second factor is probably the

399 influence of large-scale disturbance. There was a Pacific storm approaching the WFIP2 region

400 during the case period, resulting in rainfall on 6 November 2016 (Fig.6c; A2e 2015). Such

401 disturbance will definitely have an influence on the distribution of hub-height level wind speed

402 and wind shear.

403

404 Figures 12 and 13 show the results from the frozen soil case. Overall, the observations suggest

405 very limited impacts of the land surface on hub-height wind speed and wind shear. The results

406 from the morning hours are generally similar to those from the dry and wet soil cases. From noon

407 to afternoon, the observed hub-height wind speed and wind shear increase but there is very little

408 change in the SH flux. Note that the magnitude of the SH flux from the frozen soil case is also the

409 weakest among the three cases. This suggests that the land surface is very likely decoupled from

410 the overlying atmosphere during the frozen soil case period. The WFIP2 observational team has

19
411 documented a cold pool event happening during this period which could contribute substantially

412 to this decoupling.

413

414 The results from the model simulations indicate a different story. From noon to afternoon, the

415 decrease in simulated SH flux is associated with an increase in the simulated hub-height wind

416 speed but a decrease in wind shear whereas LH does not play a significant role (SFigures 4 and 5).

417 Similar to the wet soil case, this indicates a weak coupling between the land surface and hub-height

418 level wind. Overall, the observations indicate almost nonexistent impact of the land surface on

419 hub-height wind speed and near-surface wind shear during the frozen soil period whereas the

420 simulations suggest a weak impact. The differences between model and observation could be

421 partially attributed to the difficulty in simulating cold pool in the current numerical model (Olson

422 et al. 2019)

423

424 4. Discussion

425

426 The preceding evaluation indicates a better agreement of the modeled hub-height wind speed with

427 observations for the dry soil case than the wet and frozen soil cases, and that is largely insensitive

428 to the choice of LSM. Over the dry soil, there is a strong physical connection between the land

429 surface and hub-height level due to near-surface turbulent mixing. However, the relationship is

430 weaker over wet soil and almost nonexistent over frozen soil.

431

432 There are many possible reasons responsible for the discrepancy between the model simulations

433 and observations. In this paper, two factors are discussed in detail because these are the two most

20
434 obvious limiting factors that stood out from the analysis. The first is the insufficient model physics

435 description of the surface energy flux, especially the GH flux and the second is the inaccuracy of

436 soil states, such as ST and SM.

437

438 Regardless of whichever soil surface, all LSMs have difficulties in predicting the observed GH

439 flux. For the dry and wet soil cases, the simulated GH flux is either too large during daytime or

440 too small during nighttime, as compared to the observations. For the frozen soil case, the observed

441 GH flux is almost zero throughout the entire period but none of the LSMs are able to capture that.

442 Previous studies have also indicated the discrepancies between the observed and simulated GH

443 flux (Smirnova et al. 1997, Warton et al. 2016, Zhang et al. 2020). Unfortunately, most LSM

444 studies focus on discussing the impacts of LH and SH fluxes, with very little emphasis on GH flux.

445 However, atmospheric features that are relevant for wind energy are generally close to the ground.

446 Therefore, getting the correct energy partitioning is essential. Even though the GH flux does not

447 directly influence hub-height level wind speed, it can still indirectly influence wind speed

448 predictions by altering ST and thus changing the near-surface turbulence mixing.

449

450 From a modeling standpoint, the simulated GH flux is generally calculated using the gradient

451 approach (Yang et al. 1999), which is given by

452 (3)

453 where 𝜆 is the soil heat conductivity, T is soil temperature, and z is soil depth. At the surface, GH

454 flux must be balanced by the NR, LH and SH fluxes. For Noah, Noah-MP and RUC, GH flux is

21
455 computed from the temperatures at the soil surface and that at some depth close to the surface

456 (Chen and Dudhia 2001a, 2001b, Niu et al. 2009, Smirnova et al. 1997).

457

458 From an observational perspective, measuring the GH flux is not an easy task because of the large

459 temperature gradients at the surface. To solve this issue, such as that from EBBR, measurements

460 are taken at 5 cm beneath the soil layer using soil heat flow plates (adjusted with the soil heat

461 conductivity, which is determined from the soil texture and water content) and then adding the

462 energy storage in the soil above the soil heat flow plate (determined from the change in ST with

463 time in the 0-5 cm strata of the soil). The simulated and observed GH fluxes show large

464 discrepancies for all the three soil cases. Understanding the exact reasons for the causes of these

465 differences is beyond the scope of this study. However, it is possibly related to the difference

466 between the observed and simulated ST.

467

468 For the dry and wet soil cases, the simulated ST from the three LSMs manages to capture the

469 general magnitude and temporal variability of the observed ST (Figure 14). To best compare with

470 the observations, the simulated ST from the top soil layer (10 cm beneath the ground) is used for

471 Noah and NoahMP whereas the averaged ST over the top three soil layers (0 ~ 4 cm beneath the

472 ground) is used for RUC. Notice that both the maximum and minimum simulated ST are generally

473 smaller than the observed. This cold bias in surface temperature has also been documented in the

474 literature (Chen et al. 2017, Johannsen et al. 2019). As for the frozen soil case, the simulated ST

475 differs significantly from the observations. The observed ST is almost zero throughout the entire

476 period. However, model simulations show large variations, especially for the RUC. The simulated

477 ST from Noah and NoahMP match well with the observations for the first three days but still show

22
478 large variations for the last three days. In addition, the observations suggest that the correlation

479 between the GH flux and the ST is largest for the dry soil case, it is weaker for the wet soil case,

480 and is weakest for the frozen soil case (Table 3). This indicates that the GH flux has a strong

481 dependency on ST and this relationship gets weaker as temperature decreases. For Noah and

482 NoahMP, the corresponding correlation is much smaller than that from the observations, possibly

483 indicating that the observed connection between the observed GH and ST is not well reflected in

484 the parameterization. As for the RUC, the correlation is very high regardless of the soil regime.

485 Note that the high correlation during the frozen soil case can be misleading because it is caused by

486 the anomalously high ST, which is absent from the observations. This result suggests that the

487 simulated GH flux from RUC is strongly correlated with the ST, which is similar to the

488 observations. However, the parameterization is not very sensitive to soil water content which could

489 be a potential limiting factor.

490

491 Other than the challenge of simulating GH fluxes, the other limiting factor for predicting hub-

492 height wind speed could be associated with the inaccurate soil states. The above discussion has

493 demonstrated the difference between the simulated and observed ST while the following section

494 will address the issue with SM. For the dry soil case, the simulated SM from all the three LSMs is

495 smaller than the observations (Fig. 15a). Note that the observed SM is a constant line because it

496 has reached the low limitation value of the measurement technique. It is likely that the actual SM

497 is smaller than the observed values. Therefore, for this particular case, the simulated SM should

498 be closer to reality than the observations. For the wet soil case, the simulated and observed SM

499 shows large discrepancies as the observed SM is significantly greater than the simulations. For

500 instance, the observed SM is around 0.42 m3 m−3, while the simulation is only about 0.2 m3 m−3.

23
501 For the frozen soil case, no LSM is able to capture the variability of the observed SM and the

502 simulated values are significantly smaller than the observations. Note that the difference in

503 simulated SM between Noah and NoahMP is greater than that from the dry soil and wet soil cases.

504 SFigure 6 shows the timeseries of the simulated SM from Noah and NoahMP for the first set of 3-

505 day simulation of the frozen soil case. The first day (left side of the black line) is treated as spin

506 up while the next two days (right side of the black line) are used for analysis. Evidently, the

507 simulated SM from NoahMP drops significantly during the first day while that from Noah stays

508 constant. The difference might be associated with the treatment of run off and frozen soil physics

509 between these two LSMs. Since the soil moisture content is associated with soil heat conductivity,

510 the deficiency in accurately predicting the SM will certainly impact the calculation of GH flux as

511 well.

512

513 Overall, the simulated SM and ST during the dry soil case match best with the observations. During

514 the wet soil case, the simulated SM differs greatly from the observations whereas both the

515 simulated ST and SM show large discrepancies from the observations during the frozen soil case.

516 This certainly points out the importance of accurate soil conditions in predicting hub-height wind

517 speed because the analysis has also indicated that the simulated hub-height wind speeds match

518 best with the observations during the dry soil case.

519

520 As this analysis is mostly relying on observations from a single physics site, the spatial

521 representativeness of the measurements, especially in the case of ST and SM, can be questionable

522 (Bell et al. 2013, Diamond et al. 2013). To address this issue, ST and SM measurements from

523 PS01 are obtained from NOAA and compared with those from PS03. Overall, the ST

24
524 measurements between these two sites are similar. However, there is a systematic bias of about

525 0.3 m3 m−3 in the SM measurements for the wet and frozen soil periods. Note that the instruments

526 used to measure SM are different at these two sites. Even though this does introduce uncertainties

527 into our results, it does not affect the main conclusion as the measured SM at PS01 is also vastly

528 different from the model simulations for the wet and frozen soil periods.

529

530 5. Conclusions

531

532 To understand the impact of LSMs on short-term wind forecasting, this study evaluates and

533 compares the performance of three LSMs (Noah, NoahMP and RUC) coupled with WRF in

534 simulating hub-height wind speed under three distinctly different soil conditions (dry, wet and

535 frozen). The simulated hub-height wind speed, surface energy budget and soil properties are

536 compared with the observations collected from the WFIP2 field campaign to examine the LSM’s

537 capability in simulating these variables and providing potential guidance for improvements of

538 these LSMs. The primary findings are summarized as follows:

539

540 i) For the selected three case periods, the impact of LSMs on hub-height wind speed are

541 sensitive to the soil states but not so much to the choice of LSM.

542 ii) The simulated hub-height wind speed is in much better agreement with the observations

543 for the dry soil case than the wet and frozen soil cases.

544 iii) Over the dry soil, there is a strong physical connection between the land surface and hub-

545 height wind speed through near-surface turbulent mixing. Over the wet soil, the simulated

546 hub-height wind speed is less impacted by land surface because of weaker surface fluxes

25
547 and large-scale synoptic disturbance. Over the frozen soil, the LSM seems to have limited

548 impacts on hub-height wind speed because of decoupling of the land surface from the hub-

549 height level.

550 iv) Two main sources of uncertainties are identified to explain the differences between the

551 observations and model simulations. The first is the insufficient model physics representing

552 the surface energy budget, especially the ground heat flux, and the second is the inaccuracy

553 of soil states, such as ST and SM.

554

555 This work helps to document the impacts of LSMs on simulated turbine-height wind speed and

556 wind shear, which is a critical source of uncertainty on wind energy forecasting. The results point

557 out the limitations associated with the current LSMs and open up new opportunities for additional

558 research funding and topics. For instance, more research efforts should be focused on enhancing

559 our understanding of surface energy partitioning and their relative impact on hub-height level wind

560 speed by deploying more flux and SoDAR instruments at the WFIP2 site for a longer period of

561 time. In addition, data assimilation should be applied to improve the representation of initial land

562 surface conditions in the numerical simulations. All of these approaches will help to ensure a

563 sustainable growth and development of wind energy in the United States.

564

26
565 Acknowledgments

566 The authors thank the WFIP2-experiment participants who aided in the deployment and the

567 collection of remote sensing data and our colleagues who monitored, quality controlled and

568 provided data to the DAP (https://a2e.energy.gov/about/dap). The research was performed using

569 computational resources sponsored by the U.S. Department of Energy's Office of Energy

570 Efficiency and Renewable Energy and located at the National Renewable Energy Laboratory.

571

572 This work was authored in part by the National Renewable Energy Laboratory, operated by

573 Alliance for Sustainable Energy, LLC, for the U.S. Department of Energy (DOE) under Contract

574 No. DE-AC36-08GO28308. Funding was provided by the U.S. Department of Energy Office of

575 Energy Efficiency and Renewable Energy Wind Energy Technologies Office. The views

576 expressed in the article do not necessarily represent the views of the DOE or the U.S. Government.

577 The U.S. Government retains and the publisher, by accepting the article for publication,

578 acknowledges that the U.S. Government retains a nonexclusive, paid-up, irrevocable, worldwide

579 license to publish or reproduce the published form of this work, or allow others to do so, for U.S.

580 Government purposes. A portion of the research was performed using computational resources

581 sponsored by the Department of Energy's Office of Energy Efficiency and Renewable Energy and

582 located at the National Renewable Energy Laboratory.

27
583 Data Availability Statement

584 The NAM reanalysis data used in this study are publicly available from the NCAR/UCAR

585 Research Data Archive at https://rda.ucar.edu/datasets/ds609.0/. Due to privacy and ethical

586 concerns, observational data used in this study cannot be made publicly available. Further

587 information about the data is available from the Data Archive and Portal (DAP;

588 https://a2e.energy.gov/about/dap).

28
589 References:

590

591 ARM, 2015: Energy Balance Bowen Ratio Station. ARM Data Center user facility, accessed 10

592 April 2020, https://a2e.energy.gov/data/wfip2/met.z20.b0.

593 A2e, 2015: U.S. Department of Energy, Office of Energy Efficiency and Renewable Energy,

594 accessed 06 May 2021, https://a2e.energy.gov/data/wfip2/log.z01.00.

595 Bell, J.E., and Coauthors, 2013: U.S. Climate Reference Network Soil Moisture and

596 Temperature Observations. J. Hydrometeor., 14, 977-988, https://doi.org/10.1175/JHM-

597 D-12-0146.1

598 Bianco L., and Coauthors, 2019: Impact of model improvements on 80-m wind speeds during the

599 second Wind Forecast Improvement Project (WFIP2), Geosci. Model Dev. Discuss., 12,

600 4803-4821, https://doi.org/10.5194/gmd-12-4803-2019

601 Benjamin, S.G., and Coauthors, 2004a: An hourly assimilation– forecast cycle: The RUC. Mon.

602 Wea. Rev., 132, 495–518, https://doi.org/10.1175/1520-

603 0493(2004)132<0495:AHACTR>2.0.CO;2

604 Benjamin S.G., G.A. Grell, J.M. Brown, T.G. Smirnova, and R. Bleck, 2004b: Mesoscale

605 weather prediction with the RUC hybrid isentropic/terrain-following coordinate model.

606 Mon. Wea. Rev., 132, 473–494, https://doi.org/10.1175/1520-

607 0493(2004)132<0473:MWPWTR>2.0.CO;2

608 Benjamin, S.G., S.S. Weygandt, J.M. Brown, M. Hu, C.R. Alexander, T.G. Smirnova, and G.S.

609 Manikin, 2016: A North American hourly assimilation and model forecast cycle: the Rapid

610 Refresh. Mon. Wea. Rev, 144, 1669–1694. https://doi.org/10.1002/qj.3179

29
611 Cai, X., 2015: Hydrological assessment and biogeochemical advancement of the Noah-MP land

612 surface model. Doctor of Philosophy, Dept. of Geological Sciences, The University of

613 Texas at Austin, 164 pp.

614 Chen, F., and Coauthors, 1996: Modeling of land surface evaporation by four schemes and

615 comparison with FIFE observations. J Geophys Res- Atmos, 101, 7251–7268,

616 https://doi.org/10.1029/95JD02165

617 Chen, F., and J. Dudhia, 2001a: Coupling an advanced land surface hydrology model with the

618 Penn State-NCAR MM5 modeling system. Part I. Model implementation and sensitivity,

619 Mon. Wea. Rev, 129, 569–585, https://doi.org/10.1175/1520-

620 0493(2001)129<0569:CAALSH>2.0.CO;2

621 Chen, F., and J. Dudhia, 2001b: Coupling an advanced land surface–hydrology model with the

622 Penn State–NCAR MM5 modeling system. Part II: preliminary model validation. Mon.

623 Wea. Rev, 129, 587–604, https://doi.org/10.1175/1520-

624 0493(2001)129<0587:CAALSH>2.0.CO;2

625 Chen. F, C. Liu, J. Dudhia, and M. Chen, 2014: A sensitivity study of high- resolution regional

626 climate simulations to three land surface models over the western United States, J

627 Geophys Res- Atmos, 119, 7271–7291, doi:10.1002/ 2014JD021827.

628 Chen, X, Y. Liu, and G. Wu, 2017: Understanding the surface temperature cold bias in CMIP5

629 AGCMs over the Tibetan Plateau. Adv. Atmos. Sci. 34, 1447–1460.

630 https://doi.org/10.1007/s00376-017-6326-9

631 Cook, D. R., 2018: Energy Balance Bowen Ratio (EBBR) instrument handbook, Technical

632 Report Rep. DOE/SC‐ARM/TR‐037, U.S.Department of

633 Energy.https://www.arm.gov/publications/tech_reports/handbooks/ebbr_handbook.pdf

30
634 DOE, 2018: Today in energy. Energy Information Administration, accessed 10 May 2019,

635 www.eia.gov /todayinenergy/detail.php?id=35412.

636 Diamond, H.J., and Coauthors, 2013: U.S. Climate Reference Network after One Decade of

637 Operations: Status and Assessment. Bull. Amer. Meteor. Soc., 94, 485-498,

638 https://doi.org/10.1175/BAMS-D-12-00170.1.

639 Draxl, C., and Coauthors, 2021: Mountain waves impact wind power generation. Wind Energy

640 Sci., 6, 45-60, https://doi.org/10.5194/wes-6-45-2021.

641 Ek, M.B., and L. Mahrt, 1991: OSU 1-D PBL Model User's Guide, Dep. of Atmos. Sci., Oreg.

642 State Univ., Corvallis, 120pp

643 Ek, M.B., K.E. Mitchell, Y. Lin, E. Rogers, P. Grunmann, V. Koren, G. Gayno, and J.D.

644 Tarpley, 2003: Implementation of Noah land surface model advancements in the National

645 Centers for Environmental Prediction operational mesoscale Eta model, J. Geophys. Res.,

646 108, 8851, https://doi.org/10.1029/2002JD003296.

647 Grachev, A., C.W. Fairall, B.W. Blomquist, J.S.H. Fernando, L.S. Leo, F. Otárola-Bustos F, M.J.

648 Wilczak, and K.L. McCaffreya, 2020: On the surface energy balance closure at different

649 temporal scales. Agric For Meteorol, 281, 107823,

650 https://doi.org/10.1016/j.agrformet.2019.107823

651 Irina, V.D., and Coauthors, 2020: Wind ramp events validation in NWP forecast models during

652 the second Wind Forecast Improvement Project (WFIP2) using the Ramp Tool and

653 Metric (RT&M). Wea. Forecasting, 35, 2407-2421, https://doi.org/10.1175/WAF-D-20-

654 0072.1.

655 Iacono, M.J., J.S. Delamere, E.J. Mlawer, M.W. Shephard, S.A. Clough, and W.D. Collins,

656 2008: Radiative forcing by long-lived greenhouse gases: Calculations with the AER

31
657 radiative transfer models, J Geophys Res- Atmos, 113, D13103,

658 https://doi.org/10.1029/2008JD009944

659 Johannsen, F., S. Ermida, J.P.A. Martins, I.F. Trigo, M. Nogueira, and E. Dutra, 2019: Cold Bias

660 of ERA5 Summertime Daily Maximum Land Surface Temperature over Iberian

661 Peninsula. Remote Sens., 11, 2570, https://doi.org/10.3390/rs11212570.

662 Jiang, L., S. Islam, and T.N. Carlson, 2004: Uncertainties in latent heat flux measurement and

663 estimation: implications for using a simplified approach with remote sensing data. Can. J.

664 Remote. Sens., 30, 769–787. https://doi.org/10.5589/m04-038

665 Kain, J. S., and J. M. Fritsch, 1990: A one-dimensional entraining/ detraining plume model and

666 its application in convective parameterization, J. Atmos. Sci., 47, 2784-2802.

667 https://doi.org/10.1175/1520-0469(1990)047<2784:AODEPM>2.0.CO;2

668 Kain, J. S., and J. M. Fritsch, 1993: Convective parameterization for mesoscale models: The

669 Kain-Fritcsh scheme, The representation of cumulus convection in numerical models, K.

670 A. Emanuel and D.J. Raymond, Eds., Amer. Meteor. Soc., 165-170.

671 Kain, J. S., 2004: The Kain-Fritsch convective parameterization: An update. J. Appl. Meteor., 43,

672 170-181, https://doi.org/10.1175/1520-0450(2004)043<0170:TKCPAU>2.0.CO;2.

673 Koren, V., J.C. Schaake, K.E. Mitchell, Q.Y. Duan, F. Chen, and J.M. Baker, 1999: A

674 parameterization of snowpack and frozen ground intended for NCEP weather and climate

675 models, J. Geophys. Res., 104, 19569–19585, h https://doi.org/10.1029/1999JD900232.

676 Lundquist, J.K., K.K. DuVivier, and D. Kaffine, 2019: Costs and consequences of wind turbine

677 wake effects arising from uncoordinated wind energy development. Nat Energy, 4, 26–

678 34, https://doi.org/10.1038/s41560-018-0281-2.

32
679 Lee, T.R., M. Buban, D.D. Turner, T.P. Meyers, and C.B. Baker, 2019: Evaluation of the High-

680 Resolution Rapid Refresh (HRRR) Model Using Near-Surface Meteorological and Flux

681 Observations from Northern Alabama. Wea. Forecasting, 34, 635–

682 663, https://doi.org/10.1175/WAF-D-18-0184.1.

683 Liu, G., Y. Liu, and S. Endo, 2013: Evaluation of surface flux parameterizations with long‐term

684 ARM observations. Mon. Wea. Rev, 141, 773–797. https://doi.org/10.1175/mwr‐d‐12‐

685 00095.1.

686 Mahrt, L., and M. Ek, 1984: The influence of atmospheric stability on potential evaporation. J.

687 Climate Appl. Meteor., 23, 222–234. https://doi.org/10.1175/1520-

688 0450(1984)023<0222:TIOASO>2.0.CO;2.

689 Mahrt, L., and H.L. Pan, 1984: A two-layer model of soil hydrology. Bound.-Layer Meteor., 29,

690 1–20.

691 Ma, Y., X.M. Zeng, Y. Zhang, N. Wang, Y. Zheng, G. Wang, and C. Chen, 2017: Impact of the

692 Choice of Land Surface Scheme on a Simulated Heatwave Event: The Case of Sichuan-

693 Chongqing Area, China, Adv. Meteorol., 2017, 9545896,

694 https://doi.org/10.1155/2017/9545896.

695 Mellor, G.L. and T. Yamada, 1982: Development of a turbulence closure model for geophysical

696 fluid problems. Rev. Geophys., 20, 851–875, https://doi.org/10.1029/RG020i004p00851.

697 Nakanishi, M., and H. Niino, 2009: Development of an improved turbulence closure model for

698 the atmospheric boundary layer. J. Meteor. Soc. Japan, 87, 895–912,

699 https://doi.org/10.2151/jmsj.87.895.

33
700 Niu, G.Y., Z.L. Yang, R.E. Dickinson, and L.E. Gulden, 2005: A simple TOPMODEL-based

701 runoff parameterization (SIMTOP) for use in global climate models. J Geophys Res-

702 Atmos., 110, D21106, https://doi.org/10.1029/2005JD006111

703 Niu, G.Y., Z.L. Yang, R.E. Dickinson, L.E. Gulden, and H. Su, 2007: Development of a simple

704 groundwater model for use in climate models and eval- uation with Gravity Recovery and

705 Climate Experiment data. J Geophys Res- Atmos., 112, D07103,

706 https://doi.org/10.1029/2006JD007522.

707 Niu, G.Y., and Coauthors, 2011: The community Noah land surface model with multi-

708 parameterization options (Noah-MP): 1. Model description and evaluation with local-

709 scale measurements. J Geophys Res- Atmos., 116, D12109,

710 https://doi.org/10.1029/2010JD015139.

711 NOAA/NCEP, 2015: NCEP North American Mesoscale (NAM) 12 km analysis. National Center

712 for Atmospheric Research, Computational and Information Systems Laboratory, accessed

713 1 January 2020, https://doi.org/10.5065/G4RC-1N91

714 Pan, H.L. and L. Mahrt, 1987: Interaction between soil hydrology and boundary-layer

715 development. Bound.-Layer Meteor., 38, 185–202.

716 Olson, J.B., and Coauthors, 2019: Improving wind energy forecasting through numerical weather

717 pre- diction model development. Bull. Amer. Meteor. Soc., 100, 2201-2220,

718 https://doi.org/10.1175/BAMS-D-18-0040.1.

719 Powers, J.G., and Coauthors, 2017: The Weather Research and Forecasting Model: Overview,

720 system efforts, and future directions. Bull. Amer. Meteor. Soc., 98, 1717-

721 1737, https://doi.org/10.1175/BAMS-D-15-00308.1.1.

34
722 Pichugina, Y.L., and Coauthors, 2020: Evaluating the WFIP2 updates to the HRRR model using

723 scanning Doppler lidar measurements in the complex terrain of the Columbia River Basin.

724 J. Renew. Sustain. Energy, 12, 043301, https://doi.org/10.1063/5.0009138.

725 Richards, L.A. 1931: Capillary conduction of liquids through porous mediums. Physics, 1, 318-

726 333. https://doi.org/10.1063/1.1745010.

727 Sobel, A.H., E.D. Maloney, G. Bellon, and D.M. Frierson, 2008: The role of surface heat fluxes

728 in tropical intraseasonal oscillations. Nature Geosci., 1, 653–657.

729 https://doi.org/10.1038/ngeo312.

730 Smirnova, T.G., J.M. Brown, and S.G. Benjamin, 1997: Performance of different soil model

731 configurations in simulating ground surface temperature and surface fluxes. Mon. Wea.

732 Rev., 125, 1870–1884. https://doi.org/10.1175/1520-

733 0493(1997)125<1870:PODSMC>2.0.CO;2.

734 Smirnova, T.G., J.M. Brown, S.G. Benjamin, and D. Kim, 2000: Parameterization of cold-season

735 processes in the MAPS land- surface scheme. J Geophys Res- Atmos., 105, 4077–4086,

736 https://doi.org/10.1029/1999JD901047

737 Smirnova, T.G., J.M. Brown, S.G. Benjamin, and J.S Kenyon, 2016: Modifications to the rapid

738 update cycle land surface model (RUC LSM) available in the weather research and

739 forecasting (WRF) model. Mon. Wea. Rev., 144, 1851–1865,

740 https://doi.org/10.1175/MWR-D-15-0198.1.

741 Smith, T.L., S.G. Benjamin, J.M. Brown, S. Weygandt, T. Smirnova, and B. Schwartz,

742 2008: Convection forecasts from the hourly updated, 3-km High Resolution Rapid

743 Refresh (HRRR) model. 24th Conf. on Severe Local Storms, Savannah, GA, Amer.

35
744 Meteor. Soc., 11.1. [Available online

745 at https://ams.confex.com/ams/24SLS/techprogram/paper_142055.htm.]

746 Sun, X., H.A. Holmes, O.O Osibanjo, Y. Sun, and C.E. Ivey, 2017: Evaluation of Surface Fluxes

747 in the WRF Model: Case Study for Farmland in Rolling Terrain. Atmosphere, 8, 197,

748 https://doi.org/10.3390/atmos8100197

749 Sun, X., H.A. Holmes, and H. Xiao, 2020: Surface turbulent fluxes during Persistent Cold Air

750 Pool Events in the Salt Lake Valley, Utah. Part II: Simulations. J. Appl. Meteor.

751 Climatol., 59, 1029-1050, https://doi.org/10.1175/JAMC-D-19-0250.1.

752 Stull, R.B. 1998: Introduction to Boundary-Layer Meteorology, Kluwer Academic

753 Publishers, Dordrecht, p. 666

754 Shaw, W.J. and Coauthors, 2019: The Second Wind Forecast Improvement Project (WFIP 2):

755 General Overview. Bull. Amer. Meteor. Soc., 100, 1687–1699,

756 https://doi.org/10.1175/BAMS-D-18-0036.1.

757 Skamarock, W.C. and J.B. Klemp, 2008: A time-split non-hydrostatic atmospheric model for

758 weather research andforecasting applications. J. Comput. Phys., 227, 3465–3485,

759 https://doi.org/10.1016/j.jcp.2007.01.037.

760 Taylor, K.E., 2001: Summarizing multiple aspects of model performance in a single diagram, J

761 Geophys Res- Atmos., 106, 7183–7192, https://doi.org/10.1029/2000JD900719.

762 Thompson, G., P.R. Field, R.M. Rasmussen, and W.D. Hall, 2008: Explicit Forecasts of Winter

763 Precipitation Using an Improved Bulk Microphysics Scheme. Part II: Implementation of a

764 New Snow Parameterization. Mon. Wea. Rev., 136, 5095–5115.

765 https://doi.org/10.1175/2008MWR2387.1.

36
766 Thompson, G., and T. Eidhammer, 2014: A study of aerosol impacts on clouds and precipitation

767 development in a large winter cyclone. J. Atmos. Sci., 71, 3636–3658.

768 https://doi.org/10.1175/JAS-D-13-0305.1.

769 Twine T.E., W.P. Kustas, J.M. Norman, D.R. Cook, P.R. Houser, T.P. Meyers, J.H. Prueger, P.J.

770 Starks, and M.L. Wesely, 2000: Correcting eddy-covariance flux underestimates over a

771 grassland, Agric. For Meteorol., 103, 279-300, https://doi.org/10.1016/S0168-

772 1923(00)00123-4.

773 Vaisala, 2015: Triton remote sensing systems: Comparing accuracy with collocated met towers.

774 Vaisala white paper, available from: https://www.vaisala.com/en/industries-

775 innovation/renewable-energy-and-weather.

776 Verseghy, D.L., 1991: CLASS-A Canadian land surface scheme for GCMs. I Soil model. Int J

777 Climatol., 11,111–133, https://doi.org/10.1002/joc.3370110202.

778 Veers, P., and Coauthors, 2019: Grand challenges in the science of wind energy. Science, 366,

779 6464. DOI: 10.1126/science.aau2027.

780 Wilczak, J.M., and Coauthors, 2019: The Second Wind Forecast Improvement Project (WFIP 2):

781 Observational Field Campaign. Bull. Amer. Meteor. Soc., 100, 1701–1723,

782 https://doi.org/10.1175/BAMS-D-18-0035.1.

783 Wharton, S., M. Simpson, J.L. Osuna, J.F. Newman, and S.C. Biraud, 2016: Role of surface energy

784 exchange for simulating wind turbine inflow: a case study in the Southern Great Plains,

785 USA. Atmosphere, 6, 21–49, https://doi.org/10.3390/atmos6010021.

786 Xia, G., M. Cervarich, S.B. Baidya, L. Zhou, J. Minder, J.M. Freedman, and P.A. Jiménez, 2017:

787 Simulating impacts of real-world wind farms on land surface temperature using WRF

37
788 model: validation with MODIS observations. Mon. Wea. Rev., 145, 4813–4836, h

789 https://doi.org/10.1175/MWR-D-16-0401.1.

790 Xia, G., L. Zhou, J.R. Minder, R.G. Fovell, and P.A. Jimenez, 2019: Simulating impacts of real-

791 world wind farms on land surface temperature using the WRF model: physical mechanisms.

792 Clim. Dyn., 53, 1723-1739, https://doi.org/10.1007/s00382-019-04725-0.

793 Xia G., C. Draxl, A. Raghavendra, and J. K. Lundquist, 2020: Validating simulated mountain wave

794 impacts on hub-height wind speed using SoDAR observations. Renew. Energy, 163, 2220–

795 2230, https://doi.org/10.1016/j.renene.2020.10.127.

796 Yang, Z.L., Y. Dai, R.E. Dickinson, and W.J. Shuttleworth, 1999: Sensitivity of ground heat flux

797 to vegetation cover fraction and leaf area index. J Geophys Res- Atmos., 104, 19,505–

798 19,514, https://doi.org/10.1029/1999JD900230.

799 Yang, Z.L., and Coauthors, 2011: The community Noah land surface model with

800 multiparameterization options (Noah–MP): 2. Evaluation over global river basins. J

801 Geophys Res- Atmos., 116, D12110, https://doi.org/10.1029/2010JD015140.

802 Zhang, Y., and Coauthors, 2020: Comparative analysis of the meteorological elements simulated

803 by different land surface process schemes in the WRF model in the Yellow River source

804 region; Theor. Appl. Climatol., 139, 145-162, https://doi.org/10.1007/s00704-019-02955-

805 0.

806 Zhuo, L., Q. Dai, D. Han, N. Chen, B. Zhao, and M. Berti, 2019: Evaluation of Remotely Sensed

807 Soil Moisture for Landslide Hazard Assessment. IEEE J. Sel. Top Appl. Earth Obs. Remote

808 Sens,, 12, 162–173, https://doi.org/10.1109/jstars.2018.2883361.

38
809

810

811

812

813

814

815

816

817 Table 1: Comparison of the Noah, NoahMP and RUC Land Surface schemes in WRF

Vegetation
Scheme Soil Levels Snow Layers Separate Canopy
Types Per Cell

Noah 1 4 1 No

NoahMP 1 4 Up to 3 Yes

RUC 1 9 2 No

818

819

820

821

822

823

824

39
825

826

827

828

829

830

831 Table 2: Observed and simulated mean midday (11 am – 2 pm) Bowen ratio from each soil case

Observations Noah NoahMP RUC

Dry Soil 145 33 23 213

Wet Soil 1.38 3.20 2.39 0.53

Frozen Soil 1.70 0.50 1.28 0.32

832

833

834

835

836

837

838

839

840

40
841

842

843

844

845

846

847 Table 3: Correlation coefficient between ground heat flux and soil temperature from both the

848 observations and model simulation

849

Observations Noah NoahMP RUC

Dry Soil 0.80 0.20 0.24 0.72

Wet Soil 0.63 0.22 0.26 0.70

Frozen Soil 0.34 0.03 0.05 0.69

850 The bold value indicates that the correlation is statistically significant at least 95 % level.

851

852

853

41
854

855

856

857

858 Figure 1: Timeseries of observed soil temperature (℃) and soil moisture (𝑚3 𝑚−3 ) at the Physics

859 Site 03 for the three cases periods. The purple line indicates the dry soil case from 13 to 19 August

860 2016. The blue line indicates the wet soil case from 4 to 10 November 2016. The orange line

861 indicates the frozen soil case from 16 to 22 February 2017.

862

863

864

865

42
866

867

868

869

870 Figure 2: Topography (meters) over the WRF Model domain. The center black dot indicates the

871 geographical location of the Physics Site 03 and Physics Site 01.

872

873

874

43
875

876

877

878

879

880

881 Figure 3: Timeseries of observed and simulated hub-height wind speed from a) the dry soil case,

882 b) the wet soil case, and c) the frozen soil. The black, blue, orange and purple lines indicate the

883 results from the observations, Noah, NoahMP and RUC respectively.

884

885

886

887

888

44
a) b)

889

• Noah
• NoahMP
• RUC

c)

890

891 Figure 4: Taylor diagram comparing simulated hub-height wind speeds (blue for Noah, orange for

892 NoahMP, and purple for RUC) with observations for a) the dry soil case, b) the wet soil case) and

893 c) the frozen soil case. The red curved line indicates the standard deviation from the observed

894 hub-height wind speed. The closer to the reference line, the lesser difference in variances.

895

896

45
897

898

899

900

901

902 Figure 5: Measured (black) and simulated (blue for Noah, orange for NoahMP, and purple for

903 RUC) (a) net radiation (NR), (b) sensible heat (SH), (c) latent heat (LH), and (d) ground heat (GH)

904 fluxes during dry soil case (Aug/2016). Positive SH and LH fluxes indicate net energy transfer to

905 the atmosphere. Positive GH fluxes indicate net energy transfer to the ground surface. Time is

906 given in Universal Coordinated Time.

907

908

909

910

911

46
912

913

914

915

916

917 Figure 6: Same as Figure 5 but for the wet soil case (Nov/2016)

918

919

920

921

922

923

924

925

926

47
927

928

929

930

931

932

933 Figure 7: Same as Figure 5 but for the frozen soil case (Feb/2017)

934

935

936

937

938

939

48
940

941 Figure 8: Scatter plots of hub-height wind speed and surface fluxes (LH + SH) from both the

942 observations and model simulations for the dry soil case period (Aug/2016): The blue dot indicates

943 the mean value of the morning hour from 0600 to 0900 local time, the orange dot the mean value

944 of the noon hours from 1000 to 1300 local time and the purple dot the mean value of the afternoon

945 hours from 1400 to 1700 local time. The horizonal and vertical lines represent the standard

946 deviation of surface flux and wind speed for those examined hours.

49
947

948 Figure 9: Scatter plots of the near-surface wind shear (40 m to 120 m) and surface fluxes (LH and

949 SH) from both the observations and model simulations for the dry soil case period (Aug/2016):

950 The blue dot indicates the mean value of the morning hour from 0600 to 0900 local time, the

951 orange dot the mean value of the noon hours from 1000 to 1300 local time and the purple dot the

952 mean value of the afternoon hours from 1400 to 1700 local time. The horizonal and vertical lines

953 represent the standard deviation of surface flux and wind speed for those examined hours.

954

50
955

956

957 Figure 10: Similar to Figure 8 but for the wet soil case period (Nov/2016)

958

959

960

961

51
962

963

964 Figure 11: Similar to Figure 9 but for the wet soil case period (Nov/2016)

965

966

967

968

969

52
970

971 Figure 12: Similar to Figure 8 but for the frozen soil case period (Feb/2017)

972

973

974

975

976

53
977

978 Figure 13: Similar to Figure 9 but for the frozen soil case period (Feb/2017)

979

980

981

982

983

984

54
985

986 Figure 14: Timeseries of observed and simulated soil temperature (℃) from a) the dry soil case,

987 b) the wet soil case, and c) the frozen soil case. The black, blue, orange and purple lines indicate

988 the results from the observations, Noah, NoahMP and RUC respectively.

989

990

991

992

993

994

995

996

997

998

55
999

1000

1001 Figure 15: Same as Figure 14 but for soil moisture (𝑚3 𝑚−3)

1002

56

View publication stats

You might also like