Planetary Boundary Layer Heights from Cruises in Spring to Autumn Chukchi-Beaufort Sea Compared with ERA5

Abstract

The planetary boundary layer height (PBLH) is a diagnostic field related to the effective heat capacity of the lower atmosphere, both stable and convective, and it constrains motion in this layer as well as impacts surface warming. Here, we used radiosonde data from five icebreaker cruises to the Chukchi and Beaufort Seas during both spring and fall to derive PBLH using the bulk Ri method, which were then compared with results from ERA5 reanalysis. The ERA5 PBLH was similar to but slightly lower than the ship observations. Clear and consistent seasonal changes were found in both the observations and the reanalysis: PBLH decreased from mid-May to mid-June and subsequently increased after August. The comparison with ERA5 shows that, besides surface temperature, biases in PBLH are also a function of wind direction, suggesting that the availability of upwind observations is also important in representing processes active in the boundary layer over the Arctic Ocean.

1. Introduction

The Arctic has experienced a rapid warming since the 1980s in a so-called Arctic Amplification [1,2,3,4,5], and the rate of surface warming is more than twice that of the entire globe [6,7,8,9]. The planetary boundary layer (PBL) plays an important role in air-surface interactions and impacts the rate of surface warming. In addition to the retreat of sea ice [10], increased water vapor [11], increased poleward energy transport [12], and lapse-rate feedback [13], the Arctic surface warming has also been attributed to the typically shallow boundary layer in the Arctic, as a shallow boundary layer acts to amplify any surface warming [14,15,16].

The PBL height (PBLH) has been recognized as an important parameter in quantifying the role of boundary layer processes in surface processes. PBLH is closely related to the effective heat capacity of the atmosphere, both stable and convective [16], and is a primary determinant of cloud type and coverage that impacts the Earth’s radiation budget [17,18]. PBLH varies under different climate forcing; therefore, it is critical to gain an understanding of the spatiotemporal variability of the PBLH, especially in the Arctic [19,20]. It remains a challenge to parameterize the physical and chemical PBL processes in the global climate models, as these models do not resolve the shallow boundary layer [21]. There is evidence that the PBLH in both numerical weather prediction and climate models is generally higher than the observations, especially in the cases of stable conditions [21,22]. Over the central Arctic Ocean, the boundary layer usually has near-neutral stability in the summer [23,24]. During the winter, the absence of solar radiation allows for the formation of a persistent stable boundary layer during cloud-free periods, while low-level clouds tend to force a shallow but relatively well-mixed boundary layer [23,25]. This unique characteristic makes it important to study the Arctic PBL and its role in Arctic Amplification.

The recent studies on Arctic PBL focus mainly on the interactions with clouds [26,27,28] and sea ice [29]. Limited studies have attempted to derive the PBLH using climatological mean state [30] and aircraft and GPS soundings observations [31]. The Arctic is a remote, sparsely populated region with very limited infrastructure and accessibility [16] to observe atmospheric processes. As a result, there are very limited data as to the structure and temporal evolution of the PBL over the Arctic Ocean. The Surface Heat Budget of the Arctic Ocean experiment (SHEBA), which was conducted with a drifting icebreaker over multiyear ice in the Beaufort Sea 1997–1998 [23], has contributed to knowledge on the surface processes in the Arctic, including: examining different regimes of the stable boundary layer [32] and exploring a PBLH calculation method [33], as well as characterizing the diurnal cycle of PBLH [34]. Some other data, such as “North Pole” drifting ice stations [35], the Arctic Ocean Climate System Research observed on R/V Mirai [36,37], and the ASCOS measurement campaign observed on the Swedish icebreaker Oden in the summer of 2008 [25,38], have been helpful in studying boundary layer processes [39]. However, the lack of observations over space and time has limited our ability to understand the processes that determine the height of the planetary boundary layer. Here, we help address this gap by using the recently released Chukchi–Beaufort icebreaker cruises’ radiosonde data to study Arctic PBLH, and we also use a new reanalysis dataset, ERA5, to examine the processes in the Arctic PBL.

2. Materials and Methods

In this paper, radiosonde data from five icebreaker cruises to the Chukchi and Beaufort Seas were used. We named these cruises by the research vessel name and corresponding year of observation. Data are available from both the late spring (May–June) as well as the early fall (August–October). In total, 373 individual radiosonde ascents were used in this work (39 ascents in Louis 2013, from 6 August to 31 August; 76 ascents in Healy 2014, from 17 May to 20 June; 183 ascents in Mirai 2014, from 5 September to 28 September; 38 ascents in Louis 2014, from 26 September to 15 October; and 37 ascents in Louis 2015, from 24 September to 12 October, see Tables S1–S5 for details). The locations and mean sea ice concentrations of the observations are shown in Figure 1.

We used the radiosonde data to diagnose the “observed” PBLH. Seidel [22] tested several methods and found out that the bulk Ri method is the most suitable for diagnosing PBLH, as it is suitable for both stable and convective boundary layers and is not strongly dependent on vertical resolution. The “observed” PBLH was found by searching upwards from the lowest observation, with the PBLH defined as that level where the Ri equals the critical value of 0.25. The Ri at level k is calculated by:

$$Ri=z_k\frac{2g\left(s_{vk}-s_{vs}\right)}{[{\left(U_k-U_s\right)}^2+{\left(V_k-V_s\right)}^2]\left(s_{vk}+s_{vs}-g{z}_k-g{z}_s\right)}$$

(1)

$$where s_{vk}=c_p{T}_k\left(1+\epsilon q_k\right)+g{z}_k, s_{vs}=c_p{T}_s\left(1+\epsilon q_s\right)+g{z}_s, and \epsilon =\frac{R_{vap}}{R_{dry}}-1,$$

where z is the height, $g$ is the acceleration of gravity, S_v is the virtual dry static energy, U and V are zonal and meridional wind components, c_p is the specific heat at constant pressure of moist air, T is temperature, ɛ is parcel entrainment (R_vap and R_dry are the gas constant for water vapor and for dry air, respectively), q is specific humidity, and the subscript k and s represent the level and the surface (lowest level in observation), respectively. Here, we set $g$ as 9.8 m s⁻², U_s and V_s as zero, c_p as 1004.7 J kg⁻¹ K⁻¹, ɛ as 0.61. If the Ri at level k is larger than 0.25 and the Ri at level (k − 1) is smaller than 0.25, the PBLH will be linearly interpolated by Ri from the heights at level k and (k − 1).

ERA5 is the newest global reanalysis produced by ECMWF [39]. ERA5 has hourly output throughout, 31 km horizontal resolution, and 137 vertical levels from the surface up to a height of 80 km. There are already some works showing that ERA5 performs better than some other reanalysis in the Arctic [40,41]. Here, ERA5 hourly PBLH, 2-meter temperature, 100-meter U and V, and sea surface pressure are used in this paper.

Ri for radiosondes is calculated using the same method as in ERA5 (ECMWF. IFS CY41R2 Part IV, content 3.10.1, Formula 3.90, https://www.ecmwf.int/node/16648, accessed on 19 October 2021). To assist in the calculation, the ERA5 data were linearly interpolated to the locations of the radiosonde data. To compare the wind components, U and V in radiosonde data were linearly interpolated to a 100 m height.

3. Results

The observed PBLH of each cruise, as well as the lowest temperature, are shown in Figure 2, and the mean values and other parameters of observed and ERA5 PBLH are listed in

Table 1. Mean PBLH (m) from ERA5 and each cruise.

Period	ERA5 Mean (STD), m (m)	Observation Mean (STD), m (m)	Correlation Coefficient	RMSE, m	Bias Error, m
Total	485 (226)	488 (254)	0.63	207	−3
Louis 2013	248 (160)	316 (248)	0.36	251	−68
Louis 2014	353 (191)	562 (313)	0.49	345	−209
Louis 2015	499 (226)	501 (195)	0.73	157	−2
Mirai 2014	577 (197)	512 (224)	0.82	143	65
Healy 2014	446 (205)	476 (277)	0.55	238	−30
Healy 2014 (cold)	520 (237)	569(322)	0.50	292	−49
Healy 2014 (warm)	368 (123)	378 (171)	0.43	162	−10

. As shown in Table 1, the mean of observed PBLH is 488 m, and the standard deviation (STD) is 254 m. The mean of ERA5 PBLH after interpolation is 485 m, and the STD is 226 m. However, the root-mean-square error (RMSE) between ERA5 and observed PBLH is 201 m averaged over all the cruises and varied from 143 m for cruise Mirai 2014 to 345 m for cruise Louis 2014 (Table 1), indicating some large differences at individual locations.

The seasonal variations in PBLH are evident for the observational results (Figure 2a) and for the ERA5 results (Figure 3a). In the period of the observations, PBLH decreased from mid-May (~570 m) to mid-June (~280 m) and increased after August (~150 m) to October (~570 m), which is opposite to the variation of air temperature at the lowest levels (defined as surface air temperature in this study, and for ERA5, 2-meter temperatures were used) (Figure 2b and Figure 3b). This seasonal variation of PBLH is roughly consistent with the previous study using the aircraft and GPS soundings in SHEBA [31]. The storm events were found to be more numerous [42] and more intense [43] in winter than in summer over Chukchi–Beaufort Sea, which might contribute to the observed and modeled seasonal variability in PBLH.

To elucidate the impact of surface air temperatures on the PBLH, we consider the case of the Healy 2014 cruise, during which there was a marked transition from cold (~−4 °C) to warm (~0.4 °C) conditions (Figure 4). The time series of ERA5 sea ice concentration are shown in Figure 3c and are consistent with the temperature transition. This transition occurred on 2 June 2014, and we use this date to divide the cruise period into a cold and warm period. For the entire Healy period, the average observed PLBH is 475.7 m, and the RMSE between ERA5 and observed PBLH is 237.8 m. The mean PBLH for the cold period (568.8 ± 321.7 m) is much larger than that for the warm period (377.7 ± 170.8 m). The ERA5 PBLH are slightly lower than the observations but also show this clear shift (519.8 ± 321.7 m for the cold period and 367.5 ± 122.6 m for the warm period). The bias error of ERA5 in the cold period is −49.0 m and in the warm period is −10.1 m, and the RMSEs between ERA5 and observations are 292.2 m and 161.8 m, respectively. With comparison among these results in Table 1, except Mirai 2014, it can be concluded that the higher PBLH are, the higher variances and the higher simulated errors will be.

To understand the synoptic conditions that gave rise to this transition, we consider the surface flow as represented in ERA5 over the region of interest during the cold and warm periods (Figure 5). During the cold period, corresponding to the high PBLH and the large bias between the observations and ERA5, the Healy was situated to the east of a region of high pressure, with northerly winds being dominant. In contrast, during the warm period, corresponding to low PBLH and the small bias, the Healy was situated to the west of a region of high pressure, with southerly winds being dominant.

The Healy observations suggest that there may be a relationship between the bias error in PBLH and the direction of the meridional wind, as well as surface temperature (Figure 5). To test this hypothesis, we used the entire database and stratified the results by wind component and temperature (Figure 6). We used ERA5 100-meter winds here in order to be consistent with the following comparisons, because the heights of lowest level in different cruises are different and higher than 10 m. We picked the calendar time of composite analysis by the PBLH differences between ERA5 and observations, whether positive and greater than 1 STD or negative and greater than −1 STD. The results were subtracted out from the long mean for the period of interest (1979 to 2018) to avoid the seasonal and diurnal differences.

Consistent with the hypothesis noted above, the large biases mainly occurred when the northerly winds were dominant, independent of whether the bias was positive (Figure 6a) or negative (Figure 6b). ERA5 more likely holds large positive bias of PBLH at the east of the high-pressure anomalies when the high-pressure anomalies are over the Chukchi Sea. Additionally, ERA5 more likely holds large negative biases of PBLH at the southwest of the low-pressure anomalies when the low-pressure anomalies are over the Beaufort Sea.

For further verification of the relationship between the PBLH biases and the northerly winds, we show the dotted points with plus symbols by each variation, and the PBLH differences between ERA5 and observations as coordinates (Figure 7). We divided the 373 observations into six boxes with two black ±1 STD lines and one zero line (mean value line for PBLH). We circled large biases of each variation to avoid ERA5 biases in other variations influencing our estimate. According to Figure 7a,b, ERA5 holds huge negative biases when the observed PBLH are large, and huge positive biases when the observed PBLH are small. It is worth noting that the air temperature biases could cause PBLH biases. When ERA5 simulates air temperature lower than the observations, the PBLH in ERA5 is generally lower than the observed PBLH, and vice versa (Figure 7c,d). The zonal winds have no obvious and consistent influence according to Figure 7e,f. The number counts in Figure 7g,h show the preponderant and consistent results that more cases of large biases occur under the influences of northerly winds whether the bias is positive or negative, although the huge biases occur in both northerly and southerly winds.

4. Discussion and Conclusions

We analyzed the PBLH of five cruises observed in 2013~2015 from late spring to autumn in the Chukchi–Beaufort Sea and compared them with ERA5 results. It turns out that the seasonal variations observed in ERA5 PBLH are evident and parallel. The mean and STD of PBLH derived from observation and from ERA5 both decrease from late spring to the summer and then increase back when it comes to autumn. Esau [30] mentioned some aspects of the seasonal cycle of PBLH. They mainly focus on the climatology of Arctic PBL and they divided the whole Arctic into central Arctic, continental zone, and marine zone but did not separate the ice edge zone individually. For the “central Arctic” region, which included the Chukchi–Beaufort Sea in their work, the PBLH is greatest during summer months. It is the same with the seasonal cycle in continental regions, maybe because the domination of sea ice in central Arctic region makes it more like the continental zone. For the other maritime regions, the PBLH is greatest during winter and spring months when the air–sea temperature difference is greatest. Compared with their work, our results indicate that the seasonal change of PBLH at the ice edge zone is more like the marine zone instead of the continental zone.

Consistent with previous studies [19,44,45], surface temperature might have a significant influence on the magnitude and the variance of PBLH. Surface temperature might also have a significant effect on ERA5 PBLH simulation performance. In our work, when ERA5 simulates air temperature lower than observations, the PBLH in ERA5 is generally lower than the observed PBLH, and vice versa. The Arctic has a limited number of in situ observations. It is also a region where it remains a challenge to assimilate satellite observations into numerical weather prediction models [46]. It follows that the observed bias, when stratified by wind direction, may also be attributed to the fact southerly flow advects information from land-based stations into the region, thereby improving the representation of the PBLH in ERA5 [47,48]. This characteristic may contribute to our finding that large biases of the ERA5 PBLH are more common when there are northerly winds.