1. Introduction
The vertical structure of the Arctic atmosphere, together with its interactions with the clouds and surface, plays an important role in the atmospheric radiative transfer and surface energy exchange processes and, therefore, Arctic sea ice change. Both temperature inversions (TIs; e.g., Kahl 1990; Serreze et al. 1992; Curry et al. 1996) and specific humidity inversions (SHIs; e.g., Devasthale et al. 2011; Vihma et al. 2011; Nygård et al. 2014) are pervasive and dominant features of the Arctic troposphere. Additionally, SHIs often coincide with TIs (Serreze et al. 1995; Vihma et al. 2011; Tjernström et al. 2012; Sotiropoulou et al. 2016), and their interactions can be essential for cloud growth and persistence in the Arctic (Solomon et al. 2011, 2014; Nygård et al. 2014; Sedlar et al. 2012; Sedlar 2014). Moreover, the predominant Arctic wintertime TI tends to suppress infrared cooling and enhance Arctic amplification (Bintanja et al. 2011). Hence, better understanding on the characteristics of the Arctic TIs and SHIs, as well as their relationships, is needed to improve the Arctic climate prediction.
Radiosonde (RS) and tethersonde soundings are useful observation techniques for atmospheric profile measurements. RS observations can typically reach altitudes of approximately 30 km, and they have been traditionally used to monitor the Arctic TIs and/or SHIs (e.g., Kahl 1990; Tjernström and Graversen 2009). However, RS measurements suffer from sparse distribution of meteorological stations in the Arctic, as they are limited to point measurements mainly at coastal and interior stations. Tethersonde soundings are also one of the powerful tools for studying the vertical structure of the Arctic atmosphere (e.g., Beine et al. 2001; Argentini et al. 2003; Vihma et al. 2011, Kilpeläinen et al. 2012), they are fairly sparse and limited to the rare and short campaigns over the Arctic. In addition, the tethersonde system can only measure in the lowest few hundred meters of the atmosphere, which therefore may be insufficient to detect the upper-level inversions.
Spaceborne monitoring is an effective way to estimate the inversion characteristics over the entire Arctic region. The Moderate Resolution Imaging Spectroradiometer (MODIS; Liu and Key 2003) and the High Resolution Infrared Radiation Sounder (HIRS; Liu et al. 2006) are used to estimate the Arctic low-level TIs under clear-sky conditions in the cold season. The Atmospheric Infrared Sounder (AIRS) can monitor both TIs and SHIs (Devasthale et al. 2010, 2011) under clear-sky conditions, as well as under heavy (e.g., 50%) cloud cover (Devasthale et al. 2016). Besides, the microwave-based global positioning system (GPS) radio occultation (RO) observations are also reliable to detect the Arctic TIs under all weather conditions (Ganeshan and Wu 2015; Chang et al. 2017; Yu et al. 2018).
Arctic TI strengths has also been investigated based on the state-of-the-art coupled atmospheric–oceanic general circulation models (GCMs) over the entire Arctic. The strength of TI is defined in GCMs as the difference of atmospheric temperature between the 850 and 1000 hPa (Boé et al. 2009; Medeiros et al. 2011). However, the GCMs tend to overestimate the Arctic TI strength (Pavelsky et al. 2011; Pithan et al. 2014). Moreover, information on occurrence, base height, and depth of the TI as well as the SHI characteristics over the Arctic has not been studied with the climate models.
Atmospheric reanalyses are also widely applied in the Arctic to study the TIs and SHIs (e.g., Tjernström and Graversen 2009; Lüpkes et al. 2010; Wesslén et al. 2014; Brunke et al. 2015; Yu et al. 2019) since they are arguably the best representation of four-dimensional structure of the Arctic atmosphere (Screen and Simmonds 2011). However, large errors still exist in tropospheric temperature and specific humidity profiles in the reanalyses over the Arctic (e.g., Screen and Simmonds 2011; Jakobson et al. 2012), which results in the frequent absence of the low-level inversions (Serreze et al. 2012), as well as the significant biases in the inversions (Tjernström and Graversen 2009; Jakobson et al. 2012). Although the reanalyses are not necessarily realistic for all inversion characteristics over the Arctic, the main features of the temporal (both seasonal and annual) and spatial variation of the inversions are less affected and well represented in the reanalyses (Tjernström and Graversen 2009; Naakka et al. 2018). Among the previous generation of reanalyses from different agencies, such as the National Centers for Environmental Prediction (NCEP) Climate Forecast System Reanalysis (CFSR; Saha et al. 2010), the European Centre for Medium-Range Weather Forecasts (ECMWF) interim reanalysis (ERA-I; Dee et al. 2011), the Japan Meteorological Agency (JMA) Climate Data Assimilation System (JCDAS; Onogi et al. 2007), and the National Aeronautics and Space Administration (NASA) Modern-Era Retrospective Analysis for Research and Applications (MERRA; Cullather and Bosilovich 2011), the ERA-I outperforms other reanalyses in the Arctic (Jakobson et al. 2012; Lindsay et al. 2014). To further improve the quality of the reanalyses, extensive work has also recently been carried out in producing new reanalyses, such as the Japanese 55-year Reanalysis (JRA-55; Kobayashi et al. 2015) and the ERA5 (Hersbach and Dee 2016). Nevertheless, performance of the Arctic TIs and SHIs in the new reanalyses needs to be further investigated.
In this paper, occurrence and characteristics of TI and SHI (e.g., base height, depth, and strength) from ERA-I, ERA5, and JRA-55 are assessed against in situ RS measurements over the Arctic Ocean from the Arctic Summer Cloud Ocean Study (ASCOS) in 2008 (Tjernström et al. 2012, 2014) and the Norwegian young sea ICE in 2015 (N-ICE2015) (Granskog et al. 2016; Cohen et al. 2017; Kayser et al. 2017) expeditions. Moreover, the relationships between TIs and SHIs in the reanalyses are also investigated and evaluated. In section 2, the data used are described and the method on extracting the TI and SHI characteristics is introduced. In section 3, the uncertainties in temperature and specific humidity profiles in the reanalyses are analyzed, the characteristics of the Arctic TIs and SHIs are estimated and evaluated, and the relationships among TI (and SHI) characteristics in the reanalyses are validated. Section 4 is devoted to assessing the characteristics of the strongest TI and SHI in a profile and their dependencies on surface meteorological conditions in the reanalyses. The relationship between the simultaneous TIs and SHIs in the reanalyses are presented and assessed in section 5. Conclusions are drawn in section 6.
2. Data and method
a. Datasets
ERA-I is the last generation of global atmospheric reanalyses from ECMWF (Dee et al. 2011), which is included here because of its advantages over other older reanalyses in the Arctic (Jakobson et al. 2012; Lindsay et al. 2014). Besides, the more recent reanalyses from JRA-55 (Kobayashi et al. 2015) and ERA5 (Copernicus Climate Change Service 2017) are also incorporated to evaluate their accuracy in representing the vertical structure of temperature and moisture over the Arctic Ocean. All reanalyses use a four-dimensional variational data assimilation (4D-Var) system, and the analysis fields with 6-h intervals are used in this study. Variables from all reanalyses are on a regular longitude–latitude grid, with a model horizontal resolution of about 0.7° × 0.7° in ERA-I, 0.28° × 0.28° in ERA5, and 0.56° × 0.56° in JRA-55. Similar to Naakka et al. (2018), the model-level fields, rather than the pressure-level fields, from all reanalyses are used to obtain better vertical resolution in the lower troposphere. The number of model levels is typically about 22, 42, and 22 from surface to 500 hPa in ERA-I, ERA5, and JRA-55, respectively. Specifically, there are 9, 18, and 10 levels in the lowest 100 hPa in ERA-I, ERA5, and JRA-55, and the vertical resolution decreases in the reanalyses vary within 3–19, 2.5–9.3, and 1.5–19 hPa, respectively. Note also that extra assistance from surface parameters is needed to derive the height and pressure of each model level in ERA-I and ERA5 and estimate the pressure at each model level in JRA-55 (Simmons and Burridge 1981).
Moreover, in situ RS observations from two Arctic field campaigns, the 2008 ASCOS (north of 87°N; Tjernström et al. 2012, 2014) and the N-ICE2015 (north of 83°N; Granskog et al. 2016; Cohen et al. 2017; Kayser et al. 2017), are used to examine the vertical structure of the Arctic atmosphere in the reanalyses. The ASCOS took place from 2 August to 8 September 2008 on the Swedish icebreaker Oden, and the N-ICE2015 carried out from January to June 2015 on the Norwegian Polar Institute’s Research Vessel Lance. The RS measurements were recorded every 6 and 12 h during ASCOS and N-ICE2015, respectively. Given that the RS data from ASCOS and N-ICE2015 were collected in different seasons, we investigate the performance of the reanalyses in summer (June–August), winter (December–February), and spring (March–May) with the ASCOS data in August 2008 and the N-ICE2015 data in January–February 2015 and March–May 2015 (see Fig. 1). As a result, a total of 116, 82, and 131 RS profile data were collected in summer 2008, winter 2015, and spring 2015, respectively. Note that, although the RS data were also collected in June during the N-ICE2015, assessment of the reanalyses in summer 2015 is not included in this study because of the short time span (less than one month) of the RS data. Considering that the launch locations of RS were changed along with the ship track during the expeditions, two criteria are applied in this study to select the spatiotemporally synchronized RS and reanalysis data pairs. First, the selected reanalysis grid should be the nearest to the launch location of RS. Second, the time differences between RS and reanalysis data should be no more than 1.5 h. It is also worthwhile mentioning that the RS data from the ASCOS were not assimilated into the reanalyses (Wesslén et al. 2014), while those from the N-ICE2015 were contribute to the production of all reanalyses analyzed in this study (Graham et al. 2019). Although the RS measurements cannot be considered as independent reference data for the intercomparison in winter 2015 and spring 2015, they have consistent effects on all reanalyses and can therefore still be valuable to assess the performance of the reanalyses with different assimilation methods in those seasons.
Locations of RS data from the ASCOS field campaign in summer (magenta circles), as well as from the N-ICE2015 expedition in winter (green circles) and spring (blue circles) over the Arctic Ocean.
Citation: Journal of Applied Meteorology and Climatology 60, 4; 10.1175/JAMC-D-20-0079.1
b. Algorithm to extract the TI and SHI characteristics
In this study, TIs and SHIs are identified from the temperature and specific humidity profiles below 500 hPa. For each TI (or SHI) layer from RS and reanalysis data, an inversion base height, depth, and strength are all determined. The inversion base height is defined to be the level where the temperature (or specific humidity) starts to increase with altitude. Additionally, an inversion is defined to be surface based if the base height is located below 50 m altitude. The inversion depth is defined to be the difference between the base height and the level at which the temperature (or specific humidity) starts to decrease with altitude. The inversion strength is defined as the difference in temperature (or specific humidity) between the inversion top and base.
To reduce the errors in inversions detection and estimation, we exclude meteorological profiles from the surface to 500 hPa in which the temperature is below −40°C and the specific humidity is below 0.2 g kg−1 (Nygård et al. 2013). Nygård et al. (2013) also screens the TIs and SHIs with an inversion depth less than 10 m. However, considering the vertical resolution of RS profile data during the ASCOS and the N-ICE2015 is about 5 and 25 m, respectively, we extend the threshold of inversion depth to 25 m in this paper. In addition, the two inversion layers separated by thin (<100 m) noninversion layers are considered as the same inversion (Zhang et al. 2011). The minimum strength of TIs is also set to 0.3°C following Kilpeläinen et al. (2012).
3. Estimation and evaluation of TIs and SHIs in the reanalyses over the Arctic Ocean
a. Uncertainties in temperature and specific humidity profiles in the reanalyses
Figure 2 shows the comparison of temperature and specific humidity from the reanalyses against RS at 1000–500 hPa over the Arctic Ocean. The temperatures from the reanalyses generally agree well with the RS data in each season (Figs. 2a–c). However, while the near-surface atmospheric stable stratification seen in RS over the Arctic Ocean is clearly observed in the reanalyses in winter (Fig. 2b) and spring (Fig. 2c), it disappears in summer (Fig. 2a). This phenomenon may be due to the factor that the RS observations are assimilated in the reanalyses in winter and spring, which improves the presentations of temperature and humidity profiles in the reanalyses. Additionally, the near-surface atmospheric stable stratification in the reanalyses exhibit lightly different patterns to that in RS data in winter and spring, indicating uncertainties in temperature remain after data assimilation in the reanalyses.
Mean (a)–(c) temperature and (j)–(l) specific humidity from ERA-I (magenta lines), ERA5 (green lines), JRA-55 (blue lines), and RS (black lines), as well as (mean bias of d)–(f) temperature and (m)–(o) specific humidity and RMSE of (g)–(i) temperature and (p)–(r) specific humidity from the ERA-I, ERA5, and JRA-55 reanalyses against RS data at 1000–500 hPa during (top) summer of the ASCOS and (middle) winter and (bottom) spring of the N-ICE2015.
Citation: Journal of Applied Meteorology and Climatology 60, 4; 10.1175/JAMC-D-20-0079.1
ERA-I typically has the smallest bias in temperature among these reanalyses. However, ERA-I generally suffers from small cold biases at 1000–500 hPa in all analyzed seasons, except for the layers below 975 hPa and above 700 hPa in summer (Fig. 2d) and below 975 hPa in spring (Fig. 2f). Temperature bias in ERA5 presents similar pattern to that in ERA-I, except that its warm bias below 925 hPa and above 600 hPa is smaller than ERA-I, and its cold bias at 700–950 hPa is notably larger than ERA-Interim in summer (Fig. 2d). Additionally, the temperature biases from ERA5 and ERA-I are opposite below 950 hPa in winter (Fig. 2e). JRA-55 has a larger cold bias below 800 hPa in summer and below 900 hPa in winter than ERA-Interim and ERA5. ERA-I has smaller temperature root-mean-square error (RMSE) than ERA5 at 850–950 hPa in summer (Fig. 2g) as well as at 500–1000 hPa in spring (Fig. 2i). Therefore, there is no evidence that ERA5 is better than ERA-I to simulate low tropospheric temperatures over the Arctic Ocean. Among the reanalyses, JRA-55 has the largest temperature RMSE at 1000–500 hPa. Specifically, the wintertime RMSE in JRA-55 could be caused by cold bias below 900 hPa.
Reanalyses also generally capture the shape of mean specific humidity profile seen in RS data in each season (Figs. 2j–l), except for the obvious specific humidity differences in summer at the lowest layers (Fig. 2j). Additionally, all reanalyses show that the mean bias of specific humidity is typically largest in summer (Fig. 2m), followed by and spring (Fig. 2o) and winter (Fig. 2n). Note also that the relative error of mean specific humidity in the reanalyses is not remarkably larger in summer than in winter and spring (not shown), revealing that exclusion of RS data in the reanalyses in summer is not responsible for the largest mean bias of specific humidity. One possible explanation may be that warmer air in summer can “hold” more water vapor, and the biases of specific humidity are larger under warmer air. Moreover, the specific humidity bias is in general larger at bottom levels than upper levels in all reanalyses. In summer, the specific humidity in ERA-I generally agrees best with RS data at 950–750 hPa, where the mean bias in specific humidity in JRA-55 is the most remarkable (Fig. 2m). Moreover, the mean bias of specific humidity below 975 hPa is smaller in JRA-55 and ERA5 than in ERA-I in summer. In winter, both ERA-I and JRA-55 are in general drier than RS at 1000–500 hPa, while ERA5 is mostly moister than RS above 850 hPa (Fig. 2n). The largest dry bias in JRA-55 in winter is found below 825 hPa. In spring, significant specific humidity bias differences among the reanalyses are found at 975–700 hPa, and JRA-55 typically has the largest dry bias at 975–750 hPa (Fig. 2o). Considering RMSE of the specific humidity (right column in Fig. 2), ERA5 generally performs best in each season, especially at bottom levels. ERA-I and ERA5 are approximately equally good in summer, but for ERA-I the RMSE is larger than 0.6 g kg−1 at surface. JRA-55 has the largest RMSE at 1000–500 hPa in all seasons except for the layers below 975 hPa in summer. It is also worthwhile mentioning an interesting feature that the warm biases in temperature and specific humidity are decreasing, together with those cold biases are increasing in all reanalyses from 1000 to 975 hPa, which could affect the occurrence and the strength of surface-based inversions in the reanalyses.
b. Relative frequency distributions of TI characteristics
In Fig. 3, the relative frequency distributions (RFDs) of the number and characteristics (base height, depth, and strength) of TIs from RS and reanalyses are shown. Profiles with multiple TI layers are very common in both RS and reanalyses in all seasons except for the JRA-55 in winter (left column in Fig. 3 and Table 1). However, while a RS profile at 1000–500 hPa contains up to 5–6 TI layers, a reanalysis profile typically has at most 2–5 TI layers in all seasons. Additionally, the occurrence of profiles with multiple TIs seen in RS data is greatly shrunk in all reanalyses in each season, indicating that the reanalyses have difficulties presenting the vertical temperature fluctuations over the Arctic Ocean. All reanalyses generally perform better in simulating the TI number in summer and spring than winter. ERA-I and ERA5 in general have similar distributions of TI number, and their distributions are closer to the results from RS than JRA-55. Moreover, TI is detected in nearly every RS profile data (above 98%) in each season, the TI occurrence in the reanalyses is typically lower than 95% with significant variations (Table 1). On average, ERA5 has the closest TI occurrence to RS, followed by ERA-I and JRA-55.
RFDs of the (left) number, (left center) base height, (right center) depth, and (right) strength of TIs from RS (filled gold circles), ERA-I (red plus signs), ERA5 (green crosses), JRA-55 (open blue circles), RS data at ERA-I levels (purple plus signs), RS data at ERA5 levels (black crosses), and RS data at JRA-55 levels (open brown circles) in (a)–(d) summer, (e)–(h) winter, and (i)–(l) spring. Bin sizes used in the RFDs are 1 for TI number, 100 m for TI base, 200 m for TI depth, and 2°C for TI strength. Note that the RFDs are estimated on the basis of all TI layers observed in a temperature profile.
Citation: Journal of Applied Meteorology and Climatology 60, 4; 10.1175/JAMC-D-20-0079.1
Occurrences of TI, profiles with multiple TIs, SHI, and profiles with multiple SHIs in RS data, RS data at ERA-I levels, RS data at ERA5 levels, RS data at JRA-55 levels and output data from ERA-I, ERA5, and JRA-55 in summer, winter, and spring.
Most TIs are elevated (i.e., base height > 50 m) in both RS and reanalyses during summer (Fig. 3b). Nonetheless, the elevated TIs with base heights less than 2000 m are poorly simulated in all reanalyses, which may be linked to the absence of the atmospheric stable stratification from mean temperature profiles (Fig. 2a), as well as the remarkable temperature uncertainties below 800 hPa (Figs. 2d,g). In winter, the low-level TIs are more frequent in both RS and reanalyses (Fig. 3f) than in summer. Note that the lower RFD of TI base in RS data than the reanalyses is due to the higher occurrence of profiles with multiple TIs in the former, as well as the number of low-level TIs is generally higher in RS than in reanalyses (not shown). Spring is transition period between winter and summer types of TIs (Fig. 3j), with the surface-based and elevated TIs occurring 19% and 81% of 131 soundings, respectively. Generally, the reanalyses display the RFDs of TI base better in winter than spring and summer. This may be associated with better simulation of surface radiative cooling in the reanalyses in winter, which helps to produce surface-based TIs and therefore capture the correct RFDs of TI base. Among the reanalyses, the TI bases seen in RS data are closest presented by ERA5, because it typically has better ability to capture the upper-level TIs than ERA-I and JRA-55.
The frequency peaks of the TI depth are observed less than 400 m in both RS and reanalyses, except for the highest occurrence of TI depth of 1000 m in JRA-55 in winter (Fig. 3g). All reanalyses typically underestimate the frequency of shallow TIs (e.g., a depth of less than 200 m) in all seasons (third column in Fig. 3) due to their uncertainties and coarse vertical resolution, suggesting they fail to capture the small-scale vertical variability in the temperature in Arctic boundary layer. Additionally, ERA-I and JRA-55 appear to overestimate the spread of TI depth to 1400 m in summer (Fig. 3c) and spring (Fig. 3k), and present remarkable deviations in RFDs of TI depth of about 1000 m in winter (Fig. 3g). Among the reanalyses, ERA5 perform best for the TI depth in all seasons.
Similar to the TI depth, the distributions of TI strength in RS and reanalyses also generally display exponential shapes in all seasons (right column in Fig. 3). However, the frequency of TIs with the strength less than 2°C tend to be remarkably underestimated in the reanalyses in all seasons. It is noteworthy that despite the highest occurrence of low-level TIs in winter (Fig. 3f), weak and shallow TIs also occurs frequently (Figs. 3g,h). This is consistent with the analyses of Tjernström and Graversen (2009) and Kayser et al. (2017), who found that the strong surface-based TIs are only part of the feature in Arctic boundary layer in winter. Among the reanalyses, the RFDs of TI strength typically suffer from the most remarkable deviations in JRA-55, followed by ERA-I and ERA5. Therefore, the ERA5 may be a better choice to depict the RFDs of TI properties over the Arctic Ocean.
As analyzed above, both the uncertainties and the coarse vertical resolution in the reanalyses are responsible for their inaccuracy in simulating the inversion occurrence and characteristics over the Arctic Ocean. To divide the possible effects of the uncertainties and the coarse vertical resolution in the reanalyses on simulating the TIs and SHIs, the inversion occurrence and the RFDs of TI (and SHI) number and characteristics from the RS data at ERA-I, ERA5, and JRA-55 levels in each season are also estimated (see Table 1, Figs. 3 and 7). It is clear in Fig. 3, while profiles with multiple TI layers from RS data at reanalysis levels are generally observed in each season, a coarser RS profile typically detects less TI layers (i.e., at most 4). In addition, occurrences of both TI and profiles with multiple TIs from RS data at reanalysis levels are lower than those from original RS data in all seasons (Table 1). Comparisons of TI number and occurrences among the RS data at ERA-I, ERA5, and JRA-55 levels against the original RS data show that the temperature profiles at least reduced vertical resolution (i.e., at ERA5 levels) generally detect the most accurate results (Fig. 3). Moreover, RS data at ERA5 levels also show better performance than those at ERA-I and JRA-55 levels on presenting the RFDs of TI base, depth, and strength in each season. In general, temperature profiles at higher vertical resolution show better ability to capture the higher, shallower, and weaker TIs.
Comparison of TI and profiles with multiple TI occurrences between reanalyses and RS at reanalysis levels shows that all reanalyses generally have lower occurrences in each season, except for JRA-55 in summer (Table 1). While differences in RFDs of TI base between reanalyses and RS at reanalysis levels are generally greater in summer than in spring and winter, those differences in RFDs of TI depth and strength are generally greater in winter than spring and summer (Fig. 3).
Among the reanalyses, ERA5 performs better than ERA-I and JRA-55 at simulating the RFDs of TI characteristics, which may be due to the higher vertical resolution as well as the accurate representation of temperature profiles in ERA5.
c. Relationships among the TI characteristics in different seasons
To understand the performance of the reanalyses on depicting the relationships among the Arctic TI characteristics, the joint histograms of TI base and depth (Fig. 4), TI base and strength (Fig. 5), and TI depth and strength (Fig. 6) from RS and reanalyses in summer, winter and spring over the Arctic Ocean are illustrated. It is clear from the RS data, summer is dominated by elevated TIs at base height up to 2500 m, typically 100–200 m deep and less than 10°C strong (Figs. 3b–d). Additionally, the quasi-linear TI depth–strength relationship and the distribution of joint histograms shown in Fig. 6a suggests that the depth and strength of TIs are most related, as well as the shallow and weak elevated TIs occur most frequently in summer. The elevated TIs in summer are formed by the advection of warm and moist air over the cold surface (Naakka et al. 2018). All three reanalyses tend to simulate the elevated TIs with deeper depth, and they typically have difficulties capturing the TIs below 400 m and above 1000 m in summer (Figs. 3b–d). This suggests that the reanalyses typically have deeper well-mixed boundary layer (up to 400 m) than RS data in summer (see also Fig. 2a). Particularly, the elevated TIs below 400 m in summer are one of the most important features for the Arctic atmosphere near the surface, which are typically indicative of the prevalence of low-level clouds (Tjernström and Graversen 2009; Cohen et al. 2017). This is consistent with the results that the near-surface Arctic cloud properties in the reanalyses are not well represented in summer (de Boer et al. 2014). The TI depth–strength relationship in the reanalyses (Figs. 6b–d) generally present similar pattern to that from RS data in summer (Fig. 6a). However, the weak TIs with deeper depth are observed more frequent in the reanalyses, which may be related to coarser vertical resolution in the reanalyses. Among the reanalyses, ERA5 performs better than ERA-I in displaying the elevated TIs above 1000 m in summer, whereas the former fails to present the TIs below 200 m. JRA-55 shows the lowest skill simulating elevated TIs at upper levels (e.g., base > 1500 m) and tend to overestimate the depth of elevated TIs most in summer.
The joint histograms of TI base and depth from (left) RS, (left center) ERA-I, (right center) ERA5, and (right) JRA-55 in (a)–(d) summer, (e)–(h) winter, and (i)–(l) spring over the Arctic Ocean. Bin sizes of 100 m for TI base and 200 m for TI depth are used, and they are normalized by the number of observations in the bin with the most observations.
Citation: Journal of Applied Meteorology and Climatology 60, 4; 10.1175/JAMC-D-20-0079.1
As in Fig. 4, but for the joint histograms of TI base and strength and bin sizes used in the histograms are 100 m for TI base and 2°C for TI strength.
Citation: Journal of Applied Meteorology and Climatology 60, 4; 10.1175/JAMC-D-20-0079.1
As in Fig. 4, but for the joint histograms of TI depth and strength and bin sizes used in the histograms are 200 m for TI depth and 2°C for TI strength.
Citation: Journal of Applied Meteorology and Climatology 60, 4; 10.1175/JAMC-D-20-0079.1
During wintertime, both surface-based and elevated TIs are frequent in RS data (Figs. 4e and 5e). The depth and strength of the surface-based TIs are up to 1300 m and 19°C, respectively, which is typically deeper and stronger than those of the elevated TIs in winter. In addition, the elevated TIs in winter are generally lower than those in summer (Figs. 4a and 5a), which is also consistent with the analysis in Tjernström and Graversen (2009). The surface-based TIs in winter are primarily developed due to strong longwave radiative cooling over the Arctic Ocean, which is typically related to clear and calm conditions (Graham et al. 2017; Cohen et al. 2017). The mechanism for the formation of the elevated TIs in winter could be dominated by moisture and heat advection during the poleward moisture transport from lower latitudes (Woods and Caballero 2016), which is similar to those in summer. The TI depth–strength relationship in winter in RS data (Fig. 6e) exhibit similar pattern as that in summer (Fig. 6a), except that the quasi-linear relationship in winter is extended deeper and stronger due to the occurrence of surface-based TIs. All three reanalyses generally present the surface-based TIs seen in RS data well in winter (Figs. 4f–h and 5f–h). However, all reanalyses typically simulate deeper and stronger elevated TIs than RS data and fail to properly depict the TIs above 1000 m in winter. Moreover, the surfaced-based TIs with depth about 1000m and strength about 10°C are more frequent in reanalyses (Figs. 6f–h) than in RS data (Fig. 6e). Among the reanalyses, ERA5 obtains the closest relationship shapes and the most similar frequency distributions to RS in winter, followed by ERA-I and JRA-55.
In spring, the relationships among the TI base height, depth and strength typically have transitional features between winter and summer in both RS and reanalyses. Additionally, ERA5 also outperforms the other reanalyses in depicting the relationships among TIs in spring, especially for the upper-level TIs simulation (Figs. 4k and 5k). However, the superiority of ERA5 in TIs simulation is not supported by the performance of temperature profile in spring, when the uncertainties in temperature profile in ERA5 are even slightly worse than ERA-I (Figs. 2f,i). The most likely reason for the superiority of ERA5 could be that the vertical resolution of ERA5 is much higher than that of ERA-I, thus suggesting the vertical resolution and the accuracy of the temperature profile are both important to properly capture the Arctic TIs.
d. RFDs of SHI characteristics
Figure 7 illustrates the RFDs of SHI number and characteristics in RS, reanalyses, and RS data at ERA-I, ERA5, and JRA-55 levels in each season. Comparing to the TIs (Fig. 3), the number of SHI layers exhibits greater seasonal variations in RS data. In addition, the profiles with multiple SHI occurrence are much smaller than the profiles with multiple TI occurrences in RS data in nonsummer seasons (see also Table 1). The occurrence of RS profiles with multiple SHI layers is highest (about 85%) in summer with a maximum SHI number of 11 (Fig. 7a), while it is only about 18% in winter with 3 SHI layers at most (Fig. 7e).
As in Fig. 3, but for SHIs and bin sizes used in the RFDs are 1 for SHI number, 100 m for SHI base, 200 m for SHI depth, and 0.2 g kg−1 for SHI strength; note that the RFDs are estimated on the basis of all SHI layers observed in a specific humidity profile.
Citation: Journal of Applied Meteorology and Climatology 60, 4; 10.1175/JAMC-D-20-0079.1
Profiles with multiple SHI layers in spring also exhibit transitional characteristic between summer and winter, with an occurrence of about 47% (Table 1) and a maximum of 9 SHI layers (Fig. 7i). The SHI occurrence is highest in summer (95.7%), followed by spring (77.1%) and winter (58.5%) (Table 1), which therefore exhibit obvious seasonal variation. Our results, however, differ remarkably from the results presented by Nygård et al. (2014) with Nordic RS data over the land, who found that the seasonal variation in SHI occurrence was small and the SHI occurrence was typically higher than 90% throughout the year. One of the potential reasons for the difference may be that the net longwave radiation was more negative over the ice-covered Arctic Ocean than land areas in winter, resulting in more efficient surface cooling and less interaction between atmosphere and surface over the oceans. It is also clearly in Table 1, all three reanalyses appear to underestimate the SHI occurrence most in summer, followed by spring and winter. Moreover, the reanalyses generally simulate the SHI number better in winter (see Fig. 7e) than nonwinter seasons (see Figs. 7a,i), which may result from their smaller uncertainties in low-level specific humidity in winter (Fig. 2q). It should be also mentioned that difference in uncertainties of specific humidity in different seasons may be affected by the assimilation of RS observations into reanalyses. Among the reanalyses, ERA5 on average exhibits the smallest SHI occurrence difference from RS, followed by ERA-I and JRA-55. Given the SHI occurrence from RS data at ERA5 levels is typically closest to that from original RS data, together with the smallest SHI occurrence difference between ERA5 and RS data at ERA5 levels (Table 1), the superiority of ERA5 in presenting the SHI occurrence could be linked to its higher vertical resolution and better presentation of specific humidity profile than other reanalyses.
The RFDs of SHI characteristics from RS data generally present similar pattern as that of TI characteristics in each season. However, the elevated SHIs are slightly less frequent than the elevated TIs (88% and 96% of 116 soundings, respectively) in summer, together with the surface-based SHIs are not as common as the surface-based TIs from RS data (28% and 60% of 82 soundings, respectively) in winter. The remarkably reduced occurrence of surface-based SHIs may result from the Arctic surface is usually a sink for sensible heat but not for water vapor (Persson et al. 2002; Vihma et al. 2011; Nygård et al. 2013). All three reanalyses typically overestimate occurrence of the elevated SHIs with base below 1200 m, especially in nonwinter seasons (Figs. 7b,j). These overestimates are mainly related to that the overall number of SHIs is obviously smaller in reanalysis than RS data (not shown). In addition, they appear to underestimate the RFDs of the most frequently occurred shallow SHIs with depth smaller than 300 m in all seasons (Figs. 7c,g,k), and JRA-55 even fails to capture the SHIs with depth smaller than 400 m in winter and spring. This is most likely due to the lower vertical resolution in reanalyses than in RS data. The reanalyses simulate the RFDs of strengths best among the SHI characteristics, except for the weak SHI with strength less than 0.2 g kg−1 in nonwinter seasons (Figs. 7d,l). Despite the reanalyses generally perform similar in presenting the RFDs of SHI characteristics in each season, ERA5 stands out as being more consistent with results from RS data than ERA-I and JRA-55, particularly in winter.
Comparison of SHI number and occurrences among the RS data at ERA-I, ERA5, and JRA-55 levels against the original RS data also show that a coarser RS profile typically decreases the number of SHI layer per profile (Fig. 7), as well as reduces the occurrences of both SHI and profiles with multiple SHIs in all seasons (Table 1). As we can also see in Fig. 7, the RFDs of SHI base and strength from RS data at reanalyses levels generally agree well with those from the original RS data in all seasons, especially in winter. However, a remarkable difference of the RFD of SHI depth between RS data at reanalyses levels and the original RS data is observed in all seasons. Among the reanalyses, RS data at ERA5 levels generally perform better than those at ERA-I and JRA-55 levels on detecting the RFD of SHI characteristics in each season.
Comparison of SHI and profiles with multiple SHI occurrences between reanalyses and RS at reanalysis levels shows that all reanalyses generally have lower occurrences in each season, except for slightly higher occurrences in JRA-55 in summer (Table 1). Among the SHI characteristics, uncertainties in specific humidity in the reanalyses generally have greater impacts on the RFDs of SHI depth than those of SHI base and strength in each season (Fig. 7). ERA5 generally simulates the RFDs of SHI depth better than ERA-I and JRA-55 in each season, suggesting the uncertainties in specific humidity in ERA5 are smallest among the reanalyses.
e. Relationships among the SHI characteristics
In Figs. 8–10, the joint histograms of SHI base–depth (Fig. 8), base–strength (Fig. 9), and depth–strength (Fig. 10) from RS and reanalyses in summer, winter, and spring over the Arctic Ocean are illustrated, respectively. The maximum base height of SHI in RS is squeezed from above 5000 m in summer (Fig. 8a) and spring (Fig. 8i) to below 3000 m in winter (Fig. 8e). Although the elevated SHIs occur more frequently than the surface-based ones in each season from RS data (Figs. 8a,e,i), no notable differences of the depth (and strength) between elevated and surface-based SHIs are found. Seasonal variation in RS-derived SHI depth is low (Figs. 8a,e,i), with the depth typically shallower than 300 m for both elevated and surface-based SHIs in all seasons. Considering also the small seasonal variation in SHI depth over the Arctic land areas (Nygård et al. 2014), we may conclude that the seasonal cycle of the SHI depth is weak over the entire Arctic. In contrast, SHI strength from RS data exhibits remarkable seasonal variation and is clearly strongest in summer (Figs. 9a) due to the highest moisture content of the air in summer (Nygård et al. 2013), followed by spring (Fig. 9i) and winter (Fig. 9e). Additionally, no general pattern is recognized from the RS-derived SHI depth–strength relationships in each season (Figs. 10a,e,i), further suggesting the SHI depth and strength could be independent of each other.
As in Fig. 4, but for SHI base and depth and bin sizes of 100 m for SHI base and 200 m for SHI depth are used, and the bin sizes are normalized by the number of observations in the bin with the most observations.
Citation: Journal of Applied Meteorology and Climatology 60, 4; 10.1175/JAMC-D-20-0079.1
As in Fig. 8, but for the joint histograms of SHI base and strength and bin sizes used in the histograms are 100 m for SHI base and 0.2 g kg−1 for SHI strength.
Citation: Journal of Applied Meteorology and Climatology 60, 4; 10.1175/JAMC-D-20-0079.1
As in Fig. 8, but for the joint histograms of SHI depth and strength and bin sizes used in the histograms are 200 m for SHI depth and 0.2 g kg−1 for SHI strength.
Citation: Journal of Applied Meteorology and Climatology 60, 4; 10.1175/JAMC-D-20-0079.1
All reanalyses appear to dramatically overestimate the SHI depth (Fig. 8), as well as remarkably underestimate the strength of low-level SHI in all seasons (Fig. 9). While the reanalyses generally exhibit similar SHI base–strength relationships (Fig. 9) as those in RS data, they fail to simulate the relationships of SHI base–depth (Fig. 8) and depth–strength (Fig. 10) in each season. Hence, the three reanalyses typically have more difficulties simulating the relationships among the SHI characteristics than those among the TI characteristics in each season, which may be attributed to their poor performance on the presentation of SHI depth. Furthermore, the ineffective simulation of SHI depth in the reanalyses is mainly due to their uncertainties in specific humidity profiles (Figs. 7c,g,k). Typically, none of the reanalyses is clearly better than the others in simulating the relationships among the SHI characteristics.
4. Assessment of the dependencies of the strongest inversions on surface conditions
a. Surface meteorological conditions during the Arctic field campaigns
In Figs. 11–14, surface meteorological conditions and the strongest TI and SHI in a profile (herein referred to as strongest TI and SHI, respectively) from in situ RS data, ERA-I, ERA5, and JRA-55 over the Arctic Ocean are illustrated. In summer 2008, surface air temperature (SAT) from in situ RS observations was mostly between 1° and −2°C but dropped between −6° and −9°C during day of year (DOY) 234–236 (Fig. 11a). The period for DOY 234–236 was characterized by surface-based strongest TIs but neither the drop in SAT nor surface-based strongest TIs were captured by any reanalysis data (Figs. 12a, 13a, and 14a). After DOY 237 until end of campaign temperature was between −2° and −4°C associated with high surface pressure (SP) and elevated strongest TIs approximately on the altitude of 1 km (Fig. 11a). This is very well captured by ERA-I (Fig. 12a) and ERA5 (Fig. 13a), but in JRA-55 (Fig. 14a) SAT was notably lower and strongest TIs were deeper than in RS observations. The RS-derived strongest SHIs were mostly elevated during summertime; however, they exhibit little relevance to surface meteorological conditions (Fig. 11a). Additionally, consistent connections between the strongest SHIs and surface meteorological conditions also cannot be detected from reanalyses (Figs. 12a, 13a, and 14a).
The strongest TI (bars with black outline) and SHI (bars without outline) per profile, as well as the surface air temperature (magenta lines) and surface pressure (black lines) from in situ RS data over the Arctic Ocean during (a) summer of 2008 ASCOS, (b) winter of N-ICE2015, and (c) spring of N-ICE2015.
Citation: Journal of Applied Meteorology and Climatology 60, 4; 10.1175/JAMC-D-20-0079.1
As in Fig. 11, but for the time and location of in situ RS data from ERA-I.
Citation: Journal of Applied Meteorology and Climatology 60, 4; 10.1175/JAMC-D-20-0079.1
As in Fig. 12, but for in situ RS data from ERA5.
Citation: Journal of Applied Meteorology and Climatology 60, 4; 10.1175/JAMC-D-20-0079.1
As in Fig. 12, but for in situ RS data from JRA-55.
Citation: Journal of Applied Meteorology and Climatology 60, 4; 10.1175/JAMC-D-20-0079.1
Meteorological conditions in winter 2015 are mainly characterized by variation of high SP with low SAT and low SP with high SAT, which can be observed from both in situ RS observations (Fig. 11b) and reanalyses (Figs. 12b, 13b, and 14b). The coldest SAT ranging between −30° and −40°C was observed from in situ RS measurements during period of high SP between DOY 23–34, when strong surface-based TIs were formed (Fig. 11b). In addition, this feature was also well simulated by all reanalyses (Figs. 12b, 13b, and 14b). However, during the cold period for DOY 37–46, SP was relatively low and the strongest TIs were mostly weak and elevated. The strongest SHIs were mostly surface-based and shallow from RS data. Although the surface-based SHIs were generally well simulated in the reanalyses, they were much deeper than in RS data. Moreover, no general connection between the strongest SHIs and surface meteorological conditions can be recognized from both RS and reanalysis data in winter.
In spring 2015, SP from in situ RS observations was mostly higher than 990 hPa except for the periods for DOY 66–71 and DOY 74–76 (Fig. 11c). During the periods for DOY 61–62, DOY 71–74, and DOY 76–80, SAT was below −20°C associated with high SP and surface-based TIs. After DOY 107, the higher SAT under surface high pressure system typically linked to the elevated TIs. These features observed from in situ RS measurements are also in general well presented in all reanalyses, except that TIs derived from reanalyses were elevated during DOY 71–74 due to their remarkable overestimation of SAT (Figs. 12c, 13c, and 14c). During DOY 60–79, both surface-based and elevated strongest SHIs from in situ RS observations were observed under high SP with low SAT and low SP with high SAT. The strongest SHIs were mostly elevated after DOY 103 until end of campaign, when high SP with low SAT and low SP with high SAT were both detected (Fig. 11c). Therefore, the springtime strongest SHIs may be not linked to surface meteorological conditions. The strongest SHIs are typically deeper in the reanalyses than RS data, and they also exhibit little connections to SP and SAT in the reanalyses (Figs. 12c, 13c, and 14c).
b. Dependencies of the strongest inversions on SAT and SP
Nygård et al. (2014) has reported that the strongest inversions in a profile are statistically dependent on the prevailing meteorological conditions over the Arctic. To investigate the dependencies of the reanalyses on surface meteorological parameters over the Arctic Ocean, the statistically significant correlations (p < 0.01) of the strongest inversion characteristics per profile with SAT and SP in RS and reanalyses in each season are shown in Fig. 15. Dependency of the strongest TI on SAT and SP in RS data is clearly consistent among the seasons. The RS-derived strongest TI base is positively (negatively) correlated with SAT (SP), and both the strongest TI depth and strength are negatively (positively) correlated with SAT (SP) in all seasons. Typically, the strongest TIs have low base height, but they are deep and strong, when SAT is low and SP is high. This may be due to the fact that the surface high pressure system helps to cool the surface and facilitate the formation of deep and strong low-level TIs (Fig. 11). In additional, correlations between the strongest TI characteristics and SAT in RS data are weaker in summer than in winter and spring, which may result from the more frequent elevated TIs in summer and the surface temperature have smaller impacts on them. It is also worth noting that correlation r between the strongest TI strengths and SP in RS data is stronger in summer (r = 0.55) than in winter (r = 0.39) and spring (r = 0.36). However, the reason for this stronger summertime correlation is difficult to identify.
Correlations of the strongest TI and SHI characteristics with surface air temperature and surface pressure in RS, ERA-I, ERA5, and JRA-55 during (top) summer, (middle) winter, and (bottom) spring over the Arctic Ocean. Here, tBas, tDep, and tStr are TI base, depth, and strength, respectively; qBas, qDep, and qStr denote SHI base, depth, and strength, respectively. Nonwhite colors indicate a statistically significant correlation (correlation coefficient r < −0.2 or > 0.2) at the 99% level or greater.
Citation: Journal of Applied Meteorology and Climatology 60, 4; 10.1175/JAMC-D-20-0079.1
None of the reanalyses is able to strictly capture the abovementioned dependencies of the strongest TI on surface meteorological conditions. ERA-I misses the connections between the strongest TI base and SAT, as well as those between the strongest TI depth and SP in summer and winter. Additionally, the springtime correlation between the strongest TI base and SP in ERA-I is even opposite to the result in RS (Fig. 15). While SP from reanalyses generally agrees well with that from RS, remarkable differences in SAT are observed between reanalyses and RS in each season (Figs. 12–14). The differences in correlations derived by ERA-I from RS may be partly linked to the uncertainties in SAT in ERA-I. Specifically, overestimation of SAT in ERA-I under surface high pressure conditions (e.g., DOY 234–236 in 2008, DOY 23–32, and DOY 147–152 in 2015) generally associates with the strongest TI at higher altitude, with weaker strength (Fig. 12). Moreover, the coarse vertical resolution and temperature uncertainties in ERA-I profiles also contribute to the correlation differences between ERA-I and RS.
ERA5 successfully avoids the opposite correlation between the strongest TI base and SP in spring seen in ERA-I, whereas it fails to capture the summertime correlation between the strongest TI base and SP, as well as the springtime correlation between the strongest TI strength and SAT. These missing correlations in ERA5 in summer and spring could be also affected by the overestimation of SAT under surface high pressure conditions, which accompanies by the strongest TI with higher altitude and weaker strength (Fig. 13).
JRA-55 appears to have good representation of the correlations of the strongest TI characteristics to SAT in all seasons, the correlations in JRA-55 also suffers from the uncertainties in SAT. Both overestimation (e.g., DOY 234–236 in 2008 and DOY 147–152 in 2015) and underestimation (e.g., DOY 238–244 in 2008) of SAT in JRA-55 under surface high pressure conditions accompanies by the strongest TI at higher altitude, with deeper depth and weaker strength (Fig. 14). Moreover, JRA-55 does not capture the correlations between the strongest TI characteristics and SP in winter and spring, which is most likely due to the coarse vertical resolution and temperature uncertainties in JRA-55. Therefore, none of the reanalyses is clearly better than the others in simulating the dependency of the strongest TIs on surface meteorological conditions over the Arctic Ocean.
In contrast to the TI, no general dependency of the strongest SHIs on SAT and SP is detected in RS data among the seasons over the Arctic Ocean. The negative correlation between the strongest SHI depth and SP in summer, as well as that between the strongest SHI strength and SP in winter, are only observed in RS. In addition, the positive correlations seen in RS data during springtime are not reproduced in any other season. The possible explanation for this independent feature may be that the elevated SHIs are more common than surface-based ones in each season as analyzed in sections 3d and 3e, and they are less affected by the surface conditions (see also Fig. 11). The independencies of the strongest SHIs on surface conditions are also revealed in all three reanalyses. However, the patterns of the correlations revealed by the reanalyses are substantially different from those in RS data (see also Figs. 12–14). Accordingly, the correlations between the strongest SHIs and surface conditions are poorly reproduced by all reanalyses in each season. Typically, representation of the vertical structure of moisture in the reanalyses is more challenging than that of temperature over the Arctic Ocean.
5. Relationships between the simultaneous TI and SHI
In this section, we analyze the relationships between the simultaneous TI and SHI in RS and reanalyses in different seasons over the Arctic Ocean, since the Arctic SHIs often occur simultaneously with TIs (e.g., Devasthale et al. 2011; Vihma et al. 2011; Sedlar et al. 2012; Tjernström et al. 2012). The simultaneousness of TI and SHI is defined in this study as a SHI occur at least partly within the same layer of a TI (Nygård et al. 2014). In Table 2, occurrences of the simultaneous TI and SHI in each season are illustrated. The occurrence of the simultaneous TI and SHI in RS data is higher in summer than in spring and winter. In addition, while the simultaneousness of TI and SHI is frequently detected at both surface and higher altitude in winter, it is typically elevated in summer and spring. It is worth mentioning in Table 2 that the sum of occurrences of simultaneous inversions at the surface and at higher altitude is greater than 100% is due to the occurrence of both surface-based and elevated inversion layers in a profile. However, no general seasonal cycle of the occurrence of the simultaneous inversions is detected in the reanalyses, and the reanalyses typically present much lower simultaneousness than the RS in all seasons. Among the reanalyses, ERA5 and ERA-I exhibit similar performances on presenting the simultaneous inversions, whereas JRA-55 generally has the largest discrepancies.
Occurrences of the simultaneous TI and SHI in RS and reanalysis data in summer, winter, and spring. Note that when the simultaneous inversions occur more than once (e.g., both at the surface and at higher altitude) in a profile only one simultaneous inversion is counted into the overall occurrence calculation.
The statistically significant correlations (p < 0.01) between the simultaneous TI and SHI characteristics in RS and reanalyses in each season over the Arctic Ocean are illustrated in Fig. 16. The simultaneous TI and SHI bases in RS show a strong correlation of close to 1 in each season, which is not surprising due to the definition of the simultaneous inversions used in this study. Additionally, the significant negative correlations of SHI base with simultaneous TI depth and strength in RS data in each season are in fact the connections among the TI characteristics for the same reason. In summer, the simultaneous TI and SHI depths, the simultaneous TI and SHI strengths, and the simultaneous TI depth and SHI strength in RS data are positively correlated. Considering also the shallow and weak elevated TIs occur most frequently in summer (Fig. 6a), the simultaneous elevated TI and SHI are therefore typically both shallow and weak in that season. However, no significant correlation between the simultaneous TI strength and SHI depth is detected in summer, which may be linked to the small variations in SHI depth in RS data (Fig. 8a). In winter, the simultaneous TI strength also has a positive correlation with SHI strength in RS data, whereas the correlation between the simultaneous TI and SHI depths is not observed. Considering that the simultaneousness of inversions occurs at both surface and higher altitude in winter (Table 2), and that the depths are typically shallow for both elevated TI and SHI (Figs. 4e and 8e), the latter disconnection could be attributed to the deep surface-based TI and the shallow surface-based SHI in winter. During springtime, the simultaneous TI and SHI depths, the simultaneous TI strength and SHI depth in RS data are positively correlated. Hence, the correlations between the simultaneous TIs and SHIs in spring also generally have transitional features between winter and summer in RS data.
Correlations between the simultaneous TI and SHI characteristics in RS, ERA-I, ERA5, and JRA-55 during (top) summer, (middle) winter, and (bottom) spring over the Arctic Ocean. Definitions and color meanings are as in Fig. 15.
Citation: Journal of Applied Meteorology and Climatology 60, 4; 10.1175/JAMC-D-20-0079.1
In Fig. 16, correlations between the simultaneous TI and SHI bases are occasionally less than 0.8 in the reanalyses, which could be affected by their coarser vertical resolution and lower accuracy. Moreover, the negative correlations of the simultaneous SHI base to TI depth and strength seen in RS in each season are either absent or opposite in the reanalyses. All reanalyses also have difficulties in displaying the other correlations between simultaneous TI and SHI characteristics in each season, further suggesting that the relationships between TI and SHI characteristics are poorly represented by the reanalyses. The poor performance of the reanalyses could mainly result from their biases in SHI characteristics simulation (section 3e) since the TI characteristics are generally presented in the reanalyses in each season (section 3c). As such, the representation of the SHI characteristics in the reanalyses would be a greater challenge to understand the boundary layer structure as well as their interactions with cloud over the Arctic Ocean.
6. Conclusions
In this paper, the vertical structure of the Arctic atmosphere from in situ RS measurements, as well as ERA-I, ERA5, and JRA-55 at model levels is investigated in summer 2008, winter 2015, and spring 2015 over the Arctic Ocean. Taking the inversion characteristics in RS observations as the reference, the relationships among the TI (and SHI) characteristics in the reanalyses are validated in each season. Furthermore, the dependencies of the strongest TI (and SHI) per profile on surface conditions in the reanalyses are evaluated, together with the relationships between the TI and SHI characteristics in the reanalyses are assessed in each season.
Note that the characteristics of TIs and SHIs and their relationships in each season presented in this study must be viewed with the caveat of fragmentarily available data in the Arctic seasons (i.e., August for summer and January–February for winter) because of the limited life span of the Arctic expeditions. The selected data in different years (i.e., summer 2008, winter 2015 and spring 2015) could also result in flaws in the seasonal patterns of the Arctic inversions and their relationships, because the Arctic climate may have undergone huge changes between 2008 and 2015. In addition, assimilation of RS observations in 2015 but not in 2008 may affect the skill of reanalyses to simulate TIs and SHIs. Also, the bias of temperature and specific humidity profiles in RS measurements could introduce uncertainties in the results depicted in this study.
Our findings from this paper can be summarized as follows:
Reanalyses generally capture the shape of mean temperature and specific humidity profile seen in RS data in each season over the Arctic Ocean, except for the obvious differences at lower layers in summer. Mean temperature bias is typically larger in JRA-55 than in ERA5 and ERA-I, especially at the layers below 850 hPa in summer and winter. Additionally, mean specific humidity bias in the reanalyses is typically larger in summer than in spring and winter, and it is generally larger at bottom levels than upper levels. While there is no evidence that ERA5 is better than ERA-I at simulating low tropospheric temperatures, ERA5 generally performs better than ERA-I at depicting the specific humidity, especially at bottom levels. Moreover, JRA-55 typically has the largest temperature and specific humidity RMSE in all seasons.
All three reanalyses typically perform better at simulating the RFDs of TI (SHI) numbers in nonwinter seasons (winter). The occurrences of inversion and profiles with multiple inversions tend to be greatly underestimated by all reanalyses, and this underestimation is generally greater for the profiles with multiple inversions than the inversions in each season. Among the reanalyses, ERA5 on average performs best on presenting the RFDs of TI (and SHI) number and characteristics followed by ERA-I and JRA-55 in all seasons over the Arctic Ocean.
All three reanalyses generally well present the relationships among the TI characteristics seen in the RS data in all seasons. ERA5 typically performs better than ERA-I and JRA-55 at simulating the relationship among the TI characteristics, which may be related with its higher vertical resolution.
Among the SHI characteristics, the reanalyses simulate the SHI depth worst in each season. In addition, none of the reanalyses is clearly better than the others in simulating the relationships among the SHI characteristics.
The dependency of the strongest TIs on SAT and SP seen in RS data is only fractionally reproduced in all reanalyses in each season, which may be related to the inefficiency of reanalyses to present SAT as well as the strongest TIs. Moreover, no consistent correlations between the strongest SHIs and surface meteorological conditions can be detected in both RS and reanalysis data. Hence, the representation of the vertical structure of moisture is more challenging than that of temperature in the reanalyses over the Arctic Ocean.
Although ERA5 and ERA-I typically have better performance than JRA-55 in presenting the simultaneous TI and SHI, all reanalyses fail to depict the occurrences of the simultaneous inversions in each season, as well as the seasonal cycle of occurrences seen in RS data. Furthermore, the correlations between simultaneous TI and SHI characteristics seen in RS data are generally poorly represented by the reanalyses, which may be related to the inefficiency of the reanalyses on SHI characteristics simulation.
Acknowledgments
The authors thank Dr. David A. R. Kristovich and two anonymous reviewers for their constructive suggestions and insightful criticisms that substantially improved the quality of our work. This work is sponsored by the National Key Research and Development Program of China (2019YFD0901404); the Shanghai Pujiang Program (19PJ1404300); the Shanghai Science and technology innovation action plan (19DZ1207502); the National Natural Science Foundation of China (42076240 and 41606208); the open fund of State Key Laboratory of Satellite Ocean Environment Dynamics, Second Institute of Oceanography, Ministry of Natural Resources (MNR) (QNHX1909); the open fund of the Key Laboratory for Polar Science, Polar Research Institute of China, MNR (KP201701); and the open fund of the Key Laboratory for Information Science of Electromagnetic Waves, Fudan University (EMW201909).
Data availability statement
We are grateful to the European Centre for Medium-Range Weather Forecasts (ECMWF), the Japan Meteorological Agency (JMA), the Stockholm University, and the Norwegian Polar Data Centre for providing their data. The ERA-I can be obtained from https://apps.ecmwf.int/datasets/data/interim-full-daily/, the ERA5 can be obtained from https://cds.climate.copernicus.eu, the JRA-55 can be downloaded from https://jra.kishou.go.jp/JRA-55/index_en.html, the RS data from ASCOS can be downloaded from https://bolin.su.se/data/ascos/datainfo.php?n=Radiosoundings, and the RS data from N-ICE2015 can be downloaded from https://data.npolar.no/dataset/216df9b3-e2bd-5111-9c02-fea848d76670.
REFERENCES
Argentini, S., A. P. Viola, G. Mastrantonio, A. Maurizi, T. Georgiadis, and M. Nardino, 2003: Characteristics of the boundary layer at Ny-Ålesund in the Arctic during the ARTIST field experiment. Ann. Geophys., 46, 185–196, https://doi.org/10.4401/ag-3414.
Beine, H. J., S. Argentini, A. Maurizi, G. Mastrantonio, and A. Viola, 2001: The local wind field at Ny-Ålesund and the Zeppelin mountain at Svalbard. Meteor. Atmos. Phys., 78, 107–113, https://doi.org/10.1007/s007030170009.
Bintanja, R., R. G. Graversen, and W. Hazeleger, 2011: Arctic winter warming amplified by the thermal inversion and consequent low infrared cooling to space. Nat. Geosci., 4, 758–761, https://doi.org/10.1038/ngeo1285.
Boé, J., A. Hall, and X. Qu, 2009: Current GCMs’ unrealistic negative feedback in the Arctic. J. Climate, 22, 4682–4695, https://doi.org/10.1175/2009JCLI2885.1.
Brunke, M. A., S. T. Stegall, and X. B. Zeng, 2015: A climatology of tropospheric humidity inversions in five reanalyses. Atmos. Res., 153, 165–187, https://doi.org/10.1016/j.atmosres.2014.08.005.
Chang, L., S. L. Song, G. P. Feng, Y. Zhang, and G. P. Gao, 2017: Assessment of the uncertainties in Arctic low-level temperature inversion characteristics in radio occultation observations. IEEE Trans. Geosci. Remote Sens., 55, 1793–1803, https://doi.org/10.1109/TGRS.2016.2633461.
Cohen, L., S. R. Hudson, V. P. Walden, R. M. Graham, and M. A. Granskog, 2017: Meteorological conditions in a thinner Arctic sea ice regime from winter to summer during the Norwegian Young Sea Ice expedition (N-ICE2015). J. Geophys. Res. Atmos., 122, 7235–7259, https://doi.org/10.1002/2016JD026034.
Copernicus Climate Change Service, 2017: ERA5: Fifth generation of ECMWF Atmospheric reanalyses of the global climate. Copernicus Climate Change Service Climate Data Store, accessed 17 May 2020, https://cds.climate.copernicus.eu/cdsapp#!/home.
Cullather, R. I., and M. G. Bosilovich, 2011: The moisture budget of the polar atmosphere in MERRA. J. Climate, 24, 2861–2879, https://doi.org/10.1175/2010JCLI4090.1.
Curry, J. A., W. B. Rossow, D. Randall, and J. L. Schramm, 1996: Overview of Arctic cloud and radiation characteristics. J. Climate, 9, 1731–1764, https://doi.org/10.1175/1520-0442(1996)009<1731:OOACAR>2.0.CO;2.
de Boer, G., and Coauthors, 2014: Near-surface meteorology during the Arctic Summer Cloud Ocean Study (ASCOS): Evaluation of reanalyses and global climate models. Atmos. Chem. Phys., 14, 427–445, https://doi.org/10.5194/acp-14-427-2014.
Dee, D. P., and Coauthors, 2011: The ERA-Interim reanalysis: Configuration and performance of the data assimilation system. Quart. J. Roy. Meteor. Soc., 137, 553–597, https://doi.org/10.1002/qj.828.
Devasthale, A., U. Willén, K. G. Karlsson, and C. G. Jones, 2010: Quantifying the clear-sky temperature inversion frequency and strength over the Arctic Ocean during summer and winter seasons from AIRS profiles. Atmos. Chem. Phys., 10, 5565–5572, https://doi.org/10.5194/acp-10-5565-2010.
Devasthale, A., J. Sedlar, and M. Tjernström, 2011: Characteristics of water-vapour inversions observed over the Arctic by Atmospheric Infrared Sounder (AIRS) and radiosondes. Atmos. Chem. Phys., 11, 9813–9823, https://doi.org/10.5194/acp-11-9813-2011.
Devasthale, A., J. Sedlar, B. H. Kahn, M. Tjernström, E. J. Fetzer, B. Tian, J. Teixeira, and T. S. Pagano, 2016: A decade of spaceborne observations of the Arctic atmosphere novel, insights from NASA’s AIRS Instrument. Bull. Amer. Meteor. Soc., 97, 2163–2176, https://doi.org/10.1175/BAMS-D-14-00202.1.
Ganeshan, M., and D. L. Wu, 2015: An investigation of the Arctic inversion using COSMIC RO observations. J. Geophys. Res. Atmos., 120, 9338–9351, https://doi.org/10.1002/2015JD023058.
Graham, R. M., and Coauthors, 2017: A comparison of the two Arctic atmospheric winter states observed during N-ICE2015 and SHEBA. J. Geophys. Res. Atmos., 122, 5716–5737, https://doi.org/10.1002/2016JD025475.
Graham, R. M., and Coauthors, 2019: Evaluation of six atmospheric reanalyses over Arctic sea ice from winter to early summer. J. Climate, 32, 4121–4143, https://doi.org/10.1175/JCLI-D-18-0643.1.
Granskog, M. A., P. Assmy, S. Gerland, G. Spreen, H. Steen, and L. H. Smedsrud, 2016: Arctic research on thin ice: Consequences of Arctic sea ice loss. Eos, Trans. Amer. Geophys. Union, 97, 22–26, https://doi.org/10.1029/2016EO044097.
Hersbach, H., and D. Dee, 2016: ERA5 reanalysis is in production. ECMWF Newsletter, No. 147, ECMWF, Reading, United Kingdom, 7, http://www.ecmwf.int/sites/default/files/elibrary/2016/16299-newsletter-no147-spring-2016.pdf.
Jakobson, E., T. Vihma, T. Palo, L. Jakobson, H. Keernik, and J. Jaagus, 2012: Validation of atmospheric reanalyses over the central Arctic Ocean. Geophys. Res. Lett., 39, n/a, https://doi.org/10.1029/2012GL051591.
Kahl, J. D., 1990: Characteristics of the low-level temperature inversion along the Alaskan Arctic coast. Int. J. Climatol., 10, 537–548, https://doi.org/10.1002/joc.3370100509.
Kayser, M., and Coauthors, 2017: Vertical thermodynamic structure of the troposphere during the Norwegian young sea ICE expedition (N-ICE2015). J. Geophys. Res. Atmos., 122, 10 855–10 872, https://doi.org/10.1002/2016JD026089.
Kilpeläinen, T., T. Vihma, M. Manninen, A. Sjoblom, E. Jakobson, T. Palo, and M. Maturilli, 2012: Modelling the vertical structure of the atmospheric boundary layer over Arctic fjords in Svalbard. Quart. J. Roy. Meteor. Soc., 138, 1867–1883, https://doi.org/10.1002/qj.1914.
Kobayashi, S., and Coauthors, 2015: The JRA-55 reanalysis: General specifications and basic characteristics. J. Meteor. Soc. Japan, 93, 5–48, https://doi.org/10.2151/jmsj.2015-001.
Lindsay, R., M. Wensnahan, A. Schweiger, and J. Zhang, 2014: Evaluation of seven different atmospheric reanalysis products in the Arctic. J. Climate, 27, 2588–2606, https://doi.org/10.1175/JCLI-D-13-00014.1.
Liu, Y. H., and J. R. Key, 2003: Detection and analysis of clear-sky, low-level atmospheric temperature inversions with MODIS. J. Atmos. Oceanic Technol., 20, 1727–1737, https://doi.org/10.1175/1520-0426(2003)020<1727:DAAOCL>2.0.CO;2.
Liu, Y. H., J. R. Key, A. Schweiger, and J. Francis, 2006: Characteristics of satellite-derived clear-sky atmospheric temperature inversion strength in the Arctic, 1980–96. J. Climate, 19, 4902–4913, https://doi.org/10.1175/JCLI3915.1.
Lüpkes, C., T. Vihma, E. Jakobson, G. König-Langlo, and A. Tetzlaff, 2010: Meteorological observations from ship cruises during summer to the central Arctic: A comparison with reanalysis data. Geophys. Res. Lett., 37, L09810, https://doi.org/10.1029/2010GL042724.
Medeiros, B., C. Deser, R. A. Tomas, and J. E. Kay, 2011: Arctic inversion strength in climate models. J. Climate, 24, 4733–4740, https://doi.org/10.1175/2011JCLI3968.1.
Naakka, T., T. Nygård, and T. Vihma, 2018: Arctic humidity inversions: Climatology and processes. J. Climate, 31, 3765–3787, https://doi.org/10.1175/JCLI-D-17-0497.1.
Nygård, T., T. Valkonen, and T. Vihma, 2013: Antarctic low-tropospheric humidity inversions: 10-yr climatology. J. Climate, 26, 5205–5219, https://doi.org/10.1175/JCLI-D-12-00446.1.
Nygård, T., T. Valkonen, and T. Vihma, 2014: Characteristics of Arctic low-tropospheric humidity inversions based on radio soundings. Atmos. Chem. Phys., 14, 1959–1971, https://doi.org/10.5194/acp-14-1959-2014.
Onogi, K., and Coauthors, 2007: The JRA-25 Reanalysis. J. Meteor. Soc. Japan, 85, 369–432, https://doi.org/10.2151/jmsj.85.369.
Pavelsky, T. M., J. Boé, A. Hall, and E. J. Fetzer, 2011: Atmospheric inversion strength over polar oceans in winter regulated by sea ice. Climate Dyn., 36, 945–955, https://doi.org/10.1007/s00382-010-0756-8.
Persson, P. O. G., C. W. Fairall, E. L. Andreas, P. S. Guest, and D. K. Perovich, 2002: Measurements near the Atmospheric Surface Flux Group tower at SHEBA: Near-surface conditions and surface energy budget. J. Geophys. Res., 107, 8045, https://doi.org/10.1029/2000JC000705.
Pithan, F., B. Medeiros, and T. Mauritsen, 2014: Mixed-phase clouds cause climate model biases in Arctic wintertime temperature inversions. Climate Dyn., 43, 289–303, https://doi.org/10.1007/s00382-013-1964-9.
Saha, S., and Coauthors, 2010: The NCEP Climate Forecast System Reanalysis. Bull. Amer. Meteor. Soc., 91, 1015–1058, https://doi.org/10.1175/2010BAMS3001.1.
Screen, J. A., and I. Simmonds, 2011: Erroneous Arctic temperature trends in the ERA-40 reanalysis: A closer look. J. Climate, 24, 2620–2627, https://doi.org/10.1175/2010JCLI4054.1.
Sedlar, J., 2014: Implications of limited liquid water path on static mixing within Arctic low-level clouds. J. Appl. Meteor. Climatol., 53, 2775–2789, https://doi.org/10.1175/JAMC-D-14-0065.1.
Sedlar, J., M. D. Shupe, and M. Tjernström, 2012: On the relationship between thermodynamic structure and cloud top, and its climate significance in the Arctic. J. Climate, 25, 2374–2393, https://doi.org/10.1175/JCLI-D-11-00186.1.
Serreze, M. C., R. C. Schnell, and J. D. Kahl, 1992: Low-level temperature inversions of the Eurasian Arctic and comparisons with Soviet drifting station data. J. Climate, 5, 615–629, https://doi.org/10.1175/1520-0442(1992)005<0615:LLTIOT>2.0.CO;2.
Serreze, M. C., R. G. Barry, and J. E. Walsh, 1995: Atmospheric water vapor characteristics at 70°N. J. Climate, 8, 719–731, https://doi.org/10.1175/1520-0442(1995)008<0719:AWVCA>2.0.CO;2.
Serreze, M. C., A. P. Barrett, and J. Stroeve, 2012: Recent changes in tropospheric water vapor over the Arctic as assessed from radiosondes and atmospheric reanalyses. J. Geophys. Res., 117, D10104, https://doi.org/10.1029/2011JD017421.
Simmons, A. J., and D. M. Burridge, 1981: An energy and angular-momentum conserving vertical finite-difference scheme and hybrid vertical coordinates. Mon. Wea. Rev., 109, 758–766, https://doi.org/10.1175/1520-0493(1981)109<0758:AEAAMC>2.0.CO;2.
Solomon, A., M. D. Shupe, P. O. G. Persson, and H. Morrison, 2011: Moisture and dynamical interactions maintaining decoupled Arctic mixed-phase stratocumulus in the presence of a humidity inversion. Atmos. Chem. Phys., 11, 10 127–10 148, https://doi.org/10.5194/acp-11-10127-2011.
Solomon, A., M. D. Shupe, O. Persson, H. Morrison, T. Yamaguchi, P. M. Caldwell, and G. de Boer, 2014: The sensitivity of springtime Arctic mixed-phase stratocumulus clouds to surface-layer and cloud-top inversion-layer moisture sources. J. Atmos. Sci., 71, 574–595, https://doi.org/10.1175/JAS-D-13-0179.1.
Sotiropoulou, G., and Coauthors, 2016: Atmospheric conditions during the Arctic Clouds in Summer Experiment (ACSE): Contrasting open water and sea ice surfaces during melt and freeze-up seasons. J. Climate, 29, 8721–8744, https://doi.org/10.1175/JCLI-D-16-0211.1.
Tjernström, M. and R. G. Graversen, 2009: The vertical structure of the lower Arctic troposphere analysed from observations and the ERA-40 reanalysis. Quart. J. Roy. Meteor. Soc., 135, 431–443, https://doi.org/10.1002/qj.380.
Tjernström, M., and Coauthors, 2012: Meteorological conditions in the central Arctic summer during the Arctic Summer Cloud Ocean Study (ASCOS). Atmos. Chem. Phys., 12, 6863–6889, https://doi.org/10.5194/acp-12-6863-2012.
Tjernström, M., and Coauthors, 2014: The Arctic Summer Cloud Ocean Study (ASCOS): Overview and experimental design. Atmos. Chem. Phys., 14, 2823–2869, https://doi.org/10.5194/acp-14-2823-2014.
Vihma, T., T. Kilpeläinen, M. Manninen, A. Sjöblom, E. Jakobson, T. Palo, J. Jaagus, and M. Maturilli, 2011: Characteristics of temperature and humidity inversions and low-level jets over Svalbard fjords in spring. Adv. Meteor., 2011, 1–14, https://doi.org/10.1155/2011/486807.
Wesslén, C., M. Tjernström, D. H. Bromwich, G. de Boer, A. M. L. Ekman, L. S. Bai, and S. H. Wang, 2014: The Arctic summer atmosphere: An evaluation of reanalyses using ASCOS data. Atmos. Chem. Phys., 14, 2605–2624, https://doi.org/10.5194/acp-14-2605-2014.
Woods, C., and R. Caballero, 2016: The role of moist intrusions in winter Arctic warming and sea ice decline. J. Climate, 29, 4473–4485, https://doi.org/10.1175/JCLI-D-15-0773.1.
Yu, X., F. Q. Xie, and C. O. Ao, 2018: Evaluating the lower-tropospheric COSMIC GPS radio occultation sounding quality over the Arctic. Atmos. Meas. Tech., 11, 2051–2066, https://doi.org/10.5194/amt-11-2051-2018.
Yu, L. J., Q. H. Yang, M. Y. Zhou, X. B. Zeng, D. H. Lenschow, X. Q. Wang, and B. Han, 2019: The intraseasonal and interannual variability of Arctic temperature and specific humidity inversions. Atmosphere, 10, 214, https://doi.org/10.3390/atmos10040214.
Zhang, Y. H., D. J. Seidel, J. C. Golaz, C. Deser, and R. A. Tomas, 2011: Climatological characteristics of Arctic and Antarctic surface-based inversions. J. Climate, 24, 5167–5186, https://doi.org/10.1175/2011JCLI4004.1.
