Application of wetPf2 Data for Investigating Characteristics of Temperature and Humidity of Air Masses over Paracel and Spratly Islands

,


Introduction
Observational data from traditional monitoring stations (surface and radiosonde stations) are mainly located on land.Researching the characteristics of atmospheric masses in maritime areas faces many difculties due to the lack of data.With the development of remote sensing techniques, satellites enable measurements and estimation of temperature and humidity profles of atmospheric masses over the ocean, where traditional observational data are scarce.Te Radio Occultation (RO) technique uses signals emitted by the Global Positioning System (GPS) satellites to monitor the Earth's atmosphere.Tis technique was frst used in the GPS/MET (GPS Meteorology) project [1].Subsequently, the RO technique was further developed and improved by Kursinski and colleagues [2].Today, this technique has been employed in many projects [3][4][5][6][7][8].Te projects have provided a large volume of Global Positioning System Radio temperature and humidity profle data in the Vietnam region, particularly in the East Sea, where there are no monitoring radiosonde stations.
Te wetPf2 data show good quality in other regions of the world.To enhance the efectiveness of using this data over the Vietnam region, it is necessary to evaluate the quality of the wetPf2 data by comparing them with Vietnam radiosonde observations.Since radiosonde stations are not available over the East Sea, detailed investigations of the temporal and spatial variations in the characteristics of air masses over the region have not been conducted using observed data.After quality assessment, the wetPf2 data can be employed to study the characteristics of atmospheric felds over the East Sea region.
Tis article frstly assesses the quality of wetPf2 data in the Vietnam region.Te temperature and relative humidity profles of wetPf2 data are compared with radiosonde observations (RAOB) data in Hanoi, Da Nang, and Tan Son Hoa (Ho Chi Minh City).Te evaluated data are then used to study the characteristics of temperature and humidity variations with height and season for air masses in the Paracel Islands, representing the northern air mass of the East Sea, and in the Spratly Islands, representing the southern air mass of the East Sea.Te subsequent sections of the article include: Section 2 shows data and method, which outlines the data sources and calculation methods, and evaluates the data quality; Section 3 presents the main fndings of the article, analyzes the data quality, and the characteristics of temperature and humidity variations in the studied maritime area; and a summary and discussion are given in Section 4.

Data and Methodology
2.1.Data.Tis study utilizes wetPf2 data (COSMIC-2/ FORMOSAT-7) from the Central Weather Administration (CWA) during a period of 4 years from October 2019 to September 2023 (https://tacc.cwa.gov.tw/data-service/fs7rt_tdpc/daily_tar/).Te wetPf2 data includes vertical profles of temperature and relative humidity at a vertical resolution of 50 m.Radiosonde data (RAOB) for the same period are from three radiosonde stations in Vietnam, namely Hanoi (21.01 °N 105.80 °E), Da Nang (16.03 °N 108.20 °E), and Tan Son Hoa (Ho Chi Minh City) (10.81 °N 106.66 °E).Te RAOB data are obtained from the University of Wyoming (https:// weather.uwyo.edu/upperair/sounding.html).Te 30 years of data (1991 to 2020) from the ffth generation of ECMWF (European Centre for Medium-Range Weather Forecasts) reanalysis, known as ERA5 (https://cds.climate.copernicus.eu), covering air temperature at the 2-meter level and winds at the 10-meter level from the ground, are utilized to calculate long-term climatological means for analyzing the annual variation in air-mass characteristics.

Method for Comparison between wetPf2 and Radiosonde
Data.One of the most important steps in evaluating GPSRO data are the selection of corresponding pairs of GPSRO and RAOB data for comparison.Previous studies have proposed various criteria for selecting data pairs.Tese criteria take into 2 Advances in Meteorology account diferences in distance and time between the two datasets such as diferences of 100 km, 200 km, 300 km in distance, and 1 h, 2 h, 3 h in observed time [9]; diferences of 100 km and 1 h [13]; diferences of 25 km, 75 km, 125 km, 175 km, 225 km, 275 km, and 0.5 h, 1.5 h, 2.5 h, 3.5 h, 4.5 h, and 5.5 h [12].In this work, we employ criteria involving maximum time diferences of 1 h, 2 h, and 3 h, as well as maximum spatial diferences of 100 km, 200 km, and 300 km to create nine groups (GP1 to GP9) of data for comparison and verifcation, as illustrated in Table 1.
After selecting the data pairs corresponding to the above data groups, the temperature and relative humidity data of wetPf2 were interpolated to the standard isopressure levels of 925 mb, 850 mb, 700 mb, 500 mb, 400 mb, 300 mb, 250 mb, 200 mb, 150 mb, and 100 mb [13].
where T and RH are temperature and relative humidity at pressure level P; T1, RH1 is the temperature and relative humidity at the pressure level P1; T2 is the temperature; and RH2 is the relative humidity at the pressure level P2.
Due to the signifcant variability of relative humidity or any moisture parameter in the troposphere, interpolating values between diferent levels, especially in the upper troposphere (e.g., 250 mb, 200 mb, 150 mb, 100 mb, or close to the tropopause), may introduce unwanted signals (or noise).Although unwanted signals are unavoidable, to minimize the noise, the very high vertical resolution (50 m interval) of GPSRO data is utilized.Applying the weighted interpolation method (equation (2)) for relative humidity interpolation in this dataset, with a maximum vertical interpolation distance of less than 50 m, may reduce unwanted signals.
Te mean and standard deviation of the error between the RO and RAOB data were used for the evaluation.
In which, T W and RH W are the temperature and relative humidity of the wetPf2 data; T R and RH R are the temperature and relative humidity of the RAOB data; ∆T and ∆RH are the diferences in temperature and relative humidity between the wetPf2 data and the RAOB data; ∆Tm and ∆RHm are the mean errors of temperature and relative humidity; SD ΔT and SD ΔRH are the standard deviations of the temperature and relative humidity errors; n is the number of samples (pairs of data).
Te average gradient according to the altitude of the mean temperature (ΔTm (z k )) and relative humidity (ΔRHm (z k )) at level k is calculated as follows: where T W k,i and T W k+1,i are the temperature at k and k + 1 of the i th temperature profle wetPf2 data, respectively; and n is the number of data.

Results
3.1.Evaluate Data Quality of wetPf2.Figure 2 presents the mean error values (ΔTm) and the standard deviation of temperature errors (SD ΔT ) between wetPf2 data and the radiosonde data at diferent altitudes corresponding to 9 groups of data.Te values of ΔTm at all altitudes in all data groups range from −0.31 °C to 0.12 °C.At altitudes of 925 mb, 300 mb, 250 mb, and 200 mb, all ΔTm values are negative, indicating that the wetPf2 data generally have lower temperature values compared to the radiosonde data.Te mean error values of the data groups range from −0.06 °C to −0.02 °C.Te SD ΔT values at all altitudes are less than 1.7 °C.Te SD ΔT values decrease from 925 mb to 200 mb.At 200 mb, SD ΔT is the smallest, ranging from 0.42 °C to 0.67 °C.From 200 mb to 100 mb, the SD ΔT values increase gradually with average standard deviation values from 0.73 °C to 1.04 °C.Te average standard deviation in the GP1 case is the smallest, with a value of 0.73 °C.Te average standard deviation in the GP9 case is the largest, with a value of 1.04 °C (Figure 2).Previous studies reported a mean temperature diference of 0.22 K with a standard deviation of 0.95 K between wetPf2 data and RAOB data in the layer from 8 km to 11 km.In the layer from 12.5 km to 16.5 km, the mean temperature diference is zero with a standard deviation of Advances in Meteorology 3 1.10 K [15].Terefore, the calculated average temperature errors and standard deviations in the Vietnam and East Sea regions are reasonable.Table 2 presents the correlation coefcient values (R (T)) of temperature between the wetPf2 data and the radiosonde data at each altitude level for the 9 data groups.Te results show that the value of R (T) at all altitudes is equal to or greater than 0.78.At the 700 mb and 500 mb levels, the value of R (T) is lower than at other levels.Te change in R (T) values with respect to time diferences (one to three hours) is smaller than the change with respect to distances (100 km to 300 km) (Tables 1 and 2).Te GP1 case has the highest average R (T) value (0.93), and the GP9 case has the lowest average R (T) value (0.86).Tis means that the wetPf2 data closer in time and distance to the radiosonde station observation have a better temperature correlation coefcient.Te overall values of R (T) for all data groups indicate a good correlation between the temperature profle of wetPf2 data and radiosonde data.
Figure 3 presents the variations of the mean error and standard deviation of the error of relative humidity between the wetPf2 data and the radiosonde data at the corresponding altitude levels for the 9 data groups.Figure 3(a) illustrates the profle of the mean error of relative humidity (∆RHm).Te ∆RHm values below 400 mb are smaller compared to the values above 400 mb.Te ∆RHm values at altitudes below 400 mb range from −5.6% to 4.6%.For altitudes above 400 mb, the ∆RHm values vary from 10.3% to 40.5%.Te ∆RHm values do not difer signifcantly among the 9 data groups, with most diferences being less than 2.7%.Te average ∆RHm for all cases oscillates between 11.6% and 12.5%.In Figure 3(b), the variations of the standard deviation of the relative humidity error (SD ΔRH ) are presented.Te SD ΔRH values are smaller at lower altitude levels compared to higher levels, and the maximum SD ΔRH is observed at 250 mb and 200 mb for all data groups.Te minimum value of SD ΔRH is found at 925 mb.Te smallest average SD ΔRH occurs in case GP1 (15.1%), while the largest occurs in case GP9 (19.06%).Te results also indicate that the diference in the value of SD ΔRH with respect to time is smaller than the diference in the value of SD ΔRH with respect to distance.Previous studies have shown that the mean error of relative humidity is about 10% with a standard deviation of 15% to 20% in the Asian monsoon region [16].Te mean error of relative humidity in this work ranges from 11.6% to 12.5% with a standard deviation of 15.1% to 19.1%, which is consistent with previous research results.Figure 3(b) shows that the standard deviation of the humidity error is separated into three groups corresponding to the three groups of distance from the observation station: 100 km, 200 km, and 300 km.Tis suggests that the mean error (mean diference between satellite observation data and radiosonde observation data) contains three sources including (i) error due to satellite observation quality at the same location with the radiosonde observation station, (ii) diference in data due to temporal variations in the internal characteristics of the air mass (due to satellite observation time does not coincide with the observation time at the radiosonde station), and (iii) diference due to satellite observation and the spatial location of the radiosonde observation station in diferent air mass positions (Figure 3(b)).In the future, if a sufciently long dataset becomes available for fltering satellite observation data and radiosonde observation data based on time and location, the errors in the dataset will solely come from the quality of satellite observations.Tis could potentially lead to a substantial reduction in overall diferences compared to the errors present in the current dataset used in this study.
Te correlation coefcient of relative humidity (R (RH)) between wetPf2 data and radiosonde data are presented for each altitude level of the 9 data groups in Table 3. Te results indicate signifcant variation in the R (RH) values across altitudes and data groups.At altitudes of 925 mb and 850 mb, R (RH) values are relatively low, ranging from 0.43 to 0.70.For altitudes ranging from 700 mb to 250 mb, R (RH) exhibits higher values compared to other altitudes (R (RH) > 0.60).Notably, the maximum value of R (RH) is observed at the 500 mb altitude, ranging from 0.81 to 0.91, depending on the data group.At levels lower than 500 mb, active convection causes relative humidity (RH) to signifcantly change with space, resulting in more pronounced diferences between RH data at radiosonde stations and satellite sample locations, especially at 925 mb level (Figure 3(a)).Figure 3(a) also shows that the mean error of RH at 500 mb is smaller than at lower levels.At levels higher than 500 mb, the amount of atmospheric moisture (mixing ratio) is relatively small, leading to a relatively larger RH error (Figure 3(a)).In addition, a larger number of samples at 500 mb (Figure 2) makes R (RH) statistics more robust.Tese factors may be important contributors to the correlation coefcient between the two data sources at the 500 mb level being the largest compared to the remaining levels.Te average R (RH) value is highest for case GP1 (0.76) and lowest for case GP9 (0.63).A comparison between wetPf2 data and RAOB data have revealed that case GP1 yields the best results with 4 Advances in Meteorology the least number of available observations.Conversely, case GP9 produces the poorest results despite having a large number of available observations.
Overall, the mean error values, standard deviations, and average correlation coefcients between wetPf2 data and radiosonde data for both temperature and relative humidity (a-i) correspond to data groups GP1, GP2, GP3, GP4, GP5, GP6, GP7, GP8, and GP9, respectively.Te vertical axis denotes the barometric level, the lower horizontal axis indicates the temperature error value, and the upper horizontal axis represents the number of samples.Te black solid line is the mean error value, the red dashed line illustrates the standard deviation of the error, and the blue solid (+) line indicates the number of samples.winter, the winter monsoon brings cold air from the north, signifcantly lowering the temperature in this area compared to summer.Additionally, the efects of heating and nearsurface turbulence lead to a much larger annual temperature variation in the boundary layer compared to the free atmosphere above.In the layer from 3.5 km to 16 km, Tm exhibits small seasonal variations.Te amplitude of Tm seasonal variation at diferent levels ranges from 0.4 °C to 2.1 °C.Above 16 km, the seasonal variation of Tm is larger than in the lower levels, with an amplitude ranging from 1.6 °C to 5.8 °C.Te lowest Tm values are −83.1 °C (in winter) and −79.4 °C (in summer).Tese values are consistent with previous studies that stated the minimum temperature at the tropopause in the North Pacifc monsoon region was about 194.0 K (−79.0 °C) in summer and from 189.4 K to 190.6 K (−83.7 °C to −82.5 °C) in winter [19].

Advances in Meteorology
Te vertical profles of temperature (Tm) for the four seasons, derived from 7730 profles of data in the Spratly Islands region, are presented in Figure 5.It indicates a similar variation trend of Tm with altitude as observed in the Paracel Islands region, decreasing gradually from the surface to the tropopause.Regarding the annual variability of Tm, in lower levels from the surface to 2.0 km, the highest values are found in summer, while the lowest values occur in winter.Te amplitude of Tm seasonal variations at particular levels ranges from 1.2 °C to 2.2 °C.Near the surface, Tm values are 26.1 °C (winter) and 28.3 °C (summer).Compared with the Paracel region, the seasonal temperature variability in the Spratly Islands region is much smaller.Tis refects the weaker infuence of the winter monsoon winds in this region, resulting in higher winter Tm values and smaller temperature variations throughout the year over the Spratly Islands than over the Paracel region.From the 2.0 km to 15.8 km levels, Tm values are higher in spring compared to other seasons.Te variations of Tm at particular levels within the layer are small seasonally dependent, with a variation amplitude ranging from 0.3 °C to 1.4 °C.In the upper atmosphere above 15.8 km, the highest Tm values are observed in summer.Te amplitude of Tm variations at particular levels in this layer is larger than in the layer below, ranging from 1.4 °C to 6.1 °C.Te minimum values of Tm are −79.1 °C (summer) and −82.1 °C (winter) (Figure 5).
To investigate the contrast in temperature between air masses in the north and south of the East Sea at diferent times of the year, Figure 6 presents the temperature differences with altitude between the air masses over the Paracel and Spratly regions during the summer (red line) and winter (blue line).Te fgure shows that in summer, the temperature diference between the two air masses is not signifcant (Figure 6).Tere is a noteworthy cooling efect (−1.0 to −2.3 °C) in the lower atmosphere of the northern air Advances in Meteorology mass (above the Paracel Islands) compared to the southern air mass (above the Spratly Islands) during the winter season.Te substantial temperature diference in winter is primarily infuenced by the northward wind carrying cold air from higher latitudes to the Northern Vietnam and East Sea regions.Due to the characteristics of cold air in the region, which is mainly concentrated in the lower troposphere (below 3 km), the colder air mass over the northern part of the East Sea is predominantly concentrated at lower altitudes (below 3 km) (Figure 6).7 represents the vertical profle of average relative humidity (RHm) in the range from 0 km to 12 km in the Paracel Islands region.Te results show that the RHm values in the boundary layer are higher than in the free atmosphere layer.In the near-surface layer, the seasonal average humidity values range from 75.4% (summer) to 77.6% (winter).Te RHm values increase with height and reach a maximum at around 0.6 km to 0.7 km.Tis height is often the location of the lifting condensation level (LCL).Te maximum values of RHm are 80.7% in spring, 80.5% in summer, 85.5% in autumn, and 85.3% in winter.Due to the infuence of winter monsoon winds, the cold air fow brings low-temperature air from the northern Asian continent to the East Sea, causing temperatures to approach the dew point.Tis results in the highest relative humidity values in this region during winter in the air layer from the surface to 400 m (Figure 7).In the free atmosphere layer, RHm decreases with height for spring, autumn, and winter, reaching a minimum value at the mid-tropospheric level, and then increases with height.However, for the summer RHm profle, a secondary maximum is observed at around 4.8 km.Tis secondary maximum value may be associated with the freezing level within convective clouds.

Relative Humidity. Figure
Te amplitude of RHm variation with height is smallest in summer and largest in winter.Te minimum values of RHm are 28.2% (spring), 56.4% (summer), 49.2% (autumn), and 22.3% (winter) at the 8 to 9 km level.Te amplitude of the annual variation in RHm within the free atmosphere layer is much greater than in the boundary layer, reaching a maximum annual variation value of 35.8% (Figure 7).Te RHm values in summer are signifcantly higher than in winter in the free atmosphere layer.Te large decrease in humidity with height during winter, while relatively little variation in relative humidity occurs with height in summer, may be related to the important role of strong convective activity, which brings moist low-level air to higher atmospheric layers during summer.
Figure 8 represents the vertical profle of the average seasonal relative humidity (RHm) in the layer from 0 km to 12 km in the Spratly Islands region.Te results show that the average seasonal humidity values range from 74.3% (summer) to 78.5% (winter) at the near-surface layer.Similar to the Paracel Islands region, the RHm values increase with altitude and reach a maximum at 0.65 km.Te maximum values of RHm are 80.3% in spring, 81.6% in summer, 84% in autumn, and 85.4% in winter.Subsequently, RHm decreases gradually and reaches a minimum value in the middle troposphere.Te RHm values in the boundary layer are higher than those in the free-atmosphere layer.In the layer below 1.25 km, the annual variation amplitude of RHm is small (<7.5%).Mean relative humidity has the highest values in winter.Tis may be due to the infuence of winter monsoon winds and ITCZ (Te Intertropical Convergence Zone) activities.Te winter monsoon transports lowtemperature air to the East Sea in combination with the ITCZ with active convection zone moving to lower latitudes (near and over the Spratly Islands region) in the early winter months.Tis results in high relative humidity within the air mass during the winter (Figure 8).
In the free atmosphere layer, RHm is higher in summer and autumn than in winter and spring.Te amplitude of RHm variation with height in summer and autumn is smaller than that in winter and spring.Te annual variation amplitude of RHm at the same altitude reaches a maximum value of 29.2% (Figure 8), which is smaller than that in the Paracel Islands region.Te RHm values in summer and autumn do not signifcantly change with height in comparison to that in winter and spring.Tis feature may be related to the important role of strong convective activity in carrying moist low-level air to higher atmospheric layers during summer and autumn.

Te Diference between the Temperature and Humidity Fields of Maritime Air Mass and the Continental Air Mass at the Same Latitude.
To highlight the distinctive features of temperature and relative humidity variations over a maritime air mass, we compared the variations of the variables over the Paracel Islands region and that over the land area (13 °N-18 °N, 101 °E-106 °E).Te land area region is chosen at the same latitudes as the maritime (the Paracel Islands) region.Based on surface station temperature statistics, the hottest temperatures over the land region occur in late spring to early summer [20].Te three-month period of April, May, and June (AMJ) is, therefore, selected to compute the mean temperature during the hot months.Te winter period (DJF) is chosen to investigate the temperatures of the two air masses during the coldest months.
Figure 9 represents the average temperature profles in April, May, and June (AMJ) and during winter over the land and the Paracel Islands areas.It shows that the average temperature profles over the land area and the Paracel Islands area did not exhibit signifcant diferences at higher levels (above 3 km).However, at lower levels (below 2 km) during the hot months (AMJ), the mean temperature of the land air mass is 1 °C to 2 °C higher than that in the maritime air mass due to heating over the land (Figure 9).Te heating over land during the hot months can be observed more distinctly in Figure 10. Figure 10 displays the long-term (30 years) mean of 2 m air temperature and 10 m wind vectors from the ERA5 data.In April, the continental air mass exhibits noticeably higher air temperatures than the air mass over the Paracel Islands region (Figures 1 and 10).During the winter months, because the oceanic air mass is signifcantly infuenced by cold air fow from the winter monsoon (Figure 11), the air temperature at lower levels of the land air mass remains 1 °C to 2 °C warmer than in the maritime air mass (Figures 1, 9, and 11).
Figure 12 represents the average relative humidity profles in summer and winter over the land and the Paracel Islands areas.Te results show that the amplitude of the annual relative humidity variation (summer RHm-winter RHm) is approximately 33.9% in the land air mass, which is much higher than in the maritime air mass (Figure 12).
In summer, the mean RHm values for both regions are high (Figure 12).Below 0.8 km, the RHm value of the maritime air mass is higher than that of the land air mass.Te summer mean RHm of the air mass over the sea is higher in the lower part of the boundary layer, which may be associated with the role of turbulent motions that transport abundant moisture from the sea into the atmospheric boundary layer.However, in the free atmosphere layer, the summer mean RHm of the air mass over the land is higher than that over the Paracel Islands region (Figure 12).Te reason may be that during the summer season, the southwest monsoon wind transports a moist air mass from the Indian Ocean to the mainland region [21].Under the infuence of local factors and topography, convection develops strongly, transporting moisture to higher levels in the free

Figure 1 :
Figure 1: Te study area and location radiosonde monitoring stations.

Figure 2 :
Figure 2: Mean error and standard deviation of temperature error calculated between wetPf2 data and radiosonde data at various altitudes.(a-i)correspond to data groups GP1, GP2, GP3, GP4, GP5, GP6, GP7, GP8, and GP9, respectively.Te vertical axis denotes the barometric level, the lower horizontal axis indicates the temperature error value, and the upper horizontal axis represents the number of samples.Te black solid line is the mean error value, the red dashed line illustrates the standard deviation of the error, and the blue solid (+) line indicates the number of samples.

Figure 3 :
Figure 3: Te mean error and standard deviation of the error in relative humidity between wetPf2 data and radiosonde data calculated for nine data groups at corresponding altitude levels for: (a) the mean error of relative humidity and (b) the standard deviation of the error of relative humidity.

Figure 6 :Figure 7 :
Figure 6: Vertical profle of temperature diference between two air masses over Paracel and Spratly Islands regions in summer (red) and winter (blue).

Table 1 :
Data groups for pair comparison.

Table 2 :
Correlation coefcients of temperature between wetPf2 data and radiosonde data at each corresponding altitude level for each data group.

Table 3 :
Correlation coefcients of relative humidity between wetPf2 data and radiosonde data at each altitude level for the 9 data groups.Figure4illustrates the vertical profles of the seasonal mean temperature (Tm) for the four seasons based on 6215 profles of data in the Paracel region (13 °N-18 °N, 110 °E-115 °E).Te results clearly depict the annual and vertical variations of the air mass temperature over the area.Te temperature decreases gradually from the surface to the top of the troposphere (from 16.8 km to 17.5 km).Te annual variation of Tm at each level follows a distinct pattern, with the highest values observed in summer and the lowest in winter.Te Tm profle in the layer from 0 km to 3.5 km demonstrates relatively signifcant seasonal variations, with the amplitude of Tm variation ranging from 1.3 °C to 5.0 °C.Near the surface, Tm values are 24.0 °C in winter and 28.5 °C in summer.Te large seasonal variation of Tm in lower levels may relate to the strong infuence of the winter monsoon fow in this region.During