Continuous observation of Stable Isotopes of Water 1 Vapor in Atmosphere Using High-Resolution FTIR 2

Vapor in Atmosphere Using High-Resolution FTIR 2 Chang-gong Shan 1, , Wei Wang , Cheng Liu , You-wen Sun, Yuan Tian, Isamu 3 Morino 4 School of Environment science and Optoelectronic Technology, University of 5 Science and Technology of China, Hefei, 230000, China 6 2 Key Laboratory of Environmental Optics and Technology, Anhui Institute of Optics 7 and Fine Mechanics, Chinese Academy of Sciences, Hefei, 230031, China 8 3 University of Science and Technology of China, Hefei, 230000, China 9 Center for Excellence in Urban Atmospheric Environment, Institute of Urban 10 Environment, Chinese Academy of Sciences, Xiamen, 361021, China 11 Satellite Observation Center, National Institute for Environmental Studies, Tsukuba, 12 305-8506, Japan 13 Correspondence to: Cheng Liu (chliu81@ustc.edu.cn), 14 Wei Wang (wwang@aiofm.ac.cn) 15 16 Abstract 17 Observations of stable isotopes of water vapor provide important information for water 18


Introduction
Water cycle plays an important role in climate change.Water vapor plays a key role in to study the seasonal and inter-seasonal variations of water cycle (Gribanov et al., 2014).
The variation of atmospheric temperature and humidity near the surface also cause the atmospheric water recycling (Boucher et al., 2004;Destouni et al., 2010;Tuinenburg et al., 2012).Therefore, many studies reported that meteorological parameters at ground level are correlated with the stable isotopologue of water vapor.For example, δD have a positive correlation with temperature and relative humidity of the atmosphere in summer in Mediterranean coastal area (Delattre et al., 2015).Bastrikov (2014) also analyzed the relationship between δD and temperature and humidity in different seasons in West Siberia.However, these reports are based on in-situ measurements, and there are few studies about the relationship between the column averaged HDO/H2O Geosci.Instrum.Method.Data Syst.Discuss., https://doi.org/10.5194/gi-2018-43Manuscript under review for journal Geosci.Instrum.Method.Data Syst.Discussion started: 3 December 2018 c Author(s) 2018.CC BY 4.0 License.ratio δD and the meteorological parameters.
Ground-based FTIR technique is widely used to obtain long-term time series of atmospheric composition and validate satellite data (Schneising et al., 2012;Scheepmaker et al., 2015).High-resolution FTIR observations have achieved accurate detection of greenhouse and trace gases (Washenfelder et al., 2006).The Total Carbon Column Observing Network (TCCON) and the Network for the Detection of Atmospheric Composition Change (NDACC) use high-resolution FTIR instrument to accurately and precisely derive the main stable isotopologue of water vapor, HDO (Hannigan et al., 2009;Wunch et al., 2011).The total column of HDO and H2O are retrieved in the near infrared region, and the column averaged HDO/H2O ratio are calculated.Also, the Column averaged HDO derived from the high-resolution FTIR instrument have been used for comparison with model simulations and satellite data (Boesch et al., 2013;Frankenberg et al., 2013;Rokotyan et al., 2014;Dupuy et al., 2016).
Water isotopologues composition has been analyzed in Hefei with an obvious seasonal variation, only at the month scale, using in situ measurements (Wang et al., 2012).
However, so far no research has been dedicated to the water vapor and its isotopologues variation in a large spatial-temporal scale at Hefei.To better understand evapotranspiration, process and the relationship between meteorological parameters and water vapor isotopologues, the column stable isotopologues of water vapor observed by ground-based FTIR technique are presented in the paper.
The instrumentation and retrieval strategy for column averaged H2O and HDO at Hefei site are described in Section 2. The retrieval results are discussed in Section 3, also, the relationships between the isotopic composition δD and temperature, relative humidity are analyzed.Moreover, the evapotranspiration signature δET and the sources of water vapor based on the back trajectories calculation of air masses are clarified in this Section.The conclusions are given in Section 4.

Instrumentation and retrieval strategy
The ground-based high-resolution FTIR spectrometer (Bruker IFS 125 HR) and solar tracker (A547) installed on the roof of laboratory, are combined to collect the solar absorption spectra at Hefei site.Hefei (31.9︒N, 117.17︒E, about 30 m above the sea level) is a continental site, away from the southeast urban area about 10 km (Figure 1).The CaF2 beamsplitter and InGaAs detector are used to collect the near-infrared (NIR) spectra.The NIR spectral range covers 4000-11000cm -1 , and the spectral resolution is 0.02 cm -1 , corresponding to a 45 cm maximum optical path.In order to ensure the stability of the measurement, the instrument is vacuated under 10 hPa.A weather station is installed near the solar tracker on the roof of the lab building to record meteorological data.Wang (2017) described the instrumentation and the measurement routine at Hefei site.
The solar spectra collected from September 2015 to September 2016 are analyzed.We use the GGG2014 software package to retrieve the water vapor and its isotopes (Wunch et al., 2015).GGG2014 is a nonlinear least square spectral fitting algorithum (GFIT), which scales an a priori profile derived from the National Centers for Environmental Prediction and the National Center for Atmospheric Research (NCEP/NCAR) reanalysis data (Toon et al., 2014) to minimize residulas between measured and simukated spectra.GGG2014 produces the total column of trace gases, then the column-averaged dry-air mole fractions (DMF) of trace gasees are computed as: (2) The column of dry air, units of molecules/cm 2 , is computed from the oxygen (O2) column (Wunch et al, 2011) dividing by 0.2095.Figure 2 depicts the spectral fitting of the H2O and HDO in the spectral window of 4565-6470 and 4054-6400 cm -1 , respectively.The rms spectral fitting residuals are 0.16% and 0.25% for H2O and HDO respectively.Table 1 lists the spectral windows for column retrievals of H2O and HDO, which are the standard GFIT windows.Figure 3 shows the column averaging kernals of H2O and HDO.The difference of the column averaging kernals below 500 hPa between them is very small, with the value of 4.34%.

Time series of δD, water vapor and meteorological parameters
The DMFs of H2O and HDO are calculated using total columns of H2O and HDO based on equation ( 2).The δD time series at Hefei station is plotted in Figure 4

Comparison with nearby TCCON observations and satellite data
The time series of XH2O are compared with the GOSAT data (v02.72)from September 2015 to September 2016.For co-locating the GOSAT data with the ground-based FTS data, the GOSAT observations of ±5° latitude and longitude centered in the Hefei site within ± 2 hour overpass were selected (Kuze et al., 2009;Yoshida et al., 2013;Scheepmaker et al., 2015).In order to eliminate the influence of different a priori 2011; Zhou et al., 2016).The comparison results of XH2O are depicted in Figure 6.The mean bias, which is defined as the mean difference of XH2O between FTIR and satellite date, is about 11.98ppm.The XH2O observed by FTIR showed a similar variation trend with the corrected satellite data, and the variation range agrees with that of GOSAT data.
Since water vapor mainly concentrate in the lower troposphere, and the ground-based observations have high sensitivity near surface, but the satellite data are insensitive in the lower troposphere, so the FTIR data are slightly higher than the satellite data.Also, we calculated the correlation between FTIR and GOSAT data, and there is a high correlation between FTIR and GOSAT data (R = 0.98).The correlation coefficients between FTIR and GOSAT data are 0.95 and 0.93 for Japanese Tsukuba and Saga site, respectively (Dupuy et al.;2016).The slope of the scatter plot of our FTIR and GOSAT data is 0.98.It is concluded that FTIR data at Hefei site agree well with the satellite observations.
Furthermore, to verify the accuracy of our calculated data, we compare the isotopic ratios δD from Tsukuba TCCON station (Morino et al., 2014) with our δD values.
Tsukuba TCCON station (36.05°N, 140.12°E, 31m above the sea level) is a Japanese TCCON station close to our site and at a similar latitude (Figure 1).

Relationship of stable isotopes of water vapor with meteorological parameters
Atmospheric circulation strongly affects the variations of stable isotopic compositions of water vapor in the atmosphere (Deshpande et al., 2010;Guan et al., 2013).The spatiotemporal distribution of water vapor in the atmosphere is strongly correlated with the weather, and the stable isotopic ratios of water vapor change with the meteorological parameters (Noone et al., 2012, Vogelmann et al., 2015).The surface meteorological data are important for quantifying the distributions of the stable isotopes of water vapor.
The statistical data of monthly averaged δD and surface temperature are summarized in in December 2015 and February 2016, respectively, while the corresponding value is 6.3 and 8℃ in July and August, respectively.It is noted that the correlation coefficient between monthly variation amplitude of δD and temperature is 0.95.So it is concluded that the surface temperature strongly influences the variation of δD in Hefei site.
For all the data collected, the linear relationship of individual δD and the surface temperature is expressed as δD=5.30‰T-242.64‰.The correlation coefficient is 0.83 between δD and temperature at Hefei site, as shown in Figure 8(a).Bastrikov (2014) and Bonne (2014) found that there was a positive correlation between the stable isotopes of water vapor and temperature in western Siberia and southern Greenland.In Bastrikov (2014), the slope of δD and temperature in western Siberia is 3.1‰℃ -1 .The evaporation of water vapor weakens with the decrease of temperature, and heavier isotopologue, HDO, condenses more actively and evaporate less actively than the main isotopologue H2O due to their different saturation vapor pressure, so the depletion in heavy isotopes with decreasing temperature happens.A simple distillation model, Rayleigh distillation model, helps to understand the relationship between δD and H2O (Schneider et al., 2010).The variation of water vapor and δD are connected via the equation In which δD0 and  2  0 are the deuterium and water vapor of the airmass from the ocean, while α represents the fractionation coefficient between the oceanic source and the sampling site.
There is a linear relationship between ln(δD/1000+1) and ln(XH2O) , according to the equation ( 3).The slope of ln(δD/1000+1) and ln(XH2O) represents a measure of the transport pathway of water vapor.Analysis of the slope allows investigating the importance of different hydrological processes (Worden et al., 2007;Schneider et al., 2010).As shown in Figure 8(c), there is a strong correlation (R=0.88) between ln(δD/1000+1) and ln(XH2O), and the slope of linear regression is 0.081.The results prove that the stable isotopes of water vapor are highly correlated with the fraction of water remaining in the cloud.In western Siberia, the correlation coefficient of linear regression between ln(δD/1000+1)/ln(XH2O) is 0.71, and the slope of linear regression is 0.07 (Gribanov et al, 2014).

Variation sources of regional δD in Hefei
The NOAA Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model is a complete system using NCEP/NCAR reanalysis data to understand transport paths and sources of air masses (Draxler and Rolph, 2003;Stein et al., 2015).The of air parcels (Li et al., 2012).The back trajectories of 72 hours are calculated for each day, and the height of the backward trajectories is set as 500 magl.The geographic region precision is selected as 0.5°× 0.5° grid cells in the calculation.The PSCF calculated by the backward trajectories is weighted according to the method of Polissar et al. (1999) to identify the source strength (WPSCF).
Figure 9 shows the cluster analysis results and the WPSCF distribution of δD during the period from September 2015 to August 2016.The sources of air masses of Hefei area mainly originated from three regions: the Southeast China (SEC), North of China (NC) and Northwest of China (NWC).51.35% of airmass were from SEC during the observation period.Also, The WPSCF analysis indicates that the main potential sources of δD are near the Hefei site.The potential source of δD are divided into three regions: the east area with moist and warm airmass, the north area with dry and cold airmass, and the southwest area with moist and warm airmass.Especially the main airmass from the east area bring the moist and warm airmass into Hefei, which result in the enrichment of heavy isotopes.

δ-value of evapotranspiration
Keeling plot is usually applied to estimate the δ-value of evapotranspiration (Keeling et al., 1958;Wei et al., 2015).The Keeling equation assumes that the actual atmospheric water vapor is the mixing of the atmospheric background and an additional component from local evapotranspiration, and each component has distinct isotopic signature.The water vapor and its isotopes in the atmosphere can be written as (Yepez et al., 2003;Williams et al., 2004;Sun et al., 2005) Where   and   are DMF and δ-value of the water vapor, respectively.  and   are DMF and δ-value of the background, respectively.  is the δ-value of evapotranspiration.Therefore, the evapotranspiration signature (  ) is also expressed as the y-axis intercept of equation ( 4).
Keeling plot is used to calculate the δ-value of the evapotranspiration of water vapor.
The days with 4-hour continuous observations are considered to ensure that the data are keeling plot analysis during the measurement period are shown in Figure 11.Over the period, δD value of evapotranspiration varied from (15.3 ± 2.9) ‰ to (-114 ± 8.9) ‰, and the averaged δD value of evapotranspiration is -44.43 ‰.It is seen that the variation range ofδD value for evapotranspiration was large, reflecting the fact that the source isotopic signal did not keep constant over the measurement period.In the study of Wang (2012), the deuterium isotopic signature from evapotranspiration is between -113.93 ± 10.25 ‰ and -245.63 ± 17.61 ‰ in July in Hefei. Griffith (2006) found that the deuterium isotopic ratio from evapotranspiration is between -90 ‰ and -100 ‰ in a pasture.

Conclusions
The DMFs of H2O and HDO were retrieved from the spectra observed by the groundbased high resolution FTIR at Hefei site.Time series of XH2O were compared with GOSAT data.The mean relative bias was 2.85% and the correlation coefficient is 0.98 between FTIR and satellite date, showing a good agreement.XHDO/XH2O ratio expressed as δD are calculated.δD from nearby Tsukuba station with similar latitude are used to verify the accuracy of our data.It is found that the δD in Hefei showed a same trend as that in Tsukuba, with the maximum value in summer and minimum in winter.Variation Further, we used the NOAA HYSPLIT model to calculate the back trajectories of air parcels in Hefei, and performed the cluster analysis and PSCF analysis.The results of cluster and PSCF analysis showed the sources of δD and their potential contributions are mainly from the surrounding area of Hefei site and especially in the east area.
Also, the δD value of evapotranspiration is calculated based on Keeling plot analysis.
from September 2015 to September 2016.The precision of δD (1-σ precision divided by the measured value) is about 3.63%.The daily averaged δD varies from -17.02‰ to -282.3‰.δD shows an obvious seasonal variation over the observed period, with the lowest δD values occurring in mid-January and the peak in early August.The time series of XH2O and meteorological parameters from September 2015 to September 2016 at Hefei station are plotted in Figure 5.The mean relative retrieval error (1-σ precision divided by the measured value) of XH2O is about 1.11%.The variations of XH2O are similar to those of δD, with an obvious seasonal pattern.The variation of XH2O is large during the period.The daily averaged XH2O was in the peak of 8821.97 ppm in early August in summer and reduced to the minimum of 225 ppm in mid-January in winter.The variation of surface temperature is close to XH2O variation, while the relative humidity of atmosphere shows a weak seasonal variation.The peak and valley values of water vapor and δD seem to accompany with those of temperature, and the different amplitude of daily variation of δD in different seasons depends on temperature, therefore, the relationships of water vapor and δD with temperature are discussed in sec.4.2.
profiles and averaging kernels on XH2O, we use a priori profile of the ground-based FTS to correct the column-averaged mole fractions of gases from GOSAT(Reuter et  al., Geosci.Instrum.Method.Data Syst.Discuss., https://doi.org/10.5194/gi-2018-43Manuscript under review for journal Geosci.Instrum.Method.Data Syst.Discussion started: 3 December 2018 c Author(s) 2018.CC BY 4.0 License.
Figure 7 is the plot of δD in Hefei compared to those of Tsukuba from September 2015 to February 2016.It is found that the δD in Hefei showed a similar trend as that in Tsukuba, both with the maximum value in summer and the minimum in winter.During the observation period, the δD of the two sites began to fall from October 2015 and to the valley value in January 2016.Hefei and Tsukuba sites have a similar atmosphere circulation pattern due to the similar latitude, which may result in the similar variation in the stable isotopes of water vapor in the atmosphere, as shown in Figure7.However, the daily averaged δD of Hefei ranges from -36.46‰ to -282.3‰ during this period, while δD in Tsukuba is from -35.74‰ to -198.37‰, falling in the range of our δD.Scheepmaker (2015) plots the time series of δD in six TCCON stations, and the δD observed from these stations in the Northern hemisphere are in the range from about -50‰ to -300‰, which are comparable to those of our results.Geosci.Instrum.Method.Data Syst.Discuss., https://doi.org/10.5194/gi-2018-43Manuscript under review for journal Geosci.Instrum.Method.Data Syst.Discussion started: 3 December 2018 c Author(s) 2018.CC BY 4.0 License.
Geosci.Instrum.Method.Data Syst.Discuss., https://doi.org/10.5194/gi-2018-43Manuscript under review for journal Geosci.Instrum.Method.Data Syst.Discussion started: 3 December 2018 c Author(s) 2018.CC BY 4.0 License.δD of atmosphere in Hefei show a weak correlation with relative humidity, as plotted in Figure 8(b).The correlation coefficient of linear regression between δD and relative humidity is 0.45, and the slope of linear regression is 2.11‰% -1 .Wen (2010) reported that the stable isotopes of water vapor in Beijing is positively correlated with the relative humidity (R = 0.42), while the diurnal and seasonal variation of δD have a strong relationship with the relative humidity in northwest Greenland (Steen-Larsen et al., 2013).

Figure 6 :
Figure 6: The scatter plot of XH2O at Hefei site and the coincident GOSAT data

Figure 11 :
Figure 11: δD values of evapotranspiration during the measurement period.The error bars are standard deviations of value

Table 1 .
The monthly averaged surface temperature decreased from 30.18 to 4.74 ℃ between Sep.2015 and Jan.2016, and the variation of δD also dropped from -126.89‰ to -257.86‰ at the same time.Especially, the daily averaged δD reached the minimum of -282.3‰ in 25 January 2016, which is the coldest day during the period.Also, δD shows a large variation in winter, with the monthly variation amplitude of 186.38‰ and 213.66‰ in December 2015 and February 2016, respectively.However, the monthly variation amplitude of δD in summer is about one third of the corresponding values in winter.Furthermore, the monthly variation amplitude of temperature is 14.1 and 19.2℃

Table 1 .
The FTIR technique offers a new opportunity to monitor the stable isotopes of water vapor.The long time series of the stable isotopes of water vapor provide a basis of revealing the water cycle of the atmosphere.The further research work will focus on accurate retrieval of H2 18 O from solar absorption spectra, and can clearly clarify the water cycle in combination with HDO.The statistics of monthly averaged δD and surface temperature.