Factors contributing to spatial–temporal variations of observed oxygen concentration over the Qinghai-Tibetan Plateau

Oxygen (O2) is the most abundant molecule in the atmosphere after nitrogen. Previous studies have documented that oxygen concentration remains nearly constant (20.946%) at all altitudes. Here we show for the first time that oxygen concentration varies significantly from earlier consensus and shows strong spatial and seasonal differences. Field observations on the Qinghai-Tibetan Plateau (QTP) indicate oxygen concentration of 19.94–20.66% (2018, n = 80), 19.98–20.78% (2019, n = 166) and 19.97–20.73% (2020, n = 176), all statistically different from earlier reports (p < 0.001) and are lower than the nearly constant. The mean oxygen concentration in summer (20.47%) is 0.31% higher than that of winter (20.16%) (n = 53) at identical locations in 2019, sampled in the Qilian Mountains, northwest QTP. We used LMG (The Lindeman, Merenda and Gold) method to estimate the relative contributions of altitude, air temperature and vegetation index (Fractional Vegetation Cover, FVC and Leaf Area Index, LAI) to oxygen concentration, which are 47%, 32% and 3% (FVC, R2 = 82%); 45%, 30% and 7% (LAI, R2 = 82%), respectively. These findings provide a new perspective for in-depth understanding on population risk in high altitude regions in the context of global climate change, to ensure the health and safety of residents and tourists in high altitude regions and promoting the stability, prosperity and sustainable development of high-altitude regions worldwide.


Variations of oxygen concentration over the QTP.
Our results indicate significant difference of oxygen concentrations from the constant assumed (20.946%) 11 . In our sample by electrochemical oxygen meter, oxygen concentrations range from 19.94 to 20.78% and are lower than the nearly constant ( Table 1). The oxygen concentrations sampled in the margin areas of QTP (which locates in/near Tarim Basin, Qinghai Lake, Qilian   Table S1, Table S2). Moreover, the distribution of oxygen concentration for each field investigation also has high variation ( Fig. 3) as the samples were taken along with different investigation routes and landforms over the QTP.
Decreasing oxygen concentration by altitude. Oxygen concentration shows pronounced linear relationship with altitude, and the slope is negative and significant (Fig. 4, n = 369, R 2 = 0.7367, p < 0.001). For each 1000 m increase in altitude, oxygen concentration is going to drop by 0.15% in summer.
Influence on oxygen concentration from air temperature and vegetation. Air temperature is directly proportional to oxygen concentration (Fig. 5, n = 369, R 2 = 0.5661, p < 0.001), for all the samples in 2018-2020 (winter samples excluded). There are also strong seasonal variations in oxygen concentration. Data from 53 paired-samples (winter vs. summer), each pair at identical locations in Qilian Mountains, northeast of the QTP, indicate oxygen concentrations in summer (n = 53, mean: 20.47%) are statistically higher than those in winter (n = 53, mean: 20.16%) (Table S3). A continuous fixed-point observation at Fangshan Station, Beijing indicates strong seasonal and diurnal variation in oxygen concentration ( Figure S1). For monthly means, the highest oxygen concentration (21.31%) appeared in June 2019 ( Figure S2, Table S4), and the lowest appeared in January 2020 (20.29%), which indicated a month-scale range of 1.02%. The daily variation of oxygen concentration has with a minimum value between 4:00 to 7:00 h (Beijing Time, hereinafter inclusive) (mean: 20.67%) and a maximum value between 13:00 and 15:00 h (mean: 21.14%), with an average diurnal range of 0.47%. Overall, the lowest oxygen concentration (19.69%) appeared at 7:43 on December 31, 2019, and the highest oxygen concentration (21.64%) appeared at 15:28 on June 14, 2019 ( Figure S3).
Oxygen is mainly produced from the photosynthesis of green plants, the process by which plants use sunlight energy to produce organic compounds from CO 2 and water. The ability of an ecosystem to sequester CO 2 and produce O 2 is largely determined by its vegetation conditions 17,18 . During the fieldwork on the QTP, we took 33   www.nature.com/scientificreports/ survey sample area of 1 km × 1 km to estimate local FVC, which was then compared and examined to be well fitted with the estimated FVC and LAI retrieved from remote sensing data (Table S5, Figures S4 and S5). The FVCs had significantly positive correlations with oxygen concentration ( Figure S5). The LAI, which was validated with local FVC, also had a significantly positive correlation with oxygen concentration ( Fig. 6 and Fig. S4). There was a distinctive disparity in the FVCs between winter and summer in the Qilian Mountains, which intertwined   Factors influencing oxygen concentration and their relative contributions. The LMG method in total explained 82% of the variation in oxygen concentration, among which, altitude contributed to 47%, air temperature 32% and FVC 3% (Table 2). Similarly, the relative contributions of altitude, air temperature and LAI to oxygen concentration are 45%, 30%, and 7%, respectively (R 2 = 82%). Both calculations are based on all the samples except those in winter of 2019 (n = 369).

Discussion
Factors' relative contributions to oxygen concentration. In the study by Shi et al. (2019), the main factors affecting oxygen concentration on the QTP included altitude, 500 hPa air temperature and FVC. 500 hPa air temperature was negatively correlated with oxygen concentration 19 , but our results show a positive correlation between near-surface air temperature and oxygen concentration. Although the two air temperature datasets came from different sources, their relationships with the oxygen concentration should have presented a certain consistency. What caused such a mismatch can be owed to how well the two datasets represented the local conditions. The 500 hPa air temperature, which was derived from ERA-Interim reanalysis temperature data (the European Centre for Medium-Range Weather Forecasts) and has a spatial resolution of 0.25° × 0.25° (roughly 30 km × 30 km), may introduce some errors when applied directly to the sample sites due to its relatively coarse spatial and temporal resolutions 20 . The real-time measured air temperature used in this study should be much more reliable, and the controlled investigations in Qilian Mountains during winter and summer and the observations at Fangshan Station, Beijing further confirmed this positive correlation.
O 2 concentration over the QTP under global climate change. The QTP has been experiencing unprecedented warming under global climate change, largely intensified by human activities 5 . Meteorological records revealed that the warming rate on the QTP was twice of that observed globally over the past five decades 21 , although some studies suggested that the warming hiatus is also occurred over the QTP 22 , induced by internal to decadal climate variations originating from the Pacific and Atlantic. This warming trend is projected to continue in the future, based on the outputs from both global climate models and regional climate models [23][24][25] . Evidence gleaned from satellite observations, long-term ecological stations, and modeling works have all consistently shown that the QTP ecosystems have been greening under climate warming-with a clear north-south disparity 26,27 . To date, the most important limiting factor on the most part of this region for vegetation growth is temperature 28,29 . Vegetation productivity firstly increases and then decreases with temperature rises (the temperature corresponding to the maximum productivity is called optimal temperature, T opt ). Chen et al. (2021) found that a remarkable geographical heterogeneity in T opt was observed over the QTP 30 . Higher T opt values generally appeared in the north-eastern parts, while the south-western QTP had relatively lower T opt (< 10 °C), and the average T opt of non-forest vegetation on the QTP was ~ 14.7 °C. Moreover, the QTP was a source of atmospheric O 2 from 2000 to 2013, although the O 2 consumption showed an increasing trend over the QTP in recent decades 31,32 . Thus, enhanced vegetation growth under a warming climate before a certain threshold will greatly improve the capacity of oxygen production of these ecosystems. A warming climate and favorable vegetation growth conditions on the QTP over the past decades could have led to rising oxygen concentration. While it will weaken the vegetation growth if the air temperature exceeds a certain threshold in the future, even though great uncertainties in climate projections still exist. From a global perspective, much more attention should be paid to the population risks of permanent residents and short-lived visitors to high altitude regions (≥ 2500 m above sea level), which cover a total area of ~ 11 million km 2 , accounting for 7.7% of the Earth's land surface. In 2015, the total population in these areas was ~ 107 million, accounting for 1.5% of the world's total population 33 . It has been estimated that more than 100 million tourists travel to high-altitude regions by the year 2000 34 , and this number keeps increasing in many parts of these regions such as the QTP 16 , Indian Himalayan Region 35 , and the European Alps 36 . Therefore, we suggest the science community and stakeholders pay more attention to this overlooked issue and investigate the spatial and temporal heterogeneity of O 2 by sampled and fixed continuous observations. More studies with fixed continuous www.nature.com/scientificreports/ observations and meteorological, remote sensing, reanalysis, statistical, and clinical data of hypoxia-related fields, will further validate and consolidate these findings at a much more accurate spatial and temporal scale. These works will promote multi-and cross-disciplinary studies on geoscience and high-altitude medicine to improve the health of at-risk populations and ensure the sustainable development of high-altitude regions.

Data and methods
From 2017 to 2020, we conducted several field investigations on the QTP and its vicinities. We collected 487 samples located within 76.9336-104.0639 E and 24.8350-39.7990 N, with altitude varied from 645 to 5238 m above sea level (Fig. 1, Tables S1 and S2). On each site, we record geographical location, altitude, barometric pressure, air temperature and oxygen concentration (by electrochemical oxygen meter). For all sampled sites, each of these parameters was simultaneously measured with three instruments and then averaged for later analysis. 33 fractional vegetation covers (FVCs) were also measured in the field with 1 km × 1 km survey sample area 37 then the ones for other sample sites were interpolated by FVC_GLASS since the linear fit coefficient was the highest ( Figure S4) among FVC and the retrieved FVCs (FVC_GE, FVC_MODIS and FVC_GLASS) (Table S5). Continuous in situ observation is carried out at Fangshan Station, Beijing (39.6914 N and 116.0547 E, altitude: 33 m, Fig. 1) as the control group data, where oxygen concentration is automatically recorded by electrochemical oxygen meter for every ~ 2 min. The boundary data of the QTP is provided by the Resource Environmental Science and Data Center, Institute of Geographical Sciences and Natural Resources Research, Chinese Academy of Sciences (http:// www. resdc. cn). We use altitude (Alt), near-surface temperature (Tem) and fractional vegetation cover (FVC, or Leaf Area Index, LAI) to explain the variation of oxygen concentration via linear regression: where ε is the random error representing factors that are not considered but might contribute to variations in oxygen concentration 38 . To further determine the relative importance of variables, we apply the method developed by Lindemann, Merenda and Gold (LMG) 39 . The LMG method calculates the relative contribution of each variable to the R 2 with the consideration of the sequence of predictors appearing in the model. The relative contributions of the factors (altitude (Alt), air temperature (Tem), FVC/LAI) to oxygen concentration are estimated using a bootstrap technique based on the LMG method.

Data availability
Data used in this study are available from the corresponding author upon request.