**2.1 Methods and data**

Air density is derivable from air pressure, temperature, and humidity [9, 10]:

$$\rho\_a = P/[\mathcal{R}\_d T(\mathbf{1} + \mathbf{0.608q})] \tag{2}$$

where *P* is air pressure (*Pa*), *Rd* is dry air gas constant (�287 J/kg/K),*T* is absolute air temperature (K), and *q* is specific humidity (g/g). To apply Eq. (2) to near-surface level, atmospheric fields of pressure (*Ps*), temperature (*Ts*), and specific humidity *qs* at ground level (subscript "*s*" means surface) are required. These parameters fortunately are primary outputs from the coupled model intercomparison project (CMIP, e.g., https://cmip.ucar.edu/; Ref. [11]). The monthly climate model outputs are obtained from the IPCC Deutsches

Klimarechenzentrum (DKRZ) Data Distribution Centre (http://www.ipcc-data.org/ sim/gcm\_monthly/AR5/Reference-Archive.html). For models providing multiple perturbation runs, only *r*1*i*1*p*1 runs are used. To examine whether the historical runs from the climate models are close to reality in their simulated air density, NCEP/NCAR reanalyses are used as observations. The monthly NCEP/NCAR reanalysis [12] data are obtained from the Earth System Research Laboratory website: http://www.esrl.noaa.gov/psd/data/gridded/data.ncep.reanalysis.pressure. html. Specific humidity provided by reanalyses can be converted into specific humidity before applying Eq. (2).

(H2O has smaller molar mass than N2 and O2, the dominant constituents of dry air) is another effective way of reducing air density. Although from Clausius-Clapeyron equation [13] warm air has more capacity of holding moisture, it still is debatable whether earth atmosphere actually gains mass, because the hydrological cycle also tends to intensify [14–16], through facilitating interhemispherical moisture exchange [17] and destabilizing local stratification profile [15, 16]. If precipitation increases more than evaporation, there still is a net mass loss for atmosphere. Interestingly, all existing reanalysis datasets show no statistically significant changes in global total air mass during their respective reanalyses period. This implies that the net water vapor input into atmosphere is globally delicately

Applying the formula as in Eq. (2) to climate model-simulated (under RCP 8.5, a strong emission scenario) near-surface pressure, temperature, and humidity, near-surface air density is estimated over the globe. The same formula also is used on the NCEP/NCAR reanalysis data. Density variations over 1900–2100 for six global airports are shown in **Figure 1**, as representatives. All 27 climate models unanimously indicate that all six locations experienced salient density decreases. Significant inter-model spread exists but started well before the year 1900 and should be ascribed to model systematic biases/drifts. For each climate model, the amount of density decrease easily exceeds the natural interannual variability mag-

nitude. Geographically, high-latitude regions (e.g., Moscow) have larger

balanced between geographic regions.

*Climate Warming and Effects on Aviation DOI: http://dx.doi.org/10.5772/intechopen.86871*

**Figure 2.**

**179**

*GFDL-CM3-simulated near-surface air density (kg/m<sup>3</sup>*

*shown in (c), with corresponding percentage changes shown in (d).*

*(2081–2100) (b), under RCP 8.5 scenario assumption. The density differences between these two periods are*

*) averaged over two periods: (2005–2025) (a) and*
