*2.4. Atmospheric Data*

Atmospheric measurements acquired during the experimental period include columnar atmospheric parameters (i.e., AOD and total columnar water vapor, or CWV), the diffuse-to-global irradiance ratio, and the radiosonde vertical profile. A CE318 photometer was used to measure the total AOD and CWV. A total of 5 days of valid data were acquired on 25, 26, and 28 February and 4 and 7 March 2017. The Langley calibration method [30] and a modified calibration method were used for non-water and water absorption channels, respectively, to update the calibration coefficients for the sun measurement channels. These calculations were made with the measurements acquired on 7 March 2017, as the atmosphere was stable and aerosol burden was low. Then, the AOD was calculated in each channel using Beer's law and a spectral response function [31]. We used the measured pressure from a barometer and columnar ozone and nitrogen dioxide content from the Ozone Monitoring Instrument (OMI). The CWV was retrieved using a 4-parameter method [32]. These parameters, which were retrieved within the SPARK overpass period, were averaged over 15 minutes and used as inputs in the calibration process. The 550 nm channel AOD was calculated via logarithmic interpolation of the 440 nm and 675 nm channel AODs. The 550 nm channel AOD and CWV are shown in Figure 9 for the SPARK satellite overpass dates; data influenced by clouds were excluded. Stable atmospheric conditions are indicated by the AOD and water vapor content patterns on 7 March 2017.

**Figure 9.** *Cont.*

**Figure 9.** 550 nm AOD (**a**) and CWV (**b**) retrieved from CE318 measurements and Microtops II measurements on 7 March and 28 February 2017, respectively.

Measurements from a Microtops II sunphotometer were used to verify the accuracy of the CE318 observations, as shown in Figure 9. The AOD measurements are more accurate, with differences of less than 0.02 between the two instruments. However, the water vapor content differs greatly between the two instruments. We speculate that the calibration coefficients for the Microtops II 940 nm channel need to be updated. A lack of such updates would introduce additional error in water vapor retrievals. Nevertheless, the CE318 results are expected to be more reliable, as the CE318 automatic operation mode is used extensively worldwide. The 550 nm AOD and CWV were averaged over 15 min intervals within the satellite overpasses on 7 March and 28 February 2017; the average AOD and CWV values are 0.1928 and 0.3513 g/cm<sup>2</sup> for 7 March and 0.3476 and 0.5379 g/cm<sup>2</sup> for 28 February. A rural aerosol type was chosen for use in MODTRAN due to the barren Gobi Desert surroundings. Also, the angstrom exponent coefficients derived from the 440 nm to 675 nm channel AOD measurements are 0.75 and 0.3519 for the SPARK-01 and -02 overpass times, respectively. The ozone density was 299 and 305 DU on 7 March and 28 February 2017, respectively; these values were derived from NASA OMI data [33].

In addition, radiosonde balloons were released during SPARK satellite overpass periods to measure the vertical profiles of atmospheric pressure, temperature, and humidity; balloons were released at 05:49:55 and 05:05:08 UTC on 28 February and 7 March 2017, respectively. Figure 10 shows the vertical profiles measured on each date. The variations in pressure and temperature with altitude are similar between the two dates. The humidity changes little at altitudes greater than 5000 m.

**Figure 10.** *Cont.*

**Figure 10.** Vertical profiles of (**a**) pressure; (**b**) temperature; and (**c**) relative humidity measured using radiosondes released on 7 March and 28 February 2017.

To acquire the diffuse-to-global irradiance ratio data, an irradiance sphere was used with a SVC GER1500 spectrograph at the calibration site to measure the irradiance at ten minutes intervals throughout the day. Each of the measurements outlined below were repeated three consecutive times. The global solar irradiance (L1) was measured fist, followed by the sky diffuse irradiance (L2), which was assessed with a light barrier. Finally, the global solar irradiance (L3) was determined [20]. Figure 11a shows the diffuse-to-global irradiance ratios at 550 nm on the date of the SPARK satellite overpass during the Dunhuang experiment. The smooth diffuse-to-global irradiance ratio curve indicates a very stable atmosphere on 7 March 2017. Figure 11b shows diffuse-to-global irradiance ratio for the entire spectrum at the time of the SPARK satellite overpass; the lower aerosol burden on 7 March 2017 caused lower diffuse-to-global irradiance ratios in comparison to those measured on 28 February 2017. Lastly, the diffuse-to-global irradiance ratios were convolved with the spectral response functions of the corresponding SPARK satellites channels.

**Figure 11.** Diffuse-to-global irradiance ratios measured (**a**) at 550 nm on the date of the SPARK satellite overpass during the Dunhuang experiment, and (**b**) for the entire spectrum at the time of the SPARK satellite overpass.
