*5.9. Preliminary Validation*

The calibration coefficients were derived from the irradiance-based method and applied to the SPARK-01 image. Then, the ground reflectance from the 7 March 2017 SPARK-01 image was calculated using measured atmospheric parameters. The retrieved desert reflectance is compared to the in-situ measured reflectance in Figure 8 (Figure 34). Retrieved values are close to the measured values; the discrepancy is within 8% in 500–1000 nm. The difference is partly attributed to the radiometric calibration and partly attributed to the slight terrain fluctuation, inhomogeneous surface and BRDF effect of the desert.

**Figure 34.** Comparison of in-situ desert reflectance measurements with retrieved reflectance from the SPARK-01 image acquired on 7 March 2017.

Terra MODIS images acquired on 28 February and 7 March 2017 were also used to verify the vicarious calibration methods. The processes used to predict the TOA radiance were similar to those used in the SPARK calibration. Because MODIS has an on-board calibration system, and thus, its calibration accuracy is expected within 3% [44], the MODIS image radiance was taken as a reference to calculate the relative accuracy of the TOA radiance predicted using the vicarious calibration methods (Table 8). For the MODIS image acquired on 28 February, the improved irradiance-based method appears superior to the reflectance-based method in the infrared and shortwave infrared spectral bands, but inferior in the third and fourth bands. However, the weather on 28 February was poor, and thus this comparison shows only that the improved irradiance-based method may be appropriate during non-ideal conditions. On 7 March, the atmosphere was stable and aerosol burden was low. Thus, both the reflectance- and irradiance-based methods predicted values approaching those from the MODIS image, with no more than 4% error in the first four bands. In the 1.2 and 1.6 μm bands, the difference between MODIS and the reflectance-based method is larger than that between MODIS and the irradiance-based method, which is likely due to the aerosol-type assumption. The irradiance-based method shows large difference from MODIS in the 2.1 μm bands, which may be attributed to the low signal-to-noise ratio of the instrument in this band. Comparison of TOA radiances predicted by MODIS and the vicarious calibration methods used for the SPARK satellites show that the SPARK calibration methods achieved the accuracy expected.


**Table 8.** Differences between the TOA radiance predicted by the vicarious methods (i.e., the Reflectance-, Irradiance-, and Improved irradiance-based methods) and that from MODIS image radiance.
