**4. Results**

#### *4.1. Relative Radiometric Calibration*

The relative radiometric calibration coefficients were derived for the SPARK-01 and -02 satellites. The results in terms of blue, green, and red bands are shown in Figures 18 and 19. In total, 32 sub-regions were evident in the SPARK-01 and -02 dark current curves; this number coincides with the design, which features 32 electrical outputs. These coefficients were applied to SPARK images acquired over the calibration site in on 28 February and 7 March 2017. The non-uniformities and variations were largely eliminated after relative radiometric correction using the row-averaged curves (Figure 20).

**Figure 18.** *Cont.*

**Figure 18.** Relative radiometric correction coefficients for SPARK-01 satellite images at 650.4 nm, 551.5 nm, and 461.7 nm showing (**a**) gain curves (non-uniform correction coefficients) and (**b**) offset curves (dark current).

**Figure 19.** Relative radiometric correction coefficients for SPARK-02 satellite images at 638.0 nm, 549.5 nm, and 459.0 nm showing (**a**) gain curves (non-uniform correction coefficients) and (**b**) offset curves (dark current).

**Figure 20.** *Cont.*

**Figure 20.** Row-averaged values at 650.4 nm, 551.5 nm, and 461.7 nm from images acquired over the calibration site on 7 March and 28 February 2017 using SPARK-01 (**a**) before and (**b**) after relative calibration, and at 638.0 nm, 549.5 nm, and 459.0 nm using SPARK-02 (**c**) before and (**d**) after relative calibration.

#### *4.2. Absolute Radiometric Calibrations*

The MODTRAN-simulated radiance calculated using both the reflectance- and irradiance-based methods is shown in Figures 21 and 22, respectively, for the SPARK-01 and -02 satellites. The absolute radiometric calibration is simple to derive by dividing the radiance from the 6 × 6 averaged DN values. The difference between the results from reflectance- and irradiance-based methods does not exceed 6% for the SPARK-01 satellite and shows evident discrepancies in spectral bands <600 nm. However, the differences between the reflectance- and irradiance-based results are greater than 9% for the SPARK-02 satellite in spectral bands <500 nm. These large discrepancies are caused partially by the relatively large AOT (AOT at 550 nm = 0.35) and partially by the unstable weather conditions on 28 February 2017. In comparison, the improved irradiance-based method, which uses only the downward diffuse-to-global irradiance ratio, derived approximately the same radiance for the SPARK-02 on 28 February 2017 as did the reflectance-based method.

**Figure 21.** MODTRAN-simulated radiance for the SPARK-01 satellite calculated using both reflectanceand irradiance-based methods.

**Figure 22.** MODTRAN-simulated radiance for the SPARK-02 satellite calculated using reflectance-, irradiance-, and improved irradiance-based methods.
