**4. Validation**

#### *4.1. Validation of LANDSAT Visible Band Albedo Conversion Algorithm*

We used the spectrometer data to validate the Landsat albedo conversion algorithm (Equation (2)). As shown in Figure 6, the visible band albedo from the Landsat algorithm was highly correlated with the spectrometer measurement, with a linear correlation coefficient greater than 0.99, and a *p* value of less than 0.001. The mean bias error (Landsat minus spectrometer) was 0.01 under clear sky conditions. The linear correlation and the slope of the regression for the Yale Playground was slightly lower if the Landsat algorithm was applied to the band reference values obtained under overcast sky conditions (Figure 6d), but this was not a surprise because the Landsat algorithm was intended for clear skies.

The red, green, and blue bands defined for the ground targets were 620–670 nm, 540–560 nm, and 460–480 nm, respectively, in order to match those of the camera spectral sensitivity. These bands do not correspond precisely to the Landsat bands. Figure 6 shows that despite the slight mismatches, the Landsat algorithm can be used to convert camera-acquired reflectance to albedo.

**Figure 6.** Comparison of visible band albedo measured with the spectrometer and derived with the Landsat conversion algorithm for the Brooksvale Park (**<sup>a</sup>**,**b**) and the Yale Playground (**<sup>c</sup>**,**d**). Panels (**<sup>a</sup>**,**<sup>c</sup>**) are for clear sky conditions and (**b**,**d**) are for overcast sky conditions.

## *4.2. Landscape Albedo Validation*

For the Brooksvale Park, the drone-derived visible band albedo is 0.086 ± 0.110. For comparison, the Landsat visible band albedo is much lower, at 0.054 ± 0.0118. For the Yale Playground, the landscape visible band albedo was 0.037 ± 0.063 and 0.047 ± 0.012 according to the drone measurement and the Landsat measurement, respectively. Once again, because the standard deviations were computed from the individual pixel values in the scene for both the drone and the Landsat data, they indicated spatial variations of the albedo in the landscape, rather than uncertainties of estimation. The values of drone- and Landsat-derived visible band albedo for the Yale Playground were in much better agreemen<sup>t</sup> than those for the Brooksvale Park. We suggested that matching of sky conditions is the dominant factor for the different accuracies, as explained in the next section.

The Landsat-derived shortwave albedo is 0.103 ± 0.019 for the Brooksvale Park and 0.128 ± 0.013 for the Yale Playground. Compared with the landscape visible band albedo, the drone-derived shortwave albedo values (0.332 for Brooksvale Park and 0.054 to 0.061 for Yale Playground) are quite different from these reference values.
