Estimation of sunlight conditions through a drone-mounted solar irradiation sensor

Abstract

　Sunlight conditions, such as a clear or cloudy sky, can have an adverse effect on vegetation spectroscopy. Drones with mounted solar irradiation sensors have become available recently. Estimation of sunlight conditions at the observation location by using data from these solar irradiation sensors can help increase the accuracy of spectroscopy measurements. In this study, we estimated sunlight conditions using a drone-mounted solar irradiation sensor by the following two approaches: Method A, in which direct and diffuse solar radiations were separated based on the global solar radiation estimated using the solar irradiation sensor, and Method B, in which direct and diffuse solar radiations were separated using spectroscopic data from the solar irradiation sensor. For Method A, we found that in the near-infrared (NIR) wavelength range, the light was not readily scattered by the atmosphere, allowing for more accurate estimation. The results of the separation of the direct and diffuse solar radiations based on the estimated global solar radiation were compared with the results of actual measurements. The coefficient of determination (R²) and root mean square error (RMSE) were 0.971 and 0.062, respectively; however, the direct component was overestimated. Thus, the use of the empirically determined equation for the separation of the direct and diffuse solar radiations may result in a systematic bias. For Method B, the proportion of diffuse solar radiation was estimated by comparing the blue wavelength range, within which scattering occurs readily, with the red wavelength range. A comparison of the estimated and observed values revealed that R² and RMSE were 0.868 and 0.127, respectively. Thus, the estimation accuracy of Method B is lower than that of Method A, even though Method B is simpler to use.