Sensitivity Analysis Method for Spectral Band Adjustment between Hyperspectral Sensors: A Case Study Using the CLARREO Pathfinder and HISUI

Abstract

The International Space Station has become the platform for deploying hyperspectral sensors covering the solar reflective spectral range for earth observation. Intercalibration of hyperspectral sensors plays a crucial role in evaluating/improving radiometric consistency. When intercalibrating between hyperspectral sensors, spectral band adjustment is required to mitigate the effects of differences between the relative spectral responses (RSRs) of the sensors. Errors in spectral parameters used in spectral band adjustment are propagated through to the adjustment results. The present study analytically approximated the uncertainty in the spectral band adjustment for evaluating the relative contributions of uncertainties in parameters associated with the exo-atmosphere, atmosphere, and surface to the total uncertainty. Numerical simulations using the derived equations were conducted to perform a sensitivity analysis for the case of the spectral band adjustment between the Climate Absolute Radiance and Refractivity Observatory (CLARREO) Pathfinder (CPF) and the Hyperspectral Imager Suite (HISUI). The results show that the effects of errors in the solar irradiance were greater than those of other sources of error, indicating that accurate estimates of atmospheric reflectances and tranismittances are not needed for spectral band adjustment between CPF and HISUI in the atmospheric windows. The accuracy of the analytical approximation was also evaluated in the simulations. The framework of the sensitivity analysis is applicable to other pairs of hyperspectral sensors.

Combination of Cross- and Inter-Band Radiometric Calibrations for a Hyperspectral Sensor Using Model-Based Spectral Band Adjustment

“…This may relate to the blurred linearity in the scatterplot (Figure 12) and inconstant RCCCs in time series (Figure 13) on such bands. In contrast, atmospheric condition variability has a limited effect on cross-calibration, probably because of the canceled bias along the two-way radiative transfer simulation [7]. We could also confirm that the effect of the geolocation error was limited over RRVP because of the spatial homogeneity.…”

“…The spectral edge region, particularly in NIR and SWIR, also showed high uncertainty (Figure 10). In the estimated error budget (Tables 6 and 7), the uncertainty in the solar irradiance model was the second largest source of error [6,7,19] after the calibration uncertainty in the reference data (except atmospheric absorption regions). We used the solar irradiance model distributed by MCST considering the consistency with retrieved ancillary parameters in the reference sensor (MODIS) such as Esun.…”

“…Then, the RTM was used along the inverse direction (1-D in Figure 1), with a minor change of the parameter settings regarding the sensor geometry and band response, to retrieve the TOA radiance of CAL an,sim . The two-way RTM canceled the errors in the atmospheric parameters to some extent; as a result, the influence of the atmospheric conditions on the calibration result was expected to be limited [7]. We also estimated the uncertainty arising from the atmospheric conditions using the sensitivity analysis (see Appendix A).…”

“…For this purpose, researchers have developed various cross-calibration methods, including a comparison between sensors aboard the same satellite platform [6,7], those aboard different platforms but with near-simultaneous nadir overpass (SNO) [8][9][10], and those without SNO but with similar sensor and solar geometries [11], as well as a statistical comparison of pseudo-invariant calibration sites [12] and deep convective clouds [13]. Most approaches ensure that the selected image pairs are geometrically alike to minimize the uncertainty arising from the different illumination angles and bidirectional behaviors of the surface scatter.…”

“…Another approach to the band adjustment is a physical or model-based approach that makes full use of the soil line concept [18] considering the atmospheric radiative transfer. The linear relationship of the surface reflectance between two spectral bands, i.e., the soil line, enables radiometric cross-calibration between the analogous bands of two sensors [6,7] as well as inter-band calibration within a sensor [19]. Ensuring inter-band consistency within one sensor is particularly important for sensors with abundant spectral bands (i.e., hyperspectral sensors), such as Hyperion.…”