Development of an alternative hyperspectral Moon phase reddening model

It led to the development of a new lunar reflectance model that could be compared to the GSICS Implementation of the ROLO (GIRO) model as developed by EUMETSAT. The GIRO uses a fixed composite Apollo spectrum that is adjusted according to the geometrical conditions of illumination, following the approach from the USGS ROLO model. This project is a contribution to the enhancement of the GIRO. Its allows a better understanding of the phase dependence observed with some satellite datasets, such as the Meteosat lunar data acquired with the SEVIRI instruments, which cover GIRO’s full range of illumination conditions.

Objectives

This project is a contribution to the enhancement of the GIRO. The outcome of this study allows a better understanding of the phase dependence observed with some satellite datasets, such as the Meteosat lunar data acquired with the SEVIRI instruments, which cover GIRO’s full range of illumination conditions.

The objectives of the study were:

1. Improve the radiometric quality of the SCIAMACHY lunar observations, which was an input to the project. It involved the removal or at least the reduction of the impact on the data of the residual polarisation in the SCIAMACHY lunar dataset. As part of those improvement, the possibility to reduce the measurement noise beyond 1.3 microns was investigated.
2. The impact on the SCIAMACHY data of reducing the polarisation and noise artefacts was assessed and fully documented and a newly derived SCIAMACHY lunar dataset was delivered.
3. By using both the improved SCIAMACHY dataset and RELAB data, a model for the lunar soil reflectance spectrum as a function of the phase was inferred, with its associated uncertainties, at a spectral resolution expected to be higher than the current Apollo spectrum in the GIRO model (10 nm).
4. An assessment of the model validity below 350 nm was performed.

Overview

As part of the first series of activities, the original SCIAMACHY lunar dataset provided as an input to the study was improved by taking on board recent findings in the SCIAMACHY calibration results. Those findings are encompassed in the version v9 of the official SCIAMACHY calibration results and are related to a better estimate of the Bi-directional Reflectance Distribution Function of the Elevation Scanner Module (ESM) diffuser. In addition, the radiometric quality of the dataset was enhanced by reducing the noise in the data. This noise correction was obtained by identifying, flagging and excluding noisy pixels from the measurements used later on in the fitting procedure to derive the lunar reflectance model. The impact of this latter correction is particularly visible in the overlap regions between channels and in the NIR/SWIR region. Finally, further investigation allowed the implementation of a polarisation correction to account for SCIAMACHY polarisation sensitivity in the measurements.

Based on the resulting improved SCIAMACHY lunar reflectance dataset, and after a review of existing lunar reflectance parametrisations, a new model was inferred by fitting the lunar observations. This model was complemented by the use of RELAB lunar soil spectra. It allows a full spectral coverage over the range [250, 2500] nm, resolving known issues with SCIAMACHY data in the NIR/SWIR and in the instrument channel overlaps. As an example, Figure 1 shows a sample measured lunar reflectance (after instrumental effects removal) in red and the final model in black with the associated uncertainties in yellow.

In order to assess the performances of the newly developed model, several comparisons were undertaken:

1. using lunar measurements from the GOME-2 instrument aboard Metop-A and Metop-B, after accounting for the instrument temporal degradation;
2. against the GIRO model, as presented in Figure 2;
3. using measurements from the SEVIRI instrument aboard Meteosat-8 to 11, for four reflective solar bands (VIS0.6, VIS0.8, NIR1.6 and HRV).