Auxiliary Material for Paper 2010GL045896 Disentangling chlorophyll fluorescence from atmospheric scattering effects in O2A-band spectra of reflected sun-light

We perform a retrieval simulation on (1) how the fluorescence emission F s affects the retrieval of surface and scattering parameters from space-based observations of the O2A-band and on (2) how the retrieval of such parameters can mask the fluorescence signal. Here, we consider as retrieval parameters surface pressure ps, Lambertian surface albedo r, the aerosol optical density τ (AOD in main text) and the center height z of an aerosol layer. We simulate a measurement ~y by a forward model ~ f(F rel s > 0, ptrue, rtrue, τtrue, ztrue) that takes fluorescence emission into account. Given the synthetic measurement, we retrieve the target parameters by a retrieval forward model ~ f(F rel s = 0, p, r, τ, z) that neglects fluorescence. For the full physics retrieval, the forward model ~ f is based on the radiative transfer (RT) model developed by Hasekamp and Landgraf (2002, 2005) and extensively used for CO2 and CH4 retrievals by Hasekamp and Butz (2008), Butz et al. (2009), and Butz et al. (2010). The RT model calculates spectra of sunlight backscattered by the Earth’s surface and atmosphere considering absorption and (multiple) scattering by molecules and particles in a multi-layer, plane parallel atmosphere. The RT model also provides the Jacobians with respect to the parameters p, r, τ , and z and thus allows for their retrieval. Fluorescence emission is simulated by adding ~ F s as given by equation (1) to the RT modeled spectra. Thereby, we explicitly consider re-absorption of the fluorescence emission by O2 between the surface and topof-the-atmosphere (TOA), but we neglect scattering of the emitted fluorescence for simplicity. The modeled spectra and Jacobians are convolved by an instrument line shape that mimics the spectral resolution of a GOSAT-like observer (Fourier transform spectrometer with 2.5 cm optical path difference and 15.8 mrad full-angle field-of-view). Measurement noise is assumed vanishing. In analogy to equation (2), we use a non-linear weighted least-squares method to find the retrieval parameters p, r, τ , and z or a subset of these. In contrast to the simplified fluorescence retrieval, the full-physics retrieval is performed in intensity space as, due to noise and the sinc-type instrumental line-shape, negative radiances can be observed. The simulation is performed for a scene (ptrue = 1013 hPa, rtrue = 0.4) with a moderate aerosol load (τtrue = 0.15 at 765 nm) of fine aerosol particles in a layer close to the surface (Gaussian layer centered at ztrue = 2 km, full-width-at-half-maximum 1 km). The assumed composite aerosol consists of non-absorbing Mie-particles with a refractive index of 1.43 following a power-law size distribution according to Mishchenko et al. (1999) with size parameter α = 4.5. A GOSAT-like instrument observes the scene at exact nadir and solar zenith angle (SZA) 45. The input solar irradiance (Fig. S1, panel a) is modeled as described below and then normalized to a continuum value of 1. The assumed plant fluorescence emission at the surface (Fig. S1, panel b) amounts to roughly 1.5% of the reflected radiance (Fig. S1, panel c) with a smooth spectral dependence as suggested by Meroni et al. [2009]. The fluorescence signal at TOA ~ F s (Fig. S1, panel d) clearly shows re-absorption by O2 along the light-path from the surface to TOA. We undertake 5 retrieval simulations. Fig. S1, panels e through i, illustrate the results and spectral fitting residuals. In a first exercise (Fig. S1 panel e), we allow for the retrieval of the albedo parameter r only. Next, we simultaneously retrieve albedo r and aerosol optical density τ (Fig. S1 panel f). Then, we further add aerosol height z as retrieval parameter (Fig. S1 panel g). These three panels correspond to the lower three panels of Fig. 1 in the