Rutherford Appleton Laboratory Requirements Consolidation of the Near-Infrared Channel of the GMES-Sentinel-5 UVNS Instrument: FP, 25 April 2014, ESTEC Height-resolved aerosol R.Siddans (RAL)

Overview Follows work in Eumetsat, ESA Camelot, ESA S4 studies to define instrument requirements Uses same optimal estimation retrieval simulation scheme employed in the previous studies. These updated to simulate S5. Retrieval scheme based on OE Aerosol optical properties (single-scat albedo, phase fn) assumed Extinction profile retrieved (2km vertical grid) Wavelength shift retrieved Results integrated to various layer aerosol optical depths (AOD) for comparison to user requirements Requirement on BL and free-trop column AOD=0.05 km spatial resolution) A priori error on surface albedo 0.01 assumed Estimated standard deviation (ESD) from measurement noise derived from solution covariance Errors to be assessed by performing “linear mapping” of error spectra (Δy) into L2 error (Δx): Δx = G Δy

Extension of retrieval scheme Scheme applied to both oxygen A and B bands (for various instrument concepts), over much wider range of conditions than before, leading to global maps of retrieval performance. Albedo retrieved with linear wavelength dependence (2 terms) For Concept A, spectral range nm omitted and linear dependence fitted in both oxygen A and B bands (4 terms) Fluorescence included by mapping as error or including in retrieval Retrieval of ILS width implemented as option

O 2 -B BandH2OH2O O 2 -A Band Concept B 0.12nm resolution Concept A 0.39nm resolution; 2x better throughput

Simulated instruments

Geophysical scenarios Aerosol retrieval performance very dependent on view / solar geometry due to variations in aerosol phase function, light path for aerosol light-path for O 2 absorption etc. Surface albedo Assumed aerosol type + size (asymmetry, single scatter albedo etc) Earlier simulations show that 0.05 requirement cannot be met for height- resolved quantities in all (most) S5 observing conditions Difficult to concisely summarise performance and optimise inst/L1 requirements without considering many conditions Aim to provide realistic along-orbit simulations of retrieval peformance Based on S5 orbit model + linear retrievalsfor range of conditions: Solar zenith angle: 30, 45, 60, 70.,75, 80 o View zenith angle: 0, 30, 50, 60, 70 o Relative azimuth angle: 0, 30, 60, 90,120,150,180 o Surface albedo: 0.01,0.1,0.2,0.3,0.5,0.7,0.9 Aerosol profile: Camelot Mid-latitude background and “tropical dust ocean” conditions (total AOD 0.2 and 0.67, respectively)

Instrumental errors simulated Instrument noise: The impact of instrument noise on the estimated precision of aerosol layer optical depth is estimated via the ESD as described above. Sensitivity to errors in the spectral response function or instrument line shape (ILS): This is determined by linearly mapping a 1% error in the width of the assumed ILS. Sensitivity to errors in radiometric gain is determined by linearly mapping the impact of a 10% gain error, i.e. multiplying the observed spectrum by a factor 1.1 Sensitivity to an additive absolute radiometric accuracy requirement (ARA) Mapped MTRD specification as gain (~3%) Mapped proposed relaxation as a separate radiometric offset. Other representations of the error possible / more realistic ? Intra-band co-registration Mapping spatial variation in albedo associated with spatial shift, with 2 nd order wavelength dependence in band First order would be handled by linear retrieval of albedo

Co-registration requirements These identified as challenging at S4/5 MAG, proposal to relax to 0.3 inter (keep 0.1 intra for NIR)

Change in albedo at 858nm for 20% shift in Spatial response

Retrieval errors for favourable geometry (LZA=60,SZA=60,RAZ=90) With H 2 O modelled and retrieved

Retrieval errors for favourable geometry (LZA=60,SZA=60,RAZ=90) With fluorescence retrieved

Retrieval errors for favourable geometry (LZA=60,SZA=60,RAZ=90) With fluorescence and spectral response function width retrieved

Weighting functions

Retrieval errors for favourable geometry (LZA=60,SZA=60,RAZ=90)

Retrieval Simulation results Estimated Standard Deviation (ESD) = retrieval precision (random noise) for Free-tropospheric column Concept B Concept A (A Band) Concept A (A+B Bands)

Concept B Concept A (A Band) Concept A (A+B Bands) Retrieval Simulation results 1% error in instrument spectral line-shape (ILS) = spectral response function If not fitted in retrieval

Concept B Concept A (A Band) Concept A (A+B Bands) Retrieval Simulation results ARA relaxation, considered as radiometric offset error (assuming High-Lat Dark limit)

Concept B Concept A (A Band) Concept A (A+B Bands) Retrieval Simulation results Error due to 2% shift in spatial response with 2 nd order wavelength dependence Real magnitude not clear: Result highlights potential issue if spatial co-registration or spectral dependence not known.

Conclusions (Aerosol) Height resolved aerosol retrievals improve with increasing (finer) spectral resolution, even considering an instrument with fixed total throughput. Dependence on geometry large – no concept compliant over whole swath Concept B clearly preferred but retrieval remains a challenge Better performance over more of swath Option A only competitive if both O 2 A and B bands used Even then sensitivity to instrumental errors larger than concept B Introduces need to model spectral dependence of aerosol optical properties (only limited demonstrations for GOME-2 exist) For option B relaxation of ARA requirement (to HL-dark level) seems acceptable (if mapped as offset); This is not the case for concept A. Now that concept A selected, height-resolved aerosol in terms of layer optical depths meeting requirements very unlikely for S5 and would unrealistically drive requirements for the band (remains possible for S4) With concept A, could aim for less challenging aerosol layer height under high aerosol load (no quantitative user requirement or this but could be useful) – implications of this should be taken into account in future