D4.2, D4.3 EUMETSAT, UCL, BC, NPL, JRC, RF QA4ECV Mid-Term Review - PowerPoint PPT Presentation

QA4ECV WP4 T4.2 - Development of harmonised retrieval for surface BRDF/albedo, D4.2, D4.3 EUMETSAT, UCL, BC, NPL, JRC, RF QA4ECV Mid-Term Review Meeting, MSSL, 14-16 June 2016

BRDF/Albedo Plan A Overview Development of GlobAlbedo approach • Linear BRDF models (‘kernel models’) • Constraints: – Heterogeneous observations – Smoothness – Prior mean QA4ECV Mid-Term Review Meeting, MSSL, 14-16 June 2016

New setup • Stack datasets in time – So process spatial stack of time series – (e.g. 256x256 blocks) • Constraints: – Heterogeneous satellite observations – Regularisation • 2 scales: dt = 1 year; dt = 1 day – No prior • not needed if feed in MODIS observations QA4ECV Mid-Term Review Meeting, MSSL, 14-16 June 2016

Plan B Processing Chain QA4ECV Mid-Term Review Meeting, MSSL, 14-16 June 2016

GlobAlbedo/Plan B Approximate optimal solution using weighting on C obs in time QA4ECV Mid-Term Review Meeting, MSSL, 14-16 June 2016

Plan A Dx at day and year period Formal linear optimisation framework: g smoothness control QA4ECV Mid-Term Review Meeting, MSSL, 14-16 June 2016

Solution Low pass filter QA4ECV Mid-Term Review Meeting, MSSL, 14-16 June 2016

QA4ECV Plan A uncertainty QA4ECV Mid-Term Review Meeting, MSSL, 14-16 June 2016

Heterogeneous observations Linear mapping to common spectral basis Hainich QA4ECV Mid-Term Review Meeting, MSSL, 14-16 June 2016

Spectral mapping: MODIS-MERIS QA4ECV Mid-Term Review Meeting, MSSL, 14-16 June 2016

Solution QA4ECV Mid-Term Review Meeting, MSSL, 14-16 June 2016

Speed up • Regularisation in time using DCT – Convenient (reflexive) boundary conditions – Very fast • So can afford multiple passes • Outlier detection – Can integrate with uncertainty framework – Can optimise smoothing parameter • Fixed in globalbedo (GA) • Use GA value as prior and solve for this • Or, use Generalised Cross Validation (fast in DCT) • So, smoothness then adaptive QA4ECV Mid-Term Review Meeting, MSSL, 14-16 June 2016

Summary • New approach • Full formal linear optimisation system • Constraints: – Observations (inc. MODIS, MERIS, VGT, Proba-V) – Dx (1 day) Dx (1 year) – No need for prior • Options for further speedup (DCT) • Order datasets by time to make best use QA4ECV Mid-Term Review Meeting, MSSL, 14-16 June 2016

Plan B Processing : AVHRR+GEO AVHRR/GEO prior creation BRDF-Prior SDR BB SDR BB → (Spectral & BB) Orbit/disk → BBDR/MODIS by Tile (Broadband mapping ) SDR BB /tile/day/ by /sensor/tile/year/da y/ MOD/MYD09 BBDR → BB BRDF SDR/MODIS→ BBDR/MODIS Uniformity SDR → Spectral BRDF (MODIS SDR) (Spectral to BB) Weights (Weights by creation) /sensor/modis- band/time/ MERIS SDR l1b SDR → BBDR/MODIS by /sensor/tile/year/day/ti (Spectral to BB) me/ BRDF ( Spectral & BB) L1 → SDR Orbit → by /tile/year/day/ Tile VGT l1p SDR → SDR/MODIS (Spectral mapping) Tiles → Mosaics (Upscaling by energy BRDF conservation) ( Spectral & BB) Proba-V by /mosaic/year/day/ l1b Albedo ( Spectral & BB) by /tile/year/day/ & BRDF → Albedo /mosaic/year/day/ QA4ECV Mid-Term Review Meeting, MSSL, 14-16 June 2016

T4.2 Development of harmonised retrieval for surface BRDF/albedo  Development of harmonised retrieval for surface BRDF/albedo, LAI & FAPAR :  New methods for BRDF/albedo retrieval:  New instruments:  Ingestion of AVHRR-LTDR to broadband BRDF/albedo retrieval  Ingestion of GEO (MVIRI+SEVIRI) to broadband BRDF/albedo retrieval  New spectral bands:  From broadband to spectral BRDF/albedo using MODIS spectral bands QA4ECV Mid-Term Review Meeting, MSSL, 14-16 June 2016

T4.2 Development of harmonised retrieval for surface BRDF/albedo Ingestion of AVHRR-LTDR surface reflectance products  Use new AVHRR-LTDR reader module recently developed for BEAM/SNAP  Supported products are:  AVH02C1 (TOA reflectances, daily, global, 0.05deg (7200x3600 pixel))  AVH09C1 (Surface reflectances, daily, global, 0.05deg (7200x3600 pixel))  The AVHRR-LTDR products are provided by JRC and provide:  TOA or Surface reflectances at 630nm and 865nm  a quality flag, including a cloud mask  Idea: fill AVHRR gaps with Meteosat GEO (MVIRI+SEVIRI) data QA4ECV Mid-Term Review Meeting, MSSL, 14-16 June 2016

T4.2 Development of harmonised retrieval for surface BRDF/albedo Ingestion of AVHRR-LTDR surface reflectance products  From global products to MODIS SIN tiles:  generate reasonable subsets with BEAM/SNAP subsetting tool  reproject to all MODIS SIN tiles (1/20deg resolution as in GlobAlbedo) which have an intersection with the subset  The reprojection module is available in the existing GlobAlbedo processing chain and has been used for MERIS and VGT in the BBDR retrieval. Adaptation to other sensors is straightforward. QA4ECV Mid-Term Review Meeting, MSSL, 14-16 June 2016

T4.2 Development of harmonised retrieval for surface BRDF/albedo From global product to latitude subset: QA4ECV Mid-Term Review Meeting, MSSL, 14-16 June 2016

T4.2 Development of harmonised retrieval for surface BRDF/albedo Reprojection of latitude subset onto MODIS SIN tiles: QA4ECV Mid-Term Review Meeting, MSSL, 14-16 June 2016

T4.2 Development of harmonised retrieval for surface BRDF/albedo Ingestion of AVHRR-LTDR surface reflectance products QA quality flag coding: QA4ECV Mid-Term Review Meeting, MSSL, 14-16 June 2016

T4.2 Development of harmonised retrieval for surface BRDF/albedo Ingestion of Meteosat GEO MVIRI+SEVIRI surface reflectance products  We consider the Meteosat GEO MVIRI+SEVIRI BRF daily/10- day ‘disk’ products, as seen from the satellites’ Equatorial position at different longitudes  Use new Meteosat GEO reader module developed for BEAM/SNAP, which uses a special geolocation algorithm to handle off-planet pixels in the disk product.  The ‘disk’ products are reprojected onto a lat/lon grid (1/30deg) and then also onto MODIS SIN tiles (1/120deg) to be merged with AVHRR  The products provide: :  BRF from wavelength range 0.45 – 1.0 μ m and corresponding uncertainty  corresponding geometry (SZA, VZA, rel. azimuth) does not come with the products , their computation was implemented separately QA4ECV Mid-Term Review Meeting, MSSL, 14-16 June 2016

T4.2 Development of harmonised retrieval for surface BRDF/albedo From ‘disk’ to lat/lon grid: QA4ECV Mid-Term Review Meeting, MSSL, 14-16 June 2016

T4.2 Development of harmonised retrieval for surface BRDF/albedo MODIS SIN tiles within lat/lon area: QA4ECV Mid-Term Review Meeting, MSSL, 14-16 June 2016

T4.2 Development of harmonised retrieval for surface BRDF/albedo GEO BRF on 4 MODIS SIN tiles: QA4ECV Mid-Term Review Meeting, MSSL, 14-16 June 2016

T4.2 Development of harmonised retrieval for surface BRDF/albedo AVHRR + Meteosat GEO MVIRI+SEVIRI: From BRF to BBDR to BRDF/albedo  BRDF/albedo algorithm inherited from GlobAlbedo requires 3 ‘broadband’ input reflectances: BB_VIS, BB_NIR, BB_SW  AVHRR: conversion from band 1 and 2 (630nm and 865nm) to the three broadbands using coefficients provided by Liang et al. (2000). Corresponding uncertainties are computed from simple error propagation arithmetics  MVIRI+SEVIRI : the ‘shortwave broadband’ (representing 0.3 – 3.0 μ m, obtained using the method described by Loew and Govaerts (2010)) is taken as BB_SW sample. No BB_VIS, BB_NIR samples fed into the retrieval  Computation of BRDF/albedo with GlobAlbedo algorithm can then be done separately (using AVHRR or MVIRI/SEVIRI samples only) or by combining the samples QA4ECV Mid-Term Review Meeting, MSSL, 14-16 June 2016

T4.2 Development of harmonised retrieval for surface BRDF/albedo Albedo from AVHRR and/or MVIRI: examples  BRDF/albedo algorithm was used to compute albedo for selected MODIS SIN tiles for the 8-day periods in year 2005  Computations were done with samples from  AVHRR alone  MVIRI alone  AVHRR + MVIRI  Examples shown below are BHR/DHR albedo for African tiles h18v06 (Sahara desert) and h19v08 (tropical rain forest) for 2005, DoY 121. Comparisons with GlobAlbedo results retrieved from MERIS+VGT QA4ECV Mid-Term Review Meeting, MSSL, 14-16 June 2016

T4.2 Development of harmonised retrieval for surface BRDF/albedo BHR/DHR h18v06. Left: AVHRR; centre: AVHRR+MVIRI; right: GlobAlbedo QA4ECV Mid-Term Review Meeting, MSSL, 14-16 June 2016

T4.2 Development of harmonised retrieval for surface BRDF/albedo BHR/DHR h19v08. Left: AVHRR; centre: AVHRR+MVIRI; right: GlobAlbedo QA4ECV Mid-Term Review Meeting, MSSL, 14-16 June 2016

T4.2 Development of harmonised retrieval for surface BRDF/albedo BHR/DHR h18v06 (desert): AVHRR and AVHRR+MVIRI vs. GlobAlbedo MERIS+VGT QA4ECV Mid-Term Review Meeting, MSSL, 14-16 June 2016

T4.2 Development of harmonised retrieval for surface BRDF/albedo BHR/DHR h19v08 (rain forest): AVHRR and AVHRR+MVIRI vs. GlobAlbedo MERIS+V GT QA4ECV Mid-Term Review Meeting, MSSL, 14-16 June 2016

T4.2 Development of harmonised retrieval for surface BRDF/albedo BHR/DHR h18v06: MSA_Albedo standard product (left) vs AVHRR+MVIRI (right) QA4ECV Mid-Term Review Meeting, MSSL, 14-16 June 2016

T4.2 Development of harmonised retrieval for surface BRDF/albedo BHR/DHR h19v08: MSA_Albedo standard product (left) vs AVHRR+MVIRI (right) QA4ECV Mid-Term Review Meeting, MSSL, 14-16 June 2016