Product Specification Document

Transcription

1 1.rev.11 03/07/ Product Specification Document Milestone Product specification document Authors Distribution Milestone 1 Pierre Defourny (UCL), Sophie Bontemps (UCL), Céline Lamarche (UCL), Martin Böttcher (BC), Carsten Brockmann (BC), Norman Fomferra (BC), Grit Kirches (BC), Olaf Krüger (BC) ESA : Olivier Arino, Vasileios Kalogirou, Fabrizio Romaino, Frank Martin Seifert

2 1.rev.11 03/07/ Table of recorded changes Issue Record Sheet ISSUE DATE REASON /01/2011 Initial version /03/2011 Updated version according to RIDs received on 03/02/2011 and on comments from PM /04/2011 Updated version according to RIDs received on 25/03/ /07/2011 Updated version according to RIDs and comments received at PM /08/2011 Updated version before PM /05/2012 Updated version according to comments from PM /11/2012 Updated version according to comments from PM /05/2013 Updated version, which is aligned with Epoch1 v1.1, Epoch2 v1.0 and Epoch3 v /07/2013 Updated version according to comments from PM9 + to be aligned with the new DSWG naming conventions /11/2013 Updated version for the final meeting /06/2014 Updated version for Phase 1 closure /07/2014 Updated version for Phase 1 closure w.r.t RIDs Detailed Change Record Sheet ISSUE RID PROBLEM DESCRIPTION SECTION CHANGE 1.1 PSD/1 Symbols and acronyms to correct Symbos and acronyms Table updated 1.1 PSD/2 List of references to correct References Table harmonized and updated 1.1 PSD/4 Discussion of the gridding, projection, annotation flags and metadata structure 1.1 PSD/7 List all the satellite data sources that will or might be used. Provide different priority to these sources. 1.1 PSD/9 It is not clear the way the terminology that has been established before, is used here. Table 4 includes Trees, Shrubs 4, 5, 6 Issues discussed in details in the Product definition section (4), in the Quality documentation section (5) and in the Technical specifications section (6) 4.1, 4.2 Table 3 (in section 4.1) lists the dataset used in the pre-processing Table 4 (in section 4.2) recaps this information The thematic content of the global land cover map is specified by both land cover state and condition. The legend counts a

3 1.rev.11 03/07/ etc. but those are not Land Cover states. Features maybe? Same applies to Table 5 where we refer to states again but the table includes variables and characteristics. 1.1 PSD/10 Extend the discussion of quality assurance. In particular we should describe here the error budget and how the error accumulation process can be indicated for every product. These metrics should be included as layers in the final product. 1.1 PSD/11 Include gridding discussion in the technical specifications 1.1 PSD/12 Include projection discussion in the technical specifications 1.1 PSD/13 Include projection discussion in the technical specifications certain number of classes, which each corresponds to one land cover state, plus some seasonality information derived from the land cover condition. Clarifications have been brought in the text and legend has been provided (which should clarify the link between land cover state and products) 5, 6 Section 5 has been largely rewritten and extended, detailing the quality indicators associated both with the surface reflectance and the land cover products Indicators have been added as products layers in the technical specifications 6 Information specified in the sections spatial extent. Readers are referred to sections 4.1.2, 4.1.3, and 6.3 for more explanation. 6 Information specified in the sections projection. Readers are referred to sections and for more explanation. 6 Information specified in the sections projection. Readers are referred to sections and for more explanation. 1.1 PSD/14 Naming conventions 6 Naming convetion has been added in the technical specifications 1.1 PSD/16 Which bands will be included in the SR products? 4.1, 6 Table 2 has been added which lists the 15 MERIS bands that will be included in the product This information also appears in table 14 in the technical specifications 1.1 Progress 3, 4.2 Section 3 has been re-written to clarify the new concept of land cover. Section 4.2 presents how this new concept can be measured by EO data 1.1 Progress Legend expressed in LCCS and in PFT has been added (Tables 6, 7) 1.1 Progress All document Link between URD and PSD has been clarified 1.2 PSD/23 Brackets to remove 1.1 Brackets deleted 1.2 Progress Modification of the strucutre of sections 4, 5 and 6 4, 5, 6 In each of these sections, the global LC products are now presented first (before the SR

4 1.rev.11 03/07/ PSD/33 Include the assumption of stable LC state over the multi-year period. How noisy /change signatures of LC condition will be differentiated from the stable ones? 1.2 PSD/34 10-day compositing period to remove composites). It was indeed found more convenient since LC databases are the product of interest for users. 4.1 The first paragraph of section 4.1 has been rephrased and extended to emphasize this assumption Table 2 has been updated: SR 10- day time series replaced by SR composites 1.2 PSD/37 Projection description WKT description of the projection inserted as the Figure PSD/40 Simplified LCCS legend for validation 1.2 PSD/42 Please explain with another sentence the third error source. 1.2 PSD/43 Please clarify/correct the two footnotes. 1.2 PSD/44 In point 3, about the number of clear observations: inside the whole epoch (5 years)? The need for using simplified LCCS legend in the validation has been expressed, just after Table Section rephrased to account for the RID Section rephrased to account for the RID Section clarified to account for the RID 1.2 PSD/49 Units of LC condition Units of water, snow and fire have been modified (from binary flags to probabilities) 1.2 PSD/50 PSD/51 10-day interval The fact that the compositing period needs to be defined has been added in the PSD and references to the 10-day compositing period have been removed 1.3 Actions PM4-1 and PM Actions PM4-2 to PM4-4; RID PSD- 1.2/ Action PM4-23 Adjust the LC products specifications according to specific users requirements about PFTs Improve the specifications of the aggregation tool Tiling system for products distribution Annex A The list of PFTs has been aligned with the climate requirements and the correspendence between this list and the LCCS legend for each model/user is provided in Annex A 6.3 The tool now aims at reprojecting and modifying the spatial resolution. A minimum list of requirements in terms of resolution and re-projection has been specified Regional windows planned for the distribution of the LC products have been adjusted to improve the consistency with the CCI-Fire project

5 1.rev.11 03/07/ Actions PM4-2 and PM4-4; Improve the specifications of the aggregation tool 6.3 Technical information about the tool has been added at the end of the section Progress / 4 Section has been updated to reflect that only 1 LC condition product will be generated over the entire period ( ) 1.5 Progress / LCCS legend has been updated to reflect the 2-level hierarchy of the forest classes 1.5 Progress / Limits of continental tiles for the delivery of the LC maps (Figure 6 and Table 9) have been updated to ensure consistency between CCI Fire and LC projects 1.5 Progress / Spatial extent of LC maps has been updated to 90 N-90 S (based on IODD RIDs) 1.5 Progress / Section has been updated to reflect that only 1 LC condition product will be generated over the entire period ( ) 1.5 progress Simplification of file names, update of file name convention 1.5 Progress Adjustment of product definitions according to PVASR decisions 1.5 progress Update of metadata according to CCI Data Standards Working Group guidelines 1.6 Updating the data description w.r.t. the processing version 1.7 Updated version, which is aligned with Epoch1 v1.1, Epoch2 v1.0 and Epoch3 v1.0 w.r.t. pre-processing Ref. docs (deleted) & Appendix Processing time and centre dropped, sequence of naming elements harmonised 10d aggregation period replaced by 7d multi-sensor SR product removed AATSR removed SR field definition updated SR size estimation updated from experience CCI Guidelines replaces GML as reference RD-8, references updated NetCDF-4 instead of NetCDF-3 updating the definition of status_invalid updating the aggregation rules w.r.t cloud status updating the calculation of uncertainities

6 1.rev.11 03/07/ Updated version, which is aligned with Epoch1 v1.1, Epoch2 v1.0 and Epoch3 v1.0 w.r.t. LC maps 1.7 Updated version, which is aligned with Epoch1 v1.1, Epoch2 v1.0 and Epoch3 v1.0 w.r.t. LC condition B updating of the product description of the CCI LC preprocessing products w.r.t requirements of Data Standard Working Group Updating the method used to derive the 3 LC maps (i.e. first generate a baseline 10-year map and then derive the 3 maps) Updating w.r.t. the use of SPOT- VGT data (only to increase the temporal coverage of the project until 1998 not for the SWIR channel) Updating w.r.t. the legend (addition of classes 1a and 1b table 3) Updating w.r.t. the use of regional subsets ONLY for the public release Updating w.r.t. the quality flags (4 instead of 5 for the moment) Updating w.r.t. the file format (naming convention, spatial resolution and quality flags) The temporal coverage was changed from 15 to 13 years to be aligned with the new version of the products The number of quality flags where changed from 4 to 2 layers. It was indicated that the consistency assessment will only be delivered with the public release of the products The variable code identifier for the LC conditions was updated to take into account the differences between the netcdf and GTiff format. The mention of the possible 300m spatial resolution was removed from table 1.4 The example of the filename, the layers of the global LC condition products and the estimated size were updated 1.8 Comments from PM9 Updated version to be aligned with the new DSWG naming conventions & Naming conventions descriptions have been updated 1.8 Comments from PM9 Updated version to align the legend description with the (Table 3) New legend description

7 1.rev.11 03/07/ delivered products and the PUG 1.8 Progress 6.2 Modification of the characterization of the uncertainty description 1.9 Progress Updated legend and color codes for the LC maps 1.9 Progress The documentation of the quality measurements of the LC condition products has been updated 1.9 Progress Updated technical specifications for the LC condition products 1.10 Progress Legend table updated (no data are coded as 0 ) 1.10 Progress Removal of the regional subsets for the global LC map; removal of the PFT legend (as the PFT translation is now included in the user tool) 1.10 Progress Naming convention for the global LC map has been updated; reference to a PFT legend has been removed 1.10 Progress 6.3 Update of the section to reflect the last improvements 1.11 FR-01 "The link related to GlobCover Project is wrong please substitute it with the following: Page 8 / Acronym List 1.11 FR-02 Page 10 / Applicable Documents table 1.11 FR-03 Only 300m is reported as spatial resolution, but should be 300m or 1000m 1.11 FR-04 In the 'Wavelength_nm row', the values of MERIS B9 and B12 are different from the ones reported in the Table 8. Page 43, 44, 45 / Spatial resolution Page 47 / Table 17 Change made Applicable documents list updated The 1000m spatial resolution was mentionned in the text and a precision was made that Tables 15/16/17 only concern the full resolution product and that the same attributes apply to the reduced resolution product except for the spatial resolution which is 1000m. Table 8 updated

8 1.rev.11 03/07/ Symbols and acronyms AATSR : Advanced Along-Track Scanning Radiometer AMSR-E : Advanced Microwave Scanning Radiometer - Earth Observing System ASAR : Advanced Synthetic Aperture Radar ATBD : Algorithm Theoretical Basis Document BADC : British Atmospheric Data Center BA : Burnt Area BC : Brockmann Consult CCI : Climate Change Initiative CCI-LC : Climate Change Initative Land Cover CDL : Common Data Language CMC : Climate Modelling Community CMUG : Climate Modelling User Group CNES : Centre National d Etudes Spatiales CRS : Coordinate Reference System DARD : Data Access Requirement Document DUE : Data User Element ECV : Essential Climate Variable EO : Earth Observation ESA : European Space Agency FR : Full Resolution GCOS : Global Climate Observing System GFED : Global Fire Data GLC2000 : Global Land Cover 2000 GlobCover : ESA DUE project ( GML : Geography Markup Language GTOS : Global Terrestrial Observing System IFOV : Instantaneous Field-Of-View IGCO : Integrated Global Carbon Observation IGOL : Integrated Global Observations for Land LAI : Leaf Area Index LC : Land Cover LCCS : Land Cover Classification System MERIS : Medium Resolution Imaging Spectrometer NDVI : Normalized Difference Vegetation Index NIR : Near InfraRed PFT : Plant Functional Type

9 1.rev.11 03/07/ PSD PVP RR SDR SPOT SPOT-VGT SR SWE SWIR UN UR URD UCL VITO WB WGS WS : Product Specification Document : Product Validation Plan : Reduced Resolution : Surface Directional Reflectance : Satellite Pour l'observation de la Terre : SPOT-Vegetation : Surface Reflectance : Snow Water Equivalent : Short-Wave InfraRed : United Nations : User Requirement : User Requirement Document : Université catholique de Louvain : Vlaamse Instelling Voor Technologisch Onderzoek : Water Body : World Geodetic System : Wide Swath

10 1.rev.11 03/07/ Applicable and reference documents Applicable documents ID Title [AP-1] ESA Climate Change Initiative Phase I - Scientific User Consultation and Detailed Specification Statement of Work (SoW), version 1.4, [AP-2] ESA Climate Change Initiative Phase 1 - Project Guidelines V1 [AP-3] CCI-LC URD Phase I. Land Cover Climate Change Initiative - User Requirements Document (URD), V2-2.2 (23/02/2011) [AP-4] CMUG, 2010, Requirement Baseline Document MOHC, MPI-M, ECMWF, MétéoFrance. [AP-5] CCI-LC DARD Phase I. Land Cover Climate Change Initiative - Data Access Requirements Document, V1-1.9 (06/06/2012) [AP-6] CCI-LC ATBD Phase I. Land Cover Climate Change Initiative - Algorithm Theoretical Basis Document (ESA), v2-2.3 (28/11/2013) [AP-7] CCI-LC ATBD Phase I. Land Cover Climate Change Initiative - Algorithm Theoretical Basis Document (ESA), v2-2.3 (28/11/2013) [AP-8] CCI-LC PVP Phase I. Land Cover Climate Change Initiative - Product Validation Plan, V1-1.3 (04/07/2011) Reference documents ID Title [RD-1] Herold M., Woodcock C., Wulder M., Arino O., Achard F., Hansen M., Olsson H., Schmulllius C., Brady M., Di Gregorio A., Latham J. and Sessa R., 2009, GTOS ECV T9: Land Cover - Assessment of the status of the development of standards for the Terrestrial Essential Climate Variables [RD-2] GCOS, 2004, Implementation plan for the Global Observing System for Climate in Support of the UNFCCC, World Meteorological Institute. Available at: [RD-3] GCOS, 2010a, Guideline for the generation of datasets and products meeting GCOS requirements, GCOS-143, WMO/TD No. 1530, May 2010 Available at: [RD-4] GCOS, 2010b, Implementation plan for the Global Observing System for Climate in Support of the UNFCCC, World Meteorological Organisation, August 2010 (update). Available at: [RD-5] Ciais P. and Moore B., 2004, Integrated Global Carbon Observation (IGCO) Implementation Plan, IGOS theme report. Available at: [RD-6] Townshend J. R., Latham J., Arino O., Balstad R., Belward A., Conant R., Elvidge C., Feuquay J., El Hadani D., Herold M., Janetos A., Justice C.O., Liu J., Loveland T., Nachtergaele F., Ojima,D., Maiden M., Palazzo F., Schmullius C., Sessa R., Singh A., Tschirley J. and Yahamoto H., 2008, Integrated Global Observations of teh Land: an IGOS-P theme. IGOL Report No. 8, GTOS, 54. Available at: [RD-7] Arino O., Bicheron P., Achard F., Latham J., Witt R., Weber J.L. et al Globcover: the most detailed portrait of Earth. ESA Bulletin 136 [RD-8] Bennett, V, Guidelines for Data Producers - Climate Change Initiative Phase 1, CCI-PRGM-EOPS-

11 1.rev.11 03/07/ TN , issue 2, revision 1, 20/03/2012 [RD-9] Di Gregorio A., 2005, UN Land Cover Classification System (LCCS) Classification concepts and user manual for Software version 2. Available at: [RD-10] Burley T. M., Land use or land utilization? Professional Geographer, 14(5): [RD-11] GLP, 2005, Science Plan and Implementation Strategy. IGBP Report No. 53/IHDP Report No. 19. IGBP Secretariat, Stockholm. 64 p. [RD-12] Miller R.I., 1994, Mapping the diversity of the nature, London, New York: Chapman & Hall [RD-13] Bontemps S., 2010, On the monitoring of land surface change. In: Towards an automated satellitebased alarm system for global forest monitoring, PhD thesis, Université catholique de Louvain [RD-14] EEA, 2006, Land accounts for Europe , EEA Report No 11/2006 prepared by Haines- Young, R. and Weber, J.-L. Available at : [RD-15] Bontemps S., Defourny P., Van Bogaert E., Soukup T., Weber J.L., Kalogirou V. and Arino O., 2010, GlobCorine 2005 Description and Validation report, version 2.1 [RD-16] Committee on Earth Observation Satellites, 2008, Working Group on Information Systems and Services - Interoperability Handbook, February 2008 Issue 1.1. Available at: [RD-17] NetCDF Climate and Forecast (CF) Metadata Conventions, Version 1.5, 25 October, 2010 [RD-18] Patrice Bicheron, Mireille Huc, Caroline Henry, GLOBCOVER partners, GLOBCOVER - Description Products Manual GLOBCOVER_DPM_I2.0, 2 Rev. 0, Date5/09/2008 [RD-19] P. Lewis, C. Brockmann, O. Danne, J. Fischer,mL. Guanter, A. Heckel, O. Krueger, G. López, J-P. Muller, P. North, R. Preusker GlobAlbedo - Algorithm Theoretical Basis GlobAlbedo_ATBD_V1.0 Date 09/09/2010 [RD-20] Poulter B., Ciais P., Hodson E., Lischke1 E., Maignan F., Plummer S. and Zimmermann N.E., Mapping plant functional types to reduce uncertainty of earth system processes, Submitted [RD-21] Herold M., Latham J.S., Di Gregorio A. and Schmullius C., 2006, Evolving standards in land cover characterization. Journal of land use science, 1, 2-4,

12 1.rev.11 03/07/ Table of content 1. Introduction Purpose and scope Structure of the document Synthesis of user requirements Findings summary Quantitative requirements Land cover concept in the project Definition based on spatial and temporal dimensions Land cover state and condition Products definition Global LC products Information included in the global LC databases Spatial and temporal resolution of the global LC products Format, metadata, and projection Global SR composite time series Projection and tilling system Format and metadata Quality documentation Global LC products Global LC map Global LC condition product Independent validation process Global SR composite time series Attributes from the pixel identification Surface directional reflectance (SDR) products Global SR composite time series Technical specifications Global LC products Global LC map Global LC conditions Global SR composite time series CCI LC user tool Products delivery mechanisms Link between URD and PSD Appendix A. Link between LCCS legend and PFTs for each climate model and/or climate user... 55

13 1.rev.11 03/07/ Laboratoire des Sciences du Climat et de l Environnement Max Planck Institute for Meteorology Requirements for global models (ECHAM6 and JSBACH) Requirements for regional models (REMO) Met Office Hadley Center Appendix B. NetCDF data structure (CDL header) of a 7 day SR composite... 59

14 1. rev.11 03/07/ Introduction 1.1. Purpose and scope The Product Specification Document (PSD) transforms the climate user requirements described in the User Requirement Document (URD) [AP-3] into a complete and consistent set of products specifications. Clearly, not every requirement can be met and pragmatic choices have to be made when converting the outputs of the URD into detailed product technical specifications. As outlined in the Climate Change Initiative (CCI) project guidelines [AP-2], it is important to explain the rationale behind the choices made. Defining the product specifications and justifying them is the scope of this document. From the users survey carried out during the requirement analysis, the CCI Land Cover (CCI-LC) products could be of interest for users outside the climate research community. This is particularly the case for the global land cover community. There is no doubt that the primary role of the CCI led by the European Space Agency (ESA) is to develop products that meets the needs of the Global Climate Observing System (GCOS) and that could therefore be recognized as Essential Climate Variables (ECV). However, in addition to focusing on the requirements expressed by the Climate Modelling Community (CMC), the CCI-LC project also aims at making a particular effort to target the global land cover community. The output products of the CCI-LC production chain consist of global Surface Reflectance (SR) composite time series and of global Land Cover (LC) products. Within this document, these products are defined in terms of their content, legend, spatial and temporal applicability, grids, projections and quality. The convention about the products identification, format, structure, metadata and delivery is presented [AP-1, AP-2] Structure of the document After this introduction, the document is divided into 6 sections that are briefly described below: Section 2 summarizes the main findings of the user requirement analysis, as well as their quantitative requirements in terms of spatial and temporal coverage, spatial resolution, overall accuracy, temporal stability and errors characterization; Section 3 presents the new land cover concept underlying the development of the CCI-LC products; Section 4 describes the different types of products delivered by the project (global LC products and SR time series), by making links with the URD and Data Access Requirement Document (DARD) [AP-3, AP-5]; Section 5 documents the quality indicators associated with each of these products; Section 6 details the technical specifications of the CCI-LC products, using the same template than in the URD [AP-3, section 4.10] and presents the products delivery mechanisms; Section 7 provides a table which summarizes how the user requirements described in the URD have been translated in product specifications.

15 1. rev.11 03/07/ Synthesis of user requirements 2.1. Findings summary During the first three months of the CCI-LC project, a user requirement analysis has been conducted to derive the specifications for a new global LC product to address the needs of key-users from the CMC. This user assessment was built upon the general guidance from the GCOS an its related panel activities and has provided the next step to further derive more detailed characteristics and foundations to observe Land Cover as an ECV [RD-1]. As part of this analysis, a user consultation mechanism was set-up to actively involve different climate modelling groups by setting out surveys. These climate users surveys focused on three major ways land cover observations are used in climate models: 1) as proxy for several land surface parameters assigned based on Plant Functional Types (PFTs); 2) as proxy for (tracking) human activities, i.e. land use affecting land cover; 3) as datasets for validation of model outcomes (i.e. time series) or to study feedback effects. The evolution of requirements for these aspects from current models to future new modelling approaches was specifically taken into account. In addition, the broad land cover data user community, represented by users of the ESA GlobCover product [RD-7], was also surveyed. Next to the surveys, requirements from the GCOS Implementation Plan 2004 and 2010 [RD-2, RD-3, RD-4] and associated strategic Earth Observation (EO) documents for land cover (Global Terrestrial Observing System (GTOS), Integrated Global Observations for Land (IGOL), Integrated Global Carbon Observation (IGCO), Climate Modelling User Group (CMUG) in [AP-4, RD-5, RD-6]) were considered and integrated. Finally, a detailed literature review was carried out with special attention to innovative concepts and approaches to better reflect land dynamics in the next generation climate models. The outcome of the user requirements assessment has shown that although the range of requirements coming from the CMC is broad, there was a good match among the requirements coming from different user groups and the broader requirements derived from GCOS, CMUG and other relevant international panels. The findings of the User Requirement (UR) analysis highlight that: UR1: There is need for both stable land cover data and a dynamic component in form of time-series and changes in land cover; UR2: Consistency among the different model parameters is often more important than accuracy of individual datasets, and it is important to understand the relationship between land cover classifiers with the parameters and the relative importance of different land cover classes; UR3: Providing information on natural versus anthropogenic vegetation (disturbed fraction), tracking human activities and defining history of disturbance is of increasing relevance; in particular for land use affecting land cover with most details needed for focus areas with large anthropogenic effects;

16 1. rev.11 03/07/ UR4: Land cover products should provide flexibility to serve different scales and purposes both in terms of spatial and temporal resolution; UR5: The relative importance of different class accuracies varies significantly depending on which surface parameter is estimated and the need for stability in accuracy should be reflected in implementing a multi-date accuracy assessment; UR6: Future requirements for temporal resolution refer to intra-annual and monthly dynamics of land cover including also remote sensing time series signals; UR7: More than 90% of the general land cover users find the United Nations Land Cover Classification System (UN-LCCS) a suitable approach for thematic characterization; and this approach is also quite compatible with the PFT concept of many models; UR8: Quality of land cover products need to be transparent by using quality flags and controls, and including information on the probability for the land cover class or anticipated second class or even the probability distribution function for each class (coming from the classification algorithm) Quantitative requirements The user requirements assessment also resulted in quantitative outputs. At least two levels of requirement were identified: The threshold requirement, standing for the limit at which the observation becomes ineffectual and is not of use for the climate-related application. The target requirement, which is the maximum performance limit for the observation, beyond which no significant improvement would result for climate applications. These two levels of requirements are detailed in Figure 1.

17 1. rev.11 03/07/ Threshold requirement Target requirement Geographic Coverage Temporal sampling Global Best/stable map and regular updates Coverage and sampling Global with regional and local specific products Monthly data on vegetation dynamics and change Temporal extent 1-2 years, most recent 1990 (or earlier)-present Horizontal Resolution Vertical Resolution Precision Accuracy Stability Error Characteristics Resolution 1000 m 30 m - - Error/Uncertainty Thematic land cover detail sufficient to meet current modelling user needs Higher accuracy than existing datasets Higher stability than existing datasets Independent one-time accuracy assessment Thematic land cover detail sufficient to meet future model needs Errors of 5-10% either per class or as overall accuracy Errors of 5-10% either per class or as overall accuracy Operational and independent multidate validation Figure 1. Threshold and target requirements for land cover products, resulting from the URD

18 1. rev.11 03/07/ Land cover concept in the project Land cover is referred to as one of the most obvious and commonly used indicators for land surface and the associated human induced or naturally occurring processes. With the development of EO technology, it has become increasingly feasible to derive land surface information from a combination of in-situ surveys and EO satellite data at global, regional and national scales [RD-1]. While traditional in-situ surveys usually allow acquiring specific information (such as species composition, landform, soil type, land use, etc.), satellite data provide a spatially-exhaustive, georeferenced and repetitive coverage of the Earth. Building upon the increasing availability of EO satellite data, land cover mapping has progressively become one of the most popular approaches to describe the land surface. The land surface in different regions of the world has been mapped and characterized several times. Many countries have some kind of land monitoring system in place (e.g. forest, agriculture and cartographic information systems and inventories). In addition, a number of global land cover mapping activities exist. These activities have evolved with the availability of global moderate spatial resolution satellite observations since the early 1990s. Meanwhile, the development of the UN-LCCS has provided a basic level of thematic land cover standardization [RD-9]. These efforts have yielded several products in the 300 m 1 km spatial resolution range, all based on a single-sensor approach and mostly associated with legends described according to LCCS classifiers. The recent capabilities of acquiring and processing global multi-year and multi-sensor time series of EO data call for revisiting the land cover concept while capitalizing with all the experience acquired in various aspects of land cover mapping. Furthermore, the URD highlights the expectations of the climate communities for an improved land cover product which would be more integrative than the current one. Indeed, there was a clear requirement for a land cover which includes both stable and dynamic components (UR1), while making the difference between land cover change and natural variability (i.e. land cover dynamics which do not question the existence of the land cover itself) (UR2). For a long time, there was some diversity of opinion about what the land cover was and how it was distinct from the land use. In 1961, Burley defined the land cover as the vegetation and the artificial constructions covering the land [RD-10]. In the context of LCCS, land cover refers to the physical and biological cover over the surface of land, including water, vegetation, bare soil, and/or artificial structures [RD-9]. IGOL defines land cover as the observed bio-physical cover on the earth s surface [RD-6], while recognizing, in the current practices, the confusion between land cover and land use. Land use characterizes the arrangements, socio-economic activities and inputs people are undertaking on a certain land cover type. It includes both space and time dimensions and should be considered separately from land cover type to ensure internal and external consistency and comparability [RD- 11]. Yet, it must be recognized that the land cover cannot, at the same time, be defined as the physical and biological cover on the terrestrial surface [RD-1, RD-9] and remains stable and consistent over time as requested by the climate users [AP-3]. A revisited land cover concept has therefore to be introduced to reconcile these two views. Defining such new land cover concept that will address the climate users requirements in the context of the CCI-LC project is the objective of this section.

19 1. rev.11 03/07/ Definition based on spatial and temporal dimensions The land cover concept used in the CCI-LC project assumes the land cover is organized along a continuum of temporal and spatial scales [RD-12] and that each land cover type is defined by a characteristic scale, i.e. by typical spatial extent and time period over which its physical traits are observed [RD-13]. This twofold assumption requires introducing the time dimension in the land cover characterization, which contributes to define the land cover concept in a more integrative way, as requested by the CMC (UR1). Indeed, accounting for the time dimension allows distinguishing between the stable and the dynamic component of land cover. The stable component, which is named land cover state, refers to the set of land surface features which remain stable over time and thus define the land cover independently of any sources of temporary or natural variability. Conversely, the dynamic component is directly related to this temporary or natural variability that can induce some variation in land surface features over time but without changing the land cover state in its essence. This second component is referred to as land cover condition Land cover state and condition Land cover states and conditions are mapped through the use of land surface features, which consist in landscape elementary units (e.g. a house, a tree, a water body, etc.). The land cover state refers to a stable ensemble of land surface features. It is fully described by: 1) the type of the observed features, such as tree, shrub, herbaceous vegetation, moss/lichen vegetation, terrestrial or aquatic vegetation, inland water, built-up areas, permanent snow/ice, etc; 2) the structure of the observed features, like vegetation height, vegetation cover, building density, etc; 3) the nature of the observed features, such as the level of artificiality or some species information (e.g. C3/C4 distinction); 4) the homogeneity of the observed features at the level of observation, leading to pure or mosaic classes. The land cover state is well described using the LCCS, as illustrated by Figure 2. The anthropogenic dimension, included in the level of artificiality of the features nature, rather refers to the land use than to the land cover. The users surveys and the expected uses of the CCI-LC products call to include such a simple surrogate for land use. However, the Corine Land Cover and the GlobCorine experiences amongst others [RD-14, RD-15] showed that this information was proved to be quite difficult to derive from satellite observations.

20 1. rev.11 03/07/ Figure 2. Description of the land cover state mostly based on LCCS The land cover condition encompasses the inter-annual processes modifying temporally the land surface along the year. Typically driven by biogeophysical processes, it corresponds to annual time series mode of instantaneous observations of the land cover status. The land cover condition is described by different observable variables: 1) the green vegetation phenology through vegetation indices (e.g. the Normalized Difference Vegetation Index NDVI) profiles; 2) the snow coverage allowing users to derive the snow cover period; 3) the open waterpresence related to floods, water extent dynamic or irrigation; 4) the fire occurrence and the associated burnt scars; 5) possibly, the number of cropping cycles; 6) possibly, the albedo (whenever available); 7) possibly, the Leaf Area Index (LAI) (whenever available).

21 1. rev.11 03/07/ The land cover condition can be described in a relevant way through an interpolation between instantaneous statuses of the land cover. This can take the form of time profiles in the case of continuous variables (e.g. NDVI, etc.) or of temporal distribution of occurrence probabilities in the case of discrete variables (e.g. snow or water). Both possibilities are illustrated in Figure 3. The land cover condition provides a reference information depicting the seasonal land cover change which is not related to a given year (UR3). This information is obtained by averaging multiple (minimum 5) years. In the case of continuous variables, the mean time profile is associated with one standard deviation interval characterizing the inter-annual variability. Figure 3. Description of the land cover conditions through probabilistic distributions (a) or time profiles (b) Table 1 illustrates this new land cover concept (state and condition) with two distinct illustrations, the first one referring to artificial urban areas and the second one to a dense tropical forest.

22 1. rev.11 03/07/ Table 1. Illustration of the concepts of land surface, land surface feature, land cover state and land cover conditions. Land surface (what is observed): roofs Land surface features (elementary units): houses Land cover state (stable ensemble of land surface features): urban area Feature type: built-up Feature structure: high density Feature nature: artificial Feature homogeneity: urban patterns, mosaic with green areas Land cover condition (instantaneous status of land surface features): no snow, no floods No snow observed No water observed Specific albedo Land surface (what is observed): rough canopy Land surface features (elementary units): trees Land cover state (stable ensemble of land surface features): forest Feature type: woody (trees, shrubs) vegetation Feature structure: dense cover, density Feature nature: natural, broadleaved, evergreen Feature homogeneity: homogeneous (few clearings) Land cover condition (instantaneous status of land surface features): no snow, no floods, no fire No snow observed No water observed No fire or burnt scares observed Specific albedo Using this new concept of land cover made of state and condition offers the opportunity to characterize the land surface in a more integrative way than just categories (i.e. forest or open water) or continuous variables classifiers (fraction of tree canopy cover). This addresses the critical requirements of consistency between land surface characteristics (UR2), the required stability (UR1, UR5) and the dynamic dimension at the intra-annual and seasonal levels (UR6). Furthermore, it is of direct use for the climate models parameterization. Indeed, as highlighted by the user requirement analysis [AP-3], LC products can be used in climate models as a proxy for several land surface parameters. These parameters are assigned, for instance, on the basis of the PFTs. The coupling of the permanent and the dynamic dimensions of the land surface allowed by the proposed land cover concept contributes to improve the consistency of the land surface parameterization (UR2). As a consequence of this revisited land cover definition, the land cover change must be referred to as a permanent modification of the land cover state and not of the land cover conditions in comparison with a baseline status. However, further work is needed to define precisely change thresholds, i.e. the modification levels required to be considered as land cover change.

23 1. rev.11 03/07/ Products definition As introduced in section 3, the outputs of the CCI-LC project consist in global LC databases made of LC state products for three epochs centred around 2000, 2005 and 2010 and of LC condition products. The global SR composite time series for the period , which serve as input for generating the global land cover databases, are also delivered. These products are provided with associated metadata as well as with products documentation and validation reports. The deliverables required by ESA, including the system specification, are also part of the CCI-LC project outputs Global LC products Most often, LC maps are generated from few instantaneous observations of the land cover state. As a result, classification outputs are sensitive to the date(s) of observation and can reflect temporary conditions (e.g. map savannahs as burnt scars, boreal forest as snow, croplands as bare soils, etc.). An alternative could probably come from the description of the land cover state (which stands for the stable component of the land cover) from multi-year observation dataset. In this case, assuming that no land cover change even temporary has occurred over this multi-year period, the land cover is expected to be mapped in a consistent way over time. Conversely, the instantaneous observations of the land cover condition should be considered within the perspective of a time cycle (typically a year) precisely in order to reflect the above-mentioned temporary conditions. The land cover condition analysis over a multi-year period (still under the assumption of stable land cover state over this period) should permit to derive a reference land cover condition as well as information about the intra-annual variability. This approach is implemented in the CCI-LC project. The general concept is presented in Figure 4 and illustrates how multi-year observation dataset are planned to be used to characterize the land cover state and condition.

24 1. rev.11 03/07/ Figure 4. Use of multi-year dataset to characterize the land cover state and condition and therefore, generating the global land cover products Three different global LC maps (standing for the LC state products) are delivered, centred to the epochs 2000, 2005 and The global LC map from 2010, which is the more recent product, is aimed to be the best existing map, both in terms of accuracy and stability (see user requirement in Figure 1). The three maps are derived from a baseline land cover map which is generated thanks to the entire MERIS archive. This 10-year baseline land cover map is then updated using (i) SPOT-VGT time series from 1998 to 2012 to detect the annual LC changes (only those affecting the forest classes) and (ii) the MERIS time series from and from for the 2010 and 2005 epoch respectively to map at 300m the changes identified with SPOT-VGT. The CCI-LC output databases are thus made with a dominance of MERIS Full Resolution (FR) imagery. MERIS Reduced Resolution (RR) is used to compensate for a possible lack of MERIS FR acquisitions. SPOT-VGT is used for its high temporal consistency to identify the main dynamics between the 3 epochs, and to extend the temporal coverage until Table 2 lists the satellite dataset that are planned to be used in order to generate the three LC maps.

25 1. rev.11 03/07/ Table 2. Satellite data sources that are planned to be used to generate the global LC maps Global LC database Baseline 10-year global LC map Global LC map for the 2010 epoch Global LC map for the 2005 epoch Global LC database for the 2000 epoch Reference period Satellite data source MERIS FR global SR composites between 2003 and Baseline 10-year global LC map SPOT-VGT global SR composites between 2008 and 2012 to identify and date the changes MERIS FR global SR composites between 2008 and 2012to map the identified changes at 300m spatial resolution Baseline 10-year global LC map SPOT-VGT global SR composites between 2003 and 2007 to identify and date the changes MERIS FR global SR composites between 2003 and 2007 to map the identified changes at 300m spatial resolution Baseline 10-year global LC map SPOT-VGT global SR composites between 1998 and 2002 The whole archive of Advanced Synthetic Aperture Radar (ASAR) Wide Swath (WS) is also used in order to produce a global water bodies map. Remote sensing imagery is complemented by a set of auxiliary data, which mainly consist in existing LC maps, land surface products or other CCI initiatives (e.g. the CCI-glaciers). This set of auxiliary products is used as ancillary information to produce the global LC maps or to characterize the LC conditions. Indeed, with regard to the different variables which describe the LC conditions (see section 3.2), the CCI-LC project does not generate all of them. The CCI-LC team attaches a particular importance to the consistency requirement (UR2) highlighted in the URD [AP-3] and develops methods to cross-check the internal consistency among the land surface information. Besides, one of the main objectives of the project is to contribute to at least to inform about the consistency between the different datasets (including the land cover one) that are used to derive models land surface parameters. But, generating all these datasets is certainly out of the scope of the project. In terms of products delivery, the CCI-LC project only focuses on information which is directly related to the climate modellers needs, such as the vegetation seasonality. For the other land surface parameters (water, snow, burnt area, LAI or albedo), the project may integrate them if they are compatible and available on time for the cross-check analysis. An exhaustive description of all the satellite and auxiliary datasets is provided in the DARD [AP-5], including technical specifications and availability conditions Information included in the global LC databases The classification process aims at transforming multispectral SR composites generated by the preprocessing steps into meaningful global LC products for the climate modellers point of view. The product usefulness is dependent on its thematic content as well as on its accuracy. With regard to the three global LC maps, they are associated with a legend combining key elements of LC state and conditions. The legend indeed counts a certain number of classes, which each corresponds to one LC state, i.e. to a stable ensemble of land surface features, plus some seasonality information derived from the LC condition. The legend is constrained by three main issues: (i)

26 1. rev.11 03/07/ information for all desired classes must be available from input observation datasets to derive a product of sufficient accuracy, (ii) the LC should be described using a standardized approach to ensure horizontal and vertical harmonization (i.e. across local to global scales and through multi-temporal analysis) and (iii) its thematic content should meet the climate models requirements. In order to ensure spatial and temporal harmonization, the global LC maps are based on the UN-LCCS [RD-9]. Their legend is mostly built on the GlobCover legend, with extensions to better discriminate between the different forest types according to their leaf type and phenology. The classes defined using UN-LCCS classifiers are listed in Table 3. A possible refinement of the LCCS-based legend could be the discrimination between C3 and C4 plants both for the croplands and the grasslands (classes 1, 2 and 13) if this is proved feasible. Furthermore, it is worth mentioning that a simplified LCCS-based legend will be considered for the validation process, in order to ensure higher robustness. This simplification will consider on the principles for harmonizing land cover information based on LCCS [RD-21] in order to also allow comparison with other existing land cover products. Table 3. Legend of the global LC maps based on LCCS Value Label Color 0 No Data 10 Cropland, rainfed 11 Herbaceous cover 12 Tree or shrub cover 20 Cropland, irrigated or post-flooding 30 Mosaic cropland (>50%) / natural vegetation (tree, shrub, herbaceous cover) (<50%) 40 Mosaic natural vegetation (tree, shrub, herbaceous cover) (>50%) / cropland (<50%) 50 Tree cover, broadleaved, evergreen, closed to open (>15%) 60 Tree cover, broadleaved, deciduous, closed to open (>15%) 61 Tree cover, broadleaved, deciduous, closed (>40%) 62 Tree cover, broadleaved, deciduous, open (15-40%) 70 Tree cover, needleleaved, evergreen, closed to open (>15%) 71 Tree cover, needleleaved, evergreen, closed (>40%) 72 Tree cover, needleleaved, evergreen, open (15-40%) 80 Tree cover, needleleaved, deciduous, closed to open (>15%) 81 Tree cover, needleleaved, deciduous, closed (>40%) 82 Tree cover, needleleaved, deciduous, open (15-40%) 90 Tree cover, mixed leaf type (broadleaved and needleleaved) 100 Mosaic tree and shrub (>50%) / herbaceous cover (<50%) 110 Mosaic herbaceous cover (>50%) / tree and shrub (<50%) 120 Shrubland 121 Evergreen shrubland 122 Deciduous shrubland 130 Grassland 140 Lichens and mosses 150 Sparse vegetation (tree, shrub, herbaceous cover) (<15%) 160 Tree cover, flooded, fresh or brakish water

27 1. rev.11 03/07/ Tree cover, flooded, saline water 180 Shrub or herbaceous cover, flooded, fresh/saline/brakish water 190 Urban areas 200 Bare areas 201 Consolidated bare areas 202 Unconsolidated bare areas 210 Water bodies 220 Permanent snow and ice Such LCCS approach has been found compatible with the PFT concept used by many models (UR7) [AP-3, RD-20]. Cross-walking tables have been built for different climate models and are included in the CCI-LC user tool (see section 6.3). In addition to the global LC maps, the CCI-LC project also provides information about LC conditions through the observation of selected variables on an annual basis. These variables are expected to provide information about the dynamic processes which affect the land cover and which are of interest for the CMC, as shown in Table 4. The way this information is provided in the LC condition products is presented in Table 5. Variables observed Table 4. List of variables and dynamic processes addressed in the CCI-LC project. Amount of vegetation, described by the normalized difference vegetation index (NDVI) Burnt areas Water coverage** Snow coverage Dynamic processes characterized by the variables Phenology, seasonality, productivity (both for natural vegetation and for croplands), number of cropping cycles, carbon fluxes Fire regimes, vegetation succession, carbon fluxes Water seasonality, floods, irrigation, distinction between permanent and seasonal water extent Snow seasonality, distinction between permanent and seasonal snow ** This variable is not currently addressed in this version of the project. It will be in the future. Table 5. Information provided to describe the variables in the CCI-LC condition products. Information provided in the products - Annual time profile averaged over multiple years, detailed on a 7-day frequency - Standard deviation over this multi-year period, for each 7-day interval - Number of valid and cloud-free years contributing to each 7-day period - The pixel status - Percentage of Burnt Areas occurrence over multiple years and detailed on a 7-day frequency - Number of valid years contributing to each 7-day period of the occurrence series Variables NDVI Burnt areas occurrence, water bodies occurrence**, snow occurrence Consistency information between all these variables and the LC maps** generated by All variables the CCI-LC project * Variables written in italic (LAI, albedo) are currently not addressed by the CCI-LC project but could be integrated if they are compatible and available from other projects on time. ** This variable is not currently addressed in this version of the project. It will be in the future.

28 1. rev.11 03/07/ It is very important to repeat that some variables are not directly produced by the project. They are either obtained from other on-going or forthcoming CCI projects (water bodies and burnt areas) or NASA projects (Snow Product MOD10A2) and combined NASA FP7 projects (MCD64A1 integrated in GFEDv3). As for the LC products accuracy, it is evaluated both through overall accuracy and temporal stability values. The accuracy targets are presented in Table 6, which compares them to the GCOS and users (CMUG and CCI-LC users) requirements. It shall be noted that the accuracy assessment only focuses on the LC state products, the LC condition ones being not validated by independent reference dataset. Table 6. Overall accuracy and temporal stability targets of the CCI-LC project Overall accuracy Stability GCOS CMUG CCI-LC users CCI GCOS CMUG CCI-LC users CCI >85% 90-95% 90-95% 80% >85% 90-95% >85% 80-85% With regard to the requirement of land cover change information expressed by the users (need for tracking human activities and defining history of disturbance, UR3), the land cover change must first be referred to as a permanent modification of the land cover state and not of the land cover condition in comparison with a baseline status. Further investigation is required to set precisely the modification level required to be considered as a change which is achievable with the current available sensors. As no specific land cover change product has been proposed in the baseline proposal, there is currently no plan to detect the land cover state change in a systematic and consistent way. The development of a specific processing chain was only foreseen as a CCI-LC option. Indeed, simple intercomparisons of classification products are well known to highlight the land cover classification instabilities rather than the land cover change, the respective land cover accuracy being below the land cover change rate. In spite the fact that the proposed global inter-annual products are expected to be more consistent than the previous global single-year LC products (GLC2000, GlobCover 2005 and 2009), any land cover change of the same order of magnitude than the MERIS spatial resolution can be hardly mapped through classification process. The currently proposed LC maps may allow depicting some major land cover change over certain hotspot areas (UR3). These differences are expected to be reported on the three LC maps corresponding to each epoch. However, this is not achieved neither in a systematic, nor a consistent way throughout the biomes. As for the LC condition products, they inform about seasonal reference dynamics of the different LC classes over a 15-year period and the associated intra-annual variability. Such information could possibly offer the opportunity to derive indications about LC changes but here again this is mainly expected for the hot-spot areas. Those detected land cover change areas locally document some drastic cover change which should be translated as such in a PFT transformation Spatial and temporal resolution of the global LC products The spatial and temporal resolutions of the global LC products have been determined based on the CMC needs expressed in the URD [AP-3].

29 1. rev.11 03/07/ In the temporal dimension, a double resolution is foreseen. Three LC maps are generated for three epochs separated by a 5-year interval (2000, 2005 and 2010) with the aim of mapping land cover in a consistent way over time. On the other hand, the LC conditions are described on shorter intervals over a cycle (typically a year) and these descriptions are averaged over time (around 13 years) to provide information about inter-annual variability. Coupling these two temporal resolutions was found to be a suitable trade-off between the needs of a best/stable LC map and of regular information about vegetation dynamics (UR1, UR4 and UR6). Considering that there is not one spatial resolution that fits all purposes [AP-3] and that flexibility is required to address different scales and purposes (UR4), the CCI-LC project delivers LC maps at their full spatial resolution (300m or 1km, depending on the epoch for the LC map and on the variables for the LC condition products) along with an aggregation tool. This aggregation tool is developed by the CCI-LC consortium and should allow users to aggregate the LC maps to the spatial resolution which is suitable for their models. Aggregation methodology is detailed in a User Guide Document provided with the tool. More information on this tool is found in section 6.3. Spatial and temporal resolutions of the CCI-LC global LC products are summarized in Table 7, which compares them to the GCOS and users (CMUG and CCI-LC users) requirements. Table 7. Spatial and temporal resolutions of the global LC products GCOS 250m 1km (accuracy better than 1/IFOV) Spatial resolution CMUG 300m 1km LC_CCI users 300m and coarser CCI GCOS CMUG 300m 1km Yearly 2-5 years Temporal resolution LC_CCI users Best stable map + annual updates CCI 5 years for the LC maps + 15 years with a 7-day frequency for the LC conditions Format, metadata, and projection The Coordinate Reference System (CRS) used for the global LC products is a geographic coordinate system (GCS) based on the World Geodetic System 84 (WGS84) reference ellipsoid and using a Plate Carree projection. The projection makes use of an equatorial radius (also called semi-major axis) of km and of a polar radius (also called semi-minor axis) of km. The inverse flattening parameter is of m. The coordinates are specified in decimal degrees. A complete description of the CRS is given in Figure 5 as an ISO WKT representation. Figure 5. Description of the coordinate reference system defining the global LC products The CCI-LC project delivers the LC maps both as global files.

30 1. rev.11 03/07/ The LC products (maps and conditions) are delivered in the NetCDF-4 format, i.e. the "classic model" of NetCDF since most tools can handle them. The file specification follows CF conventions [RD-17]. Global LC products are also delivered in the GeoTiff format for specific users of the land cover community. In the case of the NetCDF format, the geographic location information is contained in the NetCDF data. As for the GeoTiff format, it is included in the GeoTiff raster. Metadata of the CCI LC products are conforming to the CCI guidelines for data producers specification [RD-8] Global SR composite time series The SR products delivered by the CCI-LC project consist in MERIS global SR composite time series covering the period that are the input for the classification algorithms. The spectral content encompasses the 13 (without band 11 and 15) of the 15 MERIS spectral channels (Table 8) and the spatial resolution is of 300 m for MERIS FR and 1000m for MERIS RR. Table 8. MERIS spectral channels Band number Band centre (nm) Band width (nm) Use Yellow substance and detrital pigments Chlorophyll absorption maximum Chlorophyll and other pigments Suspended sediment, red tides Chlorophyll absorption minimum Suspended sediment Chlorophyll absorption and fluorescence reference Chlorophyll fluorescence peak Fluorescence reference, atmospheric corrections Vegetation, cloud O 2 R-branch absorption band Atmosphere corrections Vegetation, water vapour reference Atmosphere corrections Water vapour, land The time series are made of temporal synthesis obtained over a specific compositing period. The compositing period most suitable for the classification chain and so the temporal resolution of the SR products delivered by the project has been decided during the round-robin procedure to be 7 days. The exact schema for the 7-day periods is to start at January 1 and go on 7-day by 7-day periods until the end of the year. In this way, it should be noted that the last period of December comprises 8 days. As for leap years, the 7-day period including February 29 comprises 8 days. There are separate time series for MERIS FR and MERIS RR.

31 1. rev.11 03/07/ Observations acquired by the SPOT-VGT sensor are also used in order to increase the temporal coverage of the project over the years As a result, one other kind of global SR composite time series may also be produced and delivered by the CCI-LC project: SPOT-VGT global SR composite time series covering the period , if compatible with the possibilities offered by the CNES/VITO data distribution policy. Table 9 lists the satellite dataset that are planned to be used in order to generate the global SR composite time series. The way these global SR composite time series serve as input for producing the land cover products for the three different epochs is detailed in Table 2. A complete specification of these dataset is provided in the DARD [AP-5]. Table 9. Satellite data that are planned to be used to generate the CCI-LC SR time series Global SR composite time series MERIS global SR composite time series SPOT-VGT global SR composite time series Reference period Satellite data source Envisat MERIS FR & RR daily images SPOT-VGT P or S1 images* Technical specifications of the satellite data source 300-m or m resolution full swath 15 spectral bands in visible and near infrared (NIR) Global coverage 1-km resolution 4 spectral bands in visible and infrared Global coverage *SPOT-VGT S1 will be used for phase I. SPOT-VGT P is planned to be used for the phase.ii Projection and tilling system The CRS used for the global LC products is a GCS based on the WGS84 reference ellipsoid and using a Plate Carrée projection. The projection makes use of an equatorial radius (also called semi-major axis) of km and of a polar radius (also called semi-minor axis) of km. The inverse flattening parameter is of m. The coordinates are specified in decimal degrees. A complete description of the CRS is given in Figure 5 (section 4.1.3) as an ISO WKT representation. In order to simplify the handling and analysis of 300m spatial resolution global datasets, the SR time series are being delivered in tiles. Global products are subdivided into 72 x 36 tiles following the tiling system already used in the GlobCover project [RD-18] (Figure 6).

32 1. rev.11 03/07/ Figure 6. Description of the tiling system used for the SR products [RD-18] Tiles are 5 degrees by 5 degrees. The tile coordinate system starts at (0,0) (85N180W) (horizontal tile number, vertical tile number) in the upper left corner and proceeds right (horizontal) and downward (vertical). The tile in the bottom right corner is (71, 35) (90S175E). A tile is physically represented by a single NetCDF file whose file name also indicates the tile south-west corner (see section 6.2 for a complete description of the naming convention). In addition, tiles having no land contribution are not delivered Format and metadata All the SR products are delivered in the NetCDF-4 format, using the "classic model" of NetCDF with compression. This is the desired target format for CCI [AP-2]. SR products are delivered with associated metadata included in the NetCDF header. It follows the CCI metadata convention [RD-8].

33 1. rev.11 03/07/ Quality documentation Tracking products uncertainties and documenting their quality are specific requirements for all the CCI projects [AP-1, AP-2], and this exercise is a critical step for the acceptance of the products. The uncertainties of the CCI-LC products are characterized on the one hand through quality control procedures and on the other hand thanks to the validation exercise (which only concerns the LC products) [AP-8]. An uncertainty budget is performed, which consist in a list of random and systematic errors, associated with estimates of the uncertainty they contribute to the measurement [AP-2]. The results of these quality assessments are reported in synthetic documents made available to the users along with the metadata (UR8) Global LC products The LC products quality depends on the input data quality and on the classification processes. The quality information available for the SR products are used and complemented by quality indicators specific to (i) the LC classification process and (ii) the consistency between the LC map and the LC conditions (UR8). This set of quality indicators is provided as additional layers in the LC maps and LC condition products. The independent validation exercise, which is achieved by JRC and which is described in the Product Validation Plan (PVP) [AP-8], also provides confidence and accuracy information about the LC map. These different sources of quality assessment are described in the following sections Global LC map For the LC maps, the quality is documented using a set of five flags and values, determined on a perpixel basis. 1. The pixel has been processed or not a. If the pixel has been processed: i. The name of the sensor from which the reflectance value has been derived; ii. An overall quality indicator describing the reflectance value (derived from the pre-processing and the SR product). b. If the pixel has not been processed, the reason is detailed (no data, cloud effect, other effect) 2. The status map of the pixel: this information is produced during the pre-processing, through the pixel identification step [AP-6]. Five statuses are foreseen: Cloud, Clear land, Clear water, Clear snow and Invalid ; 3. The number of clear observations available for the classification; 4. Overall LC assessment: summary indicator, providing an overall confidence level.

34 1. rev.11 03/07/ Global LC condition product For the LC condition products, their accuracy is not documented as it is not the objective of the CCI- LC project but rather, their internal consistency. For each LC condition product (NDVI, snow and burnt area and possibly, water, LAI and albedo), a set of flags are used to document how the reference LC condition has been determined (see section and Table 5) and its consistency with the other datasets. In total, 2 quality flags are provided: 1. The number of years of observation which has been used to calculate the reference status: this information is necessary since this reference status is derived from a multi-year dataset (either as an average annual time profile or as an occurrence probability, as explained in Table 6). The number of observations used to calculate this reference value is provided for each 7-day interval. 2. Consistency assessment between condition products: indicator of consistency between condition products according to a set of pre-defined rules. This consistency assessment will be provided only with the public release of the conditions Independent validation process The LC map is subject to an independent validation process, made of 4 major components: a confidence-building procedure, a statistical products validation, a comparison with other existing LC products and an assessment of the products temporal consistency. The general relationships between these different components are shown in Figure 7. A detailed description of the overall validation process is provided in the PVP [AP-8]. Figure 7. Organization and links between the different validation components As illustrated in Figure 7, the confidence-building is a particular procedure since it is not really part of the validation process. It should rather be considered as a component of the classification process. Indeed, the results of the analysis are employed for removing errors and improving the map. The