Land Use and Land Cover Mapping in Europe

European policy aims, wherever possible, to safeguard and improve the citizens’ quality of life through directives and regulations. Protection of the environment is a high priority, which involves the rational usage of resources through risk assessment and threat reduction with increasingly important focus areas being food and water security. The delivery of policy requires Information Services that are based on a combination of data analysis and modelling from Earth Observation satellites as well as ground-based in-situ sensor networks. These Services need to operate in an integrated manner to provide geographically wide-area and cross border harmonized products. Remote Sensing is a key element of the Service, especially with the requirement for near real time information together with monitoring changes and supporting/testing future scenarios. An overview of European initiatives is provided, primarily enacted through the Copernicus programme, focusing on the land monitoring. It outlines both the potential and limitations of existing approaches, including existing service specifications and operational requirements.

The observation of global-scale land cover (LC) is of importance to international initiatives such as the United Nations Framework Convention on Climate Change (UNFCCC) and Kyoto protocol, governments, and scientific communities in their understanding and monitoring of the changes affecting the environment, and the coordination of actions to mitigate and adapt to global change. As such, reliable and consistent global LC (GLC) datasets are being sought. For instance, GLC datasets are used as an input for many Global Circulation Models, Earth Systems Models and Integrated Assessment Models used for global and regional climate simulations, dynamic vegetation modelling, carbon (stock) modelling, ecosystem modelling, land surface modelling, and impact assessments (Hibbard et al., 2010; Herold et al., 2011).

From the different communications published by the EU, policy makers and public authorities have been defined as the major users of the GMES land initiative. In parallel, national mandated bodies and research institutions have been in charge of the development of the national land monitoring program and of the development of the CORINE Land Cover program at the EU level, which has been included as one of the GMES activities.

Defined by their geographical influence, the users in the current European land monitoring context are divided into:

– High-level European and International user organizations, which influence and support the production of uniform European wall-to-wall products;

– National and regional/local user organizations, which are more focused on determining the added value of the services being supplied, establishing what would be their general needs and provide access to in-situ data.

In the GMES program, there has been variable involvement of user organizations linked to the different projects resulting in variable levels of satisfaction with the project outcomes. Involvement ranged from partnerships between service providers and users where users were actively engaged in establishing requirements and service specifications, service reviews and final assessment of the utility of the product; to a more theoretically and political involvement, where users would have little opportunities to participate in the development of the services.

The current global economic crisis forces nations to pinch and scrape on all their expenses. At the same time the demand for more and better information on the state of environment is increasing, together with more obligations to report on development and to estimate the impacts of environmental and spatial planning policies. Both needs may motivate a much stronger use of remote sensing, as so-called “free data” covering large regions or even continents become more and more important.

CORINE Land Cover (CLC) was specified to standardize data collection on land in Europe to support environmental policy development. Since the late 1980s, three iterations of this land cover inventories have been realised (timed around 1990, 2000 and 2006). The 4th inventory recently commenced, as part of the GIO land (GMES Initial Operations Land) project. CLC is widely used in indicator development, environmental modelling and land use/land cover change analysis in the European context. The number of participating countries has increased over time – currently 39, including 32 European Environment Agency (EEA) member countries and seven cooperating countries, with a total area coverage of 5,8 Mkm². Ortho-corrected high spatial resolution satellite images provide the geometrical basis for mapping. In-situ data (e.g. topographic maps, ortho-photos, and ground survey data) are essential ancillary information. The project is mainly co-financed by the European Commission and participating countries and implemented by national teams under the management and quality control of EEA. The basic technical parameters of CLC (i.e. nomenclature, 25 ha minimum mapping unit and 100 m minimum mapping width) have not changed since the beginning of the project; therefore the results of the different inventories are comparable. The method of mapping has, however, changed significantly. Working on plastic overlay at the start was fully replaced by computer assisted photo-interpretation (CAPI). In the Scandinavian countries and recently in Germany and Ireland, labour-intensive photo-interpretation was partly substituted by a semi-automatic classification methodology based on the integration of existing land use data, satellite image processing and generalization. CORINE Land Cover Changes (CLCC) are derived from satellite imagery by the direct mapping of changes based on image-to-image comparison. Change mapping applies a 5 ha MMU to pick up much more detail in CLCC layer than in CLC status layer. Two European validation studies have shown that the achieved accuracy is above the specified minimum (85 %) for CLC, as well as for CLCC. Results of the three CLC inventories can be downloaded from the EEA Data Service free of charge for all users. One of the proposed ways to increase the value of CLC in the future is to populate landscape level objects of the CLC database with high spatial resolution land cover features. The GIO land project is making the first steps towards this.

Initiated in 2007, the Area Frame Sampling Europe subtask of the Seasonal and Annual Change Monitoring Service (SATChMo) Core Service in the geoland2 project delivered its final products in 2012. Three of these are described in this paper: (i) an Area Frame Sampling scheme design that aims at optimizing the statistical accuracy when extrapolating a land cover classification at reasonable cost, (ii) a semi-automatic classification tool that is able to discriminate 10 land cover classes on VHR samples with 0.25ha minimum mapping unit (MMU), and (iii) a highly automatized change detection tool based on Multivariate Alteration Detection (MAD) approach that additionally employs Normalised Difference Vegetation Index (NDVI) and texture characteristics. This later step, as well as giving change/no-change mask provides directions of changes in three main categories, artificialization, revegetation, and devegetation. The algorithms were cast into the form of production chains starting with data acquisition and processing, through the main processing to validation and product dissemination via Spatial Data Infrastructure servers. The whole process was tested on representative set of 114 sites from across the European Union (EU).

This article provides a comprehensive overview of Earth Observation based forest monitoring approaches in Europe, from early approaches using low- and medium-resolution satellite data to most recent operational monitoring systems developed in the frame of the European earth observation programme Copernicus/GMES, on which a special focus is laid. First, the pan-European forest-related information contained in CORINE Land Cover as well as the rich portfolio of regional, national and pan-European-scale Forest services and products implemented by the ESA-funded GMES Service Element Forest Monitoring are discussed. Together with the pan-European forest products provided by the Joint Research Centre’s European Forest Data Centre, GSE FM had laid the foundations for the subsequent pre-operational stage of European Forest Monitoring undertaken by the European Commission-funded FP7 Land core project geoland2. This has successfully demonstrated a set of Core Mapping Service and Core Information Service forest products in multi-national demonstration sites over Europe, finally leading to first-time operational mapping of a pan-European High-Resolution Forest Layer in the frame of the GMES Initial Operations Land, being implemented by the European Environment Agency since 2012. The evolution, product specifications and differences between these Copernicus Core Service forest products are discussed. Additionally, customized regional forest Downstream services as designed by the FP7 project EUFODOS are presented. Conclusions and an outlook towards future forest monitoring approaches are provided for the upcoming Sentinel-2 era.

The origin of the Urban Atlas answers to a European Commission need for detailed and comparable information related to cities. The urban information is so far provided by local and national statistics or using data extracted from land cover maps as CORINE Land Cover. CORINE Land Cover provides a resolution of 1:100,000 (25 ha) and is clearly unsatisfactory for the demand of the Commission services. As in Europe, urban areas accommodate more than three-quarters of the population and these areas have grown rapidly in recent decades, there is an urgent need for pan-European, reliable and inter-comparable urban planning data. The Urban Atlas, developed as part of GMES (Global Monitoring for Environment and Security) brings exactly that.

This chapter presents an extensive review on the techniques proposed in the recent literature for the classification of remote sensing (RS) images. Automatic classification techniques for obtaining land-cover maps from RS images are usually based on supervised learning methods. Accordingly, we focus our attention on supervised techniques for the classification of different types of RS images acquired by new generation satellite sensors. Initially we analyze the critical problems related to different types of RS data and review the classification techniques that can overcome these problems. Then, the most recent methodological developments related to classification techniques in RS are addressed by focusing on semisupervised learning, active learning and domain adaptation approaches. Finally, the most promising research directions in RS data classification are discussed.

Change detection is a key methodology in remote sensing. Despite numerous important papers dedicated to this topic no comprehensive review of the subject currently exists. Most existing reviews are limited to certain fields of application or are simply no longer current. Here, we provide an overview of the most important algorithms used for change detection and their development and refinement through time. Change detection cannot be considered as a mere algorithm. We thus describe all of the steps necessary to perform change detection as part of a process chain. We also show how such a change detection process chain may be adapted to specific individual requirements. The labeling of changes is recognized as a basic part of this process. Three change labeling categories are introduced: (1) pre-change extraction labeling, (2) concurrent labeling, and (3) post-change extraction labeling. Methods developed specifically for use with synthetic aperture radar (SAR) data are a focus of this review. SAR data provide information compatible with data produced by optical sensors, but also often require specialized processing techniques and are useful within a unique field of application. An examination of time series analysis methods is also included in this review. To date, these techniques have not been considered in reviews, but the increasing availability of remote sensing data as well as recent advances in remote sensing change detection make it essential that they are included here. Although not exhaustive, this review is intended to provide a comprehensive overview of well established change detection methods as well as recent advances in this field.

Nowadays Earth Observation (EO) systems play a major role in supporting environmental programs and monitoring compliances, such as the European Global Monitoring for Environment and Security (GMES – Copernicus) program. Copernicus aims on the provision of reliable and current information on our planet and its environmental state in three Earth system domains “Land”, “Atmosphere”, and “Marine”. Moreover, the corresponding services support the management of humanitarian crises, natural disasters and man-made crisis (Aschbacher J, Milagro-Pérez MP, Remote Sens Environ 120:3–8, 2012). Products in the Land-domain, comprise accurate and cross-border harmonized information on land cover and land cover change, including information on seasonal and annual changes, the vegetation state and the monitoring of the water cycle. Overall these products will support decision-making and various monitoring applications in context of land use and land cover change, water quality, spatial planning, and global food security (GMES, http://​www.​gmes.​info/​pages-principales/​services/​land-monitoring, Last accessed May 2012, 2012).

Since the launch of commercial software for object-oriented data analysis numerous research and application works were undertaken, in order to apply this concept and elaborate semi-automatic methods for land cover classification based on satellite images. The research works were concentrated on two main aspects of object-oriented approaches: multi-resolution segmentation to adjust objects to terrain elements in an optimal way and on classification methods, exploiting comprehensively spectral, spatial and textural features of image objects, as well as their mutual relationships. Applications range from studies using multi-resolution satellite data (Whiteside T, A multi-scale object-oriented approach to the classification of multi-sensor imagery for mapping land cover in the top end. Proceedings of NARGIS 2005 – application in tropical spatial science. 4th–7th July 2005 Charles Darwin University, Darwin, NT, Australia. http://​www.​ecognition.​com/​sites/​default/​files/​272_​18.​1_​20whiteside_​_​20tim.​pdf, 2005) to very high-resolution images (QuickBird, Ikonos), which enabled more effective analysis of texture and shape features (Wei W, Chen X, Ma A, Object-oriented information extraction and application in high-resolution remote sensing image. Proceedings of the IGARSS 2005 symposium. Seoul, Korea. July 25–29, 2005, pp 3803–3806, 2005; Kressler FP, Steinnocher K, Kim YS, Enhanced semi-automatic image classification of high-resolution data. Proceedings of the IGARSS 2005 symposium. Seoul, Korea. July 25–29, 2005. http://​www.​ecognition.​com/​sites/​default/​files/​240_​igarsskressler2.​pdf, 2005; de Kok R, Wężyk P, Principles of full autonomy in image interpretation. The basic architectural design for a sequential process with image objects. In: Blaschke Th, Lang S, Hay GJ (eds) Object-based image analysis. Series: lecture notes in geoinformation and cartography. Springer, Berlin/Heidelberg, ISSN: 1863–2246, pp 697–710, 2008; de Kok R, Wezyk P, Weidenbach M, The role of edge objects in full autonomous image interpretation. Proceedings of GEOBIA 2008 – pixels, objects, intelligence, GEOgraphic object based image analysis for the 21st century, Calgary, Alberta, Canada. http://​www.​isprs.​org/​proceedings/​XXXVIII/​4-C1/​Sessions/​Session1/​6717_​DeKok_​Proc_​pap.​pdf, 2008). A common classification approach was based on applying training areas for particular land cover classes and a Standard Nearest Neighbour Classifier to assign objects to land cover categories (Yuan F, Bauer ME, Mapping impervious surface area using high resolution imagery: a comparison of object-based and per pixel classification. Proceedings of ASPRS 2006 annual conference, Reno, Nevada; May 1–5, 2006. http://​www.​ecognition.​com/​sites/​default/​files/​184_​asprs2006_​0178.​pdf, 2006; Hajek F, Object-oriented classification of remote sensing data for the identification of tree species composition. Proceedings of ForestSat 2005 conference, May 31–June 3, 2005, Boras, Sweden. http://​www.​ecognition.​com/​sites/​default/​files/​229_​forestsat2005_​20_​_​20filip_​20hajek.​pdf, 2005; Elmqvist B, Ardo J, Olsson L, Int J Remote Sens 29(24):7129–7140, 2008). Alternative approaches for the classification process comprise the use of parametric values of spectral and texture type as well as hierarchical classification workflows, based on a decision tree method (Lewinski St, Bochenek Z, Rule-based classification of SPOT imagery using object-oriented approach for detailed land cover mapping. Proceedings of 28th EARSeL symposium “Remote sensing for a changing Europe”, Istanbul, Turkey, 2–5 June 2008, pp 197–204, 2008; Lucas R, Rowlands A, Brown A, Keyworth S, Bunting P, ISPRS J Photogramm Remote Sens 62(3):165–185, 2007; Su W, Li J, Chen Y, Liu Z, Zhang J, Low TM, Suppiah I, Hashim SAM, Int J Remote Sens 29(1):3105–3117, 2008). The presented work on an object based classification approach emerged from the needs formulated within the Geoland 2 SATChMo Core Mapping Service.

A rapidly changing environment with land use and climate as the most dynamic components causes new challenges for nature conservation and management of protected areas. Dealing with these changes requires a systematic monitoring. To date, such monitoring programs are mostly backed by expert guess or permanent observation plots. Both have their merits but the plot-based approach is certainly more objective. However, even in the case of appropriate sampling, plots provide only punctual information and changes in the area between plots are easily missed. This gap can be closed by remote sensing.

New challenges due to changing climatic and environmental conditions, economic and demographic polarization and new energy concepts require smart tools for decision makers and regional and urban planners especially in the context of growing cities. Urban growth models can be valuable tools in order to define future policy alternatives or to analyze environmental impacts of urban growth. This paper provides a concise introduction into the challenges emerging from urban sprawl and introduces methods and technologies that enable a deeper understanding of the processes of urban sprawl. It is focused specifically on European urban areas and demonstrates how empirical research and remote sensing can contribute to the development of improved urban growth models. An example application is presented in order to provide a summary of current research activities in the field of integrated urban modeling.

The project Land Information System Austria (LISA) demonstrates the feasibility of an Austrian-wide homogeneous land cover and land use monitoring solution. State of the art technologies in remote sensing and GIS combined with the effective integration of national spatial data infrastructure (orthophotos, airborne laser scanning data and other countrywide available geodata) are used in combination with European GMES/Copernicus data infrastructure to meet the requirements of a modern land monitoring system, specified by governmental institutions on state, provincial and municipality level. The conceptual basis consists of an object oriented data model that covers the two main land monitoring issues separately: land cover and land use. Whereas land cover is derived solely based on remote sensing data, the land use data are assessed as synoptic view from various thematic databases (spatial zoning plans, IACS, nature conservation, road network etc.) and land cover classifications. The applied technical approach is the result of various research activities. Currently, the countrywide realization of LISA in form of an administrative national collaboration across various hierarchical administration levels (regional and national) as well as across various thematic contributions (cadastral mapping, spatial planning, agriculture, etc.) is discussed.

National land monitoring activities are moving towards a harmonized integration of national datasets with pan-European LC/LU initiatives. A growing number of European countries derive their contributions to pan-European land cover and land use datasets following the principles of a bottom-up approach. In the case of Germany this concept is put into practice by using the topographic reference data of the Federal Stateś land surveying authorities (ATKIS ® Basis-DLM) as the main input to create a homogeneous LC/LU dataset for the entire country, supported by remote sensing methods. For the reason of temporal consistency the reference year of DLM-DE production has been synchronized with the Corine Land cover (CLC) phase 2012.

DLM-DE is a cooperative project between the Federal Environment Agency (UBA) in the role of the National Reference Center for land cover (NRC) and the Federal Agency for Cartography and Geodesy (BKG) as the data producer. A key criterion in the DLM-DE update procedure is the separated mapping of land cover and land use information. Through a semantic transformation the ATKIS feature types are translated to a preliminary coding of land cover and land use. This intermediate classification distinction between land cover and land use is assigned to the polygon features of Basis-DLM as LC and LU codes. During the update process these codes are verified or changed through semi-automated analysis and interpretation of multi-temporal satellite imagery, which is used as the main source of information.

The result of the DLM-DE is a high resolution vector data set describing LC and LU with a minimum mapping unit of 1 ha. The intermediate description is applied in a bidirectional way. On the one hand, it serves for deriving CLC classes by combining the encoded LC and LU information. On the other hand, it is also flexible enough to partly transform the updating results back to the original vector data source of Basis-DLM. On the long-term run the aim is to promote further applications of DLM-DE in the field of land monitoring and beyond.

The United Kingdom (UK) has a long heritage in the development and application of land cover mapping approaches based on Earth Observation (EO) data. This began in 1990 with a simple pixel-based classification and the latest iteration in 2007 was an object-based map fully aligned with the national cartographic dataset. During this time the UK has adopted innovative processing methods and data specifications and applied them operationally at the national scale. The UK mapping is now being integrated with the European activity through the use of GMES data and services and coordinated delivery of European mapping products.

This chapter deals with long-term Dutch national land cover and land use mapping activities. Land cover and land use are often difficult to separate 1:1. For practical reasons in this chapter only the term land cover will be used with the exception of direct translations of the names of database. Four main databases can be discerned:

All four databases differ in spatial detail, semantics and update frequency. Also the format of the data, the history and application focus of the databases vary between each other. All databases use aerial photography and/or satellite imagery for their classifications. The LGN database is presented in more detail, i.e. the history, the production methodology and the land cover changes. The LGN database has a long history offering a unique possibility to follow developments in land cover for more than 20 years. Real land cover changes are monitored and can be separated from methodological changes. The differences between LGN and CORINE Land Cover database for the Netherlands are discussed here regarding spatial detail, temporal frequency and semantic mapping. The semantic relations between both databases are of the 1:1, 1:m and m:1 type.

Land cover and land use form baseline data in order to understand environmental changes in space and time and to make informed decisions. Categorisations assist in the communication of ideas and concepts related to land cover and land use. Few categorisations take formal shape or any formal algorithm, even fewer are standardised. The Land-Cover Classification System (LCCS) is a formal system with a parameterised approach, i.e. the parameters used to define categories and classes are explicit.

LCCS has been used in a number of Eastern European countries in the period of transition when spatial developments were rapid as a consequence of the land reform choices made in these countries. With time LCCS was applied not only for the countrywide inventory of land-cover/use types but for dedicated studies at detailed level in specific areas. LCCS has the advantage that the same concept is used to generate more detailed classes and these have an intrinsic hierarchical order. For monitoring and evaluation purposes the parameterised approach furnishes the parameters to be measured over time. Thus, the applications in which LCCS was used became more refined with the increasing experience and capabilities in the countries.

The experiences in Eastern Europe contributed to a certain extent to the fact that LCCS is now being used at European level in the Land Parcel Identification System Quality Assessment. Furthermore, the fundamental structure of LCCS became the ISO standard structure for classification systems in 2009 and LCCS was further used as the basis for the Land Cover Meta Language established as an ISO standard in 2012.

Since 2000 the image analysis in remote sensing made a move towards an object-oriented approach, databases became object-oriented, but what about categorisation? The way forward is to develop a new generation of categorisation systems adopting the Parametric Object-Oriented Model because this allows unprecedented flexibility and capability in the design and use of very complex information systems that are needed to understand the multifaceted environmental changes.

The implementation of productive and sustainable cultivation procedures is a major effort regarding the agricultural production in the European Community. However, political, economic and environmental factors impact the cultivation strategies directly and indirectly, and therewith strongly determine the condition and transformation of the cultivated and natural landscape. To assess the actual status, identify basic trends and mitigate major threats with respect to the agricultural production and its impact on the cultural and natural landscape, a frequent and area-wide monitoring of cropland and grassland is required. Satellite-based earth observation (EO) provides ideal capabilities for the area-wide and spatially detailed provision of up-to-date geo-information on the agricultural land use and the properties of the cultivated landscape. A specific benefit of EO is given by analysing multi-seasonal data acquisitions. Intra-annual time series facilitate the analysis of the phenological behaviour of the main crop and grassland types – key information with respect to the characterisation of the land use intensity and its impacts on the environment.

Experimental results for a test area in Germany assess the effectiveness of the proposed approach and demonstrate that a multi-scale and multi-temporal analysis of satellite data can provide spatially detailed and thematically accurate geo-information on crop types and the cropland-grassland distribution, respectively.

An enhanced long term remote sensing based data set for Green Vegetation Fraction (GVF) was created for the Mediterranean area. The dataset contains 10-day composites of GVF for the time period 1989–2005 on a scale of 0.01°, covering the Mediterranean basin. The MEDOKADS data set was employed to create mixture triangles of NDVI and surface temperature, of which three abundances, the “vegetated”, “non-vegetated” and “cold” abundance were derived. The vegetated abundance was eventually converted to GVF. Compared to NDVI, clear improvements have been made for GVF, in particular in respect to the mitigation of undesired effects of bad atmospheric conditions. GVF can be derived in an almost fully operational way, which enables it as base data for monitoring vegetation and related purposes. The data has been successfully employed in two case studies on olive farming intensity and rural land abandonment.

This chapter provides an overview of methods used for the extraction of biophysical vegetation variables from remote sensing imagery. It starts with the description of the main spectral regions in the optical window of the electromagnetic spectrum based on typical spectral signatures of land surfaces. Subsequently, the merit and problems of using radiative transfer models to describe the relationship between spectral measurements and biophysical and chemical variables of vegetation are described. Next, the use of statistical methods by means of vegetation indices for the same purpose gets attention. An overview of different types of indices is given without having the ambition in being exhaustive. Subsequently, an overview is provided of the biogeophysical vegetation variables that can directly be estimated from optical remote sensing observations, with emphasis on using vegetation indices. These vegetation variables are: (1) chlorophyll and nitrogen, (2) vegetation cover fraction and fAPAR, (3) leaf area index, and (4) canopy water. Finally, an outlook for a major research direction in the near future in this context is provided.

Understanding the impact of land transformation processes on ecosystem services (ESS) is an essential prerequisite for drafting and implementing sustainable land management concepts. This study presents an analysis of land transformation processes in Horqin Sandy Lands, one of the dry areas in Inner Mongolia (China). It aims at demonstrating the impacts of governmental management policies on land use change and its impact on the long-term availability of important ecosystem services. Spectral mixture analysis is applied to a calibrated time series of Landsat-TM/ETM+ images which covers a period of 20 years (1987–2007); the mixture model comprises three spectral end-members (Green Vegetation, Mobile Sand, Water) which are conceived as surrogates for important ecosystem services. Changing land surface conditions are identified through linear trend analysis of end-member proportions and by mapping the spatial extension of specific surface types at subsequent dates within the observation period. For translating the derived change rates into readjustments of selected ESS-indicators a simple linear model is proposed. Fuelled by long-term satellite observations, the synoptic representation of changing ecosystem services forms the basis for addressing synergies and trade-offs between ecological and societal well-being. The case of Horqin Sandy Lands, where new land use concepts are implemented by promoting selected ecosystem services at the cost of others, provides a striking example for these mechanisms.

Tropical peat swamp forests are among the most carbon rich ecosystems and are therefore in the focus of the program Reducing Emissions from Deforestation and Forest Degradation plus (REDD+), which requires accurate aboveground biomass (AGB) estimations for emission assessments. The present study evaluates different SAR frequencies and polarizations as well as spectral mixture analysis (SMA) of optical satellite data for estimating AGB on the basis of forest inventory data. Furthermore, two different approaches, continuous and discrete AGB estimation, were compared for their performance to retrieve large scale AGB estimations on the basis of multispectral data. The continuous approach relates satellite signals to field derived AGB values, while the discrete approach is based on a land cover classification by linking AGB values to each land cover class (stratify & multiply). Using the continuous approach, a correlation between AGB and SAR backscatter was found with TerraSAR-X HH and ALOS PALSAR HV data while no significant correlation was found with RADARSAT-2 HH, HV and ALOS PALSAR HH imagery. RapidEye matched filtering (MF) fractions derived from reference spectra of green vegetation (GV), soil and non-photosynthetic vegetation (NPV) were determined and all three MF fractions showed a correlation to AGB. The combined TerraSAR-X HH and ALOS PALSAR HV polarized AGB model was more accurate (r² = 0.68) than the single-frequency models. Similarly, the combined multispectral MF fractions model was also more accurate (r² = 0.92) than the single MF fractions models, while a synergistic use of SAR and multispectral data produced no improvements in AGB estimation. The comparison of the continuous and discrete approach based on multispectral imagery showed that the continuous AGB estimation approach depicted the spatial variability within one land cover class in contrast to the discrete approach, while it suffered from saturation either of the SAR backscatter or of the MF fractions in the higher biomass ranges. Such evaluation of AGB estimations using different satellite data and approaches considering the accuracy is considered as a relevant step for recommendations on defining MRV (Monitoring, Reporting and Verification) methods with regard to REDD+.