Zisis I. Petrou

The Earth Observation Data for Habitat Monitoring (EODHaM) System

To support decisions relating to the use and conservation of protected areas and surrounds, the EU-funded BIOdiversity multi-SOurce monitoring System: from Space TO Species (BIO_SOS) project has developed the Earth Observation Data for HAbitat Monitoring (EODHaM) system for consistent mapping and monitoring of biodiversity. The EODHaM approach has adopted the Food and Agriculture Organization Land Cover Classification System (LCCS) taxonomy and translates mapped classes to General Habitat Categories (GHCs) from which Annex I habitats (EU Habitats Directive) can be defined. The EODHaM system uses a combination of pixel and object-based procedures. The 1st and 2nd stages use earth observation (EO) data alone with expert knowledge to generate classes according to the LCCS taxonomy (Levels 1 to 3 and beyond). The 3rd stage translates the final LCCS classes into GHCs from which Annex I habitat type maps are derived. An additional module quantifies changes in the LCCS classes and their components, indices derived from earth observation, object sizes and dimensions and the translated habitat maps (i.e., GHCs or Annex I). Examples are provided of the application of EODHaM system elements to protected sites and their surrounds in Italy, Wales (UK), the Netherlands, Greece, Portugal and India.

A rule-based classification methodology to handle uncertainty in habitat mapping employing evidential reasoning and fuzzy logic☆

Abstract Habitat mapping is a core element in numerous tasks related to sustainability management, conservation planning and biodiversity monitoring. Land cover classifications, extracted in a timely and area-extensive manner through remote sensing data, can be employed to derive habitat maps, through the use of domain expert knowledge and ancillary information. However, complete information to fully discriminate habitat classes is rarely available, while expert knowledge may suffer from uncertainty and inaccuracies. In this study, a rule-based classification methodology for habitat mapping through the use of a pre-existing land cover map and remote sensing data is proposed to deal with uncertainty, missing information, noise afflicted data and inaccurate rule thresholds. The use of the Dempster–Shafer theory of evidence is introduced in land cover to habitat mapping, in combination with fuzzy logic. The framework is able to handle lack of information, by considering composite classes, when necessary data for the discrimination of the constituting single classes is missing, and deal with uncertainty expressed in domain expert knowledge. In addition, a number of fuzzification schemes are proposed to be incorporated in the methodology in order to increase its performance and robustness towards noise afflicted data or inaccurate rule thresholds. Comparison with reference data reveals the improved performance of the methodology and the efficient handling of uncertainty in expert rules. The further scope is to provide a robust methodology readily transferable and applicable to similar sites in different geographic regions and environments. Although developed for habitat mapping, the proposed rule-based methodology is flexible and generic and may be well extended and applied in various classification tasks, aiming at handling uncertainty, missing information and inaccuracies in data or expert rules.

Expert knowledge for translating land cover/use maps to General Habitat Categories (GHC)

Monitoring biodiversity at the level of habitats and landscape is becoming widespread in Europe and elsewhere as countries establish international and national habitat conservation policies and monitoring systems. Earth Observation (EO) data offers a potential solution to long-term biodiversity monitoring through direct mapping of habitats or by integrating Land Cover/Use (LC/LU) maps with contextual spatial information and in situ data. Therefore, it appears necessary to develop an automatic/semi-automatic translation framework of LC/LU classes to habitat classes, but also challenging due to discrepancies in domain definitions. In the context of the FP7 BIO_SOS (www.biosos.eu) project, the authors demonstrated the feasibility of the Food and Agricultural Organization Land Cover Classification System (LCCS) taxonomy to habitat class translation. They also developed a framework to automatically translate LCCS classes into the recently proposed General Habitat Categories classification system, able to provide an exhaustive typology of habitat types, ranging from natural ecosystems to urban areas around the globe. However discrepancies in terminology, plant height criteria and basic principles between the two mapping domains inducing a number of one-to-many and many-to-many relations were identified, revealing the need of additional ecological expert knowledge to resolve the ambiguities. This paper illustrates how class phenology, class topological arrangement in the landscape, class spectral signature from multi-temporal Very High spatial Resolution (VHR) satellite imagery and plant height measurements can be used to resolve such ambiguities. Concerning plant height, this paper also compares the mapping results obtained by using accurate values extracted from LIght Detection And Ranging (LIDAR) data and by exploiting EO data texture features (i.e. entropy) as a proxy of plant height information, when LIDAR data are not available. An application for two Natura 2000 coastal sites in Southern Italy is discussed.

Remote sensing for biodiversity monitoring: a review of methods for biodiversity indicator extraction and assessment of progress towards international targets

Recognizing the imperative need for biodiversity protection, the convention on biological diversity (CBD) has recently established new targets towards 2020, the so-called Aichi targets, and updated proposed sets of indicators to quantitatively monitor the progress towards these targets. Remote sensing has been increasingly contributing to timely, accurate, and cost-effective assessment of biodiversity-related characteristics and functions during the last years. However, most relevant studies constitute individual research efforts, rarely related with the extraction of widely adopted CBD biodiversity indicators. Furthermore, systematic operational use of remote sensing data by managing authorities has still been limited. In this study, the Aichi targets and the related CBD indicators whose monitoring can be facilitated by remote sensing are identified. For each headline indicator a number of recent remote sensing approaches able for the extraction of related properties are reviewed. Methods cover a wide range of fields, including: habitat extent and condition monitoring; species distribution; pressures from unsustainable management, pollution and climate change; ecosystem service monitoring; and conservation status assessment of protected areas. The advantages and limitations of different remote sensing data and algorithms are discussed. Sorting of the methods based on their reported accuracies is attempted, when possible. The extensive literature survey aims at reviewing highly performing methods that can be used for large-area, effective, and timely biodiversity assessment, to encourage the more systematic use of remote sensing solutions in monitoring progress towards the Aichi targets, and to decrease the gaps between the remote sensing and management communities.

Synergy of airborne LiDAR and Worldview-2 satellite imagery for land cover and habitat mapping: A BIO_SOS-EODHaM case study for the Netherlands

Abstract A major challenge is to develop a biodiversity observation system that is cost effective and applicable in any geographic region. Measuring and reliable reporting of trends and changes in biodiversity requires amongst others detailed and accurate land cover and habitat maps in a standard and comparable way. The objective of this paper is to assess the EODHaM (EO Data for Habitat Mapping) classification results for a Dutch case study. The EODHaM system was developed within the BIO_SOS (The BIOdiversity multi-SOurce monitoring System: from Space TO Species) project and contains the decision rules for each land cover and habitat class based on spectral and height information. One of the main findings is that canopy height models, as derived from LiDAR, in combination with very high resolution satellite imagery provides a powerful input for the EODHaM system for the purpose of generic land cover and habitat mapping for any location across the globe. The assessment of the EODHaM classification results based on field data showed an overall accuracy of 74% for the land cover classes as described according to the Food and Agricultural Organization (FAO) Land Cover Classification System (LCCS) taxonomy at level 3, while the overall accuracy was lower (69.0%) for the habitat map based on the General Habitat Category (GHC) system for habitat surveillance and monitoring. A GHC habitat class is determined for each mapping unit on the basis of the composition of the individual life forms and height measurements. The classification showed very good results for forest phanerophytes (FPH) when individual life forms were analyzed in terms of their percentage coverage estimates per mapping unit from the LCCS classification and validated with field surveys. Analysis for shrubby chamaephytes (SCH) showed less accurate results, but might also be due to less accurate field estimates of percentage coverage. Overall, the EODHaM classification results encouraged us to derive the heights of all vegetated objects in the Netherlands from LiDAR data, in preparation for new habitat classifications.

Discrimination of Vegetation Height Categories With Passive Satellite Sensor Imagery Using Texture Analysis

Vegetation height is a crucial factor in environmental studies, landscape analysis, and mapping applications. Its estimation may prove cost and resource demanding, e.g., employing light detection and ranging (LiDAR) data. This study presents a cost-effective framework for height estimation, built around texture analysis of a single very high-resolution passive satellite sensor image. A number of texture features are proposed, based on local variance, entropy, and binary patterns. Their potential in discriminating among classes in a wide range of height values used for habitat mapping (from less than 5 cm to 40 m) is tested in an area with heath, tree, and shrub vegetation. A number of missing data handling, outlier removal, and data normalization methods are evaluated to enhance the proposed framework. Its performance is tested with different classifiers, including single and ensemble tree ones and support vector machines. Furthermore, dimensionality reduction (DR) is applied to the full feature set (192 features), through both data transformation and filter feature selection methods. The proposed approach was tested in two WorldView-2 images, representing the peak and the decline of the vegetative period. Vegetation height categories were accurately distinguished, reaching accuracies of over 90% for six height classes, using the images either individually or in synergy. DR achieved similarly high, or higher, accuracies with even a 3% feature subset, increasing the processing efficiency of the framework, and favoring its use in height estimation applications not requiring particularly high spatial resolution data, as a cost-effective surrogate of more expensive and resource demanding approaches.

A vegetation height classification approach based on texture analysis of a single VHR image

Vegetation height is a crucial feature in various applications related to ecological mapping, enhancing the discrimination among different land cover or habitat categories and facilitating a series of environmental tasks, ranging from biodiversity monitoring and assessment to landscape characterization, disaster management and conservation planning. Primary sources of information on vegetation height include in situ measurements and data from active satellite or airborne sensors, which, however, may often be non-affordable or unavailable for certain regions. Alternative approaches on extracting height information from very high resolution (VHR) satellite imagery based on texture analysis, have recently been presented, with promising results. Following the notion that multispectral image bands may often be highly correlated, data transformation and dimensionality reduction techniques are expected to reduce redundant information, and thus, the computational cost of the approaches, without significantly compromising their accuracy. In this paper, dimensionality reduction is performed on a VHR image and textural characteristics are calculated on its reconstructed approximations, to show that their discriminatory capabilities are maintained up to a large degree. Texture analysis is also performed on the projected data to investigate whether the different height categories can be distinguished in a similar way.

Land cover to habitat map translation: Disambiguation rules based on Earth Observation data

Earth Observation (EO) images have been extensively used to provide a synoptic view of land cover/use (LC/LU) patterns and land cover/use changes. Land covers are not as clearly relatable to biodiversity in comparison to habitat classifications which can provide more scope for biodiversity monitoring. The main purpose of the paper is to provide an automatic general framework for translating LC maps (in LCCS taxonomy) into habitat maps (in GHC taxonomy) by means of VHR remote sensing data.

Multi-modal knowledge base generation from very high resolution satellite imagery for habitat mapping

Abstract Monitoring of ecosystems entails the evaluation of contributing factors by the expert ecologist. The aim of this study is to examine to what extent the quantitative variables, calculated solely by the spectral and textural information of the space-borne image, may reproduce verified habitat maps. 555 spectral and texture attributes are extracted and calculated from the image. Results reached an overall accuracy of 65% per object, 76% per pixel, and 77% in reproducing the original objects with segmentation. Taking into consideration the large number of different habitats queried and the lack of any ancillary information the results suggest the discriminatory power of the finally selected attributes. Potential and limitations are discussed.

LAND COVER TO HABITAT MAP CONVERSION USING REMOTE SENSING DATA: A SUPERVISED LEARNING APPROACH

The derivation of habitat maps is enhanced if land cover maps are used as basis for the mapping procedure. In this study, a supervised learning framework is proposed to perform object-based classification to General Habitat Categories. A Land Cover Classification System map is used as basis, and an approach to generate numerical features from the object land cover class names and attributes is introduced. An additional number of spectral, morphological, and topological features are extracted from very high resolution satellite imagery and classification accuracies up to 80.4% for 14 classes are reached. Inclusion of LiDAR (Light Detection And Ranging) data or proposed texture analysis features, improve accuracies to 86% and around 83%, respectively, with the latter proving as promising surrogates of LiDAR data features. The method outperformed rule-based approaches, indicating its potential in accurate and labor- and time-efficient habitat classification.