Toon Spanhove

Remote sensing for mapping natural habitats and their conservation status – New opportunities and challenges

Safeguarding the diversity of natural and semi-natural habitats in Europe is one of the aims set out by the Habitats Directive (Council Directive 92/43/EEC on the conservation of natural habitats and of wild fauna and flora) and one of the targets of the European 2020 Biodiversity Strategy, and is to be accomplished by maintaining a favourable conservation status. To reach this aim a high-level understanding of the distribution and conditions of these habitats is needed. Remote sensing can considerably contribute to habitat mapping and their observation over time. Several European projects and a large number of scientific studies have addressed the issue of mapping and monitoring natural habitats via remote sensing and the deriving of indicators on their conservation status. The multitude of utilized remote sensing sensors and applied methods used in these studies, however, impede a common understanding of what is achievable with current state-of-the-art technologies. The aim of this paper is to provide a synthesis on what is currently feasible in terms of detection and monitoring of natural and semi-natural habitats with remote sensing. To focus this endeavour, we concentrate on those studies aimed at direct mapping of individual habitat types or discriminating between different types of habitats occurring in relatively large, spatially contiguous units. By this we uncover the potential of remote sensing to better understand the distribution of habitats and the assessment of their conservation status in Europe.

Burned area detection and burn severity assessment of a heathland fire in Belgium using airborne imaging spectroscopy (APEX)

Uncontrolled, large fires are a major threat to the biodiversity of protected heath landscapes. The severity of the fire is an important factor influencing vegetation recovery. We used airborne imaging spectroscopy data from the Airborne Prism Experiment (APEX) sensor to: (1) investigate which spectral regions and spectral indices perform best in discriminating burned from unburned areas; and (2) assess the burn severity of a recent fire in the Kalmthoutse Heide, a heathland area in Belgium. A separability index was used to estimate the effectiveness of individual bands and spectral indices to discriminate between burned and unburned land. For the burn severity analysis, a modified version of the Geometrically structured Composite Burn Index (GeoCBI) was developed for the field data collection. The field data were collected in four different vegetation types: Calluna vulgaris-dominated heath (dry heath), Erica tetralix-dominated heath (wet heath), Molinia caerulea (grass-encroached heath), and coniferous woodland. Discrimination between burned and unburned areas differed among vegetation types. For the pooled dataset, bands in the near infrared (NIR) spectral region demonstrated the highest discriminatory power, followed by short wave infrared (SWIR) bands. Visible wavelengths performed considerably poorer. The Normalized Burn Ratio (NBR) outperformed the other spectral indices and the individual spectral bands in discriminating between burned and unburned areas. For the burn severity assessment, all spectral bands and indices showed low correlations with the field data GeoCBI, when data of all pre-fire vegetation types were pooled (R2 maximum 0.41). Analysis per vegetation type, however, revealed considerably higher correlations (R2 up to 0.78). The Mid Infrared Burn Index (MIRBI) had the highest correlations for Molinia and Erica (R2 = 0.78 and 0.42, respectively). In Calluna stands, the Char Soil Index (CSI) achieved the highest correlations, with R2 = 0.65. In Pinus stands, the Normalized Difference Vegetation Index (NDVI) and the red wavelength both had correlations of R2 = 0.64. The results of this study highlight the superior performance of the NBR to discriminate between burned and unburned areas, and the disparate performance of spectral indices to assess burn severity among vegetation types. Consequently, in heathlands, one must consider a stratification per vegetation type to produce more reliable burn severity maps.

Classification of heathland vegetation in a hierarchical contextual framework

Heathlands in Western Europe have shown dramatic declines over the last century and therefore have been given a high conservation priority in the Habitats Directive of the European Union (EU). Accurate surveying and monitoring of heathland habitats is essential for appropriate conservation management, but the large heterogeneity of vegetation types within habitats as well as the occurrence of similar vegetation across habitat types hinders a straightforward, automated mapping based on aerial images. In such a case, a context-dependent classification algorithm is expected to be superior to traditional classification techniques. This article presents a novel approach to map the conservation status of heathland vegetation by using a hierarchical classification scheme that describes the structural dependencies in the field between the basic vegetation and the land-cover types that habitats are composed of. These dependency relationships are included as contextual information in the classification process, using a tree-structured Markov random field (TS-MRF) technique with a tree that reflects the hierarchy of the classification scheme. Results of this approach for a heathland area in Belgium were compared with results from more conventional classification approaches. Validation of the results showed that the structure of the scheme contained important spatial relationships, which were further reinforced by using the contextual classification strategy, especially for the most detailed level of the classification scheme. Accuracy increased and the classification results were more suitable for visual interpretation.

Accuracy assessment of contextual classification results for vegetation mapping

Abstract A new procedure for quantitatively assessing the geometric accuracy of thematic maps, obtained from classifying hyperspectral remote sensing data, is presented. More specifically, the methodology is aimed at the comparison between results from any of the currently popular contextual classification strategies. The proposed procedure characterises the shapes of all objects in a classified image by defining an appropriate reference and a new quality measure. The results from the proposed procedure are represented in an intuitive way, by means of an error matrix, analogous to the confusion matrix used in traditional thematic accuracy representation. A suitable application for the methodology is vegetation mapping, where lots of closely related and spatially connected land cover types are to be distinguished. Consequently, the procedure is tested on a heathland vegetation mapping problem, related to Natura 2000 habitat monitoring. Object-based mapping and Markov Random Field classification results are compared, showing that the selected Markov Random Fields approach is more suitable for the fine-scale problem at hand, which is confirmed by the proposed procedure.

Mapping wetland loss and restoration potential in Flanders (Belgium): an ecosystem service perspective

With the case of Flanders (northern part of Belgium) we present an integrated approach to calculate accurate losses of wetlands, potentials for restoration, and their ecosystem services supplies and illustrate how these insights can be used to evaluate and support policy making. Flanders lost about 75% of its wetland habitats in the past 50–60 years, with currently only 68,000 ha remaining, often in a more or less degraded state. For five different wetland categories (excluding open waters) we calculated that restoration of lost wetland is still possible for an additional total area of about 147,000 ha, assuming that, with time and appropriate measures and techniques, the necessary biophysical and ecological conditions can more or less be restored or created. Wetland restoration opportunities were mapped according to an open and forested landscape scenario. Despite the fact that for 49,000 ha wetland restoration is justifiable by the actual presence of an appropriate spatial planning and/or protection status, the official Flemish nature policy only foresees 7,400 to 10,600 ha of additional wetland (open waters excluded) by 2050. The benefits of a more ambitious wetland restoration action program are underpinned by an explorative and quantified analysis of ecosystem service supply for each of the two scenarios, showing that the strongly increased supply of several important regulating and cultural ecosystem services might outweigh the decrease of food production, especially if extensive farming on temporary wet soils remains possible. Finally, we discuss the challenges of wetland restoration policies for biodiversity conservation and climate change.

Towards a Mature Age of Remote Sensing for Natura 2000 Habitat Conservation: Poor Method Transferability as a Prime Obstacle

Over the past decades, remote sensing has been repeatedly identified as a promising and powerful tool to aid nature conservation. Many methods and applications of remote sensing to monitor biodiversity have indeed been published, and continue to be at an increasing rate. As such, remote sensing is seemingly living up to its expectations; yet, its actual use in nature conservation (or rather the lack thereof) contradicts this. We argue that, at least for the practical implementation of regular vegetation monitoring, including within protected areas (e.g., European Natura 2000 sites), a lack of transferability of remote sensing methods is an overlooked factor that hinders its effective operational use for nature conservation. Among the causes of poor method transferability is the large variation in objects of interest, user requirements, ground reference data, and image properties, but also the lack of consideration of transferability during method development. To stimulate the adoption of remote sensing based techniques in vegetation monitoring and conservation, we recommend that a number of actions are taken. We call upon remote sensing scientists and nature monitoring experts to specifically consider and demonstrate method transferability by using widely available image data, limiting ground reference data dependence, and making their preferably open-source programming code publicly available. Furthermore, we recommend that nature conservation specialists are open and realistic about potential outcomes by not expecting the replacement of current in-place methodologies, and actively contributing to the thought process of generating transferable and repeatable methods.

Mapping natura 2000 heathland in Belgium - an evaluation of ensemble classifiers for spaceborne angular CHRIS/Proba imagery

Natura 2000 is an ecological network of protected areas in the territory of the European Union (EU). With the introduction of the Habitats Directive in 1992, EU member states are obligated to report every six years the status of the Natura 2000 habitats so that better conservation policy can be formulated. This paper examines the use of angular hyperspectral CHRIS/Proba image for the mapping of heathland at a Belgian Natura 2000 site. We find that the use of angular images increases the overall classification rate as compared to using only the nadir image; with the incorporation of angular images the final mapping is also more homogenous with less salt and pepper effect. While the class accuracy of Calluna- and Erica-dominated heathlands are still low, class accuracy of Molinia-dominated heathland is generally more encouraging. Two tree-based ensemble classifiers, Random Forest (RF) and Adaboost, were compared with Support Vector Machines (SVM). When only the nadir image was used, SVM attained the highest accuracy. When angular images were included, all three classifiers obtained comparable accuracies though in general RF and Adaboost had faster training time. We also adopted an assessment approach which repeats the accuracy assessment in ten independent trials, instead of the common practice of having only one trial. Our results show that accuracy attainment can vary significantly among different trials and hence it is recommendable to have more than one trial in order that a more objective characterization of the classifiers is obtained. 1