This article presents an environmental remote sensing application using a UAV that is specifically aimed at reducing the data gap between field scale and satellite scale in soil erosion monitoring in Morocco. A fixed-wing aircraft type Sirius I (MAVinci, Germany) equipped with a digital system camera (Panasonic) is employed. UAV surveys are conducted over different study sites with varying extents and flying heights in order to provide both very high resolution site-specific data and lower-resolution overviews, thus fully exploiting the large potential of the chosen UAV for multi-scale mapping purposes. Depending on the scale and area coverage, two different approaches for georeferencing are used, based on high-precision GCPs or the UAV’s log file with exterior orientation values respectively. The photogrammetric image processing enables the creation of Digital Terrain Models (DTMs) and ortho-image mosaics with very high resolution on a sub-decimetre level. The created data products were used for quantifying gully and badland erosion in 2D and 3D as well as for the analysis of the surrounding areas and landscape development for larger extents.

This study aims at the detection of gully-affected areas by applying object-based image analysis in the region of Taroudannt, Morocco, which is highly affected by gully erosion while simultaneously representing a major region of agro-industry with a high demand of arable land. As high-resolution optical satellite data are readily available from various sensors and with a much better temporal resolution than 3D terrain data, an area-wide mapping approach to extract gully-affected areas using only optical satellite imagery was developed. The methodology additionally incorporates expert knowledge and freely-available vector data in a cyclic object-based image analysis approach. This connects the two fields of geomorphology and remote sensing. The classification results show the successful implementation of the developed approach and allow conclusions on the current distribution of gullies. The results of the classification were checked against manually delineated reference data incorporating expert knowledge based on several field campaigns in the area, resulting in an overall classification accuracy of 62%. The error of omission accounts for 38% and the error of commission for 16%, respectively. Additionally, a manual assessment was carried out to assess the quality of the applied classification algorithm. The limited error of omission contributes with 23% to the overall error of omission and the limited error of commission contributes with 98% to the overall error of commission. This assessment improves the results and confirms the high quality of the developed approach for area-wide mapping of gully-affected areas in larger regions. In the field of landform mapping, the overall quality of the classification results is often assessed with more than one method to incorporate all aspects adequately.

Landform mapping is more important than ever before, yet the automatic recognition of specific landforms remains difficult. Object-based image analysis (OBIA) steps out as one of the most promising techniques for tackling this issue. Using the OBIA approach, in this study, a multiscale mapping workflow is developed and applied to two different input data sets: aerial photographs and digital elevation models. Optical data are used for gully mapping on a very local scale, while terrain data are employed for drumlin mapping on a slightly broader scale. After a multiresolution segmentation, a knowledge-based classification approach was developed for the multiscale mapping of targeted landforms. To identify well-suited scale levels for data segmentation, the estimation-of-scale-parameter tool was applied. Contrast information and shape properties of segments were implemented for gully classification. Contextual and shape information was utilized for mapping drumlins. An accuracy assessment was performed by comparing classification results with independent reference data sets that were delineated manually from the input data. We achieved satisfactory agreements between mapped and reference landforms. Knowledge-based identification of segment features improves both accuracy and transferability of the classification system.

In this paper, we report the outcomes of the 2015 data fusion contest organized by the Image Analysis and Data Fusion Technical Committee (IADF TC) of the IEEE Geoscience and Remote Sensing Society. As for previous years, the IADF TC organized a data fusion contest aiming at fostering new ideas and solutions for multisource studies. The 2015 edition of the contest proposed a multiresolution and multisensorial challenge involving extremely high resolution RGB images (with a ground sample distance of 5 cm) and a 3-D light detection and ranging point cloud (with a point cloud density of approximatively 65 pts/m ² ). The competition was framed in two parallel tracks, considering 2-D and 3-D products, respectively. In this Part B, we report the results obtained by the winners of the 3-D contest, which explored challenging tasks of road extraction and ISO containers identification, respectively. The 2-D part of the contest and a detailed presentation of the dataset are discussed in Part A.