@article{BrandmeierKuhlemannKrumreietal., author = {Brandmeier, Melanie and Kuhlemann, J. and Krumrei, I. and Kappler, A. and Kubik, P.W.}, title = {New challenges for tafoni research. A new approach to understand processes and weathering rates}, series = {Earth Surface Processes and Landforms}, volume = {36}, journal = {Earth Surface Processes and Landforms}, number = {6}, doi = {10.1002/esp.2112}, pages = {839 -- 852}, abstract = {Cavernous tafoni-type weathering is a common and conspicuous global feature, creating artistic sculptures, which may be relevant for geochemical budgets. Weathering processes and rates are still a matter of discussion. Field evidence in the type locality Corsica revealed no trend of size variability from the coast to subalpine elevations and the aspect of tafoni seems to be governed primarily by the directions of local fault systems and cleavage, and only subordinately by wind directions or the aspect of insulation. REM analysis of fresh tafone chips confirmed mechanical weathering by the crystallization of salts, as conchoidal fracturing of quartz is observed. The salts are only subordinately provided by sea spray, as calcium and sodium sulfates rather than halite dominate even close to the coast. Characteristic element ratios compare well with aerosols from mixed African and European air masses. Sulfates are largely derived from Sahara dust, indicated by their sulfur isotopic composition. Salt crystals form by capillary rise within the rock and subsequent crystallization in micro-cracks and at grain boundaries inside rain-protected overhangs. Siderophile bacteria identified by raster electron microscopy (REM) analysis of tafone debris contribute to accelerated weathering of biotite and tiny sulfide ore minerals. By applying 10Be-exposure dating, weathering rates of large mature tafone structures were found to be about an order of magnitude higher than those on the exposed top of the affected granite blocks.}, language = {en} } @article{BrandmeierErasmiHansenetal., author = {Brandmeier, Melanie and Erasmi, S. and Hansen, C. and H{\"o}weling, A. and Nitzsche, K. and Ohlendorf, T. and Mamani, M. and W{\"o}rner, G.}, title = {Mapping patterns of mineral alteration in volcanic terrains using ASTER data and field spectrometry in Southern Peru}, series = {Journal of South American Earth Sciences}, volume = {48}, journal = {Journal of South American Earth Sciences}, issn = {0895-9811}, doi = {10.1016/j.jsames.2013.09.011}, pages = {296 -- 314}, abstract = {Because formation of ore deposits is linked to volcanic and post-volcanic processes, an understanding of alteration style in volcanic regions has important applications in economic geology. We use ASTER (Advanced Spaceborne Thermal Emission and Reflection Radiometer) data and field spectrometry for mineral mapping in selected Miocene to Quaternary volcanic areas in Southern Peru to better characterize and understand the Tertiary volcanic evolution in this region. Our goal is to characterize volcanic regions near Puquio (Ayacucho) by correlating areas of intense alteration and related ignimbrite outflow sheets. In particular, we spectrally and mineralogically map different types and intensities of alteration based on remote sensing and ground-truth data. ASTER ratio images, alteration indices and false color composites were used to select ground-training areas for sample collection and field spectrometry. Alteration samples were characterized geochemically, mineralogically and spectrally. Absorption features correlate with chemical properties (e.g. iron content). Hyperspectral data from field spectrometry allow identification of important alteration minerals such as kaolinite and smectite. Alteration mineral assemblages range from silicic to argillic to "zeolite-type". Using a support vector machine classification (SVM) algorithm on ASTER data, we mapped the different types and intensities of alteration, along with unaltered ignimbrite and lava flows with an accuracy of 80\%. We propose a preliminary model for the interpretation of alteration settings, discuss the potential eruption sites of the ignimbrites in the region and, propose pH and temperature estimates for the respective classes based on the mineral assemblages identified.}, language = {en} } @article{SzekelyKomaKaratsonetal., author = {Sz{\´e}kely, Bal{\´a}zs and Koma, Zs{\´o}fia and Kar{\´a}tson, D{\´a}vid and Dorninger, Peter and W{\"o}rner, Gerhard and Brandmeier, Melanie and Nothegger, Clemens}, title = {Automated recognition of quasi-planar ignimbrite sheets as paleosurfaces via robust segmentation of digital elevation models: an example from the Central Andes}, series = {Earth Surface Processes and Landforms}, volume = {39}, journal = {Earth Surface Processes and Landforms}, number = {10}, issn = {1096-9837}, doi = {10.1002/esp.3606}, pages = {1386 -- 1399}, abstract = {Quasi-planar morphological surfaces may become dissected or degraded with time, but still retain original features related to their geologic-geomorphic origin. To decipher the information hidden in the relief, recognition of such features is required, possibly in an automated manner. In our study, using Shuttle Radar Topography Mission (SRTM) digital elevation model (DEM), an existing algorithm has been adapted to recognize quasi-planar features fulfilling specified criteria. The method has been applied to a study area of the Central Andes with Miocene to Quaternary volcanic edifices, tilted ignimbrite surfaces, and basin-filling sediments. The result is a surface segmentation, whereas non-planar features (gullies, tectonic faults, etc.) are sorted out. The main types of geomorphic features that can be distinguished and interpreted are as follows. (1) The west-dipping western margin of the Altiplano is differentiated into segments of the lower sedimentary cover that of increased erosion by tectonic steepening at intermediate levels, and an upper plane with limited erosion. (2) In the central part of the Western Cordillera, the Oxaya ignimbrite block shows a 'striped' bulging pattern that results from a smoothly changing surface dip. This pattern is due to continuous folding/warping of the ignimbrite block possibly related to gravitational movements. (3) To the west, large, uniform planes correspond to flat, smooth, tectonically undisturbed surfaces of young sedimentary cover of the Central Basin. (4) The evolution of Taapaca volcanoes with sector collapse events and cone-building phases is shown by several segments with overlapping clastic aprons. (5) To the east, on the western margin of the Altiplano, young intermontane basins filled by Upper Miocene sediments show progressively increasing dip toward basin margins, reflected by a circular pattern of the segmentation planes. We show that the segmentation models provide meaningful images and additional information for geomorphometric analysis that can be interpreted in terms of geological and surface evolution models.}, language = {en} } @article{FreymuthBrandmeierWoerner, author = {Freymuth, Heye and Brandmeier, Melanie and W{\"o}rner, Gerhard}, title = {The origin and crust/mantle mass balance of Central Andean ignimbrite magmatism constrained by oxygen and strontium isotopes and erupted volumes}, series = {Contributions to Mineralogy and Petrology}, volume = {169}, journal = {Contributions to Mineralogy and Petrology}, number = {6}, issn = {1432-0967}, doi = {10.1007/s00410-015-1152-5}, pages = {58 -- 58}, abstract = {Volcanism during the Neogene in the Central Volcanic Zone (CVZ) of the Andes produced (1) stratovolcanoes, (2) rhyodacitic to rhyolitic ignimbrites which reach volumes of generally less than 300 km3 and (3) large-volume monotonous dacitic ignimbrites of up to several thousand cubic kilometres. We present models for the origin of these magma types using O and Sr isotopes to constrain crust/mantle proportions for the large-volume ignimbrites and explore the relationship to the evolution of the Andean crust. Oxygen isotope ratios were measured on phenocrysts in order to avoid the effects of secondary alteration. Our results show a complete overlap in the Sr-O isotope compositions of lavas from stratovolcanoes and low-volume rhyolitic ignimbrites as well as older (>9 Ma) large-volume dacitic ignimbrites. This suggests that the mass balance of crustal and mantle components are largely similar. By contrast, younger (<10 Ma) large-volume dacitic ignimbrites from the southern portion of the Central Andes have distinctly more radiogenic Sr and heavier O isotopes and thus contrast with older dacitic ignimbrites in northernmost Chile and southern Peru. Results of assimilation and fractional crystallization (AFC) models show that the largest chemical changes occur in the lower crust where magmas acquire a base-level geochemical signature that is later modified by middle to upper crustal AFC. Using geospatial analysis, we estimated the volume of these ignimbrite deposits throughout the Central Andes during the Neogene and examined the spatiotemporal pattern of so-called ignimbrite flare-ups. We observe a N-S migration of maximum ages of the onset of large-volume "ignimbrite pulses" through time: Major pulses occurred at 19-24 Ma (e.g. Oxaya, Nazca Group), 13-14 Ma (e.g. Huaylillas and Altos de Pica ignimbrites) and <10 Ma (Altiplano and Puna ignimbrites). Such "flare-ups" represent magmatic production rates of 25 to >70 km3 Ma-1 km-1 (assuming plutonic/volcanic ratios of 1:5) which are additional to, but within the order of, the arc background magmatic flux. Comparing our results to average shortening rates observed in the Andes, we observe a "lag-time" with large-volume eruptions occurring after accelerated shortening. A similar delay exists between the ignimbrite pulses and the subduction of the Juan Fernandez ridge. This is consistent with the idea that large-volume ignimbrite eruptions occurred in the wake of the N-S passage of the ridge after slab steepening has allowed hot asthenospheric mantle to ascend into and cause the melting of the mantle wedge. In our model, the older large-volume dacitic ignimbrites in the northern part of the CVZ have lower (15-37 \%) crustal contributions because they were produced at times when the Central Andean crust was thinner and colder, and large-scale melting in the middle crust could not be achieved. Younger ignimbrite flare-ups further south (<10 Ma, >22°S) formed with a significantly higher crustal contribution (22-68 \%) because at that time the Andean crust was thicker and hotter and, therefore primed for more extensive crustal melting. The rhyolitic lower-volume ignimbrites are more equally distributed in the CVZ in time and space and are produced by mechanisms similar to those operating below large stratovolcanoes, but at times of higher melt fluxes from the mantle wedge.}, language = {en} } @article{ZimmermannBrandmeierAndreanietal., author = {Zimmermann, Robert and Brandmeier, Melanie and Andreani, Louis and Mhopjeni, Kombada and Gloaguen, Richard}, title = {Remote Sensing Exploration of Nb-Ta-LREE-Enriched Carbonatite (Epembe/Namibia)}, series = {Remote Sensing}, volume = {8}, journal = {Remote Sensing}, number = {8}, issn = {2072-4292}, doi = {10.3390/rs8080620}, pages = {620 -- 620}, language = {en} } @article{WesselBrandmeierTiede, author = {Wessel, Mathias and Brandmeier, Melanie and Tiede, Dirk}, title = {Evaluation of Different Machine Learning Algorithms for Scalable Classification of Tree Types and Tree Species Based on Sentinel-2 Data}, series = {Remote Sensing}, volume = {10}, journal = {Remote Sensing}, number = {9}, issn = {2072-4292}, doi = {10.3390/rs10091419}, pages = {1419 -- 1419}, language = {en} } @article{HamdiBrandmeierStraub, author = {Hamdi, Zayd Mahmoud and Brandmeier, Melanie and Straub, Christoph}, title = {Forest Damage Assessment Using Deep Learning on High Resolution Remote Sensing Data}, series = {Remote Sensing}, volume = {11}, journal = {Remote Sensing}, number = {17}, issn = {2072-4292}, doi = {10.3390/rs11171976}, pages = {1976 -- 1976}, language = {en} } @article{DeigeleBrandmeierStraub, author = {Deigele, Wolfgang and Brandmeier, Melanie and Straub, Christoph}, title = {A Hierarchical Deep-Learning Approach for Rapid Windthrow Detection on PlanetScope and High-Resolution Aerial Image Data}, series = {Remote Sensing}, volume = {12}, journal = {Remote Sensing}, number = {13}, issn = {2072-4292}, doi = {10.3390/rs12132121}, pages = {2121 -- 2121}, abstract = {Forest damage due to storms causes economic loss and requires a fast response to prevent further damage such as bark beetle infestations. By using Convolutional Neural Networks (CNNs) in conjunction with a GIS, we aim at completely streamlining the detection and mapping process for forest agencies. We developed and tested different CNNs for rapid windthrow detection based on PlanetScope satellite data and high-resolution aerial image data. Depending on the meteorological situation after the storm, PlanetScope data might be rapidly available due to its high temporal resolution, while the acquisition of high-resolution airborne data often takes weeks to a month and is, therefore, used in a second step for more detailed mapping. The study area is located in Bavaria, Germany (ca. 165 km2), and labels for damaged areas were provided by the Bavarian State Institute of Forestry (LWF). Modifications of a U-Net architecture were compared to other approaches using transfer learning (e.g., VGG19) to find the most efficient architecture for the task on both datasets while keeping the computational time low. A custom implementation of U-Net proved to be more accurate than transfer learning, especially on medium (3 m) resolution PlanetScope imagery (intersection over union score (IoU) 0.55) where transfer learning completely failed. Results for transfer learning based on VGG19 on high-resolution aerial image data are comparable to results from the custom U-Net architecture (IoU 0.76 vs. 0.73). When using both architectures on a dataset from a different area (located in Hesse, Germany), however, we find that the custom implementations have problems generalizing on aerial image data while VGG19 still detects most damage in these images. For PlanetScope data, VGG19 again fails while U-Net achieves reasonable mappings. Results highlight the potential of Deep Learning algorithms to detect damaged areas with an IoU of 0.73 on airborne data and 0.55 on Planet Dove data. The proposed workflow with complete integration into ArcGIS is well-suited for rapid first assessments after a storm event that allows for better planning of the flight campaign followed by detailed mapping in a second stage.}, language = {en} } @incollection{BrandmeierWessel, author = {Brandmeier, Melanie and Wessel, M.}, title = {Workflows f{\"u}r Bilddaten und Big Data Analytics - Das Potenzial von Sentinel-2-Daten zur Baumartenklassifizierung}, series = {Fl{\"a}chennutzungsmonitoring. IX: Nachhaltigkeit der Siedlungs- und Verkehrsentwicklung?}, booktitle = {Fl{\"a}chennutzungsmonitoring. IX: Nachhaltigkeit der Siedlungs- und Verkehrsentwicklung?}, number = {Band 73}, editor = {Meinel, Gotthard and Schumacher, Ulrich and Schwarz, Steffen and Richter, Benjamin and f{\"u}r {\"O}kologische Raumentwicklung, Leibniz-Institut}, publisher = {Rhombos}, address = {Berlin}, isbn = {978-3-944101-73-6}, pages = {135 -- 142}, language = {de} } @article{Brandmeier, author = {Brandmeier, Melanie}, title = {Remote sensing of Carhuarazo volcanic complex using ASTER imagery in Southern Peru to detect alteration zones and volcanic structures - a combined approach of image processing in ENVI and ArcGIS/ArcScene}, series = {Geocarto International}, volume = {25}, journal = {Geocarto International}, number = {8}, issn = {1010-6049}, doi = {10.1080/10106049.2010.519787}, pages = {629 -- 648}, language = {en} } @article{BrandmeierCabreraZamoraNykaenenetal., author = {Brandmeier, Melanie and Cabrera Zamora, Irving Gibran and Nyk{\"a}nen, Vesa and Middleton, Maarit}, title = {Boosting for Mineral Prospectivity Modeling: A New GIS Toolbox}, series = {Natural Resources Research}, volume = {29}, journal = {Natural Resources Research}, number = {1}, issn = {1520-7439}, doi = {10.1007/s11053-019-09483-8}, pages = {71 -- 88}, language = {en} } @article{BrandmeierChen, author = {Brandmeier, Melanie and Chen, Y.}, title = {Lithological classification using multi-sensor data an convolutional neural networks}, series = {The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences}, volume = {XLII-2/W16}, journal = {The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences}, doi = {10.5194/isprs-archives-XLII-2-W16-55-2019}, pages = {55 -- 59}, language = {en} } @article{ScharvogelBrandmeierWeis, author = {Scharvogel, Daniel and Brandmeier, Melanie and Weis, Manuel}, title = {A Deep Learning Approach for Calamity Assessment Using Sentinel-2 Data}, series = {Forests}, volume = {11}, journal = {Forests}, number = {12}, issn = {1999-4907}, doi = {10.3390/f11121239}, pages = {1239 -- 1239}, language = {en} } @incollection{Brandmeier, author = {Brandmeier, Melanie}, title = {The Anatomy of Supervolcanoes}, series = {GIS for Science: Applying Mapping and Spatial Analytics}, booktitle = {GIS for Science: Applying Mapping and Spatial Analytics}, publisher = {Esri}, language = {en} } @article{CherifHellBrandmeier, author = {Cherif, Eya and Hell, Maximilian and Brandmeier, Melanie}, title = {DeepForest: novel deep learning models for land use and land cover classification using multi-temporal and -modal sentinel data of the amazon basin}, series = {Remote Sensing}, volume = {14}, journal = {Remote Sensing}, number = {19}, issn = {2072-4292}, doi = {10.3390/rs14195000}, abstract = {Land use and land cover (LULC) mapping is a powerful tool for monitoring large areas. For the Amazon rainforest, automated mapping is of critical importance, as land cover is changing rapidly due to forest degradation and deforestation. Several research groups have addressed this challenge by conducting local surveys and producing maps using freely available remote sensing data. However, automating the process of large-scale land cover mapping remains one of the biggest challenges in the remote sensing community. One issue when using supervised learning is the scarcity of labeled training data. One way to address this problem is to make use of already available maps produced with (semi-) automated classifiers. This is also known as weakly supervised learning. The present study aims to develop novel methods for automated LULC classification in the cloud-prone Amazon basin (Brazil) based on the labels from the MapBiomas project, which include twelve classes. We investigate different fusion techniques for multi-spectral Sentinel-2 data and synthetic aperture radar Sentinel-1 time-series from 2018. The newly designed deep learning architectures—DeepForest-1 and DeepForest-2—utilize spatiotemporal characteristics, as well as multi-scale representations of the data. In several data scenarios, the models are compared to state-of-the-art (SotA) models, such as U-Net and DeepLab. The proposed networks reach an overall accuracy of up to 75.0\%, similar to the SotA models. However, the novel approaches outperform the SotA models with respect to underrepresented classes. Forest, savanna and crop were mapped best, with F1 scores up to 85.0\% when combining multi-modal data, compared to 81.6\% reached by DeepLab. Furthermore, in a qualitative analysis, we highlight that the classifiers sometimes outperform the inaccurate labels.}, language = {en} } @article{HellBrandmeierBriechleetal., author = {Hell, Maximilian and Brandmeier, Melanie and Briechle, Sebastian and Krzystek, Peter}, title = {Classification of Tree Species and Standing Dead Trees with Lidar Point Clouds Using Two Deep Neural Networks: PointCNN and 3DmFV-Net}, series = {PFG - Journal of Photogrammetry, Remote Sensing and Geoinformation Science}, volume = {90}, journal = {PFG - Journal of Photogrammetry, Remote Sensing and Geoinformation Science}, issn = {2512-2819}, doi = {10.1007/s41064-022-00200-4}, pages = {103 -- 121}, abstract = {Knowledge about tree species distribution is important for forest management and for modeling and protecting biodiversity in forests. Methods based on images are inherently limited to the forest canopy. Airborne lidar data provide information about the trees' geometric structure, as well as trees beneath the upper canopy layer. In this paper, the potential of two deep learning architectures (PointCNN, 3DmFV-Net) for classification of four different tree classes is evaluated using a lidar dataset acquired at the Bavarian Forest National Park (BFNP) in a leaf-on situation with a maximum point density of about 80 pts/m2. Especially in the case of BFNP, dead wood plays a key role in forest biodiversity. Thus, the presented approaches are applied to the combined classification of living and dead trees. A total of 2721 single trees were delineated in advance using a normalized cut segmentation. The trees were manually labeled into four tree classes (coniferous, deciduous, standing dead tree with crown, and snag). Moreover, a multispectral orthophoto provided additional features, namely the Normalized Difference Vegetation Index. PointCNN with 3D points, laser intensity, and multispectral features resulted in a test accuracy of up to 87.0\%. This highlights the potential of deep learning on point clouds in forestry. In contrast, 3DmFV-Net achieved a test accuracy of 73.2\% for the same dataset using only the 3D coordinates of the laser points. The results show that the data fusion of lidar and multispectral data is invaluable for differentiation of the tree classes. Classification accuracy increases by up to 16.3\% points when adding features generated from the multispectral orthophoto.}, language = {en} } @inproceedings{ErbeBrandmeierSchmittetal., author = {Erbe, Karin and Brandmeier, Melanie and Schmitt, Michael and Donbauer, Andreas and Liebscher, Jan-Andreas and Kolbe, Thomas}, title = {Detektion von Fahrradst{\"a}ndern in Luftbildern mittels Deep Learning}, series = {42. Wissenschaftlich-Technische Jahrestagung der DGPF. 5.-6. Oktober 2022 in Dresden}, volume = {30}, booktitle = {42. Wissenschaftlich-Technische Jahrestagung der DGPF. 5.-6. Oktober 2022 in Dresden}, editor = {Kersten, Thomas P. and Tilly, Nora}, issn = {0942-2870}, doi = {10.24407/KXP:1795622415}, pages = {27 -- 39}, language = {de} } @article{HellBrandmeier, author = {Hell, Maximilian and Brandmeier, Melanie}, title = {Identifying Plausible Labels from Noisy Training Data for a Land Use and Land Cover Classification Application in Amaz{\^o}nia Legal}, series = {remote sensing}, volume = {26}, journal = {remote sensing}, number = {12}, doi = {10.3390/rs16122080}, url = {http://nbn-resolving.de/urn:nbn:de:bvb:863-opus-57532}, pages = {24}, abstract = {Most studies in the field of land use and land cover (LULC) classification in remote sensing rely on supervised classification, which requires a substantial amount of accurate label data. However, reliable data are often not immediately available, and are obtained through time-consuming manual labor. One potential solution to this problem is the use of already available classification maps, which may not be the true ground truth and may contain noise from multiple possible sources. This is also true for the classification maps of the MapBiomas project, which provides land use and land cover (LULC) maps on a yearly basis, classifying the Amazon basin into more than 24 classes based on the Landsat data. In this study, we utilize the Sentinel-2 data with a higher spatial resolution in conjunction with the MapBiomas maps to evaluate a proposed noise removal method and to improve classification results. We introduce a novel noise detection method that relies on identifying anchor points in feature space through clustering with self-organizing maps (SOM). The pixel label is relabeled using nearest neighbor rules, or can be removed if it is unknown. A challenge in this approach is the quantification of noise in such a real-world dataset. To overcome this problem, highly reliable validation sets were manually created for quantitative performance assessment. The results demonstrate a significant increase in overall accuracy compared to MapBiomas labels, from 79.85\% to 89.65\%. Additionally, we trained the L2HNet using both MapBiomas labels and the filtered labels from our approach. The overall accuracy for this model reached 93.75\% with the filtered labels, compared to the baseline of 74.31\%. This highlights the significance of noise detection and filtering in remote sensing, and emphasizes the need for further research in this area.}, language = {en} } @article{SpeckenwirthBrandmeierPaczkowski, author = {Speckenwirth, S{\"o}nke and Brandmeier, Melanie and Paczkowski, Sebastian}, title = {TreeSeg - A Toolbox for Fully Automated Tree Crown Segmentation Based on High-Resolution Multispectral UAV Data}, series = {remote sensing}, volume = {16}, journal = {remote sensing}, number = {19}, publisher = {MDPI}, doi = {10.3390/rs16193660}, url = {http://nbn-resolving.de/urn:nbn:de:bvb:863-opus-57590}, pages = {17}, abstract = {Single-tree segmentation on multispectral UAV images shows significant potential for effective forest management such as automating forest inventories or detecting damage and diseases when using an additional classifier. We propose an automated workflow for segmentation on high-resolution data and provide our trained models in a Toolbox for ArcGIS Pro on our GitHub repository for other researchers. The database used for this study consists of multispectral UAV data (RGB, NIR and red edge bands) of a forest area in Germany consisting of a mix of tree species consisting of five deciduous trees and three conifer tree species in the matured closed canopy stage at approximately 90 years. Information of NIR and Red Edge bands are evaluated for tree segmentation using different vegetation indices (VIs) in comparison to only using RGB information. We trained Faster R-CNN, Mask R-CNN, TensorMask and SAM in several experiments and evaluated model performance on different data combinations. All models with the exception of SAM show good performance on our test data with the Faster R-CNN model trained on the red and green bands and the Normalized Difference Red Edge Index (NDRE) achieving best results with an F1-Score of 83.5\% and an Intersection over Union of 65.3\% on highly detailed labels. All models are provided in our TreeSeg toolbox and allow the user to apply the pre-trained models on new data.}, language = {en} } @article{BrandmeierHessdoerferSiebenlistetal., author = {Brandmeier, Melanie and Heßd{\"o}rfer, Daniel and Siebenlist, Philipp and Meyer-Spelbrink, Adrian and Kraus, Anja}, title = {Time Series Analysis of Multisensor Data for Precision Viticulture}, series = {remote sensing}, volume = {16}, journal = {remote sensing}, number = {8}, publisher = {MDPI}, doi = {10.3390/rs16081419}, url = {http://nbn-resolving.de/urn:nbn:de:bvb:863-opus-57527}, pages = {18}, abstract = {In the context of climate change, vineyard monitoring to better understand spatiotemporal patterns of grapevine development is of utter importance for precision viticulture. We present a time series analysis of hyperspectral in situ and multispectral UAV data for different irrigation systems in Lower Franconia and correlate results with sensor data for soil moisture, temperature, and precipitation. Analysis of Variance (ANOVA) and a Tukey's HSD test were performed to see whether Vegetation Indices (VIs) are significantly different with respect to irrigation systems as well as topographic position in the vineyard. Correlation between in situ measurements and UAV data for selected VIs is also investigated for upscaling analysis. We find significant differences with respect to irrigation, as well as for topographic position for most of the VIs investigated, highlighting the importance of adapted water management. Correlation between in situ and UAV data is significant only for some indices (NDVI and CIRedEdge, 𝑟2 of 0.33 and 0.49, respectively), while shallow soil moisture patterns correlate well with in situ-derived VIs such as the CIRedEdge and RG index (𝑟2 of 0.34 and 0.46).}, language = {en} } @article{AnwanderBrandmeierPaczkowskietal., author = {Anwander, Julia and Brandmeier, Melanie and Paczkowski, Sebastian and Neubert, Tarek and Paczkowska, Marta}, title = {Evaluating Different Deep Learning Approaches for Tree Health Classification Using High-Resolution Multispectral UAV Data in the Black Forest, Harz Region, and G{\"o}ttinger Forest}, series = {Remote Sensing}, volume = {16}, journal = {Remote Sensing}, number = {3}, doi = {10.3390/rs16030561}, pages = {561 -- 561}, abstract = {We present an evaluation of different deep learning and machine learning approaches for tree health classification in the Black Forest, the Harz Mountains, and the G{\"o}ttinger Forest on a unique, highly accurate tree-level dataset. The multispectral UAV data were collected from eight forest plots with diverse tree species, mostly conifers. As ground truth data (GTD), nearly 1500 tree polygons with related attribute information on the health status of the trees were used. This data were collected during extensive fieldwork using a mobile application and subsequent individual tree segmentation. Extensive preprocessing included normalization, NDVI calculations, data augmentation to deal with the underrepresented classes, and splitting the data into training, validation, and test sets. We conducted several experiments using a classical machine learning approach (random forests), as well as different convolutional neural networks (CNNs)—ResNet50, ResNet101, VGG16, and Inception-v3—on different datasets and classes to evaluate the potential of these algorithms for tree health classification. Our first experiment was a binary classifier of healthy and damaged trees, which did not consider the degree of damage or tree species. The best results of a 0.99 test accuracy and an F1 score of 0.99 were obtained with ResNet50 on four band composites using the red, green, blue, and infrared bands (RGBI images), while VGG16 had the worst performance, with an F1 score of only 0.78. In a second experiment, we also distinguished between coniferous and deciduous trees. The F1 scores ranged from 0.62 to 0.99, with the highest results obtained using ResNet101 on derived vegetation indices using the red edge band of the camera (NDVIre images). Finally, in a third experiment, we aimed at evaluating the degree of damage: healthy, slightly damaged, and medium or heavily damaged trees. Again, ResNet101 had the best performance, this time on RGBI images with a test accuracy of 0.98 and an average F1 score of 0.97. These results highlight the potential of CNNs to handle high-resolution multispectral UAV data for the early detection of damaged trees when good training data are available.}, language = {en} } @inproceedings{HellBrandmeierNuechter, author = {Hell, Maximilian and Brandmeier, Melanie and N{\"u}chter, Andreas}, title = {Transferability of Deep Learning Models for Land Use/Land Cover Classification}, series = {43. Wissenschaftlich-Technische Jahrestagung der DGPF in M{\"u}nchen. 22.-23. M{\"a}rz 2023 in M{\"u}nchen - Publikationen der DGPF}, volume = {31}, booktitle = {43. Wissenschaftlich-Technische Jahrestagung der DGPF in M{\"u}nchen. 22.-23. M{\"a}rz 2023 in M{\"u}nchen - Publikationen der DGPF}, editor = {Kersten, Thomas P. and Tilly, Nora}, doi = {10.24407/KXP:1841078182}, pages = {142 -- 149}, language = {en} } @article{VahrenholdBrandmeierMueller, author = {Vahrenhold, Jan Richard and Brandmeier, Melanie and M{\"u}ller, Markus Sebastian}, title = {MMTSCNet: Multimodal Tree Species Classification Network for Classification of Multi-Source, Single-Tree LiDAR Point Clouds}, series = {Remote Sensing}, volume = {17}, journal = {Remote Sensing}, number = {7}, publisher = {MDPI AG}, issn = {2072-4292}, doi = {https://doi.org/10.3390/rs17071304}, abstract = {Trees play a critical role in climate regulation, biodiversity, and carbon storage as they cover approximately 30\% of the global land area. Nowadays, Machine Learning (ML)is key to automating large-scale tree species classification based on active and passive sensing systems, with a recent trend favoring data fusion approaches for higher accuracy. The use of 3D Deep Learning (DL) models has improved tree species classification by capturing structural and geometric data directly from point clouds. We propose a fully Multimodal Tree Species Classification Network (MMTSCNet) that processes Light Detection and Ranging (LiDAR) point clouds, Full-Waveform (FWF) data, derived features, and bidirectional, color-coded depth images in their native data formats without any modality transformation. We conduct several experiments as well as an ablation study to assess the impact of data fusion. Classification performance on the combination of Airborne Laser Scanning (ALS) data with FWF data scored the highest, achieving an Overall Accuracy (OA) of nearly 97\%, a Mean Average F1-score (MAF) of nearly 97\%, and a Kappa Coefficient of 0.96. Results for the other data subsets show that the ALS data in combination with or even without FWF data produced the best results, which was closely followed by the UAV-borne Laser Scanning (ULS) data. Additionally, it is evident that the inclusion of FWF data provided significant benefits to the classification performance, resulting in an increase in the MAF of +4.66\% for the ALS data, +4.69\% for the ULS data under leaf-on conditions, and +2.59\% for the ULS data under leaf-off conditions. The proposed model is also compared to a state-of-the-art unimodal 3D-DL model (PointNet++) as well as a feature-based unimodal DL architecture (DSTCN). The MMTSCNet architecture outperformed the other models by several percentage points, depending on the characteristics of the input data.}, language = {en} }