TY - JOUR A1 - Wurm, Michael A1 - Stark, Thomas A1 - Zhu, Xiao Xiang A1 - Weigand, Matthias A1 - Taubenböck, Hannes T1 - Semantic segmentation of slums in satellite images using transfer learning on fully convolutional neural networks JF - ISPRS Journal of Photogrammetry and Remote Sensing N2 - Unprecedented urbanization in particular in countries of the global south result in informal urban development processes, especially in mega cities. With an estimated 1 billion slum dwellers globally, the United Nations have made the fight against poverty the number one sustainable development goal. To provide better infrastructure and thus a better life to slum dwellers, detailed information on the spatial location and size of slums is of crucial importance. In the past, remote sensing has proven to be an extremely valuable and effective tool for mapping slums. The nature of used mapping approaches by machine learning, however, made it necessary to invest a lot of effort in training the models. Recent advances in deep learning allow for transferring trained fully convolutional networks (FCN) from one data set to another. Thus, in our study we aim at analyzing transfer learning capabilities of FCNs to slum mapping in various satellite images. A model trained on very high resolution optical satellite imagery from QuickBird is transferred to Sentinel-2 and TerraSAR-X data. While free-of-charge Sentinel-2 data is widely available, its comparably lower resolution makes slum mapping a challenging task. TerraSAR-X data on the other hand, has a higher resolution and is considered a powerful data source for intra-urban structure analysis. Due to the different image characteristics of SAR compared to optical data, however, transferring the model could not improve the performance of semantic segmentation but we observe very high accuracies for mapped slums in the optical data: QuickBird image obtains 86–88% (positive prediction value and sensitivity) and a significant increase for Sentinel-2 applying transfer learning can be observed (from 38 to 55% and from 79 to 85% for PPV and sensitivity, respectively). Using transfer learning proofs extremely valuable in retrieving information on small-scaled urban structures such as slum patches even in satellite images of decametric resolution. KW - slums KW - FCN KW - convolutional neural networks KW - deep learning KW - transfer learning Y1 - 2019 U6 - http://nbn-resolving.de/urn/resolver.pl?urn:nbn:de:bvb:20-opus-233799 VL - 150 ER - TY - JOUR A1 - Taubenböck, H. A1 - Weigand, M. A1 - Esch, T. A1 - Staab, J. A1 - Wurm, M. A1 - Mast, J. A1 - Dech, S. T1 - A new ranking of the world's largest cities—Do administrative units obscure morphological realities? JF - Remote Sensing of Environment N2 - With 37 million inhabitants, Tokyo is the world's largest city in UN statistics. With this work we call this ranking into question. Usually, global city rankings are based on nationally collected population figures, which rely on administrative units. Sprawling urban growth, however, leads to morphological city extents that may surpass conventional administrative units. In order to detect spatial discrepancies between the physical and the administrative city, we present a methodology for delimiting Morphological Urban Areas (MUAs). We understand MUAs as a territorially contiguous settlement area that can be distinguished from low-density peripheral and rural hinterlands. We design a settlement index composed of three indicators (settlement area, settlement area proportion and density within the settlements) describing a gradient of built-up density from the urban center to the periphery applying a sectoral monocentric city model. We assume that the urban-rural transition can be defined along this gradient. With it, we re-territorialize the conventional administrative units. Our data basis are recent mapping products derived from multi-sensoral Earth observation (EO) data – namely the Global Urban Footprint (GUF) and the GUF Density (GUF-DenS) – providing globally consistent knowledge about settlement locations and densities. For the re-territorialized MUAs we calculate population numbers using WorldPop data. Overall, we cover the 1692 cities with >300,000 inhabitants on our planet. In our results we compare the consistently re-territorialized MUAs and the administrative units as well as their related population figures. We find the MUA in the Pearl River Delta the largest morphologically contiguous urban agglomeration in the world with a calculated population of 42.6 million. Tokyo, in this new list ranked number 2, loses its top position. In rank-size distributions we present the resulting deviations from previous city rankings. Although many MUAs outperform administrative units by area, we find that, contrary to what we assumed, in most cases MUAs are considerably smaller than administrative units. Only in Europe we find MUAs largely outweighing administrative units in extent. KW - city size KW - urban agglomeration KW - rank-size distribution KW - remote sensing KW - global urban footprint KW - urban morphology Y1 - 2019 U6 - http://nbn-resolving.de/urn/resolver.pl?urn:nbn:de:bvb:20-opus-240634 VL - 232 ER - TY - THES A1 - Kanmegne Tamga, Dan Emmanuel T1 - Modelling Carbon Sequestration of Agroforestry Systems in West Africa using Remote Sensing T1 - Modellierung der Kohlenstoffbindung von agroforstwirtschaftlichen Systemen in Westafrika mittels Fernerkundung N2 - The production of commodities such as cocoa, rubber, oil palm and cashew, is the main driver of deforestation in West Africa (WA). The practiced production systems correspond to a land managment approach referred to as agroforestry systems (AFS), which consist of managing trees and crops on the same unit of land.Because of the ubiquity of trees, AFS reported as viable solution for climate mitigation; the carbon sequestrated by the trees could be estimated with remote sensing (RS) data and methods and reported as emission reduction efforts. However, the diversity in AFS in relation to their composition, structure and spatial distribution makes it challenging for an accurate monitoring of carbon stocks using RS. Therefore, the aim of this research is to propose a RS-based approach for the estimation of carbon sequestration in AFS across the climatic regions of WA. The main objectives were to (i) provide an accurate classification map of AFS by modelling the spatial distribution of the classification error; (ii) estimate the carbon stock of AFS in the main climatic regions of WA using RS data; (iii) evaluate the dynamic of carbon stocks within AFS across WA. Three regions of interest (ROI) were defined in Cote d'Ivoire and Burkina Faso, one in each climatic region of WA namely the Guineo-Congolian, Guinean and Sudanian, and three field campaigns were carried out for data collection. The collected data consisted of reference points for image classification, biometric tree measurements (diameter, height, species) for biomass estimation. A total of 261 samples were collected in 12 AFS across WA. For the RS data, yearly composite images from Sentinel-1 and -2 (S1 and S2), ALOS-PALSAR and GEDI data were used. A supervised classification using random forest (RF) was implemented and the classification error was assessed using the Shannon entropy generated from the class probabilities. For carbon estimation, different RS data, machine learning algorithms and carbon reference sources were compared for the prediction of the aboveground biomass in AFS. The assessment of the carbon dynamic was carried between 2017 and 2021. An average carbon map was genrated and use as reference for the comparison of annual carbon estimations, using the standard deviation as threshold. As far as the results are concerned, the classification accuracy was higher than 0.9 in all the ROIs, and AFS were mainly represented by rubber (38.9%), cocoa (36.4%), palm (10.8%) in the ROI-1, mango (15.2%) and cashew (13.4%) in ROI-2, shea tree (55.7%) and African locust bean (28.1%) in ROI-3. However, evidence of misclassification was found in cocoa, mango, and shea butter. The assessment of the classification error suggested that the error level was higher in the ROI-3 and ROI-1. The error generated from the entropy was able to reduced the level of misclassification by 63% with 11% of loss of information. Moreover, the approach was able to accuretely detect encroachement in protected areas. On carbon estimation, the highest prediction accuracy (R²>0.8) was obtained for a RF model using the combination of S1 and S2 and AGB derived from field measurements. Predictions from GEDI could only be used as reference in the ROI-1 but resulted in a prediction error was higher in cashew, mango, rubber and cocoa plantations, and the carbon stock level was higher in African locust bean (43.9 t/ha), shea butter (15 t/ha), cashew (13.8 t/ha), mango (12.8 t/ha), cocoa (7.51 t/ha) and rubber (7.33 t/ha). The analysis showed that carbon stock is determined mainly by the diameter (R²=0.45) and height (R²=0.13) of trees. It was found that crop plantations had the lowest biodiversity level, and no significant relationship was found between the considered biodiversity indices and carbon stock levels. The assessment of the spatial distribution of carbon sources and sinks showed that cashew plantations are carbon emitters due to firewood collection, while cocoa plantations showed the highest potential for carbon sequestration. The study revealed that Sentinel data could be used to support a RS-based approach for modelling carbon sequestration in AFS. Entropy could be used to map crop plantations and to monitor encroachment in protected areas. Moreover, field measurements with appropriate allometric models could ensure an accurate estimation of carbon stocks in AFS. Even though AFS in the Sudanian region had the highest carbon stocks level, there is a high potential to increase the carbon level in cocoa plantations by integrating and/or maintaining forest trees. N2 - Die Produktion von Rohstoffen wie Kakao, Kautschuk, Ölpalmen und Cashew ist die Hauptursache fur die Entwaldung in Westafrika (WA). Die verwendeten Produktionssyteme entsprechen einem Landbewirtschaftungskonzept, welches als Agroforstsysteme (AFS) bezeichnet wird und darin besteht, Baume und Nutzpflanzen auf der gleichen Landeinheit zu bewirtschaften. Aufgrund der kohlenstoffbindung durch Baumen sind AFS als praktikable Losung fur den Klimaschutz anerkannt, die Vielfalt der AFS in Bezug auf ihre Zusammensetzung, Struktur un raumliche Verteilung erschwert jedoch eine genaue Schatzung der Kohlenstoffvorrate. Hier konnen Daten und Methoden der satellitenbasierten Erdbeobachtung ansetzten. Ziel dieser Forschungsarbeit ist es daher, einen fernerkundungs-basierten Ansatz fur die Schatzung der Kohlenstoffbindung in AFS in den Klimaregionen von WA vorzuschlagen. Die Hauptziele waren (i) die Erstellung einer genauen Klassifizierungskarte von AFS durch Modellierung der raumlichen verteilung des Klassifizierungsfehlers; (ii) die Shatzung des Kohlenstoffbestands von AFS in den wichtigsten Klimaregionen von WA unter Verwendung von Fernerkundungs-daten (RS); (iii) die Bewertung der raumlichen Verteilung von Kohlenstoffquellen und -senken innerhalb von AFS in ganz WA. Fur jede Klimaregion in West Afrika wurden drei Regionen von Interesse (ROI) festgelegt, namlich die guineisch-kongolesische (ROI 1), die guineische (ROI 2) und die sudanesische Region (ROI 3) in Côte d'Ivoire und Burkina Faso, und es wurden drei Feldkampagnen zur Datenerhebung durchgefuhrt. Die gesammelten Daten bestanden aus Referenzpunkten fur die Bildklassifizierung und biometrischen Messungen (Durchmesser, Hohe, Artname) zur Schatzung der Biomasse. Insgesamt wurden 261 Proben in 12 AFS in ganz WA gesammelt. Fur die RS-Daten wurden jahrliche Komposite von Sentinel-1 und -2 (s1 und S3), ALOS-PALSAR und GEDI-Daten verwendet. Es wurde eine uberwachte Klassifizierung mit Random Forest (RF) algorithmus durch gefuhrt, und der Klassifizierungsfehler wurde anhand der aus den Klassenwahrscheinlichkeiten generierten Shannon-Entropie bewertet. Fur die Kohlenstoffschatzung wurden verschiedene RS-Daten, Algorithmen fur maschinelles Lernen und Kohlenstoff-Referenzquellen fur die Vorhesage des Kohlenstoffs in AFS verglichen. Die Bewertung der raumlichen Verteilung von Kohlenstoffsenken und -quellen basierte auf der Bewertung von Anomalien in der Kohlenstoffdynamik zwischen 2017 und 2021. Es wurde eine Karte zum durchschnittliche gebundenen Kohkenstoff erstellt, und die jahrliche Differenz wurde verwendet, um Kohlenstoffsenken und -quellen zu identifizieren. Die Klassifizierungsgenauigkeit war in allen ROI hoher als 0.9, in der Region dominierten Kautschuk (38.9%), Kakao (36.4%), Palme (10.8%) in ROI-1, Mango (15.2%) und Cashew (13.4%) in ROI-2, Sheabaum (55.7%) und Johannisbrot (28.1%) in ROI-3. Hinweise auf eine Fehlklassifizierung wurden vor allem bei Kakao, Mango un Sheabutter gefunden. Die Bewertung des Klassifizierungsfehlers ergab, dass das Fehlerniveau in ROI-3 und ROI-1 hoher war. Der aus der Entropie generiete Fehler konnte das Ausmass der Fehlklassifizierung reduzieren, ohne die gut klassifizierten Pixel zu beeintrachtigen. Ausserdem war der Ansatz in der Lage, Eingriffe in Schutzgebiete zuverlassig un akkurat zu erkennen. Was die Kohlenstoffschatzung betrifft, so wrude die hochste Vorhersagegenauigkeit (R²> 0.8)bei der Kombination von S1 und S2 mit Random Forest und AGB aus Feldmessungen erzielt. Vorhersagen von GEDI konnten nur als Referenz in der ROI verwendet werden, fuhrten aber zu einem Vorhersagefehler bei Cashew-, Mango-, Kautschuk- und Kakaoplantagen hoher war und der Kohlenstoffbestand bei Johannisbrot (43.9t/ha), Sheabutter (15 t/ha), Cashew (13.8 t/ha), Mango (12.8t/ha), Kakao (7.51 t/ha) und Kautschuk (7.33 t/ha) hoher war. Die Analyse zeigte, dass der Kohlenstoffbestand hauptsachlich durch den Durchmesser (R²=0.45) und die Hohe (R²=0.13) der Baume beeinflusst wird. Zudem wurde festgestellt, dass Plantagenkulturen die geringste Biodiversitat aufweisen, und es wurde kein signifikanter Zusammenhang zwischen Biodiversitatsindizes und Kohlenstoffvorraten festgestellt. Die Bewertung der raumlichen Verteilung von Kohlenstoffquellen und -senken zeigte, dass Cashew ein Kohlenstoffemittent ist, da in dieser Region Brennholz gesammelt wird, wahrend Kakaoplantagen wichtige Kohlenstoffsenken sind. Die Studie ergab zudem, dass Sentinel-Daten zur Unterstutzung eines RS-basierten Ansatzes fur die Modellierung der Kohlenstoffbindung in AFS verwendet werden konnten. Die Entropie konnte zur Kartierung von Anbauplantagen und zur uberwachen von Schutzgebiete verwendet werden. Daruber hinaus gewahrleisten feldmessungen mit geeigneten allometrischen Modellen eine genaue Schatzung der Kohlenstoffvorrate in AFS. Die AFS in der sudanesischen Region weisen die hochsten Kohlenstoffvorrate auf, aber es besteht die Moglichkeit, den Kohlenstoffgehalt in Kakaoplantagen durch die Integration und/oder Erhaltung von Waldbaumen zu erhoehen. KW - Sequestrierung KW - Fernerkundung KW - Westafrika KW - carbon sequestration KW - agroforestry systems KW - remote sensing KW - West Africa Y1 - 2024 U6 - http://nbn-resolving.de/urn/resolver.pl?urn:nbn:de:bvb:20-opus-369269 ER - TY - JOUR A1 - Kaspar-Ott, Irena A1 - Hertig, Elke A1 - Kaspar, Severin A1 - Pollinger, Felix A1 - Ring, Christoph A1 - Paeth, Heiko A1 - Jacobeit, Jucundus T1 - Weights for general circulation models from CMIP3/CMIP5 in a statistical downscaling framework and the impact on future Mediterranean precipitation JF - The International Journal of Climatology N2 - This study investigates the projected precipitation changes of the 21st century in the Mediterranean area with a model ensemble of all available CMIP3 and CMIP5 data based on four different scenarios. The large spread of simulated precipitation change signals underlines the need of an evaluation of the individual general circulation models in order to give higher weights to better and lower weights to worse performing models. The models' spread comprises part of the internal climate variability, but is also due to the differing skills of the circulation models. The uncertainty resulting from the latter is the aim of our weighting approach. Each weight is based on the skill to simulate key predictor variables in context of large and medium scale atmospheric circulation patterns within a statistical downscaling framework for the Mediterranean precipitation. Therefore, geopotential heights, sea level pressure, atmospheric layer thickness, horizontal wind components and humidity data at several atmospheric levels are considered. The novelty of this metric consists in avoiding the use of the precipitation data by itself for the weighting process, as state-of-the-art models still have major deficits in simulating precipitation. The application of the weights on the downscaled precipitation changes leads to more reliable and precise change signals in some Mediterranean sub-regions and seasons. The model weights differ between sub-regions and seasons, however, a clear sequence from better to worse models for the representation of precipitation in the Mediterranean area becomes apparent. KW - CMIP3 KW - CMIP5 KW - Mediterranean area KW - precipitation KW - statistical downscaling KW - weights Y1 - 2019 U6 - http://nbn-resolving.de/urn/resolver.pl?urn:nbn:de:bvb:20-opus-325628 VL - 39 ER -