Refine
Has Fulltext
- yes (10) (remove)
Is part of the Bibliography
- yes (10)
Document Type
- Journal article (9)
- Doctoral Thesis (1)
Keywords
- machine learning (10) (remove)
Institute
- Institut für Geographie und Geologie (10) (remove)
Sonstige beteiligte Institutionen
The fast and accurate yield estimates with the increasing availability and variety of global satellite products and the rapid development of new algorithms remain a goal for precision agriculture and food security. However, the consistency and reliability of suitable methodologies that provide accurate crop yield outcomes still need to be explored. The study investigates the coupling of crop modeling and machine learning (ML) to improve the yield prediction of winter wheat (WW) and oil seed rape (OSR) and provides examples for the Free State of Bavaria (70,550 km2), Germany, in 2019. The main objectives are to find whether a coupling approach [Light Use Efficiency (LUE) + Random Forest (RF)] would result in better and more accurate yield predictions compared to results provided with other models not using the LUE. Four different RF models [RF1 (input: Normalized Difference Vegetation Index (NDVI)), RF2 (input: climate variables), RF3 (input: NDVI + climate variables), RF4 (input: LUE generated biomass + climate variables)], and one semi-empiric LUE model were designed with different input requirements to find the best predictors of crop monitoring. The results indicate that the individual use of the NDVI (in RF1) and the climate variables (in RF2) could not be the most accurate, reliable, and precise solution for crop monitoring; however, their combined use (in RF3) resulted in higher accuracies. Notably, the study suggested the coupling of the LUE model variables to the RF4 model can reduce the relative root mean square error (RRMSE) from −8% (WW) and −1.6% (OSR) and increase the R
2 by 14.3% (for both WW and OSR), compared to results just relying on LUE. Moreover, the research compares models yield outputs by inputting three different spatial inputs: Sentinel-2(S)-MOD13Q1 (10 m), Landsat (L)-MOD13Q1 (30 m), and MOD13Q1 (MODIS) (250 m). The S-MOD13Q1 data has relatively improved the performance of models with higher mean R
2 [0.80 (WW), 0.69 (OSR)], and lower RRMSE (%) (9.18, 10.21) compared to L-MOD13Q1 (30 m) and MOD13Q1 (250 m). Satellite-based crop biomass, solar radiation, and temperature are found to be the most influential variables in the yield prediction of both crops.
Supraglacial lakes can have considerable impact on ice sheet mass balance and global sea-level-rise through ice shelf fracturing and subsequent glacier speedup. In Antarctica, the distribution and temporal development of supraglacial lakes as well as their potential contribution to increased ice mass loss remains largely unknown, requiring a detailed mapping of the Antarctic surface hydrological network. In this study, we employ a Machine Learning algorithm trained on Sentinel-2 and auxiliary TanDEM-X topographic data for automated mapping of Antarctic supraglacial lakes. To ensure the spatio-temporal transferability of our method, a Random Forest was trained on 14 training regions and applied over eight spatially independent test regions distributed across the whole Antarctic continent. In addition, we employed our workflow for large-scale application over Amery Ice Shelf where we calculated interannual supraglacial lake dynamics between 2017 and 2020 at full ice shelf coverage. To validate our supraglacial lake detection algorithm, we randomly created point samples over our classification results and compared them to Sentinel-2 imagery. The point comparisons were evaluated using a confusion matrix for calculation of selected accuracy metrics. Our analysis revealed wide-spread supraglacial lake occurrence in all three Antarctic regions. For the first time, we identified supraglacial meltwater features on Abbott, Hull and Cosgrove Ice Shelves in West Antarctica as well as for the entire Amery Ice Shelf for years 2017–2020. Over Amery Ice Shelf, maximum lake extent varied strongly between the years with the 2019 melt season characterized by the largest areal coverage of supraglacial lakes (~763 km\(^2\)). The accuracy assessment over the test regions revealed an average Kappa coefficient of 0.86 where the largest value of Kappa reached 0.98 over George VI Ice Shelf. Future developments will involve the generation of circum-Antarctic supraglacial lake mapping products as well as their use for further methodological developments using Sentinel-1 SAR data in order to characterize intraannual supraglacial meltwater dynamics also during polar night and independent of meteorological conditions. In summary, the implementation of the Random Forest classifier enabled the development of the first automated mapping method applied to Sentinel-2 data distributed across all three Antarctic regions.
Supraglacial meltwater accumulation on ice sheets can be a main driver for accelerated ice discharge, mass loss, and global sea-level-rise. With further increasing surface air temperatures, meltwater-induced hydrofracturing, basal sliding, or surface thinning will cumulate and most likely trigger unprecedented ice mass loss on the Greenland and Antarctic ice sheets. While the Greenland surface hydrological network as well as its impacts on ice dynamics and mass balance has been studied in much detail, Antarctic supraglacial lakes remain understudied with a circum-Antarctic record of their spatio-temporal development entirely lacking. This study provides the first automated supraglacial lake extent mapping method using Sentinel-1 synthetic aperture radar (SAR) imagery over Antarctica and complements the developed optical Sentinel-2 supraglacial lake detection algorithm presented in our companion paper. In detail, we propose the use of a modified U-Net for semantic segmentation of supraglacial lakes in single-polarized Sentinel-1 imagery. The convolutional neural network (CNN) is implemented with residual connections for optimized performance as well as an Atrous Spatial Pyramid Pooling (ASPP) module for multiscale feature extraction. The algorithm is trained on 21,200 Sentinel-1 image patches and evaluated in ten spatially or temporally independent test acquisitions. In addition, George VI Ice Shelf is analyzed for intra-annual lake dynamics throughout austral summer 2019/2020 and a decision-level fused Sentinel-1 and Sentinel-2 maximum lake extent mapping product is presented for January 2020 revealing a more complete supraglacial lake coverage (~770 km\(^2\)) than the individual single-sensor products. Classification results confirm the reliability of the proposed workflow with an average Kappa coefficient of 0.925 and a F\(_1\)-score of 93.0% for the supraglacial water class across all test regions. Furthermore, the algorithm is applied in an additional test region covering supraglacial lakes on the Greenland ice sheet which further highlights the potential for spatio-temporal transferability. Future work involves the integration of more training data as well as intra-annual analyses of supraglacial lake occurrence across the whole continent and with focus on supraglacial lake development throughout a summer melt season and into Antarctic winter.
In most countries, freight is predominantly transported by road cargo trucks. We present a new satellite remote sensing method for detecting moving trucks on roads using Sentinel-2 data. The method exploits a temporal sensing offset of the Sentinel-2 multispectral instrument, causing spatially and spectrally distorted signatures of moving objects. A random forest classifier was trained (overall accuracy: 84%) on visual-near-infrared-spectra of 2500 globally labelled targets. Based on the classification, the target objects were extracted using a developed recursive neighbourhood search. The speed and the heading of the objects were approximated. Detections were validated by employing 350 globally labelled target boxes (mean F\(_1\) score: 0.74). The lowest F\(_1\) score was achieved in Kenya (0.36), the highest in Poland (0.88). Furthermore, validated at 26 traffic count stations in Germany on in sum 390 dates, the truck detections correlate spatio-temporally with station figures (Pearson r-value: 0.82, RMSE: 43.7). Absolute counts were underestimated on 81% of the dates. The detection performance may differ by season and road condition. Hence, the method is only suitable for approximating the relative truck traffic abundance rather than providing accurate absolute counts. However, existing road cargo monitoring methods that rely on traffic count stations or very high resolution remote sensing data have limited global availability. The proposed moving truck detection method could fill this gap, particularly where other information on road cargo traffic are sparse by employing globally and freely available Sentinel-2 data. It is inferior to the accuracy and the temporal detail of station counts, but superior in terms of spatial coverage.
In Earth observation (EO), large-scale land-surface dynamics are traditionally analyzed by investigating aggregated classes. The increase in data with a very high spatial resolution enables investigations on a fine-grained feature level which can help us to better understand the dynamics of land surfaces by taking object dynamics into account. To extract fine-grained features and objects, the most popular deep-learning model for image analysis is commonly used: the convolutional neural network (CNN). In this review, we provide a comprehensive overview of the impact of deep learning on EO applications by reviewing 429 studies on image segmentation and object detection with CNNs. We extensively examine the spatial distribution of study sites, employed sensors, used datasets and CNN architectures, and give a thorough overview of applications in EO which used CNNs. Our main finding is that CNNs are in an advanced transition phase from computer vision to EO. Upon this, we argue that in the near future, investigations which analyze object dynamics with CNNs will have a significant impact on EO research. With a focus on EO applications in this Part II, we complete the methodological review provided in Part I.
Deep learning (DL) has great influence on large parts of science and increasingly established itself as an adaptive method for new challenges in the field of Earth observation (EO). Nevertheless, the entry barriers for EO researchers are high due to the dense and rapidly developing field mainly driven by advances in computer vision (CV). To lower the barriers for researchers in EO, this review gives an overview of the evolution of DL with a focus on image segmentation and object detection in convolutional neural networks (CNN). The survey starts in 2012, when a CNN set new standards in image recognition, and lasts until late 2019. Thereby, we highlight the connections between the most important CNN architectures and cornerstones coming from CV in order to alleviate the evaluation of modern DL models. Furthermore, we briefly outline the evolution of the most popular DL frameworks and provide a summary of datasets in EO. By discussing well performing DL architectures on these datasets as well as reflecting on advances made in CV and their impact on future research in EO, we narrow the gap between the reviewed, theoretical concepts from CV and practical application in EO.
Snow is a vital environmental parameter and dynamically responsive to climate change, particularly in mountainous regions. Snow cover can be monitored at variable spatial scales using Earth Observation (EO) data. Long-lasting remote sensing missions enable the generation of multi-decadal time series and thus the detection of long-term trends. However, there have been few attempts to use these to model future snow cover dynamics. In this study, we, therefore, explore the potential of such time series to forecast the Snow Line Elevation (SLE) in the European Alps. We generate monthly SLE time series from the entire Landsat archive (1985–2021) in 43 Alpine catchments. Positive long-term SLE change rates are detected, with the highest rates (5–8 m/y) in the Western and Central Alps. We utilize this SLE dataset to implement and evaluate seven uni-variate time series modeling and forecasting approaches. The best results were achieved by Random Forests, with a Nash–Sutcliffe efficiency (NSE) of 0.79 and a Mean Absolute Error (MAE) of 258 m, Telescope (0.76, 268 m), and seasonal ARIMA (0.75, 270 m). Since the model performance varies strongly with the input data, we developed a combined forecast based on the best-performing methods in each catchment. This approach was then used to forecast the SLE for the years 2022–2029. In the majority of the catchments, the shift of the forecast median SLE level retained the sign of the long-term trend. In cases where a deviating SLE dynamic is forecast, a discussion based on the unique properties of the catchment and past SLE dynamics is required. In the future, we expect major improvements in our SLE forecasting efforts by including external predictor variables in a multi-variate modeling approach.
Forecasting spatio-temporal dynamics on the land surface using Earth Observation data — a review
(2020)
Reliable forecasts on the impacts of global change on the land surface are vital to inform the actions of policy and decision makers to mitigate consequences and secure livelihoods. Geospatial Earth Observation (EO) data from remote sensing satellites has been collected continuously for 40 years and has the potential to facilitate the spatio-temporal forecasting of land surface dynamics. In this review we compiled 143 papers on EO-based forecasting of all aspects of the land surface published in 16 high-ranking remote sensing journals within the past decade. We analyzed the literature regarding research focus, the spatial scope of the study, the forecasting method applied, as well as the temporal and technical properties of the input data. We categorized the identified forecasting methods according to their temporal forecasting mechanism and the type of input data. Time-lagged regressions which are predominantly used for crop yield forecasting and approaches based on Markov Chains for future land use and land cover simulation are the most established methods. The use of external climate projections allows the forecasting of numerical land surface parameters up to one hundred years into the future, while auto-regressive time series modeling can account for intra-annual variances. Machine learning methods have been increasingly used in all categories and multivariate modeling that integrates multiple data sources appears to be more popular than univariate auto-regressive modeling despite the availability of continuously expanding time series data. Regardless of the method, reliable EO-based forecasting requires high-level remote sensing data products and the resulting computational demand appears to be the main reason that most forecasts are conducted only on a local scale. In the upcoming years, however, we expect this to change with further advances in the field of machine learning, the publication of new global datasets, and the further establishment of cloud computing for data processing.
In the past decades, various Earth observation-based time series products have emerged, which have enabled studies and analysis of global change processes. Besides their contribution to understanding past processes, time series datasets hold enormous potential for predictive modeling and thereby meet the demands of decision makers on future scenarios. In order to further exploit these data, a novel pixel-based approach has been introduced, which is the spatio-temporal matrix (STM). The approach integrates the historical characteristics of a specific land cover at a high temporal frequency in order to interpret the spatial and temporal information for the neighborhood of a given target pixel. The provided information can be exploited with common predictive models and algorithms. In this study, this approach was utilized and evaluated for the prediction of future urban/built-settlement growth. Random forest and multi-layer perceptron were employed for the prediction. The tests have been carried out with training strategies based on a one-year and a ten-year time span for the urban agglomerations of Surat (India), Ho-Chi-Minh City (Vietnam), and Abidjan (Ivory Coast). The slope, land use, exclusion, urban, transportation, hillshade (SLEUTH) model was selected as a baseline indicator for the performance evaluation. The statistical results from the receiver operating characteristic curve (ROC) demonstrate a good ability of the STM to facilitate the prediction of future settlement growth and its transferability to different cities, with area under the curve (AUC) values greater than 0.85. Compared with SLEUTH, the STM-based model achieved higher AUC in all of the test cases, while being independent of the additional datasets for the restricted and the preferential development areas.
Grünflächen stellen einen der wichtigsten Umwelteinflüsse in der Wohnumwelt der Menschen dar. Einerseits wirken sie sich positiv auf die physische und mentale Gesundheit der Menschen aus, andererseits können Grünflächen auch negative Wirkungen anderer Faktoren abmildern, wie beispielsweise die im Laufe des Klimawandels zunehmenden Hitzeereignisse. Dennoch sind Grünflächen nicht für die gesamte Bevölkerung gleichermaßen zugänglich. Bestehende Forschung im Kontext der Umweltgerechtigkeit (UG) konnte bereits aufzeigen, dass unterschiedliche sozio-ökonomische und demographische Gruppen der deutschen Bevölkerung unterschiedlichen Zugriff auf Grünflächen haben. An bestehenden Analysen von Umwelteinflüssen im Kontext der UG wird kritisiert, dass die Auswertung geographischer Daten häufig auf zu stark aggregiertem Level geschieht, wodurch lokal spezifische Expositionen nicht mehr genau abgebildet werden. Dies trifft insbesondere für großflächig angelegte Studien zu. So werden wichtige räumliche Informationen verloren. Doch moderne Erdbeobachtungs- und Geodaten sind so detailliert wie nie und Methoden des maschinellen Lernens ermöglichen die effiziente Verarbeitung zur Ableitung höherwertiger Informationen.
Das übergeordnete Ziel dieser Arbeit besteht darin, am Beispiel von Grünflächen in Deutschland methodische Schritte der systematischen Umwandlung umfassender Geodaten in relevante Geoinformationen für die großflächige und hochaufgelöste Analyse von Umwelteigenschaften aufzuzeigen und durchzuführen. An der Schnittstelle der Disziplinen Fernerkundung, Geoinformatik, Sozialgeographie und Umweltgerechtigkeitsforschung sollen Potenziale moderner Methoden für die Verbesserung der räumlichen und semantischen Auflösung von Geoinformationen erforscht werden. Hierfür werden Methoden des maschinellen Lernens eingesetzt, um Landbedeckung und -nutzung auf nationaler Ebene zu erfassen. Diese Entwicklungen sollen dazu beitragen bestehende Datenlücken zu schließen und Aufschluss über die Verteilungsgerechtigkeit von Grünflächen zu bieten.
Diese Dissertation gliedert sich in drei konzeptionelle Teilschritte. Im ersten Studienteil werden Erdbeobachtungsdaten der Sentinel-2 Satelliten zur deutschlandweiten Klassifikation von Landbedeckungsinformationen verwendet. In Kombination mit punktuellen Referenzdaten der europaweiten Erfassung für Landbedeckungs- und Landnutzungsinformationen des Land Use and Coverage Area Frame Survey (LUCAS) wird ein maschinelles Lernverfahren trainiert. In diesem Kontext werden verschiedene Vorverarbeitungsschritte der LUCAS-Daten und deren Einfluss auf die Klassifikationsgenauigkeit beleuchtet. Das Klassifikationsverfahren ist in der Lage Landbedeckungsinformationen auch in komplexen urbanen Gebieten mit hoher Genauigkeit abzuleiten. Ein Ergebnis des Studienteils ist eine deutschlandweite Landbedeckungsklassifikation mit einer Gesamtgenauigkeit von 93,07 %, welche im weiteren Verlauf der Arbeit genutzt wird, um grüne Landbedeckung (GLC) räumlich zu quantifizieren.
Im zweiten konzeptionellen Teil der Arbeit steht die differenzierte Betrachtung von Grünflächen anhand des Beispiels öffentlicher Grünflächen (PGS), die häufig Gegenstand der UG-Forschung ist, im Vordergrund. Doch eine häufig verwendete Quelle für räumliche Daten zu öffentlichen Grünflächen, der European Urban Atlas (EUA), wird bisher nicht flächendeckend für Deutschland erhoben. Dieser Studienteil verfolgt einen datengetriebenen Ansatz, die Verfügbarkeit von öffentlichem Grün auf der räumlichen Ebene von Nachbarschaften für ganz Deutschland zu ermitteln. Hierfür dienen bereits vom EUA erfasste Gebiete als Referenz. Mithilfe einer Kombination von Erdbeobachtungsdaten und Informationen aus dem OpenStreetMap-Projekt wird ein Deep Learning -basiertes Fusionsnetzwerk erstellt, welche die verfügbare Fläche von öffentlichem Grün quantifiziert. Das Ergebnis dieses Schrittes ist ein Modell, welches genutzt wird, um die Menge öffentlicher Grünflächen in der Nachbarschaft zu schätzen (𝑅 2 = 0.952).
Der dritte Studienteil greift die Ergebnisse der ersten beiden Studienteile auf und betrachtet die Verteilung von Grünflächen in Deutschland unter Hinzunahme von georeferenzierten Bevölkerungsdaten. Diese exemplarische Analyse unterscheidet dabei Grünflächen nach zwei Typen: GLC und PGS. Zunächst wird mithilfe deskriptiver Statistiken die generelle Grünflächenverteilung in der Bevölkerung Deutschlands beleuchtet. Daraufhin wird die Verteilungsgerechtigkeit anhand gängiger Gerechtigkeitsmetriken bestimmt. Abschließend werden die Zusammenhänge zwischen der demographischen Komposition der Nachbarschaft und der verfügbaren Menge von Grünflächen anhand dreier exemplarischer soziodemographischer Gesellschaftsgruppen untersucht. Die Analyse zeigt starke Unterschiede der Verfügbarkeit von PGS zwischen städtischen und ländlichen Gebieten. Ein höherer Prozentsatz der Stadtbevölkerung hat Zugriff das Mindestmaß von PGS gemessen an der Vorgabe der Weltgesundheitsorganisation. Die Ergebnisse zeigen auch einen deutlichen Unterschied bezüglich der Verteilungsgerechtigkeit zwischen GLC und PGS und verdeutlichen die Relevanz der Unterscheidung von Grünflächentypen für derartige
Untersuchungen. Die abschließende Betrachtung verschiedener Bevölkerungsgruppen arbeitet Unterschiede auf soziodemographischer Ebene auf.
In der Zusammenschau demonstriert diese Arbeit wie moderne Geodaten und Methoden des maschinellen Lernens genutzt werden können bisherige Limitierungen räumlicher Datensätze zu überwinden. Am Beispiel von Grünflächen in der Wohnumgebung der Bevölkerung Deutschlands wird gezeigt, dass landesweite Analysen zur Umweltgerechtigkeit durch hochaufgelöste und lokal feingliedrige geographische Informationen bereichert werden können. Diese Arbeit verdeutlicht, wie die Methoden der Erdbeobachtung und Geoinformatik einen wichtigen Beitrag leisten können, die Ungleichheit der Wohnumwelt der Menschen zu identifizieren und schlussendlich den nachhaltigen Siedlungsbau in Form von objektiven Informationen zu unterstützen und überwachen.