In the past, different approaches for automated landslide identification based on multispectral satellite remote sensing were developed to focus on the analysis of the spatial distribution of landslide occurrences related to distinct triggering events. However, many regions, including southern Kyrgyzstan, experience ongoing process activity requiring continual multi-temporal analysis. For this purpose, an automated object-oriented landslide mapping approach has been developed based on RapidEye time series data complemented by relief information. The approach builds on analyzing temporal NDVI-trajectories for the separation between landslide-related surface changes and other land cover changes. To accommodate the variety of landslide phenomena occurring in the 7500 km² study area, a combination of pixel-based multiple thresholds and object-oriented analysis has been implemented including the discrimination of uncertainty-related landslide likelihood classes. Applying the approach to the whole study area for the time period between 2009 and 2013 has resulted in the multi-temporal identification of 471 landslide objects. A quantitative accuracy assessment for two independent validation sites has revealed overall high mapping accuracy (Quality Percentage: 80%), proving the suitability of the developed approach for efficient spatiotemporal landslide mapping over large areas, representing an important prerequisite for objective landslide hazard and risk assessment at the regional scale.

In this article, we propose a deep learning-based algorithm for the classification of crop types from Sentinel-1 and Sentinel-2 time series data which is based on the celebrated transformer architecture. Crucially, we enable our algorithm to do early classification, i.e., predict crop types at arbitrary time points early in the year with a single trained model (progressive intra-season classification). Such early season predictions are of practical relevance for instance for yield forecasts or the modeling of agricultural water balances, therefore being important for the public as well as the private sector. Furthermore, we improve the mechanism of combining different data sources for the prediction task, allowing for both optical and radar data as inputs (multi-modal data fusion) without the need for temporal interpolation. We can demonstrate the effectiveness of our approach on an extensive data set from three federal states of Germany reaching an average F1 score of $0.92$ using data of a complete growing season to predict the eight most important crop types and an F1 score above $0.8$ when doing early classification at least one month before harvest time. In carefully chosen experiments, we can show that our model generalizes well in time and space.

Reliable multi-temporal landslide detection over longer periods of time requires multi-sensor time series data characterized by high internal geometric stability, as well as high relative and absolute accuracy. For this purpose, a new methodology for fully automated co-registration has been developed allowing efficient and robust spatial alignment of standard orthorectified data products originating from a multitude of optical satellite remote sensing data of varying spatial resolution. Correlation-based co-registration uses world-wide available terrain corrected Landsat Level 1T time series data as the spatial reference, ensuring global applicability. The developed approach has been applied to a multi-sensor time series of 592 remote sensing datasets covering an approximately 12,000 km² area in Southern Kyrgyzstan (Central Asia) strongly affected by landslides. The database contains images acquired during the last 26 years by Landsat (E)TM, ASTER, SPOT and RapidEye sensors. Analysis of the spatial shifts obtained from co-registration has revealed sensor-specific alignments ranging between 5 m and more than 400 m. Overall accuracy assessment of these alignments has resulted in a high relative image-to-image accuracy of 17 m (RMSE) and a high absolute accuracy of 23 m (RMSE) for the whole co-registered database, making it suitable for multi-temporal landslide detection at a regional scale in Southern Kyrgyzstan.

Large areas in southern Kyrgyzstan are subjected to high and ongoing landslide activity; however, an objective and systematic assessment of landslide susceptibility at a regional level has not yet been conducted. In this paper, we investigate the contribution that remote sensing can provide to facilitate a quantitative landslide hazard assessment at a regional scale under the condition of data scarcity. We performed a landslide susceptibility and hazard assessment based on a multi-temporal landslide inventory that was derived from a 30-year time series of satellite remote sensing data using an automated identification approach. To evaluate the effect of the resulting inventory on the landslide susceptibility assessment, we calculated an alternative susceptibility model using a historical inventory that was derived by an expert through combining visual interpretation of remote sensing data with already existing knowledge on landslide activity in this region. For both susceptibility models, the same predisposing factors were used: geology, stream power index, absolute height, aspect and slope. A comparison of the two models revealed that using the multi-temporal landslide inventory covering the 30-year period results in model coefficients and susceptibility values that more strongly reflect the properties of the most recent landslide activity. Overall, both susceptibility maps present the highest susceptibility values for similar regions and are characterized by acceptable to high predictive performances. We conclude that the results of the automated landslide detection provide a suitable landslide inventory for a reliable large-area landslide susceptibility assessment. We also used the temporal information of the automatically detected multi-temporal landslide inventory to assess the temporal component of landslide hazard in the form of exceedance probability. The results show the great potential of satellite remote sensing for deriving detailed and systematic spatio-temporal information on landslide occurrences, which can significantly improve landslide susceptibility and hazard assessment at a regional scale, particularly in data-scarce regions such as Kyrgyzstan.

This study combines spaceborne multitemporal and hyperspectral data to analyze the spatial distribution of surface evaporite minerals and changes in a semi-arid depositional environment associated with episodic flooding events, the Omongwa salt pan (Kalahari, Namibia). The dynamic of the surface crust is evaluated by a change-detection approach using the Iterative-reweighted Multivariate Alteration Detection (IR-MAD) based on the Landsat archive imagery from 1984 to 2015. The results show that the salt pan is a highly dynamic and heterogeneous landform. A change gradient is observed from very stable pan border to a highly dynamic central pan. On the basis of hyperspectral EO-1 Hyperion images, the current distribution of surface evaporite minerals is characterized using Spectral Mixture Analysis (SMA). Assessment of field and image endmembers revealed that the pan surface can be categorized into three major crust types based on diagnostic absorption features and mineralogical ground truth data. The mineralogical crust types are related to different zones of surface change as well as pan morphology that influences brine flow during the pan inundation and desiccation cycles. These combined information are used to spatially map depositional environments where the more dynamic halite crust concentrates in lower areas although stable gypsum and calcite/sepiolite crusts appear in higher elevated areas.