Search

Since its launch in 2007, TerraSAR-X observations have been widely used in a broad range of scientific applications. Particularly in wetland research, TerraSAR-X’s shortwave X-band synthetic aperture radar (SAR) possesses unique capabilities, such as high spatial and temporal resolution, for delineating and characterizing the inherent spatially and temporally complex and heterogeneous structure of wetland ecosystems and their dynamics. As transitional areas, wetlands comprise characteristics of both terrestrial and aquatic features, forming a large diversity of wetland types. This study reviews all published articles incorporating TerraSAR-X information into wetland research to provide a comprehensive study of how this sensor has been used with regard to polarization, and the function of the data, time-series analyses, or the assessment of specific wetland ecosystem types. What is evident throughout this literature review is the synergistic fusion of multi-frequency and multi-polarization SAR sensors, sometimes optical sensors, in almost all investigated studies to attain improved wetland classification results. Due to the short revisiting time of the TerraSAR-X sensor, it is possible to compute dense SAR time-series, allowing for a more precise observation of the seasonality in dynamic wetland areas as demonstrated in many of the reviewed studies.

This is the final version of the article. It first appeared from Wiley via http://dx.doi.org/10.1002/rse2.27, As remote sensing (RS) applications and resources continue to expand, their importance for ecology and conservation increases – and so does the need for effective and successful training of professionals working in those fields. Methodological and applied courses often form part of university curricula, but their practical and long-term benefits only become clear afterwards. Having recently received such training in an interdisciplinary master’s programme, we provide our perspectives on our shared education. Through an online survey we include experiences of students and professionals in different fields. Most participants perceive their RS education as useful for their career, but express a need for more training at university level. Hands-on projects are considered the most effective learning method. Besides methodological knowledge, soft skills are clear gains, including problem solving, self-learning and finding individual solutions, and the ability to work in interdisciplinary teams. The largest identified gaps in current RS training concern the application regarding policy making, methodology and conservation. To successfully prepare students for a career, study programmes need to provide RS courses based on state-of-the-art methods, including programming, and interdisciplinary projects linking research and practice supported by a sound technical background.

Cancer is a heterogeneous disease that demands precise molecular profiling for better understanding and management. Recently, deep learning has demonstrated potentials for cost-efficient prediction of molecular alterations from histology images. While transformer-based deep learning architectures have enabled significant progress in non-medical domains, their application to histology images remains limited due to small dataset sizes coupled with the explosion of trainable parameters. Here, we develop SEQUOIA , a transformer model to predict cancer transcriptomes from whole-slide histology images. To enable the full potential of transformers, we first pre-train the model using data from 1,802 normal tissues. Then, we fine-tune and evaluate the model in 4,331 tumor samples across nine cancer types. The prediction performance is assessed at individual gene levels and pathway levels through Pearson correlation analysis and root mean square error. The generalization capacity is validated across two independent cohorts comprising 1,305 tumors. In predicting the expression levels of 25,749 genes, the highest performance is observed in cancers from breast, kidney and lung, where SEQUOIA accurately predicts the expression of 11,069, 10,086 and 8,759 genes, respectively. The accurately predicted genes are associated with the regulation of inflammatory response, cell cycles and metabolisms. While the model is trained at the tissue level, we showcase its potential in predicting spatial gene expression patterns using spatial transcriptomics datasets. Leveraging the prediction performance, we develop a digital gene expression signature that predicts the risk of recurrence in breast cancer. SEQUOIA deciphers clinically relevant gene expression patterns from histology images, opening avenues for improved cancer management and personalized therapies.