|Paper title||Offline! Visualization and labeling of multi-sensor time series in the field: the EO Time Series Viewer|
|Form of presentation||Poster|
Underpinning EO-based findings with field-based evidence is often indispensable. However, especially in field work, there are countless situations where access to web-based services like Collect Earth or the Google Earth Engine (GEE) is limited or even impossible, such as in rainforests or deserts across the globe. Being able to visualize Earth observation (EO) time series data offline “in the field” improves the understanding of environmental conditions on the spot, and supports the implementation of field work, e.g., during planning of day trips and communication with local stakeholders. More broadly, there are various cases where EO time series, derived products, and additional geospatial information, like VHR images and cadastral data, exists in local data storages and needs to be visualized. For example, to better understand land cover and timing of land use changes, such as deforestation or agricultural management events, or gradual changes associated with degradation and regrowth.
Several specialized software tools have been developed to support the visualization of EO time series data. However, most of these tools work only on single platforms, with selected input data sources, and with specific response designs. There is a need for flexible tools that can visualize multi-source satellite time series consistently and aid reference data collection, e.g., for training and validation of supervised approaches.
To overcome these limitations, we developed the EO Time Series Viewer, a free and open-source plugin for QGIS (Jakimow et al. 2020). It provides a graphical user interface for an integrated and interactive visualization of the spectral, spatial, and temporal domains of raster time series from multiple sensors. It allows for a very flexible visualization of time series data in multiple image chips relating to (i) different observation dates, (ii) different band combinations, and (iii) across sensors with different spatial and spectral characteristics. This spatial visualization concept is complemented by (iv) spectral- and (v) temporal profiles that can be interactively displayed and compared between different map locations, sensors and spectral bands or derived spectral index formulations.
The EO Time Series Viewer accelerates the collection (“labeling”) of reference information. It provides various short-cuts to focus on areas and observation dates of interest, and to describe them based on common vector data formats. This helps, for example, to create training data for supervised mapping approaches, or to label large numbers of randomly selected points required for accuracy assessments.
Being a QGIS plugin, the EO Time Series Viewer supports a wide range of data formats and can be used across different platforms, offline or in cloud-services, in commercial and none-commercial applications, and together with other QGIS plugins, like the GEE Timeseries explorer, that is specialized on accessing cloud-based GEE datasets.
We will demonstrate the EO Time Series Viewer and its visualization & labeling concepts using a multi-sensor time series of Sentinel-2, Landsat, RapidEye and Pleiades observations for a field study site in the Brazilian Amazon. Furthermore, we will share our experiences in developing within the QGIS ecosystem and give an outlook on future developments of the EO Time Series Viewer.