Earth observation and remote sensing refer to satellites which orbit our planet and measure its surface. To be specific, passive sensors measure the radiation that is emitted from the Earth’s surface. The measurement of this radiation can give us useful information on the physical characteristics of land and water surfaces around the world at different spatial scales and over considerable time periods – of up to more than four decades!
The most prominent satellite mission for Earth Observation is likely Landsat by NASA, while only recently the European Space Agency (ESA) has launched the promising Sentinel mission. This reflects the growing popularity of remote sensing data, which can be used for manifold purposes. We can see satellite images frequently on the news, revealing information from otherwise inaccessible battlefields or reminding us of the ongoing forest loss in the World’s tropical forests. The latter is an example of the growing popularity of remote sensing imagery among environmental scientists. It allows us to map and monitor land cover and land use change, which in turn can serve as input for species distribution modelling.
Nevertheless, the usage of satellite images for scientific purposes has not always been easy. First of all, satellite imagery came with a considerable price tag for a long time. Only since 2008, Landsat imagery has become freely available and can be downloaded directly from the internet. Before, one had to order the specific desired image, which would cost around 600$ (and would cover an extent of 185 x 185 km). Thus, studying large areas and even conducting monitoring over a longer time period depended very much on the economic resources of the respective research institute.
Yet, even after freely accessible satellite imagery there still remained another obstacle for many interested users of remote sensing data around the world – computation power. To perform analysis over large areas or long periods, large amounts of data need to be ingested. With the improved spatial and spectral resolution of ESA’s Sentinel-2, this amount is even increasing!
Working on local servers with high computation power is therefore a huge advantage, yet definitely a privilege – that is again highly associated with the economic resources of the respective research institute. While universities, research groups or NGOs from the Global North may have access to these economic resources more frequently, entities from the Global South have had to face more restrictions in their possibility of analysis. This is in itself problematic, however especially since remote sensing data is often utilised to map and monitor the ongoing land use change and forest loss in the Global South (e.g. Amazonia). Yet, these analyses could not be conducted by local research entities.
When Google launched the Google Earth Engine (GEE), this issue could finally be addressed. GEE is a cloud-based platform, which means that all the relevant satellite data is already available in a central cloud and can be accessed via the internet. Moreover, satellite data can be filtered for specific dates, seasons, regions and atmospheric conditions. Subsequently, users can create composites and mosaics, which can be downloaded to Google Drive and then used for further analysis on one’s personal machine. Another alternative is to directly conduct further analysis in GEE, which includes classifications and time series analysis (e.g. LandTrendr). This allows users around the globe to process large amounts of satellite data and conduct sophisticated analyses with no other hardware condition but: internet access!
GEE Timelapse of Katavi National Park (1984-2020).
Want to check out Google Earth Engine? Find it here.
A final remark for all enthusiastic environmental researchers: Although we can now request and visualize satellite data in the blink of an eye (literally), cloud computing still requires energy. So be mindful when exploring the “magic” of GEE.