2: Image Processing Lab; Universitat de València, Spain
3: Brockmann Consult GmbH, Germany
4: Wageningen University & Research (WUR), Laboratory of Geo-Information Science and Remote Sensing, The Netherlands
5: Karlsruhe Institute of Technology (KIT), Campus Alpin IMK-IFU, Land Use Change and Climate, Germany
Global change, deforestation, pollution, and other human interventions transform our planet. This transformation needs to be monitored at the global scale. Over the last decades, ESA has made major efforts to provide comprehensive data streams describing land-surface processes. In particular the Climate Change Initiative (CCI) has produced key variables to monitor essential climate variables (ECVs) such as soil moisture, vegetation states, and land surface temperature, amongst other key processes on land, in the ocean, and the atmosphere that respond to climate variability and interact with it. Today, researchers all over the world can access this treasure chest for answering questions on dynamics and changes of planet Earth such as impacts of droughts on productivity, trends in dryness, and others.
But data is not information! Data will only tell us comprehensive stories when all variables and processes are interpreted jointly and with proper methodologies to let them speak out. This is why researchers recently put a variety of data products together and tried to come up with a unified description of processes in the Earth system – The Earth System Data Lab – extensively described in Mahecha et al. (2020, https://www.earth-syst-dynam.net/11/201/2020/). Based on this data-analytics approach, Guido Kraemer and colleagues have now showed that we can actually condense a multitude of Earth surface data into two-to-three dimensions that retain most of the information on the variance in the Earth System Data Cube: (https://www.biogeosciences.net/17/2397/2020/). The data speak a clear language: Land ecosystems are primarily shaped 1) along primary productivity patterns and 2) along water and energy availability. These emerging intrinsic dimensions of surface variability can now be used to trace the imprint of climate extremes, land cover change, and trends all alike in a single framework.
These indicators for the state of the land surface tell us, for instance, where big anomalies occur. The severe 2005 and 2010 Amazon droughts, or the Russian heat wave of 2010 are clearly visible in the emerging spatiotemporal indicator maps. In the light of the last two summers during which Europe was hit (again) by unprecedented heat waves, and recent weeks, where the Amazon fires are being widely discussed in the media, the relevance of this work becomes clear:
Only if we succeed in putting these impacts into a global perspective, we will be able to objectively judge their impacts. And, even more importantly, understand and thereby anticipating their impacts under future climate conditions. Figure 2 shows how these two events lead to “trajectories” in the representation data space that differ fundamentally from “normal”, i.e average, years. Early warning systems will soon be able to incorporate data-driven alerts of this kind to inform the public about changes even in the remotest places on Earth.
However, while the question of climate extremes is certainly among the most ardent ones for society, we should not forget about long-term trends. The indicators derived by in this publication do preserve these and offer scientists with long-needed tools to attain an integrative view on also on slowly happening changes of the land surface.
The next figure reveals such a long-term change on a continental scale: Large parts of South America have become less productive and drier over the past decade. A trend that needs to be monitored closely and continuously and verified with other remote sensing and modelling efforts and attributed to change or “just” decadal variability.
This synoptic exploitation of data streams was unforeseen at the time of their production. But this is exactly the idea behind the ESDL: it aims to bring together a multitude of data streams in the cloud and make them easily and freely accessible to the wider scientific community. These data are transformed to a consistent spatio-temporal grid and available in a virtual laboratory. Users can interact with these data streams and implement their workflows in the emerging programming language Julia (https://julialang.org), but also in Python (https://www.python.org) and other languages.
to get dryer also tend to get less productive and vice versa.