A research team from the University of Valencia has developed an Artificial Intelligence (AI) methodology that allows global data to generate only maps of causal interaction between regions and climate variables. The paper, published in Scientific Reports, brings to light specific hitherto unknown relationships that will help improve understanding of the Earth system and its evolution.
Unlike the well-known converged cross-mapping (CCM) technique, the new methodology developed at the Image Processing Laboratory (IPL) of the University of Valencia – the Robust CMM (RCMM) – solves the weaknesses of the previous methodology in carrying on a global scale what could only be applied at local scales.
The team has thus managed to generate the first global maps of causal interaction between regions and variables such as temperature or vegetation status, and conclude, for example, that in boreal ecosystems the soil moisture is due more to the evapotranspiration than to precipitation themselves; or that, in tropical forests, neither air temperature nor soil moisture are especially limiting factors of plant production. Among other results, the study also shows that in certain areas of the planet, there is a vicious circle of cause and effect, in which not only radiation is the cause of the level of photosynthesis, but the level of photosynthesis also has an effect on radiation.
“The Earth system is complex and has many variables that interact spatially and temporally at different scales. This new tool is able to generate global maps on these causal interactions, amplifying the traditional study based simply on correlation”, comments Emiliano Diaz, IPL researcher and main author of the work.
As explained in the article published in Scientific Reports – a journal of the Springer Nature group – before looking for and interpreting unknown climatic relationships, the study began to validate, based only on satellite data, certain relationships already known by the climate theory. The results were consistent with known patterns in earth and climate sciences and demonstrated the effectiveness of this technique in quantifying and understanding the interactions of carbon and water flows.
This gives rise to a key tool for understanding the current state of the planet, as well as its evolution in the context of climate change. “In addition, the methodology is general and can be applied to other branches of knowledge, such as social, economic and environmental sciences”, adds Gustau Camps-Valls, Professor of Electronic Engineering, owner of two ERC projects in this field and also signer of the article. “The applications are innumerable; from checking scientific hypotheses, model validity, or effects of species adaptation on ecology, to optimising treatments in the medical clinic, or identifying variables such as those that dominate changes in the economic system or those that cause the climate crisis”, add. “One possibility in the future is to make interventions in the causal model to see the effect that different emission scenarios could have on the planet”, concludes the scientist.
Related publications
Inferring causal relations from observational long-term carbon and water fluxes records. Diaz, E. and Adsuara, J. E. and Moreno-Martinez, A. and Piles, M. and Camps-Valls, G. Scientific Reports 12 :1610, 2022