Network motifs shape distinct functioning of Earth's moisture recycling hubs

Earth's hydrological cycle critically depends on the atmospheric moisture flows connecting evaporation to precipitation. Here we convert a decade of reanalysis-based moisture simulations into a high-resolution global directed network of spatial moisture provisions. We reveal global and local network structures that offer a new view of the global hydrological cycle. We identify four terrestrial moisture recycling hubs: the Amazon Basin, the Congo Rainforest, South Asia and the Indonesian Archipelago. Network motifs reveal contrasting functioning of these regions, where the Amazon strongly relies on directed connections (feed-forward loops) for moisture redistribution and the other hubs on reciprocal moisture connections (zero loops and neighboring loops). We conclude that Earth's moisture recycling hubs are characterized by specific topologies shaping heterogeneous effects of land-use changes and climatic warming on precipitation patterns. Life on land relies on the precipitation flows that provide a steady source of freshwater for most areas on Earth. This source enables a resilient Earth system and a safe operating space for humanity1. The origins of precipitation over land are almost equally distributed between Earth's land areas and oceans2. This implies that not only atmospheric changes such as climate warming, but also modifications at the land surface can affect precipitation patterns on Earth: land cover transformations, including deforestation alter evapotranspiration flows, which may subsequently propagate across continents through evaporation-precipitation cycles3. Considering how critical they are for the Earth System, we have a surprisingly shallow understanding of how atmospheric moisture flows are arranged. Recent improvements in atmospheric reanalysis data4,5 and methodological advances to track the flows of moisture around the planet at high spatial and temporal resolutions6 now allow for a significantly more detailed picture of these flows. Previously, we developed a dataset of monthly mean atmospheric moisture flows between each pair of 0.5° (ca. 55 km at the equator) grid cells across the globe based on all evaporation (including transpiration) on Earth during 2008–20173 (see Methods). One promising approach to analyze these planet-encompassing data is as a spatial network in which each grid cell represents a node that may be connected either uni- or bidirectionally to any other node10–13. A directed link represents a mass flow from source to target, that is, a flow of moisture from its location of evaporation to that of precipitation. Such a network representation can reveal several local and global features of the underlying complex (climate) system14–16. Specifically, so-called motifs are local network structures that control how transitions may cascade across the network, and that enable assessing the sensitivity of the network to changes at certain nodes (i.e., grid cells) or edges (i.e., their links)17–19. Motifs are significantly overexpressed in real-world networks relative to random networks17,20, enhancing information transport in many different types of complex networks such as the world-wide-web, gene expression networks, or food webs18,20,21. The enhanced information transport processes by the motifs can be positive in some contexts (in this case, for example, directed moisture transport allowing the forest to grow further downwind, which otherwise may not9), but also negative (deforestation causing loss of atmospheric moisture transport to downwind forests). Therefore, motifs in the moisture

Nico Wunderling, Frederik Wolf, Obbe A. Tuinenburg, Arie Staal