Understanding the Causalities between Multiple Environmental Variables and Droughts in Amazon Basin

Establishing an early-warning system for droughts in the Amazon Basin holds paramount importance due to the region's critical role in global climate regulation and biodiversity. Droughts in the Amazon not only impact local ecosystems and communities but also have far-reaching effects on global weather patterns and carbon storage capabilities. To fully understand the drought mechanism and improve early-warning monitoring, it is important not only to detect drought conditions by creating indicators but also to extract signals that could describe the risk of drought outbreaks. To reach this goal, our research characterizes pre-drought signals from multiple environmental variables using causal inference and information theory. This study focuses on environmental variables, such as temperature, precipitation, vapor pressure deficit, evapotranspiration rate, and relative humidity from three perspectives, spatiotemporal characteristic, anomalies, and accumulation. Environmental variables are obtained from satellite observations and reanalysis datasets. We harness the potential of these characteristics, exploring their intricate connections as precursors to drought formation and propagation. Expanding on simple association, we introduce causal inference techniques to discover causalities among environmental variables, and between environmental variables and droughts, while information theory helps us capture non-linear relationships among environmental variables. Thereby, we identify critical thresholds and pre-drought signals where these characteristics contribute to drought onset. This causality-based approach marks a departure from traditional indices, integrating temporal dynamics with a detailed understanding of system interactions. Our findings aim to contribute to sustainable land and water management in the Amazon, ultimately aiding in the preservation of its unique ecosystems and the services they provide.