Two people talk at once, and by the loudness of their voices you can't tell who is answering whom. It's the same with climate: warm water off the coast of Peru and rains in Australia often change together, but who influences whom? The usual method, correlation, simply notes that both events occur almost at the same time. But that doesn't say anything about cause and doesn't show a delay. Yet the ocean warms the air, and the response may come months later.

The new method, Granger causality, works like an interpreter who watches the order of utterances. It asks: if you know what the ocean temperature was in the past, can you predict rainfall elsewhere more accurately? If yes, then a signal is indeed being transmitted. The scientists tested both methods on data from 1980 to 2018. Both found known connections: droughts in Australia, rains in Amazonia. But the new method also spotted a link to rains in central India and southern California, where the old method was silent.

The new method proved robust: conclusions didn't change when the data were smoothed in different ways, from a week to a month. It's like an interpreter who understands meaning whether the speech is fast or slow. Now scientists can search for links between any climate phenomena and better predict rains and droughts. And the newly found connections, for instance with India, call for explanation, which could lead to discoveries.