Original title: A Graph Neural Network Framework for Characterizing Rainfall Variability Regimes across India
Pradyumnan Raghuveeran, Gaurav Chopra, Ajay Bankar, R. I. Sujith · August 2026
Monsoon rainfall changes a lot across India. Past studies focused on predicting rainfall amounts, but they rarely asked if a place follows the same seasonal rainfall pattern each year. To solve this, researchers used a computer model to sort locations by this year-to-year consistency. For thousands of locations, they mapped each year as a point connected to other years by how similar their seasonal rainfall looked. A special neural network then labeled each spot as consistent or erratic with 96.8% accuracy. The model showed that the Western Ghats, Northeast India, and parts of central India have consistent patterns. The team also discovered that areas with higher rainfall volumes are also the most repeatable from year to year.
Imagine you are a farmer deciding which crops to plant each year. You look at how much rain your field got in past years, but you also need to know if the pattern repeats reliably. If one field gets lots of rain but the amount varies wildly, you cannot plan. Another field gets moderate rain but the same amount every year, so you can count on it. In this paper, the researchers built a system that automatically identifies which fields are reliable and which are erratic, using a network that compares year-to-year rainfall patterns.
Farmers and water managers in India depend on monsoon rains, but they have long assumed that high-rainfall areas are also unpredictable. This study shows the opposite: the wettest regions are actually the most consistent year-to-year. That means planning for water resources in places like the Western Ghats can rely on more stable patterns than previously thought.
The study uses data from 2001 to 2022 (excluding 2012), so the time span is limited. It only classifies locations into two categories (consistent or erratic), not finer gradations. The model is applied only to India, so findings may not generalize elsewhere. The observed coupling between rainfall volume and consistency is correlational, not causal.
Ask anything about this paper. Answers come from its abstract, and say so when the abstract doesn't cover your question.
A Graph Neural Network Framework for Characterizing Rainfall Variability Regimes across India — Pradyumnan Raghuveeran, Gaurav Chopra, Ajay Bankar et al.
Abstract licence: CC0
Explanation generated by Synthesis from the abstract — it is not the original paper. This is a preprint, so it hasn't necessarily been peer reviewed.