The Variable Desert

Dryland vegetation self-organizes into striking spatial patterns — stripes, spots, and labyrinths visible from satellite imagery. These patterns are not decorative. They are functional: vegetation bands on gentle slopes harvest stormwater from upslope bare zones, concentrating water where plants can use it. The patterns are a sign of ecosystem resilience — the vegetation has organized itself to survive water stress.

Gandhi, Oline, and Silber (arXiv:2501.01569) model these patterns using a flow-kick framework: storms arrive as instantaneous kicks of soil water, and between storms, the vegetation-soil system evolves according to reaction-diffusion dynamics. The approach lets them compare idealized periodic rainfall against realistic stochastic rainfall with the same average.

The key finding: introducing randomness in storm timing and magnitude reduces the parameter range that supports pattern formation, even when average annual rainfall stays constant. More variable rainfall kills the patterns.

This is not about drought. The total water input is unchanged. The damage comes from variability itself — the irregularity of when water arrives and how much arrives per event. Patterns require a rhythm: water arrives, vegetation absorbs and grows, the system reaches a state that can receive the next pulse. When the rhythm becomes unpredictable, the vegetation can’t sustain the spatial organization that lets it survive.

The implication for climate change is direct: you don’t need to reduce precipitation to destroy arid ecosystems. You only need to make the timing less predictable. Climate models consistently project increased precipitation variability in many dryland regions. The average might hold, but the pattern — the ecosystem’s survival strategy — might not.

Gandhi, Oline, & Silber, “A flow-kick model of dryland vegetation patterns: the impact of rainfall variability on resilience,” arXiv:2501.01569 (2025).