Extended Data Fig. 4: Province-level stability of optimal inter-provincial pairing across meteorological years (2000–2022).

a, Top-1 partner retention under strategies B and C (flexibility = 0.8). Maps show the fraction of years (2000– 2022) in which the 2022 optimal partner of each province remains the top-ranked partner. Strategy B (left) exhibits generally high retention, indicating that most provinces repeatedly select the same neighbouring partner or a very small set of adjacent partners. Strategy C (right) shows lower top-1 retention, not because complementarity is unstable but because several partner provinces often offer near-equivalent complementarity for a given province, so small year-to-year meteorological variations can shift the single best partner within this limited candidate set. b, Heatmaps of partner-selection frequency under strategies B and C (flexibility = 0.8). Each cell reports the share of years in which a partner province is the top-ranked match for the source province. Under strategy B (left), most provinces concentrate weight on one or two clearly dominant partners, consistent with the high retention in panel a. Under strategy C (right), many load provinces—particularly in central and eastern China—draw repeatedly from a small group of high-resource western and northwestern partners, leading to several partners sharing intermediate frequencies rather than a single persistent top-1 choice. Together, these results show that the structure of inter-provincial complementarity is highly robust across distinct meteorological years, even though the identity of the single best partner is not always unique.