Risk–service mismatches reveal conservation and restoration priorities in China’s major urban agglomerations

Rapid urbanization can expose high-value ecosystems to disturbance while degrading ecological functions elsewhere. We combined the landscape ecological risk index ( LERI ), ecosystem service value ( ESV ), spatially validated machine learning and bivariate zoning to assess risk–service mismatches in the Beijing–Tianjin–Hebei region (BTH), Yangtze River Delta (YRD) and Guangdong–Hong Kong–Macao Greater Bay Area (GBA) at five time points from 2000 to 2020. Mean LERI declined by 2.25%, 8.32% and 6.48%, respectively. On a fixed region-specific 2020 valuation basis, ESV declined until 2015 and then recovered in BTH, peaked in 2010 in the YRD, and decreased steadily in the GBA. Sensitivity-zone shares fell from 27.00% to 23.51%, 26.68% to 20.21% and 31.75% to 20.15%, respectively, but all four zoning types remained present. Six full-variable models reconstructed ESV in spatially withheld cells ( R 2 = 0.690–0.837), whereas contextual-only models did not outperform the mean-response benchmark. This contrast reflects structural coupling, not independent ecological mechanisms. The combined assessment distinguishes protection of valuable but exposed assets from restoration of low-value, high-risk landscapes. It provides a temporally comparable basis for screening priorities while accounting for the different constraints of mountain–plain, river–lake and coastal–estuarine systems.