Historical and future trends in wetting and drying in 291 catchments across China

Abstract. The "dry gets drier, wet gets wetter" (DDWW) pattern is a popular catchphrase to summarize hydrologic changes under global warming. However, recent studies based on simulated data have failed to obtain a feasible DDWW pattern for runoff trends. This study tested the DDWW pattern using observed streamflow and meteorological data from 291 catchments in China from 1956 to 2000, interpreted it using a simple method derived from the Budyko hypothesis, and explored its future evolution according to the projections of five global climate models (GCMs) from the Coupled Model Intercomparison Project Phase 5 (CMIP5). Similar to the DDWW pattern, the results show that catchments with an aridity index of φ 1 become drier, with nearly 80 % of the studied catchments following this pattern. However, the pattern does not hold in glacier regions due to the effects of melting ice and snow. Based on precipitation and potential evapotranspiration changes, the first-order differential of the Budyko hypothesis can provide a good estimate of runoff changes (R2 = 0.70). Therefore, the atmospheric forcing of water and energy is the key factor in interpreting the DDWW pattern. Over 80 % of the estimated trends have signs coincident with those of the measured trends, implying that the DDWW pattern can be assessed with estimated data. Precipitation is the controlling factor that leads to the DDWW pattern in nearly 90 % of catchments where observed and estimated signs are consistent. In the three tested scenarios (RCP2.6, RCP4.5 and RCP8.5), the different models produce significantly different predicted changes, even under the same scenario, whereas a given model yields similar results under different scenarios. Based on the projected results, the DDWW pattern no longer provides a reliable prediction. However, this conclusion remains tentative due to the large uncertainty of the simulations. The considerable differences between the observed and modelled meteorological data for the same period suggest that this conclusion should be adopted with caution.