Figure 2. Trends in (a) QMAX, (b) PMAX, (c) Psm.MAX and (d) Peff.MAX across each of the 671 CAMELS catchments based on the normalized Thiel-Sen slope (Tc ). Scatter plots between T values of QMAXand Tc values of (f) PMAX, (g) Psm.MAX, (h) Peff.MAX are also shown. (d) Boxplot of the R2 between T values of annual floods and Tc ­values of annual precipitation extremes (see Figure S2 in the Supporting Information for disaggregation of results across water resources regions).
3.2 Co-occurrence probability of precipitation extremes and floods across the CONUS
The low correlation between the spatial pattern of changes in precipitation extremes and that of floods (discussed in Section 3.1) is potentially attributable to a weak causal relationship, as there are other mechanisms that could trigger floods [Blöschl et al. , 2019; Merz and Blöschl , 2003]. To assess this hypothesis, we assessed the co-occurrence probability between precipitation extremes and floods over individual catchments and the results are shown in Figure 3 (See Figure S3-S5 in the Supporting Information for QDOYMAX, PDOYMAX, Psm.DOYMAX, and Peff.DOYMAX across all catchments). The averaged co-occurrence probability across all catchments is 32%, 30%, and 37% for PDOYMAX, Psm,DOYMAX, and Peff.DOYMAX extremes respectively. This number is consistent with a previous investigation [Ivancic and Shaw , 2015], indicating that annual precipitation extremes are only responsible for about one-third of the annual flood population.