Estimating Drought Hydrographs for Different Return Periods Using Hydrological Records: Regulated and Natural Flow Regimes

This study proposes a probabilistic methodology to estimate drought hydrographs from observed discharge data recorded at a river gauging station under natural hydrological conditions and under reservoir operation for hydropower generation, with data fitted to statistical probability distributions. The methodology evaluates model calibration performance using RMSE, its normalized form (NRMSE), and percent bias (PBIAS) for return periods of 5, 10, 20, and 50 years. The natural hydrological conditions before the beginning of reservoir operation were represented by the 1972–1999 period, whereas the post-operation period after the start of the Urrá I reservoir operation (2000–2025), including the subperiods 2000–2010 and 2011–2025, represented different operational conditions regarding the minimum flows recorded in the river. The RMSE values obtained were 1.16, 6.67, 2.70, and 22.44 hm3; the NRMSE values were 0.59%, 5.14%, 2.31%, and 16.25%; and the PBIAS values were 0.08%, 1.15%, 0.27%, and 2.03%, respectively, for the periods analyzed. The 2011–2025 period showed the largest differences between recorded and modelled volumes across return periods. The valley time for the unregulated period was 42.86 days, while for the regulated periods it was 10.81 days for 2000–2025, 10.64 days for 2000–2010, and 8.06 days for 2011–2025. For the subperiods, interpret the 50-year estimates with greater statistical uncertainty. RMSE should be understood as a representation of model calibration rather than as a validation of independent estimates.