Assessing Extreme Monthly Runoff Over an Arid Basin Through Reanalysis Datasets

Water availability in arid basins is a serious concern and its quantification remains uncertain. Climate variability as the El Niño Southern Oscillation phenomenon (ENSO) can worsen this scenario, exacerbating flooding in some regions such as the Piura basin on the northern Peruvian Pacific coast. We analyzed four geospatial runoff datasets covering a 7622 km² basin area at ~50 km of resolution. They are two global datasets: Global Runoff (GRUN) and Linear Optimal Runoff Aggregate (LORA) based on both hydrological reanalysis; a Peruvian dataset named PISCO-Hym-GR2M based on satellite and in-situ conceptual water balance; and a semi empirical method called Rindex based only on in-situ conceptual water balance. Those monthly datasets were compared to historical observations over four selected ENSO events which accelerated and retarded the peak monthly runoff and reached up to 317 mm/month. GRUN, LORA, PISCO-Hym-GR2M and Rindex reached month predictions of 42%, 21%, 58% and 50%, respectively with an acceptable relative error below 50% during the December–May wet period. Using hydroclimatic match indices in a Taylor diagram, correlations coefficients over 0.75 were obtained for all products, the highest corresponding to the GRUN dataset. A reliable root mean square difference near 40 mm/month for GRUN and Rindex; and a reliable amplitude of the variations for PISCO-Hym-GR2M and Rindex around 80 mm/month were obtained. Most datasets underestimated runoff with shifts in peak monthly runoff timing related to spatial aggregation of input precipitation, being LORA the most accurate in matching peaks. Those products could help to study monthly runoff fluctuations, knowing their difficulty in prediction.