Observational source choice outweighs the emission scenario in projected dependable precipitation: evidence from 44 rain gauges in southern Brazil
Abstract Dependable precipitation, defined as a lower quantile of the distribution of monthly totals, is the quantity that guides irrigation design and the definition of sowing windows, yet it is seldom projected for future conditions. This study assembled five precipitation datasets over 44 sites in northwestern Rio Grande do Sul, Brazil, from 2002 to 2025, fitted an occurrence-intensity model for each site, dataset and month, and applied change factors derived from five NEX-GDDP-CMIP6 models under three shared socioeconomic pathways and two future windows. The datasets agree on annual volume, with a range of 12% between the highest and the lowest, and disagree on the daily structure that produces it, with 23.8% of wet days at 22.3 mm in the rain gauge network against 38.9% at 11.8 mm in the reanalysis. Across 9504 combinations of site, month and dataset, no projected change exceeded the natural variability band, the maximum signal-to-noise ratio being 0.831 and a signal 4.5 to 10.5 times larger being required for detection. The variance partition assigned 68.9% to 77.6% to natural variability, 12.7% to 18.8% to model choice and 1.5% to 1.9% to the emission pathway. For the absolute value that the end user applies, dataset choice accounted for 6.57% of the variance against 1.51%, a ratio of approximately 4.4 times.
Authors
- Alessandro Dal’Cól Lúcio (ORCID: https://orcid.org/0000-0003-0761-4200)
- Willyan Júnior Adorian Bandeira (ORCID: https://orcid.org/0000-0002-9430-3664)
- Ivan Ricardo Carvalho (ORCID: https://orcid.org/0000-0001-7947-4900)
- João Joel Ziotti
Institutions
- Universidade Federal de Santa Maria (BR)
- Universidade Regional do Noroeste do Estado do Rio Grande do Sul (BR)
Publication Details
- Journal
- Research Square
- Published
- 2026-11-12
- DOI
- https://doi.org/10.21203/rs.3.rs-11442487/v1
- Primary Topic
- Climate variability and models
- Type
- preprint