Paleoclimatic drought: rainfall variability analysis

• Autor(a) principal: Carneiro, Renato Quinderé
• Data de Publicação: 2021
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Institucional da Universidade Federal do Ceará (UFC)
• Texto Completo: http://www.repositorio.ufc.br/handle/riufc/58099
• topic: Último milênio; Comparação de modelos; Modos oscilatórios

Climate variability knowledge is key information in water resources planning. This information, however, could be misleading if based only on observed series since these usually are not long enough to represent the variability in its full range. Climate reconstitution by a proxy variable is very successful in extending these series, but it finds difficulties when dealing with tropical regions such as NEB. The approach adopted in this study is to use Paleoclimatic model output from PMIP´s last millennium experiment to assess Climate variability, which leads to the problem of assessing the model´s ability to represent the climate of the studied region. In its first part, this study proposes a method for model evaluation which decomposes the series in seasonal and pluriannual components and compares the result to observed series using measures from Information Theory. The model ranking was obtained by TOPSIS in three different scenarios. In the second part, the top-ranked models for the pluriannual variability scenario were assessed through the distribution-free CUSUM test and Wavelet analysis to detect change points and oscillatory modes. Models were also compared to each other and to both solar irradiation and volcanic forcing to assess whether the detected modes were due to internal variability or to external forcings. Model comparison simulations in the seasonality scenario showed that CSIRO-Mk3L-1-2, bcc-csm1-1, and CCSM4 were the models with the best adherence to the observed series, while simulations in the pluriannual scenario showed that HadCM3, MPI-ESM-P, and EC-Earth3-Veg-LR were the models with the best adherence. In the third scenario, the best performance was due to models bcc-csm1-1, MRI-ESM2-0, and HadCM3. In the second part, no change point was detected in the series with a confidence level of 95%., while in Wavelet Analysis modes with a period range of 4 to 8 years, 16 to 32 years, and 128 to 256 years were detected. Comparisons between models and external forcings analysis led to evidence that the mode with a period range of 128 to 256 years could be caused by external forcings.