Quantification of multi-source uncertainty contribution in drought prediction across Beijing area

Drought prediction is often fraught with significant uncertainties due to various factors. It is crucial to identify the key contribution source of uncertainty, and therefore enhancing the reliability of drought prediction and risk assessment. This study focuses on Beijing as a case study, employing multivariate analysis of variance (MVANOVA) to assess the contributions of uncertainties from three primary sources: Global Climate Models (GCMs), Shared Socioeconomic Pathway (SSP) scenarios, and drought indices. These factors were analyzed to predict meteorological drought characteristics such as duration, peak, and intensity, while also examining their spatiotemporal variations. Additionally, the minimum number of GCMs was explored by using the systematic sampling and derivative function methods to ensure the uncertainty triggered by GCMs reasonably estimated. The results indicated that, GCMs, and the interactions between SSPs and GCMs, are the two most significant sources of uncertainty affecting drought characteristic prediction in Beijing area. The main sources of uncertainty remain consistent across spatial and temporal scales. Furthermore, when the number of GCMs is less than seven, the uncertainty caused by the GCMs may be underestimated in Beijing area. These findings underscore the importance of reducing GCM-related uncertainties for reliable drought predictions and highlight the need to account for the interactions among factors in uncertainty research.