<p>Accurately downscaling runoff in ungauged sub-basins remains a fundamental challenge in hydrology, with critical implications for water resource management, flood forecasting, and drought mitigation. This study proposes a framework to enhance runoff specification in data-sparse or ungauged regions. The approach integrates multiple principal ancillary variables, including soil moisture, antecedent precipitation index, humidity, rainfall, and soil temperature, along with their corresponding weights derived from both linear regression and artificial intelligence models. These weighted predictors are then incorporated into spatial specification techniques, namely dasymetric and pycnophylactic mapping. The proposed performance-informed multivariate framework significantly outperforms traditional methods that rely on a single, ad-hoc ancillary variable. By leveraging highly correlated variables, the framework effectively captures spatial heterogeneity and addresses the Modifiable Areal Unit Problem (MAUP), thereby demonstrating superior downscaling performance. This study highlights the substantial potential of the proposed multivariate framework to improve runoff estimation in ungauged basins and offers a promising tool to support sustainable water resource planning in data-limited environments.</p>