Accurate reconstruction of high-resolution hydrological forcing fields from sparse station observations remains challenging, particularly in poorly instrumented regions. In addition, most existing interpolation methods ignore temporal consistency, potentially introducing artificial jitter that propagates as significant errors when fed into hydrological models. To address this, we introduce Temporally-constrained Spatial Pattern Regression (T-SPR), an extension of the Spatial Pattern Regression (SPR) framework that combines anomaly-based interpolation with temporal regularization of the reconstructed forcing fields. An intermediate anomaly-based variant (A-SPR) is also considered to isolate the contribution of the climatological decomposition. These three variants (SPR, A-SPR and T-SPR) are evaluated using high-resolution regional climate model simulations and real station observations for daily precipitation, minimum temperature, and maximum temperature over southern Québec, Canada. Results show that anomaly-based interpolation substantially improves reconstruction accuracy under sparse observational conditions, particularly for temperature forcings. Temporal regularization further enhances robustness and temporal consistency by reducing spurious day-to-day variability and sensitivity to station configuration. The greatest benefits are observed at extremely low station densities, where information is most limited. These findings demonstrate that T-SPR, which combines climatological information and temporal coherence, provides an effective and computationally efficient approach for applications to hydrological forcing reconstruction in data-sparse catchments.