Omitting Frozen‐Soil Effects in Hydrological Models Introduces Biases in Estimated Runoff Sensitivity to Climate Change in Cold Basins

Abstract Proper representation of cryospheric processes is essential for accurate hydrological simulation in cold mountainous catchments. Frozen soil, a key cryospheric element, is sometimes overlooked in hydrological models, and its hydrological influence remains insufficiently explored. This study compares two versions of the Tsinghua Representative Elementary Watershed‐Tracer‐aided model (THREW‐T): THREW‐FS (with a frozen soil module) and THREW‐NoFS (without a frozen soil module), to evaluate runoff‐projection biases caused by omitting frozen‐soil effects. Results showed: (a) Both models accurately simulated streamflow, but they differed in isotope behavior and subsurface runoff partitioning. THREW‐FS estimated a much higher annual subsurface runoff contribution than THREW‐NoFS (68.5% vs. 41.3%), providing the process basis for their different runoff sensitivities. (b) Both models estimated decreasing runoff with rising temperatures and increasing runoff with higher precipitation, but THREW‐NoFS model showed a higher changing rate. On average, the runoff change estimated by THREW‐NoFS was 2.26 times that estimated by THREW‐FS in temperature perturbation scenarios and 1.37 times in precipitation perturbation scenarios. (c) Divergences in runoff sensitivity arose from differences in surface‐subsurface runoff partitioning and dynamic soil property. THREW‐NoFS produced a better water supply condition for evaporation, making it more sensitive to temperature increases than to precipitation increases. These findings suggest that while a model without a frozen soil module can perform well in streamflow simulation, it may overestimate runoff sensitivity to climate perturbations. This study also highlights the value of tracer‐aided approach in constraining uncertainties in runoff component partitioning and producing credible runoff projections in cold mountainous basins.