| 04 Dec 2025
Calibrating on Downscaled Satellite Soil Moisture Data Can Improve Watershed Model Performance in Predicting Soil Moisture Variability
Abstract. Watershed streamflow is often the focus of hydrological model calibration and evaluation, despite other potential objectives, including water quality management, flood protection, and agricultural management. When hydrological models are calibrated on streamflow, intermediate processes such as those affecting soil moisture are not necessarily well represented. This research evaluated the performance of downscaled and bias corrected soil moisture calibrated models against streamflow calibrated models both under single and multi-objective scenarios on field scale soil moisture estimation performance. Downscaled satellite soil moisture data and streamflow data are used to calibrate a Soil and Water Assessment Tool – Variable Source Area model initialized using topographic index classes to create hydrologic response units. In-situ soil moisture measurements at 25 locations across a 4-ha mixed-grass pasture located in southwestern Virginia were used to estimate field scale average soil moisture variability for model evaluation. Leveraging downscaled satellite soil moisture data substantially improved estimation of temporal soil moisture variability without affecting the model performance in estimating streamflow. The multi-objective calibration using streamflow and satellite soil moisture improved overall model performance both in estimating streamflow and soil moisture. A three-class topographic index hydrologic response unit definition allowed for adequate representation of saturation excess runoff process. Downscaling enabled calibration in a small 14 km2 watershed using coarse satellite soil moisture data.
This manuscript examines the use of streamflow and downscaled and bias-corrected satellite soil moisture data for calibrating a SWAT-VSA model in a small, saturation-excess dominated watershed. The topic is relevant and timely; however, some issues need to be addressed before the manuscript can be considered for publication.
1. The Introduction does not sufficiently clarify the specific novelty of this study relative to existing literature. If the main contribution lies in calibrating hydrological models using downscaled and bias-corrected satellite soil moisture, the authors should more clearly review and position their work within existing studies on satellite soil moisture downscaling, bias correction, and reported performance improvements. Similarly, if the key contribution is the application of SWAT-VSA in saturation-excess dominated watersheds, the importance and added value of this choice should be more explicitly articulated. At present, the objectives and innovation of the study are not clearly distinguished from prior work.
2. The study is conducted in a single small watershed (~14 km²) dominated by saturation-excess runoff and relies on the SWAT-VSA model. While this choice is physically appropriate for the study site, it substantially limits the generality of the conclusions. The manuscript should more explicitly discuss the limits of applicability of the proposed approach, and the conclusions should be clearly framed as primarily applicable to small, saturation-excess runoff dominated catchments.
3. The study relies on a pretrained mlhrsm downscaling model and an empirical, precipitation-threshold-based bias correction scheme. However, the regional applicability of the pretrained downscaling model is not evaluated, and no comparison is shown between original, downscaled, and bias-corrected soil moisture. The authors should explicitly show these differences and discuss how the downscaling and bias correction steps influence the soil moisture signal used for calibration and, consequently, the model results.
4. A key concern relates to the representativeness of the soil moisture observations. The 25 soil moisture sensors are located within a 4.2 ha pasture, whereas the modeled watershed covers approximately 14 km². It is unclear whether such a limited area is sufficient for evaluating model performance at the watershed scale.
5. An additional concern is that the evaluation of soil moisture performance throughout the manuscript is based primarily on soil moisture data that have undergone both downscaling and bias correction by the authors, without independent validation. While this preprocessing may be necessary, it raises the question of how the reliability of the calibrated model can be independently assessed.
The manuscript would benefit from a clearer explanation of how the authors ensure that the improved soil moisture performance reflects genuine model skill rather than agreement with a processed target dataset. In particular, the role of independent in-situ soil moisture observations in validating the calibration results should be more explicitly discussed, and the potential circularity introduced by calibrating and evaluating against processed soil moisture data should be acknowledged.
6. In addition, the authors should clarify why downscaling to 500 m resolution is necessary if the soil moisture data are ultimately averaged to the watershed scale for calibration. It should be discussed whether this averaging undermines the advantages of using a spatially explicit, VSA-based distributed model.
7. The Discussion section could be streamlined by reducing repetitive literature comparisons and would benefit from a more critical assessment of which aspects of soil moisture and streamflow dynamics remain poorly captured. The authors should discuss whether these limitations reflect structural constraints of the SWAT model (e.g., simplified representation of unsaturated flow processes), and whether the observed performance gains justify the additional complexity introduced by downscaling, bias correction, and multi-objective calibration. A clearer distinction between statistical improvement and process-level improvement would strengthen the interpretation of the results.
Minor comments
1. Figures in the Appendix (e.g., Fig. A1 and Fig. A2) are essential for understanding the workflow and study area and should be moved to the main text.
2. The description in Section 2.2 and Fig. A2 may give the misleading impression that the study area is the full 57 km² Stroubles Creek watershed rather than the upper ~14.5 km² watershed.
3. The number of calibration iterations is reported in Section 2.5.4 (Parameter Uncertainty) but should be clearly stated in the Calibration Strategy section.
4. The sensitivity analysis period (2015–2019) differs from the calibration and evaluation periods without sufficient justification. In addition, results in Fig. A3 are shown only for 2018, which requires explanation.
5. Figure 1 and Figure 4 do not clearly convey differences among the three calibration strategies and could be improved for clarity.