PulseExploreJournal ClubDebatesTrendingResearchersJournals
Instagram
HomeExploreJournal ClubTrending
Synapse
⌘+K
Synapse
May 31, 2026Hydrological Processes0 citationsOpen Access

Influence of Soil and Land‐Use/Land‐Cover Data Quality on SWAT Parameter Transferability in Small Subtropical Agricultural Watersheds

View Full Paper
ERElzon C. RippelDPDaniel G. A. PiccilliLKLatif Kalin

Key Points

  • This research aims to assess how the quality of soil and land-use/land-cover data affects the transferability of SWAT model parameters in agricultural watersheds.
  • Evaluated soil and land-use data quality using a factorial proxy-basin experiment in two subtropical watersheds in southern Brazil.
  • Compared scenarios with different soil data quality and land-use representation to assess calibration performance and parameter transferability.
  • Analyzed model performance metrics such as NSE and PBIAS under different data quality scenarios.
  • Calibration performance improved significantly in WS80 from NSE = 0.16 to NSE = 0.75 by using dynamic land-use data.
  • Under fine-soil and dynamic land-use conditions, the combined parameter transfer scenario achieved satisfactory NSE values of 0.56 and 0.55.
  • PBIAS improved from a range of -67.2% to +48.8% under coarse soil scenarios to -39.9% and +28.7% with fine soil and dynamic land-use.

Abstract

ABSTRACT Regionalisation of hydrological model parameters is a cornerstone of Prediction in Ungauged Basins (PUB), yet the level of input‐data detail required for reliable parameter transfer remains unclear, particularly in intensively managed agricultural landscapes with intra‐annual land‐use/land‐cover (LULC) dynamics associated with multi‐cropping systems. While recent studies have primarily focused on refining transfer techniques—such as proximity, similarity, regression, fuzzy clustering, or machine learning—the role of watershed descriptor quality has rarely been explored. Here, we evaluated how soil and LULC data quality control the transferability of SWAT parameters through a factorial proxy‐basin experiment in two small subtropical agricultural watersheds in southern Brazil (WS80: 0.80 km 2 , WS140: 1.39 km 2 ). Four scenarios crossed two levels of soil data quality (global HWSD at 1:5000000 vs. locally sampled at approximately 1:50000) with two levels of LULC representation (static default management vs. dynamic, EVI2‐derived crop–livestock rotations). Calibration performance improved in WS80 from NSE = 0.16 under coarse soil and static LULC to NSE = 0.75 when dynamic LULC was included. Under direct parameter transfer, the combined fine‐soil and dynamic‐LULC scenario achieved satisfactory NSE in both target directions (0.56 and 0.55), compared with NSE = 0.06–0.25 under the baseline global‐soil/static‐LULC scenario. Post‐transfer PBIAS ranged from −67.2% to +48.8% across coarse‐soil scenarios, whereas the fine‐soil/dynamic‐LULC scenario reduced PBIAS to −39.9% and + 28.7%. Our findings show that, in small managed agricultural watersheds, improving the physical realism of watershed descriptors can substantially increase SWAT parameter transferability, and that both soil and dynamic LULC upgrades are needed to achieve satisfactory model performance with transferred parameters.

Ask AI
Helpful
Bookmark
Share
View Full Paper

Cite This Study

Rippel et al. (2026) studied this question.

synapsesocial.com/papers/6a1bd1f65783ba022b6fd6cdhttps://doi.org/10.1002/hyp.70570
Ask AI
Helpful
Bookmark
Share
View Full Paper