EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

From Soil Moisture Observations to Governing Flow Equations: What Data‐Driven Equation Discovery Reveals

Wenxiang Song, Liangsheng Shi, Yujie Wang, Leilei He, Mehdi Rahmati, Harry Vereecken

Water Resources Research · 2026-11 · 62 (11)

DOI: 10.1029/2026wr045026 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Water Resources Research, online erschienen am 06.11.2026. Verfasst von Wenxiang Song und 5 weiteren Autor:innen. Die Arbeit nennt 52 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: National Natural Science Foundation of China.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Traditional derivations of soil moisture flow equations are based on conservation laws and physical principles, but the resulting equations still depend on how individual processes and constitutive relationships are represented. Data‐driven equation discovery offers a complementary way to identify governing terms from observations, yet previous applications to soil moisture flow have relied mainly on numerical experiments. Here, we develop Spline‐SR, a spline‐regularized sparse regression framework for discovering nonlinear soil moisture flow equations from sparse fixed‐depth observations. Spline‐SR reconstructs continuous water‐content dynamics, evaluates candidate derivatives through automatic differentiation, and selects governing terms with state‐dependent spline coefficients. Numerical experiments show that the method reliably recovers the known three‐term structure of the water‐content form of the Richards equation under sparse and noisy sampling. We then apply Spline‐SR to measurements from sand and loamy‐sand columns under wetting, drying, and alternating conditions. For both soils, the experimental analysis identifies a parsimonious structure containing the three terms expected from the water‐content form of the Richards equation. The inferred coefficient functions vary across forcing conditions, while the resulting equations reconstruct the main features of soil moisture dynamics. These results show that sparse experimental observations can contain sufficient information to identify the governing structure of soil moisture flow equations, while coefficient relationships remain more sensitive to observational coverage and flow conditions.

Förderung

  • National Natural Science Foundation of China

Zitieren

APA

Song, W., Shi, L., Wang, Y., He, L., Rahmati, M., & Vereecken, H. (2026). From Soil Moisture Observations to Governing Flow Equations: What Data‐Driven Equation Discovery Reveals. Water Resources Research, 62(11). https://doi.org/10.1029/2026wr045026

Vancouver

Song W, Shi L, Wang Y, He L, Rahmati M, Vereecken H. From Soil Moisture Observations to Governing Flow Equations: What Data‐Driven Equation Discovery Reveals. Water Resources Research. 2026;62(11). doi:10.1029/2026wr045026

BibTeX
@article{Song2026,
  author = {Song, Wenxiang and Shi, Liangsheng and Wang, Yujie and He, Leilei and Rahmati, Mehdi and Vereecken, Harry},
  title = {From Soil Moisture Observations to Governing Flow Equations: What Data‐Driven Equation Discovery Reveals},
  journal = {Water Resources Research},
  year = {2026},
  volume = {62},
  number = {11},
  doi = {10.1029/2026wr045026}
}
Kostenlos starten →