Evolution, Attribution, and Multi-Scale Lag Prediction of Liulin Spring Discharge
Water · 24.10.2026 · 18 (21) · 2659
DOI: 10.3390/w18212659 ↗Kurz gefasst
Fachartikel in der Zeitschrift Water, online erschienen am 24.10.2026. Verfasst von Zhenxing Jia und 5 weiteren Autor:innen. Die Arbeit nennt 32 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: the Fundamental Research Program of Shanxi Province.
Abstract (Originalsprache)
Karst springs are vital water sources in arid and semi-arid regions, with discharge governed by natural recharge conditions and increasingly influenced by human activities. This study investigates Liulin Spring on the eastern margin of the Ordos Basin using hydro-meteorological data (1956–2024), groundwater extraction records (1974–2024), and a restored discharge series that removes the direct effect of abstraction. Evolutionary patterns, drivers, and lagged responses were analysed using the trend-free pre-whitening Mann–Kendall test, Pettitt and moving t-tests, double-mass curve analysis, and detrended and prewhitened cross-correlation, and a parsimonious multiple linear regression (MLR) model with a persistence benchmark and independent temporal validation was constructed for prediction. The results reveal a significant seven-decade decline in four stages—stability, sharp decline, gradual attenuation, and rebound—with a historic minimum of 1.17 m3/s in 2021. Double-mass analysis showed the precipitation–response slope falling to 18.2% of the natural baseline (1956–1973), indicating that the precipitation–discharge relationship has been severely weakened under sustained anthropogenic disturbance; the contrast between the observed and restored double-mass curves identifies groundwater abstraction as the principal stressor, while the individual contributions of coal mine dewatering, land-use change, and other factors cannot be separated by this analysis alone. The restored discharge is strongly persistent (lag-1 ACF = 0.9392); after linear detrending and AR prewhitening, the robust lagged response to precipitation and river runoff concentrates at approximately 2 years, whereas the previously reported multi-decadal correlation peaks are artefacts of the common trend and autocorrelation. The MLR model (training R2 = 0.925) is outperformed by a persistence baseline for one-step-ahead prediction but provides skilful multi-step scenario forecasts; hindcast errors for 2023–2024 are below 0.24 m3/s. Scenario forecasts for 2025–2027 under dry, normal, and wet precipitation conditions (1.33–1.63 m3/s) suggest a stable-to-weak stage rather than a sustained recovery. This study provides a methodological reference for attribution and lag analysis of analogous karst springs, for which the model parameters require site-specific recalibration.
Förderung
- the Fundamental Research Program of Shanxi Province
Zitieren
APA
Jia, Z., Zang, H., Zhao, X., Liu, Y., Wang, Z., & Li, D. (2026). Evolution, Attribution, and Multi-Scale Lag Prediction of Liulin Spring Discharge. Water, 18(21), 2659. https://doi.org/10.3390/w18212659
Vancouver
Jia Z, Zang H, Zhao X, Liu Y, Wang Z, Li D. Evolution, Attribution, and Multi-Scale Lag Prediction of Liulin Spring Discharge. Water. 2026;18(21):2659. doi:10.3390/w18212659
BibTeX
@article{Jia2026,
author = {Jia, Zhenxing and Zang, Hongfei and Zhao, Xuetong and Liu, Yuxiang and Wang, Zhenxing and Li, Dong},
title = {Evolution, Attribution, and Multi-Scale Lag Prediction of Liulin Spring Discharge},
journal = {Water},
year = {2026},
volume = {18},
number = {21},
pages = {2659},
doi = {10.3390/w18212659}
}
EUVIMED