Comparative study of deep learning models for daily karst spring discharge forecasting: LSTM Versus Hybrid VMD–LSTM
The discharge process of karst spring is controlled by complex hydrogeological structure and multi-scale dynamic mechanisms,exhibits characteristics of non-stationary and strongly nonlinearity, posing significant challenges for high-precision predictive modelings. Conventional physics-based models rely heavily on extensive hydrogeological parameters and often struggle to accurately capture non-stationary processes during extreme events. Consequently,integrating signal decomposition techniques with deep learning has emerged as a promising approach to enhance both prediction accuracy and physical interpretability. This study develops a hybrid model that combines Variational Mode Decomposition (VMD) with a Long-Short-Term Memory (LSTM) networks, referred to as VMD-LSTM, aiming to address non-stationarity and multi-scale coupling issues in daily karst spring discharge forecasting. Through a system comparison of the performance of the VMD-LSTM model and standard LSTM model during training, validation, and testing phases, the study evaluates improvements in prediction accuracy, extreme events characterization, and model stability, while elucidating the mechanisms by which VMD enhances the modeling performance of LSTM. This study utilized daily spring discharge and corresponding precipitation data from the Zhaidi karst system in Guilin, Guangxi, spanning from 2013 to 2023. The original spring discharge series was decomposed using VMD into ten Intrinsic Mode Functions (IMFs). Among them,six modes(Mode 5 to Mode10), which collectively represent the dominant karst hydrodynamic processes, were selected as inputs to a two-layer LSTM network for modeling. The model was trained using the Adam optimizer and the mean squared error loss function. The dataset was chronologically partitioned into a training set (80%), a validation set (10%), and a testing set (10%). Model performance was comprehensively evaluated using metrics including RMSE, MAE, NSE, KGE, and peak RMSE (caculated for the top 5% of high-flow events). Results demonstrated that the VMD-LSTM model consistently outperformed the standard LSTM across all three phases of the training,validation and testing. During the testing phase, the VMD-LSTM achieved an NSE of 0.951 and an RMSE of 0.524, representing improvements of 57.0% and 64.6%, respectively, compared to the LSTM model. Notably, the hybrid model exhibited substantially enhanced capability in predicting extreme discharges, reducing the peak RMSE from 6.208 to 1.542(a decrease of 75.2%). The VMD decomposition effectively mitigated non-stationarity and mode mixing present in the original series, enabling the LSTM network to more reliably identify and simulate the "rapid response–slow recession" characteristcs inherent in karst hydrological processes. This approach significantly syppressed the systematic underestimation and error dispersion during high-flow events. The VMD-LSTM hybrid model not only markedly enhanced the forecasting accuracy and extreme event modeling for karst spring discharge but also exhibited strong cross-phase consistency and physical interpretability, demonstrating high generalizability and robustness in practical forecasting scenarios. Future research could focus on incorporating hydrological physical constraints, conducting uncertainty quantification analysis, and extending the framework to higher spatiotemporal resolutions and multivariable coupling to further enhance model applicability and forcasting reliability across diverse karst systems.
Evaluation results, issues, and prospects of investigations into groundwater, mineral water, and geothermal (water) resources in Yunnan Province
At present, China is deploying the basic survey of water resources. Provinces have initiated surveys of surface reserves and investigations into the quantity and quality of groundwater resources based on their specific circumstances. The goal is to accurately assess the surface water reserves as well as the quantity and quality of groundwater resources within each province. Over the years, Yunnan Province has carried out small-scale evaluation work, including regional hydrogeological surveys at scales of 1:200,000 and 1:250,000, county-level hydrogeological surveys at 1:100,000, and hydrogeological environment geological surveys at key river basins at 1:50,000. Since the 1980s, specialized groundwater dynamic monitoring has been initiated, and the construction of the national groundwater monitoring project has been launched in 2017. Based on extensive investigations, explorations, and thematic research findings, this paper comprehensively discusses and evaluates the current status of exploitation and utilization of groundwater, mineral water, and geothermal water in Yunnan Province through summary, statistical analysis, and comprehensive evaluation. The evaluation results show that Yunnan's distinctive plateau landforms, stratigraphic structures, neotectonic movements, and hydrogeological conditions have contributed to abundant groundwater, mineral water, and geothermal resources. On the basis of regional hydrogeological investigations and evaluations, the natural groundwater resources in Yunnan were estimated using the multi-year average underground runoff modulus method, yielding a total of 75.244 billion m3·a−1. The recoverable resources amount to 19.035 billion m3·a−1. Mineral water resources are generally underexplored, making accurate resource quantification challenging. The evaluation uses summaries of spring flow during the dry season and calculations of water inflow from drilling and pumping tests. The total mineral water resources is estimated at 154.37×104 m3·d−1, with a recoverable volume of 123.50×104 m3·d−1. Yunnan mainly utilizes natural hot springs and geothermal fluids extracted by artificial drilling. The province is rich in geothermal resources, with numerous hot springs distributed across almost all counties and cities. There are 851 naturally exposed hot springs with temperatures of 25°C or higher. Calculation results of geothermal resources through the heat storage method indicate that the heat flow is 9,216.3 L·s−1, and the natural heat release is 1,056,443.27 kJ·s−1. The reserves of geothermal resources amount to 197.77×1015 kJ. This is equivalent to 6,748.50×106 t of standard coal. The mining heat is 435,116.5×108 kJ·a−1, with an additional 120,377×108 kJ·a−1 extracted. The development and utilization rate is about 27.66%. According to the monitoring results of groundwater, the overall water quality is high. Compared to other regions across Yunnan Province, the proportion of water classified as Class I–III is slightly higher. The quality of geothermal water is superior to that of pore water, which in turn is better than that of bedrock water. Groundwater quality in the red bed area of bedrock water is more complex and variable. The water quality of the weathered fissure water in the red bed area is generally higher in the mountainous and hilly areas. In low-lying areas such as the intermountain basins and valley bottoms, water quality is highly variable and often poor due to the slow water circulation. For example, the third member of the Jiangdihe Formation (K2j3) of the Upper Cretaceous in Middle Yunnan contains gypsum and rock salt, and well-developed dissolution fissures facilitate groundwater accumulation. When recharge conditions are favorable, groundwater circulation is rapid, and drainage is efficient, resulting in generally high water quality. However, in areas with slow groundwater runoff, chloride and sulfate concentrations often exceed the permitted levels significantly, leading to poor water quality. Mineral water is classified into five categories: metasilicic mineral water, carbonated mineral water, trace element mineral water, salt mineral water, and compound mineral water. There are 24 hydrochemical types of geothermal fluids, including HCO3-Ca, HCO3-Na, HCO3-Ca·Na, HCO3-Ca·Mg, HCO3·SO4-Ca, HCO3·SO4-Na, and HCO3·SO4- Ca·Na, etc. This paper also summarizes and analyzes the existing issues and deficiencies in current investigations and evaluations of water resources, such as low overall investigation accuracy, insufficient systematic consideration of water system delineation, limited parameters for groundwater resources, and a lack of integrated systems for the investigation, evaluation, development, utilization, and management of geothermal resources. For the future investigations of groundwater, mineral water, and geothermal water, it is suggested that a comprehensive application of field investigations, drilling, geophysical exploration, tracing, and other technical methods, combined with advanced techniques such as hydrodynamics, hydrochemistry, environmental isotope analysis, and modeling, should be used to scientifically delineate the boundaries of water systems in karst areas. Pilot work and scientific research should be conducted in small river basins to address critical parameters in groundwater evaluation. Research should focus on water balance in small watersheds and the processes and mechanisms of water transformation among precipitation, surface water, and ground water. The groundwater monitoring network in Yunnan Province should be improved by leveraging the National Groundwater Monitoring Project (Phase II) to complete monitoring of major basins. Monitoring stations should be strategically established in areas with significant geological issues in karst environments, ecologically sensitive and fragile zones, nature reserves, important wetlands, geothermal water-rich areas, and mineral water sources, thereby strengthening multi-dimensional systematic monitoring. Establishing a comprehensive water resources monitoring system will enhance the accuracy of groundwater resource evaluations and support the rational development and sustainable utilization of groundwater resources.
Articles in press have been peer-reviewed and accepted, which are not yet assigned to volumes /issues, but are citable by Digital Object Identifier (DOI).
2025, 44(6): 1158-1172.
doi: 10.11932/karst2025y021
2025, 44(6): 1173-1185, 1197.
doi: 10.11932/karst2025y009
Founded in 1982 bimonthly
ISSN: 1001-4810
CN 45-1157/P
Editor-in-chief: Jiang Zhongcheng
Sponsor: Chinese Academy of Geological Sciences
Sponsored by: Institute of Karst Geology, Chinese Academy of Geological Sciences
Cosponsor: International Research Center on Karst under the Auspices of UNESCOKarst Geology Committee of Geological Society of ChinaSpeleology Committee of Geological Society of China
Wechat
