Modelling karst spring flow in Southwest China based on machine learning
Karst springs are important for ecosystem and economic development in Southwest China.Controlled by the unique karst hydrogeological structure and multiple water flow processes,the karst spring flow has complex dynamic characteristics,thus posting a great challenge to simulate and predict the dynamic process of karst spring flow which can reflect the characteristics of rainfall-spring flow in the karst basin.As a data-driven model,the machine learning model omits the necessity of considering complex physical processes,showing its significant advantages in the simulation and prediction of nonlinear system variables.Therefore,it provides an effective approach for simulation and prediction of karst spring discharge.However,the influence of the flow processes and hydrogeological conditions on the structure and simulation accuracy of machine learning model is still unclear.Among the machine learning algorithms,LSTM,as the most popular algorithm in recent years,is widely used in the simulation and prediction of various long-time series data.LSTM adds a cell state similar to"conveyor belt"in the hidden layer,and the cell state is adjusted by forgetting gate,input gate and output gate.This structure can effectively solve the long-term transportation and memory problems of time series data,and is more suitable for runoff simulation and prediction than the traditional neural network algorithm.In this study,for the processes of rainfall-spring flow in karst areas,two typical karst springs(hillside karst spring and outlet karst spring)representing different geomorphic units in Southwest China are selected.Through hyper parameter optimization,a double hidden layer and double input LSTM model are adopted to build a machine learning model of typical karst spring flow.The measured meteorological and hydrological data from 0:00 on January 1,2017 to 24:00 on December 31,2019 are used.2017-2018 is the training period and 2019 is the validation period.The model has been trained and verified.Based on the simulation results,the influence of different rainfall-spring flow forming processes on the simulation accuracy of karst spring flow and the influence of karst hydrogeological structure on the response time lag of rainfall-spring flow are compared and analyzed.The results show that the Nash efficiency coefficients(NSE)for the hillside karst spring and the outlet karst spring were 0.942 and 0.951 in the training period,and 0.831 and 0.834 in the validation period,respectively.The model can well simulate the whole dynamic process of karst spring discharge in different geomorphic units,but there are significant errors in the simulation of flood peak in the rainy season.The formation process and variation of rainfall-spring discharge in the karst spring area have an important influence on the simulation accuracy of machine learning model.Compared with the model trained by the annual measured sequence,the model trained by the measured sequence in the rainy season can significantly improve the simulation accuracy of the karst spring flow in this season.The NSE of the hillside karst spring increases from 0.793 to 0.956,and the outlet karst spring increases from 0.798 to 0.962.The difference is most significant in May,June and July when rainstorms and flood are concentrated.Under the same precision of simulating spring flow,the time step of LSTM model of hillside karst spring is obviously smaller than that of outlet karst spring.When the simulation precision of the two types of spring flow is the highest,the time step of the model is 15 h and 28 h,respectively.This result can be combined with the actual geological structure.In karst areas,the development degree of the epikarst zone is closely related to the terrain,and its thickness usually decreases with the increase of the slope.Generally,the thickness of the epikarst zone on steep hillsides is less than that in flat depressions,and gradually increases from top to bottom along the hillsides.Compared with the water flow in the depression,the water flow in the aquifer of the hillside unit is rapid because it is controlled by the shallow epikarst zone,and it is easier to form the fast water flow with the large fissure as the channel,which may further show the rapid flow and rainfall response characteristics of the hillside karst spring.Therefore,affected by the development characteristics of the epikarst zone,the response lag time of rainfall-spring flow in the hillside karst spring area is less than that in the depression,and the time step of LSTM model of hillside karst spring flow is significantly smaller than that of the karst spring at the outlet of the basin.
machine learningLSTMkarst spring flowflow hysteresis to rainfallrainfall-spring flow process in the karst area
万方数据
维普
