Abstract: The data linkage relationship between mine pressure and water yield in longwall face is one of the basic contents to realize the dynamic decision-making of mining technical parameters and the intelligent early-warning of mine disasters on the basis of multi-source information fusion. Based on the field data of support resistance and water yield in longwall face, a time series GARCH model was established, and the linkage relationship and data structure characteristics between mine pressure and water inflow were mined from both static and dynamic points of view. A calculation method of lag linkage time between mine pressure and water yield based on the Almon distributed model was developed. The results revealed that the relation between peak water yield and support resistance is highly time-dependent during longwall mining operation. The variation rates of support resistance and water yield in working face presented a non-normal distribution, volatility clustering and self-correlation. An AR-MA-GARCH-t-Copula edge distribution combined model on the field data of support resistance and water yield was established from a static point of view, based on the ADF unit root test for time series stationarity, the Granger causality test, and autocorrelation test. The Clayton-Copula function was employed to describe the asymmetric data structure and data linkage, which can capture that the data linkage is stronger when the support resistance presents a downward trend or violent fluctuation, that is, the peak water yield occurred along with the roof weighting. A DCC-GARCH model, which was established from a dynamic point of view, shows that the dynamic coefficients of the data of support resistance and water yield fluctuates between -0.264 9 and 0.264 9. There is a positive linkage and a long-term cointegration relationship between the fluctuations of support resistance and water yield. A method for calculating the time lag between support resistance and water yield was developed based on the parameter determination of the Almon distributed model through PDL estimation model. The results can provide a basic data analysis model for the dynamic prediction and intelligent early warning of water yield in longwall mining.