This project aims to investigate the impact of complex gas-water interaction occurring in coal medium on coal seam gas (CSG) production through integrated experimental and numerical simulation methods, and develop improved mathematical models to better represent the underlying physics of the process. The main contributions drawn are listed below:First, a suite of gas adsorption and desorption experiments for three different types of coal samples were conducted in a pressure chamber under unconfined conditions. The experiments were carried out under both dry and wet (i.e., water-immersed) conditions to quantify the impact of capillary trapping on gas desorption and its implications to the ultimate CSG recovery. Coal transient directional deformation was measured using water-proof strain gauges, and the dynamic formation of gas bubbles from coal surfaces was observed through one of the four transparent sapphire windows on the pressure chamber and was recorded using a high-resolution camera. The gas adsorption and desorption data was collected and analysed to quantify the influence of the capillary trapping. The results show that the existence of water in coal can reduce the gas recovery rate evidently, demonstrating that the capillary pressure in small coal pores hinders the gas flow capability and thus causes the gas to be trapped. The percentage of the trapped gas is directly affected by the average pore size and coal wettability. The strain values for the water-immersion conditions is smaller compared to the dry condition, showing the capillary induced trapping effect also influences the matrix shrinkage amount. The gas bubbles that emerged during desorption captured by the camera present that more and larger bubbles are formed on coal matrix surface at lower pressures, indicating that the gas slug is more easily generated at lower reservoir pressures, which could hinder the gas-water two-phase flow in cleats and is detrimental to gas production. The results of the gas recovery rate, sorption-induced strain, bubbles generated on matrix surface, and the existence of capillary trapping offer a new set of laboratory-based information for better understanding what may occur in CSG reservoirs during gas depletion. The results can assist in evaluating the impact of water on gas desorption and can be used to improve the reliability of existing reservoir simulation models.Second, the numerical study for the capillary trapping effect during the gas transfer from matrices to cleats was conducted from pore scale to field scale. For the pore-scale simulation, the conceptual model describing the formation mechanism of capillary trapping effect in coal matrix was first established, followed by the implementation of the two-phase flow numerical simulation. Various parametric sensitivity studies on multiple factors were then conducted to quantify the amount of gas trapped under different reservoir conditions and production controls. Our results show that pore size, wettability, Langmuir volume and pressure parameters are critical for the extent of the trapping effect. For the field scale study, the change rule of contact angle with pressure was first gained by laboratory measurements, which was then fed into a fully coupled field scale numerical model, together with a classic relative permeability model to describe gas-water two-phase flow in the cleats system. A series of sensitivity studies were conducted to assess the factors affecting gas production performance. The results show that capillary trapping can evidently hinder the gas production curve, with both the production rate and the cumulative production being reduced, and caused the peak gas production rate to appear later. The decrease of the mean pore size leads to earlier attenuation of production rate and reduces the recovery rate. Pore size distribution (PSD) determines the percentage of pores in which gas breakthrough is able to happen, thus affecting the gas transfer efficiency. Higher Langmuir pressure constant delivers a lower and later peak gas rate, and the trapped gas percentage is also larger. While a higher Langmuir volume constant, a critical parameter in determining in-situ gas content, leads to higher but delayed peak value. A noticeable impact of contact angle on gas production has also been observed. The developed framework and findings from this work are expected to help better understand the capillary trapping effect on unconventional gas reservoirs and propose a way to quantify capillary trapping effects.Third, an improved capillary pressure model incorporating the cleat aperture change was developed and validated. The model was then extended to the reservoir scale to understand the impact of cleat aperture change on capillary pressure and CSG production performance. The results from the sensitivity study show that cleat aperture change can alter the capillary pressure and gas production rate considerably. Larger Young’s modulus and Langmuir volumetric strain constant can lead to smaller capillary pressure and higher gas rate as they can raise the cleat aperture. Higher Poisson’s ratio delivers larger capillary pressure and smaller gas rate as it gives less cleat aperture increase. Influence of Langmuir pressure constant on capillary pressure and production rate transit from positive to negative along the production process due to its complex impact on the change of cleat aperture. In this work, we also introduced two parameters: ω, defined as cleat geometry factor, and nwr, defined as residual saturation fitting factor. Results reveal that the cleat geometry factor can reduce capillary pressure and raise the production rate, while a larger residual saturation fitting factor gives smaller capillary pressure and production rate. In terms of relative importance, the comparison results of capillary pressure and relative production difference curves suggest that Young’s modulus and Langmuir volumetric strain constant are most important, followed by the cleat geometry factor. The importance of Poisson’s ratio is higher than Langmuir pressure constant at first but is surpassed with the depressurization of reservoirs. The residual saturation fitting factor is the least important. The improved capillary pressure model is more physics-driven, thus better capturing the underlying mechanisms of capillary pressure change during CSG production and is expected to deliver a more reliable estimation of gas production.