Abstract: In response to the technical challenge of accurately reconstructing the distribution field of mine gas content in scenarios with limited computing resources, this paper proposes a coupled algorithm of universal kriging interpolation (UK) and empirical variation function (EVF) (UK-EVF), aiming to achieve high-precision spatial distribution representation of gas content through lightweight computing, and provide technical support for coal mine gas control. The specific research process is as follows: Firstly, gas content data from different positions of the working face are collected through on-site measurements, and a measured dataset covering spatial coordinates and corresponding gas content is constructed; Secondly, the EVF algorithm is used to quantitatively analyze the spatial structural characteristics of the gas content distribution field, accurately extracting key spatial correlation parameters such as range, nugget effect, and baseline value, laying a structural foundation for subsequent interpolation prediction; Under this premise, the UK algorithm is introduced, with the spatial parameters obtained by EVF as constraints, combined with the measured gas content of the collected sample points, to perform spatial interpolation prediction on the gas content of the unsampled area, ultimately achieving a complete reconstruction of the gas content distribution field of the entire working face. To verify the performance of the algorithm, this paper selects four classic spatial interpolation algorithms and five mainstream machine learning and neural network models as comparisons. The core evaluation indicators are mean absolute error (MAE), root mean square error (RMSE), and coefficient of determination (R 2), and a comprehensive analysis is conducted from two aspects: prediction accuracy and computational complexity. The research results indicate that in small sample scenarios, the core advantage of spatial interpolation algorithms is that they "do not require complex parameter training and can directly rely on the inherent laws of spatial data nearest neighbor similarity for modeling", which can ensure basic prediction accuracy even with limited sample sizes; The UK-EVF coupling algorithm proposed in this article performs the best among all compared algorithms, with low computational resource consumption, fast response speed, and distribution field reconstruction running time not exceeding 10 seconds, fully adapting to on-site scenarios with limited computing resources; High precision reconstruction has been achieved, and the average relative error absolute value of gas content prediction results is stably controlled within 5%, meeting the strict requirements for data accuracy on site. This algorithm can effectively provide accurate spatial distribution data support for key gas control processes such as optimizing the layout of coal mine gas extraction boreholes and dynamically predicting gas outburst rates, and has significant on-site application value.