Reliable, low-cost and accurate monitoring of soil respiration is an important challenge that must be solved to fully understand the contribution of soil dynamics to climate change; however, accuracy obtained by single-chamber is insufficient. This project proposes a multi-chamber fusion method for integrating multi-source information measured using low-cost sensors. The proposed algorithm initially uses Fick’s first diffusion law to calculate the soil carbon flux values for five chambers, followed by multi-layer decomposition of a wavelet packet transform (WPT) to eliminate high-frequency noise. Then, the basic probability assignment (BPA) of each sensor is calculated via the Biggest-smallest Approach Degree and used to assign the Dempster-Shafer (D-S) fusion subjected BPA to determine the distribution weight of each gas chamber. Finally, the decision layer fusion is defined as the product of the chamber weights and feature signals obtained by wavelet multi-layer decomposition. The performance of the proposed algorithm was evaluated against existing algorithms using real data collected using a low-cost prototype device in an evergreen broad-leaved forest environment and compared to the data generated by an expensive commercial device. The proposed algorithm significantly improved the accuracy of soil respiration monitoring for the low-cost prototype device.