In order to reveal the general characteristics of electromagnetic wave signal transmission in farmland soil and scientifically deploy sensor nodes, taking the farmland soil in Guanzhong region as the research object, adopted the modular design idea, integrated sensor, wireless data transmission, processor and energy supply module into one. Wireless underground sensor networks (WUSN) nodes and sink nodes were designed. The influences of soil moisture content, buried depth of WUSN nodes and horizontal distance between nodes on signal transmission of WUSN nodes were analyzed by single factor test method, and the prediction model of received signal strength and bit error rate was established. The experimental results showed that when the WUSN node signal was transmitted vertically underground, the soil moisture content was increased by 2.5 percentage points, the received signal strength was decreased by 4~6dBm, and the bit error rate was increased by 3~5 percentage points. The buried depth of WUSN node was increased by 5cm, the received signal strength was decreased by 3~5dBm, and the bit error rate was increased by 3~4.5 percentage points. When the WUSN node signal was transmitted in the horizontal direction underground, the soil moisture content was increased by 2.5 percentage points, the received signal strength was decreased by 5~7dBm, and the bit error rate was increased by 4~5 percentage points. When the horizontal distance between nodes was within the range of 10~90cm, the received signal strength was decreased by 6~8dBm when the horizontal distance between nodes was increased by 10cm, and the bit error rate was increased by 6.5~8 percentage points. When the horizontal distance between nodes was within the range of 90~190cm, the received signal strength was decreased by about 1dBm when the horizontal distance between nodes was increased by 10cm. The bit error rate would be increased by 1~1.5 percentage points. The best fit R2 and root mean square error of WUSN node bit error rate and signal receiving strength prediction model in two transmission directions were 0.982 and 1.7%, respectively, and the best fit R2 and root mean square error were 0.942 and 5.136dBm, respectively. In conclusion, the signal transmission of WUSN nodes in soil was seriously affected by soil moisture content, buried depth of WUSN nodes and horizontal distance between nodes. The received signal strength and bit error rate can be accurately calculated by using this prediction model, and then the maximum buried depth and maximum horizontal distance of WUSN nodes can be obtained. At the same time, the general conclusion of WUSN signal transmission in farmland soil was given.