Abstract:The nitratenitrogen (NO-3N) of wet soil samples was detected through sensor fusion method. The prediction model of NO-3N was established by ionselective electrode (ISE) and soil moisture sensor. The interference of water content on fast detection of wet soil was analyzed. The statistical linear transformation between soil mass water content and volumetric water content was determined through calibration experiments. The algebra relationship among soil water content, volume of soil extractant and mass of dry material was derived. Compared with the standard spectrometric detection, the poor prediction accuracy of root mean square error (RMSE) of more than 17.00mg/L was obtained in bare ISEbased NO-3N detections. Meanwhile, the average absolute error (AE) was 3.79~44.81mg/L, which indicated a nonnegligible predicted deviation from real soil NO-3N condition. Through the sensor fusion approach, the NO-3N prediction results indicated significantly good consistency with standard spectrometric detection results. The adjusted determination coefficient (R2) of regression equation was 0.97, RMSE was less than 7.05mg/L, and AE was 0.17~17.95mg/L. The sensor fusion method obviously improved the detection accuracy of nitratenitrogen detection compared with the bare ISE detection. Meanwhile, it demonstrated the advantage of higher efficiency than that of conventional laboratory soil detections which required soil pretreatment. The method and experimental results could provide references for ISE applications to online soil macronutrient fast detections.