Abstract: Observing the dynamic changes of root growth under soil conditions is challenging. In this study, a new type of rhizobox for non-invasively observing root growth under soil conditions is presented. Variations in tobacco seedling root growth were studied as an example of its application. The apparatus consisted of a growth chamber, a nutrient solution-supplying system and an image capture-analysis system. Three subchambers with outer dimensions of 60×30×3 cm were assembled in the shape of a "Y" as the growth chamber. A tobacco seedling was transplanted to the central space of the growth chamber filled with sandy soil, and its root could extend to the soil of three subchambers. Therefore, the growth chamber could both induce two-dimensional root development and facilitate root observation. The soil water content and nutrient concentration in the growth chamber were controlled by supplying a nutrient solution in a designed concentration and volume through a nutrient solution-supplying system, independently. The nutrient solution-supplying system consisted of a solution storage bottle, a pipe, a flux controller and a dripper buried in the soil. The dripper was a J-shape pipe with a funnel to prevent it from being blocked by soil particles. This kind of dripper can make a nutrient solution spread uniformly and avoid clay particle eluviation and illuviation in the soil of the growth chamber. A camera was used to capture images of tobacco roots through a transparent pane during the course of the experiment and root parameters such as root number, root length, root width and root depth were analyzed by Image J software. In our experiment, the roots were observed to appear on the transparent panes of the growth chamber on the 10th day after transplanting (DAT), and the lateral roots appeared on the 24th DAT. Results indicated that tobacco roots had two growth peaks after transplanting. The maximal values of root growth rate were 54.58 cm/d and 185.69 cm/d, respectively. The roots reached the maximum depth on the 46th DAT, while the root width still showed a nearly linear increment on the 53rd DAT. Interestingly, the relation between root depth/root width and growth time showed a "V" feature. The ratio of root depth and root width reached the minimal value of 1.47 on the 24th DAT. It was also found that most of the root distributed in the 0-10 cm soil layer before the 35th DAT. After that, the most-rapid-elongation area of the root moved downward constantly. All seedling roots were excavated from the rhizobox in order to analyze the root parameters at the end of the experiment. The data acquired from the transparent pane were compared with those obtained from the rhizobox excavation. It was found that most of the tobacco roots were distributed in the >20~40 cm soil layer and the root length distribution had a similar pattern with the two methods. The root length data of the different soil layers acquired from the transparent pane was significantly correlated with those obtained by the rhizobox excavation. The correlation coefficient of each soil layer was over 0.9. Our results indicated the apparatus can be used for non-invasive observation of root growth dynamics under soil conditions. In the end, advantages and disadvantages of the new type of rhizobox are discussed.