Abstract: Understanding droplet distribution inside tree canopy has a great significance on optimizing spray parameters and improving spray effect. In this article, distribution of droplet penetration rate inside tree canopy during air-assisted spraying was studied by experimental method. The droplet penetration rate was defined as the ratio of droplets inside canopy along unit vertical face to that before penetrating into the canopy. The change of droplet penetration rate with tree leaf density, outlet air velocity of sprayers and sampling depth was analyzed. Disc sprayer and multi air pipe sprayer were used to spray 4 types of tree canopies of pear trees (big, medium and small sizes), wintersweet, cerasus subhirtella and punica granatum. Droplet deposition was determined in real time. The results showed that the droplet penetration rate decreased as the leaf density and sampling depth increased while it increased with increased outlet air velocity of sprayer. Among the 3 test variables of tree leaf density, outlet air velocity of sprayer and sampling depth, the sampling depth greatly affected the droplet penetration rate. Following the changing pattern of droplet penetration rate with tree leaf density, outlet air velocity of sprayer and sampling depth, 5 types of models (linear polynomial, quadratic polynomial, cubic polynomial, single exponent and quadratic exponent) were assumed to fit the changing pattern of droplet penetration rate. The quadratic exponent model had the highest accuracy with R2 higher than 0.95 and the RMSE was the least from 4.1% to 5.0%. By experimental validation, the model still had relatively reliable accuracy. Thus, the quadratic exponential model was finally chosen as the suitable model. By using this model, it had to be built based on canopy of each tree. By mixing the data from different tree canopies, we tested the feasibility to estimate droplet penetration rate by a quadratic exponential model. The results showed that the R2 was still higher than 0.95 and root mean square error was 5.3%-5.7%. By validation, the relative error could be lower than 20% for mixture of wintersweet, cerasus subhirtella and pear trees but reach up to about 29% for punica granatum tree. It was because the branch structure of punica granatum was different from the other trees. The quadratic exponential model was then extended to application in calculating droplet drift rate after tree canopy in order to analyze its influencing factor. The droplet drift rate after Cerasus subhirtella tree canopy of different leaf density under different sprayer outlet air velocities was calculated using model and validated by test. The mean relative error was 16.73%, indicating that the model could be used to determine the suitable outlet air velocities for tree canopies of different growing stages and reduce the droplet drift rate. The variation coefficients of droplet distribution inside tree canopy were also calculated by the model, which was used to better explain the principle of double-side spraying pattern. In the end, the model limitation and subsequent research assumption to improve model were illustrated: spray pressure, flow rate, environmental temperature and humidity were approximately regarded as constants in this article, and in further researches, the influence on droplet penetration rate of these factors should be investigated. A unit model for estimating droplet penetration rate could be established if the tree types had similar branches and leaf structures, and in further researches, more types of trees should be chosen and classified to build mixture model to improve the model accuracy. This paper is helpful to understanding droplet distribution in tree canopy during air-assisted spraying.