Abstract: Abstract: Gravel-mulched field is typical in Northwest of China for watermelon production. Few studies are on dry matter and yield models establishment of watermelon in gravel-mulched field, which is good for prediction of watermelon growth and yield. This study aimed to establish the dry matter and yield models for watermelon in gravel-mulched field based on photosynthesis and water-N coupling. Two experiments were designed in 2009. One was for photosynthesis-based dry matter model validation (Exp. 1) and the other was for water-N coupling model validation (Exp. 2). The experiments were conducted under a rain shelter in order to simulate different types of rainfall years.The experimental site was located in Hongquanzi Village, Zhongwei City of Ningxia (37°56′N, 105°15′E). The soil and gravel were taken from 0-40 cm depth of a typical 2-a gravel-mulched field. Based on experience and previous published papers, volumetric soil moistures of 11%-20%, 13%-24% and 16%-27% simulated dry, normal and wet years, respectively. The organic fertilizer application rate was designed at 4 levels at 0, 600, 1200 and 1800 kg/hm2. In Exp.2, five levels of organic fertilizer application rate were 600, 750, 1 200, 1 560 and 1 800 kg/hm2 and 5 levels of urea at 24, 33, 48, 61.5, 72 kg/hm2. All the experiment was conducted in barrels. A photosynthesis-based dry matter accumulation model of watermelon was established based on daily total photosynthetic assimilation amount calculated with Gauss integration method. The dry matter and yield models were also established using multiple regression method. The models were evaluated by indexes of correlation coefficient (r), root mean square of error (RMSE), and normalized RMSE (nRMSE). The results showed that: 1) the photosynthesis-based dry matter model had RMSE of 22.5 kg/hm2, nRMSE of 14.5% and r of 0.893, which indicate that the model was reliable in watermelon dry matter accumulation simulation; 2) Water-N coupling model of watermelon dry matter had low accuracy (R2=0.30) for ET≥150.2-513.8 mm and available N≥172.1-241.7 kg/hm2, and the accuracy became higher (R2=0.78, P<0.01) for ET≥213-513.8 mm and available N≥172.1-226.9 kg/hm2. The water-N coupling model of watermelon economic yield also had higher accuracy (R2=0.88, P<0.01) for ET≥213-513.8 mm and available N≥172.1-226.9 kg/hm2. Meanwhile, the water-N coupling model of watermelon dry matter and economic yield had low RMSE (25.2 and 668 kg/hm2) and nRMSE (15.4% and 14.1%), which indicated the water-N coupling model was reliable in dry matter and yield simulation of watermelon for ET≥150.2-513.8 mm and available N≥172.1-241.7 kg/hm2. The models here could be used for dry matter and yield prediction of watermelon in gravel-mulched field and provide useful tool for improving the utilization efficiency of water and fertilizer.