In order to estimate the flood discharge and to control it in the design of agricultural hydraulic structure, espacially to derive the unit hydrograph from small watersheds, Banweol G-1 (5.317㎢), G-2 (4.096㎢) in Korea and the Hupselse Beek (6.5㎢) in the Netherlands, the author analysed and considered in this paper the data observed and calculated in total eight models such as Matrix Inversion, Nash-Model, J-Model, 2-Parallel Unequal Reservoir and 2,3,4,5-Parallel Equal Reservoir. The unit hydrograph to be deduced by their models was tested whether to be significant or not. It was found out that the runoff to be estimated from the observed precipitation was calculated by the Convolution Integral. The results were as follows. 1. The shape of the precipitation distribution in Banweol and Hupselse watershed was applied to χ²-test and Kolmogorov-Smirnov test, and double exponential distribution function was acknowledged to the reasonable distribution. 2. Whether there is homogeneity between precipitation and watershed among Banweol, Suwon and Nammyeon site was examined by Bartlett’s test, and those watersheds were identified to be within the same boundary of homogeneity watershed. 3. The objective function values of the observed and calculated runoff for unit hydrograph of Matrix Inversion, Nash-Model, and J-Model in Hupselse watershed were 0.2925, 0.3734 and 0.4054 respectively and were optimized, but only Matrix Inversion was not proper because it was shown to have minus values with DUH. 4. Also the objective function values of 2-parallel unequal reservoir and 2,3,4,5-parallel equal reservoir were estimated to be 0.4423, 0.2817, 0.2508, 0.2859 and 0.2516 respectively and they all are considered to be optimized. 5. As the objective function values of Nash-Model and J-Model for the unit hydrograph in Banweol G-1 watershed in Korea were marked to be 0.0054 and 0.0557, they were considered to be optimized. But, as the values of G-2 watershed were 0.0102 and 1.1511 respectively, they were too large comparatively to be considered as not optimized. 6. J-values which are expressed the characteristics of the watershed were 245.4 and 533.5 respectively by J-Model in Banweol G-1 and G-2 watershed. The G-1 watershed having smaller J-values was judged to be more optimized than that of G-2. 7. The objective function values between the observed runoff and the calculated runoff that was obtained from unit hydrograph and precipitation in Hupselse watershed by Convolution Integral were estimated to be 15.75, 12.21 and 13.04 for Matrix Inversion, Nash-Model and J-Model respectively. Among them, Nash-Model and J-Model were shown to be satisfactory and optimized. 8. The objective function values between the observed runoff and the calculated runoff that was estimated from the unit hydrograph and precipitation in Hupselse watershed by Convolution Integral were marked to be 13.04, 12.73, 12.87 and 13.27 for 2-parallel unequal reservoir and 2,3,4,5-parallel equal reservoir model respectively. They all are expressed to be good and optimized. 9. The objective function values between the observed runoff and the calculated runoff that was obtained the unit hydrograph and precipitation data in Banweol G-1 watershed of Korea by Convolution Integral were estimated to be 1.5539 and 6.6749 for Nash-Model and 1.9365 and 6.5086 for J-Model respectively, and they all are good and optimized. 10. The objective function values between the observed runoff and the calculated runoff that was computed from the unit hydrograph and precipitation data in Banweol G-2 watershed by Convolution Integral were shown to be 1.5198 and 1.0769 for Nash-Model, and 1.6798 and 19.4302 for J-Model. The former values were taken to be optimized but the latter ones were considered not to be optimized and proper. 11. In the case of Banweol watershed, the Nash-Model was taken to be proper and to be applicable to the same cases, but as the J-Model has J-value subject to the charaterictics of watershed and the J-value governs the optimiztion of J-Model, continuous studies and analyses are required with long-term and reliable observed data. 12. The error between the observed and the calculated peak discharge for unit hydrograph was shown to be 0. 00% in Banweol G-1 watershed and 0.94% in Banweol G-2 watershed in the case of Nash-Model as shown in Table 25, and 9.09% in Banweol G-1 and 14.28% in Banweol G-2 watershed in the case of J-Model respectively. All three other values but the last one are accepted as to be within the boundary of the allowable error. 13. Generally Nash-Model was considered to be optimized in small watershed of Korea and J-Model was judged to be applicable to the watershed having the shape of leaf comparatively gentle slope as in the case of Banweol G-1 watershed.