Seite - 269 - in Book of Full Papers - Symposium Hydro Engineering

closer to Jinjiangjie than that to Panzhihua. In the view of watershed scale, the nonstationary impact on flood peak series is not taken into account in this article, and the conventional sampling method for P-III distribution are employed. 3.2.2 Sampling method (1) Parameters for sampling    v 1 s 2C a EX (1 ) C [2]    2 v s 2 a EX C C [3]  3 2 s 4 a C [4] p = a3 - n，where, n is an integer and 0 < p <1 [5] (2) Three uniform random number u1, u2, and u3 in [0,1] are generated. (3)     n 4 i i 1 a lnu [6] (4) Three uniform random number un+1, un+2, and un+3 in [0,1] are generated. (5)      1 1 p 1 p 1 n 1 2 n 2 s u ,s u [7] (6) if s1 + s2 ≤ 1， then   1 1 2 s t s s ，     n 3 x t lnu ，   4 z' x a ，   1 2 z' z a a [8] Thus, z is a random variable following P-III distribution. Else if, s1 + s2 > 1, then return to (4). 40000 samples are generated following the program mentioned above and should be undertaken hypothesis testing to confirm the results. 3.3. FLOOD INFLOW FOR EACH RESERVOIR The 1st, 2nd and 3rd cascade reservoirs were located between the 1st and 2nd hydrological station, while the 4th reservoir lies between the 2nd and 3rd hydrological station (see Fig 1). Flood peaks of the four cascade reservoirs are calculated by hydrological analogy method based on 40000 flood peak samples of the three hydrological stations and drainage area ratios. 269