Comparison of some statistical methods in estimation of suspended load in a watwrshed with sowy-rainy hydrological regime
2003
Arab Khedri, Mahmood | Hakimkhani, Shahrokh | Nikkami, Davood
The appearance of differences among sediment yield estimations by different people and organizations are common. A part of these differences are related to the quality of data, wile the use of statistical method, such as sediment rating curve (SRC) and the method of using discharges, are another sources of difference. This research, which was performed on Gazagly hydrometric station of Gorganrud Riveranalyzed the effect of data quality and statistical methods on accuracy and precision of suspended sediment load (SSL) estimation. In this station, about 2000 suspended sediment concentration record were measured especially in flooding periods within 6 years. First sampling was done with about 90 methods, which were different in terms of sampling size, number of data from flood periods, sampling methods (randomized or systematic or combination), and separation of rainy and snowy periods. Then 18 statistical methods were used for estimating SSL consisting of 6 Sediment Rating Curves methods (one curve, two curves, mean load within discharge classes, and FAO correction factor, parametric and nonparametric correction factors) in combination with 3 methods of using discharge data (flow duration curve, daily discharge and monthly discharge). The overall estimations were 750 and actual SSL was determined from observed data. To find accuracy and precision of estimations, they compared with actual SSL, by three indices of estimation error, mean error of group, and standard deviation of group. The first index was used for accuracy appraisal of the estimation, and the next two indices were used for evaluating the accuracy and precision of group estimations alternatively. The results show that the quality of primary data is the most important parameter in accuracy and precision of estimations. Existing of 300 concentration measurements in whole range of discharge is sufficient. Adding 150 records from flooding period improves the quality of primary data. Also using 500 concentration records (without additional mean and high flow data) leads to acceptable estimations. Systematic, random, or combined sampling method has not any significant effect on result improvement. All estimations had more sensitivity to SRC type. If the primary data have high quality, the andldquomean load within discharge classes methodandrdquoSRC, had the most suitable results. Also the SRC with parametric correction factor has suitable estimations in some special conditions with lower quality data. Daily and monthly discharges respectively show better results than flow duration curve method. Separations of rainy and snowy periods haven't positive effect on the results and the accuracy and precision of estimations in rainy periods was very lower than snowyrainy periods.
Показать больше [+] Меньше [-]