Hydrologic model calibration using discontinuous data: an example from the upper Blue Nile River Basin of Ethiopia

Abstract

Abstract:Hydrologic models using water balance approaches typically use continuously observed streamflow data for calibration. Many large river basins in developing countries such as the upper Blue Nile River Basin of Ethiopia have discontinuous hydrographs that contain short continuous periods. Therefore, the efficient use of observed hydrographs for calibration of a hydrologic model is important to improve model performance. The goal of this study is to assess how limitations of continuity and duration in data affect hydrologic model calibration. A previously developed water balance model for the upper Blue Nile River basin was calibrated here using continuous and discontinuous (randomly sampled) data of different lengths. The performance of both methods was then compared each other in terms of parameter uncertainty and model efficiency. The results revealed that randomly sampled data require a shorter calibration length than continuous data to reach good model performance, about 36 and 120 months, respectively. This fact implies that discontinuous hydrographs can be useful in calibration. However, the number of high flow months included in the calibration data greatly affects model efficiency. This study suggests that randomly sampled calibration data should include at least 30% of high flow months of sufficient quality. The findings of this study will be essential to develop a hydrologic monitoring strategy for remote basins of the upper Blue Nile River basin and other similar basins where continuous observations are limited.

Physically based distributed hydrological model calibration based on a short period of streamflow data: case studies in four Chinese basins

“…If there are calibrations using 6-month data that could achieve performances similar to the benchmark calibration, stage two of the experiment was initialized, in which the subsets of 6-month data records were used for calibration to explore the performance of a calibration period shorter than six months. Kim and Kaluarachchi (2009) and Yapo et al (1996) showed that data from high-flow periods are more informative than data from low-flow periods for model calibration. As our study explored the possibility of the highest performance of certain lengths of records for calibration, the 3-month, 1-month data and 1-week datasets with highest average streamflow in the 6-month records were employed as the representatives to calibrate the model and conduct the evaluation at these three temporal scales.…”

“…Generally in the two dry basins, if the model performance in the calibration period is good, 6-month data from wet years or wet periods make more reliable simulations in the validation period than the ones from dry years or dry periods. Kim and Kaluarachchi (2009) demonstrated that data from high-flow periods have greater control on model calibration because they are more informative with regard to parameter identification. In this context, our suggestion is in line with those made by Yapo et al (1996) and Melsen et al (2014): data from wetter periods are preferred for model calibration.…”

“…If such data are to be used for calibration, it is important to know how many observations are needed to calibrate model parameter. It is usually suggested that streamflow records covering several years are necessary (Yapo et al, 1996); however, several researchers have attempted to challenge this common understanding using discontinuous or short-period records of less than 1 year in basins within different climatic regions (e.g., Perrin et al, 2007;Kim and Kaluarachchi, 2009;Seibert and Beven, 2009;Tada and Beven, 2012). For conceptual models, these studies indicated that with observations of the order of several scores, reasonable parameter estimates could be derived.…”

Modeling of Sediment Yield From Anjeni‐Gauged Watershed, Ethiopia Using SWAT Model1

“…, 2008a,b; Setegn et al, 2009a,b; Abtew et al. , 2009; Kim and Kaluarachchi, 2009). In this study, the physically based SWAT model was applied to Anjeni‐gauged watershed for prediction of soil erosion and sediment yield.…”

Comprehensive Management and Modeling of Water Quantities: Utilizing an Existing Watershed Model for Comprehensive Water Management and Planning, and Optimal Design of Outlet Control Structures for Ecological Detention Ponds

“…The Nash-Sutcliffe efficiency index, the final value reported as a result of the Nash-Sutcliffe Model efficiency test (Nash and Sutcliffe 1970), is a widely used statistic for this purpose (Oudin et al 2006;Bardossy and Das 2008;Kim and Kaluarachchi 2009;van der Heijden and Haberlandt 2010).…”