Quantifying the Performances of the Semi-Distributed Hydrologic Model in Parallel Computing—A Case Study

Abstract

The research features how parallel computing can advance hydrological performances associated with different calibration schemes (SCOs). The result shows that parallel computing can save up to 90% execution time, while achieving 81% simulation improvement. Basic statistics, including (1) index of agreement (D), (2) coefficient of determination (R2), (3) root mean square error (RMSE), and (4) percentage of bias (PBIAS) are used to evaluate simulation performances after model calibration in computer parallelism. Once the best calibration scheme is selected, additional efforts are made to improve model performances at the selected calibration target points, while the Rescaled Adjusted Partial Sums (RAPS) is used to evaluate the trend in annual streamflow. The qualitative result of reducing execution time by 86% on average indicates that parallel computing is another avenue to advance hydrologic simulations in the urban-rural interface, such as the Boise River Watershed, Idaho. Therefore, this research will provide useful insights for hydrologists to design and set up their own hydrological modeling exercises using the cost-effective parallel computing described in this case study.

A Study on Climate-Driven Flash Flood Risks in the Boise River Watershed, Idaho

“…Thus, the regional climate models are evolving with additional information and new approaches to better increase the predictability using any large-scale driving data, including aerosols and chemical species [56]. Additionally, the fast-moving technologies and applications, such as high-performance computing, computer parallelism in hydrological modeling [57], and unmanned aerial system (UAS) for flood mapping would be another avenue to improve predictability by mitigating uncertainty and risks associated with other foreseen factors [13] (e.g., population growth, urbanization, and economic development). For example, Figures 7-9 illustrate the time series of ensemble 3D flows at OBS1, OBS2 and OBS3 respectively from HSPF associated with each of the climate projections.…”

Decision-Making of LID-BMPs for Adaptive Water Management at the Boise River Watershed in a Changing Global Environment

“…HSPF is a watershed scale, process-based, and semi-distributed model. This model can effectively simulate streamflow and water quality associated with land management practices and climate variability at the urban-rural interface, such as BRW [26][27][28][29][30]. The HSPF model consists of the main three modules (PERLND, IMPLND, and RCHERS).…”

“…The parameter estimation in PEST is accomplished using the Gauss-Marquardt-Levenberg algorithm (GML) to minimize the user-defined objective function (e.g., minimization of root mean squares between simulated and observed values) [38]. The detailed information of computer parallelism in the HSPF is available in the literature [28].…”

Modeling Hydrological Response to Climate Change in a Data-Scarce Glacierized High Mountain Astore Basin Using a Fully Distributed TOPKAPI Model

“…The parameters were grouped into (L, K s , V s ) and (n o , n c ) because they were independent of each other, i.e., the change in the first group was affecting the quantity and change in second was affecting the timing of runoff at Astore basin (Table 1). Three different statistical measures as recommended through literature, i.e., Index of Agreement (d), R 2 and Nash-Sutcliffe Efficiency (NSE) coefficients were used to assess the efficiency of calibration by comparing modeled and observed discharge volumes aggregated on a daily and monthly basis [68][69][70][71]. Before further application of a hydrological model for future forecasting, it is recommended that the validation of the model should be done for a better understanding of spatio-temporal performance of the model.…”