Hydrological Simulation Study in Gansu Province of China Based on Flash Flood Analysis

Hydrological Simulation Study in Gansu Province of China Based on Flash Flood Analysis

3.1. Hydrological Model Construction

The first-level partition (cluster) of the CNFF model takes the second-level water system in Gansu Province as the main reference basis and comprehensively considers the effects of the second-level hydrological partition, three major steps, underlying surface characteristics, and construction projects to achieve rapid flood parallel simulation calculations. Based on the abovementioned model cluster construction basis, Gansu Province is divided into 11 model clusters. The inland river basin includes the Shule River model cluster, Sugan Lake model cluster, Heihe model cluster, and Shiyang River model cluster; the Yellow River basin includes the Taohe model cluster, Huangshui model cluster, Yellow River model cluster, Weihe model cluster, and Jinghe model cluster. Finally, the Beiluo model cluster and the Jialing River model cluster are in the Yangtze River basin (Table 3). The scope of each model cluster is shown in Figure 4. On the basis of model clusters, the model range is further divided based on the similarity of underlying surface characteristics. Each distributed system is established through seven types of hydrological elements (small watershed, river section, node, reservoir, water distribution, water source, and depression) and six major hydrological processes (rainfall, evaporation, runoff, confluence, river evolution, and reservoir regulation and storage). A total of 289 distributed hydrological models were constructed based on the topological relationships of the basin water system within the hydrological model.
There are no hydrological stations in the Sugan Lake and Beiluo River systems, and parameter calibration and verification cannot be carried out, so this study did not consider the simulation of this water system. In order to reflect the simulation conditions of the nine major river systems, basins were selected for research and analysis based on conditions such as even distribution of rainfall stations in the basin and complete hydrological data. The specific locations are shown in Figure 5. Through the analysis of meteorological and hydrological conditions in Gansu, the average annual potential evaporation in the study area is large, rainfall is relatively low, and the distribution within and between years is uneven. The soil moisture index in most areas of the study area is in the range of 0.20 ≤ SWI 29], which mainly includes plane mixing, vertical mixing, and period mixed runoff.

3.2. Sensitivity Analysis

CNFF model parameters are divided into three categories: input parameters, initial parameters, and calibration parameters. Among them, input parameters refer to input data that can be obtained directly through data or land use, soil type, and hydrological zoning maps. Such parameters are input at the initial stage and remain unchanged during the subsequent model parameter adjustment process; the second type of parameters are initial parameters, which are initial values set according to the actual situation before the model is run; and the third type of parameters are those with physical meaning model parameters. Such parameters are determined through corresponding data, including soil type, land use, hydrogeological zoning, terrain moisture index map, water holding capacity, plant root depth, evaporation limit burial depth, soil permeability coefficient, etc., and then the parameters are calibrated. Since the CNFF model involves a large number of parameters that need to be calibrated, it is difficult to ensure the accuracy of all parameters at the same time, so it is very necessary to conduct parameter sensitivity analysis. By evaluating the sensitivity of parameters, the number of model parameters that need to be adjusted for parameter rate timing is reduced, and sensitive parameters that affect simulation accuracy are selected for parameter calibration [31,32,33,34].
This paper adopts the RSA method [35], selects nine watersheds in the study area for parameter sensitivity analysis, and selects three flood events for analysis. Three indicators, NSE, REQ, and ET, were selected to conduct multiple sensitivity analyses on the parameters. Since the results of the sensitivity analysis of model parameters for different objective functions (C) are different, this paper not only considers the objective function individually but also uses a trade-off method to construct a combined objective function, and it sensitizes the model parameters by considering the characteristics of each objective function. The sensitivity analysis is specifically as follows (4):

C = ω 1 ( 1 N S E ) + ω 2 R E Q

In the above formula, C is the combined objective function, and w1 and w2 are the trade-off coefficients of different objective functions. Since the above two objective functions are of equal importance, the values of w1 and w2 are both 1/2.

In this study, the nonlinear flow generation method was selected for model simulation, and the calibration parameters associated with linear flow generation were not considered. Therefore, there are 12 model parameters involved in the calibration. Assuming that the model parameters are uniformly distributed, use Monte Claro in CNFF

The model parameters were sampled within the feasible range (1000 times), and 1000 sets of parameter values were obtained. Then, the values were imported into the CNFF model for simulation calculation, and the objective function value and combined objective function value were calculated based on the simulation results. The objective function standard is used as the basis for distinguishing between samples that have done something and those that have done nothing, specifically: NSE > 0.7, REQ < 0.3, ET < 3 h, C < 0.3. According to the sample data results of action and inaction, the cumulative distribution curve of each parameter is calculated, respectively, and G(x) and U(x) are obtained. Kolmogorov–Smirnov (K-S) is used to test the fitting degree of the cumulative function of samples with and without action [35], specifically as follows (5):

D = m a x | G ( x ) U ( x ) |

In the above formula, G(x) and U(x), respectively, represent the cumulative distribution curve of parameter x in the sample and the cumulative distribution curve of parameter x in the sample without. D represents the maximum vertical distance of the x parameter between the two cumulative distribution curves. In the K-S test, the selected significance level (α) is 0.05, the null hypothesis can be expressed as H 0 G ( x ) = U ( x ) , and the alternative hypothesis is expressed as H 1 G ( x ) U ( x ) . If p ≤ 0.05, the null hypothesis (H0) is rejected; that is, the parameter is sensitive. According to the relative sensitivity, it can be divided into four categories: highly sensitive (D > 0.2, p ≤ 0.05), medium sensitive (0.1 ≤ D ≤ 0.2, p ≤ 0.05), relatively sensitive (D < 0.1, p ≤ 0.05), and insensitive (p ≥ 0.05) [36].
According to the above method, a table of model parameter sensitivity results for different objective functions was obtained (Table 4) in which three asterisks indicate high sensitivity, two asterisks indicate moderately sensitive parameters, one asterisk indicates mild sensitivity, and no asterisks indicate insensitive. When the objective function is the statistical objective function NSE, the parameters with high sensitivity are Satv, and the parameters without sensitivity are Expnonliner, SG Exp, Cfast_sq, CG, and CGSink. When the objective function is the flood characteristic objective function REQ, there are no highly sensitive parameters, and the insensitive parameters are SoilMax, SoilGMax, Cslowsq, CG, and CGSink. When the objective function is the flood characteristic objective function ET, there are no highly sensitive parameters, and the insensitive parameters are Cnonliner, RsoilMax, Soilmax, SoilCMax, SGEexp, Cslowsq, CG, and CGSink. When the objective function is based on C, the highly sensitive parameter is Satv, and the insensitive parameters are SoilGMax, Cslowsq, CG, and CGSink. In summary, it can be seen from the above results that the parameters with higher sensitivity in the CNFF hydrological model are gravity and priority flow reservoir water storage capacity, coefficients in the nonlinear runoff generation algorithm, the maximum possible runoff generation area ratio, etc., indicating that the type of vegetation and soil in the study area has a greater impact on the simulation results of the CNFF hydrological model. From the perspective of each hydrological module in the CNFF hydrological model, the sensitivity of the relevant parameters of the soil module is higher than that of the groundwater module. This shows that the groundwater in the study area has little influence on the simulation of the overall runoff process, the base flow is relatively stable, and the changes in runoff are mainly reflected in surface runoff and mid-soil flow.

3.4. Model Discussion

Based on the analysis of the application effects of 318 floods in the above nine basins, the average certainty coefficients of all basin rates in the study area during the regular and verification periods are 0.76 and 0.73; the average peak flow errors are 9.1% and 12.6%, respectively; and the average peak flow time errors are 1.2 h and 1.7 h. Table 5 shows the results of the flood simulation effect evaluation in nine basins. Figure 5 and Figure 6 show the regular results of partial basin rates and the verification period results of partial basins, respectively.
It can be seen from the results that the simulation results in the Jinghe River, Yellow River, Weihe River, and Jialing River basins are relatively good. The certainty coefficients in the regular rate and verification periods are both around 0.8, and the relative error of the peak flow is controlled at around 10%. The peak time error is within 1 h. The simulation results in the Taohe, Huangshui, Shiyanghe, and Heihe River basins are average. The certainty coefficients in the rate period and verification period are around 0.7, the relative errors in peak flow are around 10–15%, and the peak time errors are around 1.5 h. The simulation effect of the Shule River Basin is not ideal. The deterministic data of the rate period and the verification period are about 0.5. The relative error of the flood peak flow varies greatly. The rate period is 12%, the verification period is 19%, and the peak time error is about 2 h. Part of the calibration and verification results are shown in Figure 7 and Figure 8. The floods in Jingchuan on 4 July 2018, Qin’an on 2 July 2018, Huangluba on 11 July 2018, and Xiangtang on 28 June 2019 were all during periods of heavy rainfall. The concentration of heavy rainfall in the basin was high, and generate high discharge. Except for Qin’an, the other three river basins all produced flood peak superposition effects. Among them, the maximum rainfall in the Jingchuan River Basin occurred at 18:00 on 4 July, one hour earlier than the maximum peak flow. The remaining sites also have high fitting degrees, indicating that the model has good practicability in this region.
According to the simulation results, generally speaking, the CNFF model shows good applicability at different time and spatial scales in Gansu Province. The stations in the Jinghe River Basin, Yellow River Basin, Weihe River Basin, and Jialing River Basin are densely deployed. The rainfall data can reflect the actual rainfall conditions, and the model simulation effect is ideal. The simulation results in this area are basically consistent with the conclusions drawn by Li Changbin [36]. The Taohe River Basin and the Huangshui River Basin are high in altitude and are dominated by plateaus and mountains. The rainfall varies greatly with the terrain, causing the simulation effect to be affected to a certain extent, but overall, it is good. Snowmelt runoff exists in some areas of the Shiyang River and Heihe River basins. For seasonal inland river basins, snowmelt is one of the main source of runoff. Since the snowmelt model was not considered, the simulation results in this area were not accurate enough. However, the general trends of the simulated flow hydrograph and the actual flow hydrograph are similar. This conclusion is basically consistent with the results of Shang Ling [37], who used a hydrological model (HIMS) to simulate the Shiyang River Basin.
The simulation results of the Shule River Basin are not ideal, with a certainty coefficient of only 0.51. The degree of fit between the simulated flow process and the actual flow process is low, and the relative error deviations of peak time and flood peak are also relatively high. The main reason is that rainfall stations are sparsely distributed in the Shule River Basin, and rainfall data cannot truly reflect the rainfall conditions in the Shule River Basin. The Shule River Basin covers an area of 41,300 km2, accounting for 9.7% of the total area of Gansu Province. There are only 107 rainfall stations, with a station coverage density of 386 km2/station, and the proportion of rainfall stations is 1.8% of the rainfall stations in Gansu Province. Other watersheds in the Shule River, such as Dangcheng Bay, Panjiazhuang, Shuangtabao, etc., also have problems with varying degrees of unsatisfactory simulation results. The research results of Sun Boyang [38] and others using the distributed hydrological model (DTVGM) are relatively consistent with the simulation results of the CNFF flash flood hydrological model in the Shule River Basin in Gansu Province. However, DTVGM is a daily runoff simulation and cannot carry out the simulation of short-term flash floods and heavy rains.

Disasters Expo USA, is proud to be supported by Inergency for their next upcoming edition on March 6th & 7th 2024!

The leading event mitigating the world’s most costly disasters is returning to the Miami Beach

Convention Center and we want you to join us at the industry’s central platform for emergency management professionals.
Disasters Expo USA is proud to provide a central platform for the industry to connect and
engage with the industry’s leading professionals to better prepare, protect, prevent, respond
and recover from the disasters of today.
Hosting a dedicated platform for the convergence of disaster risk reduction, the keynote line up for Disasters Expo USA 2024 will provide an insight into successful case studies and
programs to accurately prepare for disasters. Featuring sessions from the likes of The Federal Emergency Management Agency,
NASA, The National Aeronautics and Space Administration, NOAA, The National Oceanic and Atmospheric Administration, TSA and several more this event is certainly providing you with the knowledge
required to prepare, respond and recover to disasters.
With over 50 hours worth of unmissable content, exciting new features such as their Disaster
Resilience Roundtable, Emergency Response Live, an Immersive Hurricane Simulation and
much more over just two days, you are guaranteed to gain an all-encompassing insight into
the industry to tackle the challenges of disasters.
By uniting global disaster risk management experts, well experienced emergency
responders and the leading innovators from the world, the event is the hub of the solutions
that provide attendees with tools that they can use to protect the communities and mitigate
the damage from disasters.
Tickets for the event are $119, but we have been given the promo code: HUGI100 that will
enable you to attend the event for FREE!

So don’t miss out and register today: https://shorturl.at/aikrW

And in case you missed it, here is our ultimate road trip playlist is the perfect mix of podcasts, and hidden gems that will keep you energized for the entire journey


This website uses cookies to improve your experience. We'll assume you're ok with this, but you can opt-out if you wish. Accept Read More