Flood risk mapping for the area with mixed floods and human impact: a case study of Yarkant River Basin in Xinjiang, China

Flooding has been causing severe consequences worldwide, including loss of human life and damage to property. Flood risk mapping, as a nonstructural measure, is efficient for flood protection and disaster alleviation. This study aims at completing the flood risk mapping of the region located at the middle reaches of the Yarkant River Oasis in western China, which has a dry climate and suffers from mixed flooding consisting of glacial outburst floods (GLOFs) and many other floods. In view of the complexity of flooding in the area, the study adopts two typical types of scenarios, namely overflow scenarios and dike-break scenarios, to complete the flood risk mapping. The MIKE FLOOD 1D/2D coupled model is used for two-dimensional flood flow simulation to compute the inundation depths and duration for flood risk assessment. The spatial overlay analysis was then used to combine the modeling results and land use/land cover layers with socioeconomic data to generate flood risk maps and damage losses under different scenarios. It is noted that evaporation and infiltration losses in the study area are not negligible because of the long flood process, the low precipitation, and dry surface/subsurface conditions. Due to the insufficient evaporation and infiltration data, a new method of synthesis loss rate is proposed to compute the evaporation/infiltration loss rate. Based on the water balance principle, the upstream and downstream flow data is utilized to calculate the water attenuation, which is then used to estimate the evaporation/infiltration loss rate. The proposed method can solve the problem of calculating evaporation/infiltration loss rates during the flooding process in such data-scarce areas. The flood risk mapping results indicate that the flood risk is high along the Yarkant River and that floods can cause severe inundation losses.


Introduction
Flooding is the most frequent natural disaster, as it accounts for approximately 40% of natural disasters on Earth and has caused considerable losses in terms of human life, social economy, and ecosystems [1,2]. To mitigate the adverse impacts of flooding, various flood mitigation measures are implemented. These measures can be broadly classified into two categories: structural and non-structural measures, both of which play crucial roles in flood disaster control and mitigation. Structural measures involve the construction of physical infrastructure to manage and control floodwaters. Examples include embankments, levees, floodwalls, dams, and reservoirs. These measures are designed to contain and divert floodwaters, reducing the risk of inundation and protecting communities and infrastructure in flood-prone areas. While structural measures provide immediate protection, they have limitations as they may be overwhelmed by severe flooding events that exceed their design capacity, potentially leading to damage. On the other hand, non-structural measures focus on strategies that complement or substitute physical infrastructure. These measures, guided by flood risk analysis, emphasize preparedness, early warning systems, flood forecasting, land-use planning, zoning regulations, and community-based approaches. Recently, a significant change from traditional structural measures to a combination of structural and nonstructural measures has taken place in flood protection, conforming to the trend that nonstructural far away from the sea and surrounded by high mountains. Coupled with the influence of the nearby Taklimakan Desert, the basin presents a typical arid continental climate with few rainfall events, high evaporation rates and considerable monthly temperature variations. Furthermore, the famous glacier-dammed lakes, Taramkangri and Kyagar, are located at the upper reaches of the Yarkant River. According to the records at the Kaqun hydrological station of the Yarkant River, the repeatedly burst of these two glacier lakes led to frequent GLOFs.
The study area in the Yarkant River basin encompasses the stretch between the head of the Kaqun canal and the head of the canal at the middle reaches, covering an area of 563 km 2 . The head of the Kaqun canal is located at the mountain pass, and no other tributaries flow into the stretch (figure 1). Field work has identified sudden drainage of lakes dammed by Taramkangri glacier and Kyagar glacier in the upper reaches as the primary cause of the flood disasters. Flooding from precipitation and snow melting, combined with GLOF, can result in significant damage. Moreover, the study area is located in the Yarkant River Oasis, which support the majority of the population and economic activities of the whole Yarkant River basin. Generating flood risk mapping for the oasis is essential due to the destructive nature of floods and the population and property in the oasis. However, the construction of human activities, including diversion canals for irrigation, has increased the complexity of hydrodynamic simulations and can reduce the accuracy of the simulation for the entire oasis. Thus, a pilot area of 563km 2 in the Yarkant River Oasis is selected to provide a reference for flood risk mapping in similar areas.
The meteorological data and hydrological data were abtained from Kashgar Hydrological and Water Resources Survey Bureau. Rainfall observations of 5 stations were included to capture the the rainfall characteristics in the Yarkant River Basin. It is revealed that the region receives extremely low precipitation, leading to minimal flood risk attributed to rainfall. Discharge and water level observations were available at Kaqun and Yiganqi stations,as shown in figure 1. The channel cross-sections were obtained from Kashgar Hydrological Bureau. The flood inundation was simulated using 90 m Digital Elevation Model (DEM), downloaded from http://srtm.csi.cgiar.org/. Locations of meteorological and hydrological stations, river network layout, glacier, Yarkant River Basin and research area are shown in figure 1. The data used in the simulations is shown in table 1.

Flood characteristics
From the perspective of the causes of floods, the floods can be categorized as glacial outburst floods(GLOFs), snowmelt floods, rainfall induced floods and mixed floods. Figure 2 show the process of different types of floods. Figure 2(a) is snowmelt flood process in July, 1973, which is relatively gentle and closely related to temperature. Figure 2(b) shows the glacial outburst flood(GLOF) process in the year 1984. It can be seen that GLOFs have the following characteristics: high peak discharge, small volume compared with peak discharge, rapid rising rate, short duration and single peak hydrograph. Field investigations have demonstrated that GLOFs can be very destructive. Figure 2(c) presents a rainfall induced floods process in July 1987. It can be found that the rainfall induced floods process has a low flood peak and lasts for a short time due to the low precipitation.
When the aforementioned floods coincide, a mixed flood can occur. The mixed flood has the characteristics of the floods that compose it, but it is more destructive than either of the individual floods. Figure 2(d) shows the mixed flood process in August 1971, which was a 'glacial outburst-snowmelt' flood. It has a high peak discharge. However, the flood duration is extended and the total amount of floods increases significantly in comparison to GLOFs.

Man-made structures
Due to the intense human activities in this area, varieties of man-made structures have been constructed, which change the characteristics of the underlying surface. Specifically, they reshape the calculation boundaries and influence the flooding through impermeable structures. In light of the spatial distribution of the water sources, diversion canal, roads and other impermeable buildings, the calculation boundaries are generalized as follows.
(1) The upstream boundary is the head of the Kaqun canal.
(2) The downstream boundary is the head of the canal at the middle reaches.
(3) The right bank boundary comprises the head of the Kaqun canal, the Dongan main canal, the Poskam main canal, the Yiganqi Reservoir intake canal, the Yiganqi Reservoir waste canal, and the head of the canal at the middle reaches.
(4) The left bank boundary comprises the head of the Kaqun canal, the Wupu main canal, the Zilchak canal, the Elixku Reservoir waste canal, and the head of the canal at the middle reaches.
The calculation boundaries generalized for the flood risk analysis and the main impermeable buildings inside the calculation boundaries are shown in figure 3. There are totally 35 generalized linear impermeable buildings within the study area, which contain one railway plus nine highways at the county level, a number of township roads and many water diversion channels. Considering the readability of the diagram, only the main linear impermeable buildings are shown in the figure.

Data Description
The water level data and discharge data The data at two stations, Kaqun station (upstream) and Yiganqi station (downstream) River cross section data Data of river cross sections at an average interval of 1 km Digital maps Maps with the scale of 1:10,000 SRTM-DEM The DEM data with 90m-spatial resolution Infrastructure information Information of embankments,hydraulic structures, irrigation channels, etc.

Framework of flood risk mapping
Flood risk mapping shows the spatial distribution of the damage risk of floods with different exceedance probabilities. In order to identify the areas vulnerable to flooding and to determine the degree of damage caused by floods, it is necessary to conduct flood risk mapping that contains three major steps.
(1) Derivation of the design flood. The Pearson Type-III probability distribution function [33] is utilized to get the design peak discharges, then the flood hydrographs for different return years are calculated based on the observed flood hydrograph.
(2) Hydrodynamic simulation. Based on the synthetic events with different exceedance probabilities, the hydrodynamic model is used to calculate the flood risk features such as flood inundation area and flood depth.
(3) Flood risk mapping and loss assessment. Based on the simulation result of the hydrodynamic model, the flood risk maps can be established via the GIS tools. In addition, the flood risk features are superimposed with geographic data and socioeconomic data to compute economic losses.

Derivation of design flood
In the study area, the probability distribution function Pearson Type-III [33] performs well in estimating the design peak discharges for different return periods. With the results of design peak discharge, the design flood hydrographs of different return periods are developed by using nondimensional hydrographs deduced from the flood hydrograph of a typical flood. The deduction of the design flood hydrograph is briefly described here. First, the Pearson Type-III distribution function is utilized to provide a design peak discharge for each return period. Then, the observed discharge for a selected flood is divided by maximizing the observed discharge to normalize the nondimensional hydrograph, in which the peak of nondimensional hydrograph should be equal to one. Finally, the design flood hydrograph is obtained by multiplying the ordinates of nondimensional hydrograph with design peak discharge. These procedures are repeated for return periods 5-, 10-, 20-, 50-, and 100-year, respectively.

Hydrodynamic simulation
A river network model (MIKE11), a flood routing model (MIKE21) and a 1D-2D coupling model (MIKE FLOOD) are utilized for simulating the flood process under various scenarios [24][25][26]. The simulation results include maximum water depth, duration of depth above threshold, maximum current speed, etc.
(1) 1D River Network Hydrodynamic Model The MIKE11 model is a one-dimensional (1D) hydraulic hydrodynamic model, which uses an implicit and finite difference scheme to simulate the unstable, non-uniform flows in rivers and floodplains [24]. This model can describe the flow on multiple hydraulic structures (e.g. weirs, culverts and road overtopping), which makes it one of the most popular river network models.
(2) 2D Routing Hydrodynamic Model The MIKE 21 model is a 2D hydrodynamic model that uses numerical solutions of the 2D shallow water equations to calculate surface flow [25]. It describes the flow over hydraulic structures based on the mesh, and it is widely adopted for modeling in flood-prone areas. In this research, the flexible mesh known as MIKE 21 HD FM is used. On the basis of the unstructured triangular mesh, it is allowed to adjust the size and shape of the triangular elements to suit the local conditions.
(3) 1D-2D Coupling Hydrodynamic Model MIKE FLOOD is used to dynamically couple the MIKE 11 model with the MIKE 21 model for flood propagation and inundation calculations [26]. The lateral link in the MIKE FLOOD refers to the connection between MIKE 11 and MIKE 21 models, allowing a series of MIKE 21 cells to be linked laterally to a specific reach in MIKE 11. The lateral link can exchange information between the MIKE11 model and MIKE 21 model, enabling a more accurate simulation of the flow dynamics in both the river and floodplain. It can be established through a coupling algorithm or a water exchange method. The coupling algorithm considers hydraulic processes of the main river channel and lateral links simultaneously, ensuring water flow continuity and mutual interactions. The water exchange method simulates flow exchange between the main channel and lateral links using water level or flow boundary conditions, simplifying calculations for smaller links or weak hydraulic dynamics. The choice depends on model requirements and link characteristics. The coupling algorithm suits complex systems, considering hydraulic interactions. The water exchange method suits simplified modeling. Considering the simulation objectives and system complexity of this study, the water exchange method is employed for the lateral link.

Parameters calibration
According to the historical flood data, the hydrological parameters are calibrated to bridge the gap between observed and simulated discharges and water levels. Additionally, a new parameter, namely synthesis loss, is proposed to address the problem of lack of evaporation and infiltration data. The Manning's roughness coefficient and the flooding and drying threshold parameters are also calibrated.
(1) Synthesis loss The complex climatic and topographic condition in the study area lead to large evaporation and infiltration. Therefore, the evaporation loss and infiltration loss during the flooding process cannot be ignored. However, the lack of evaporation and infiltration data makes it difficult to calculate the loss. To address this problem, a new method is proposed to calculate the synthesis loss that consists of evaporation loss and infiltration loss. In the proposed method, the upstream and downstream flow data is utilized to compute the water attenuation. Then, the synthesis loss rate is estimated based on the calculated attenuation. The details are as follow.
(1) Calculate the water surface width b i , i = 1, 2, L ,N on the basis of the corresponding water levels and the channel cross-sections data. Select N channel cross-sections and record the highest and lowest water levels of each river cross-section in each flood. Take the average of all the highest and lowest water levels of the ith channel cross-section as the mean water level and calculate the water surface width b i , i = 1, 2, L ,N at the mean water level based on the data of the channel cross-section data.
(2) Estimate the average infiltration area S, given by å å where l i represents the distance between ith section and i + 1 th section.
(3) Calculate the water attenuation ΔV according to the observed discharge process at the upstream station and downstream station, denoted by where F up gives the upstream discharge process and F down denotes the downstream discharge process.
(4) Compute the evaporation/infiltration rate μ via where Δt is the duration of the flood.
(2) Manning's roughness coefficient Manning's roughness coefficient characterizes the resistance to flood flows in channels and flood plains. It is one of the most important parameters for both hydraulic calculation and hydrodynamic model. Due to the lack of observation data, Manning's roughness coefficient in the study region is determined by topographic, channel constraints and empirical analysis.
(3) Flooding and drying threshold parameters The flooding and drying scheme is an approach of MIKE 21 HD FM for handling problems related to moving boundaries, the flooding and drying fronts. There are three parameters: drying depth, flooding depth and wetting depth. When the water depth of a unit is less than the wetting depth, the flow calculation of that unit is adjusted accordingly. If the water depth is less than the drying depth, the corresponding unit is frozen and is not calculated until it is submerged again. The determination of these parameters depends on the research region.
Calibration is a process of modifying model parameters to reduce the error between the simulated and observed streamflow. In this study, a manual way of model calibration is practiced. In manual calibration, a trialand-error parameter adjustment is performed, based on a visual judgment by comparing the measured and the simulated water level and discharge. After multiple manual calibrations, a final calibration was conducted with minor adjustments.
For the MIKE 11 model, calibration was performed using Manning's roughness coefficient. Initially, the model was simulated using the default value of Manning's roughness coefficient (n=0.033). During the calibration process, Manning's roughness coefficient was calibrated to a uniform value (0.025∼0.04) to achieve the best agreement between observed and simulated water levels and discharges at the downstream Yiganqi station. To assess the calibration results, the maximum water level error, relative error of the maximum discharge, and graphical analysis of simulated and observed runoff agreement were conducted.
In the MIKE 21 model, three parameters were calibrated: Manning's roughness coefficient, flooding and drying threshold parameters, and synthesis loss. For determining the Manning's roughness coefficient outside the riverbed, digital maps were utilized. These maps contained information on land use, topography, and landforms The Manning's roughness coefficient corresponding to different surface features was determined based on the Hydraulic Manual [34]. Drying, flooding, and wetting values of 0.05, 0.08, and 0.1, respectively, were adopted according to the DHI standards. However, during the calibration process, it was observed that regardless of parameter adjustments, the flooding persisted and covered the entire study area, which did not reflect the actual conditions. Considering the significant influence of evaporation and infiltration, which could not be calculated separately due to data limitations, the synthesis loss parameter was introduced to address this issue. For detailed information on the determination of this parameter, please refer to section 3.3 (1).

Calculation scenarios
According to the field research, due to the unique flood characteristics and underlying conditions of the Yarkant River, the flood risk pattern varies with different flood magnitudes. During a sustained low flow rate flood event, characterized by a flow rate of approximately 2000 m 3 /s, the flood progresses downstream within the deep channel of the river without completely occupying the entire channel. Consequently, the flood velocity increases significantly, along with the force of impact and scouring, which pose a heightened risk of eroding the base of permanent embankments and ultimately resulting in embankment failure. According to erosion tests conducted on the Yarkant River basin, even if a permanent embankment is buried as deep as 19 m, it can still be breached by small floods. On the contrary, when the flood discharge is higher, the flood will occupy the entire river channel, and the scouring forces on both banks will decrease, thus not causing catastrophic failures of permanent flood control structures. Considering this characteristic of the area, apart from the overflow scenarios, the dike-break scenarios were also considered to evaluate flood risks and losses in this research.
The flood risk in the study area is mainly caused by the propagation of floods in the upper region at the head of the Kaqun canal and dike breaking. In this paper, nine flood calculation scenarios are formulated, including five overflow scenarios plus a real year (1999) overflow scenario, and three dike-breaking scenarios. The downstream boundary condition for both overflow scenarios and dike-breaking scenarios is water leveldischarge relation at the head of canal at the middle reaches. The Altash Reservoir was constructed in the upstream main stem of the river in the year 2020. This reservoir plays a vital role in regulating natural flood processes, leading to an impact on downstream flood risk. According to the investigation, the Altash Reservoir is designed to withstand a flood with a return period of 1000 years. When the return period of the inflow flood is less than or equal to 20 years, the reservoir will release water at the current safe discharge of 1750 km 3 /s (2.5-year return period). When the return period of the inflow flood exceeds 20 years, the reservoir will control the outflow based on a 10-year return period flood (4272 km 3 /s). Based on this, in scenarios considering the influence of the Altash Reservoir, for floods with return periods of 5 years, 10 years, and 20 years, the Altash Reservoir will release water at a rate of 1750 km 3 /s. No overbank flooding will occur in the study area. For floods with a return period of more than 20 years (such as 50-year, 100year return period floods), the Altash Reservoir will release water based on a design flood with a return period of 10 years (peak flow of 4272 km 3 ). In this case, the flood risk in the study area will be the same as that during a 10year return period flood without considering the reservoir's influence.

Flood risk mapping and loss assessment 3.5.1. Flood risk mapping
A qualitative method is adopted for computing the flood risk [35,36]. In the method, a hydrodynamic model is used to obtain the flood risk features. A flood risk classification system can be developed based on the calculated risk features. Table 3 shows the classification of flood risk, where five different classes are presented based on the flood depth.

Flood loss assessment
Given the results in the hydrodynamic model, the impact of flooding in the study area is analyzed. Based on the GIS tools, the submerged surface layers are superimposed with land use/land cover layers via a spatial When floods with a 10-year return period occur, two sides overflow OF_03 When floods with a 20-year return period occur, two sides overflow OF_04 When floods with a 50-year return period occur, two sides overflow OF_05 When floods with a 100-year return period occur, two sides overflow OF_1999 Flood in 1999 under current conditions(real year) Dike-break DB_01 The eastern and western dikes burst DB_02 The eastern dike bursts DB_03 The western dike bursts Extreme high geographic relation to obtain the submerged objects. On this basis, combining with socioeconomic data and flood features such as flood depth and flood duration, the flood loss ratio is utilized to compute the flood loss. Flood loss contains direct losses and indirect losses. The former, known as asset loss, is caused by the physical contact of floodwater with damageable property. The latter, results from the interruption of social production and activities caused by flooding, can be regarded as output loss. Both of them are related to the flooding inundation characteristics such as inundation area and depth. In this study, we compute the flood loss from different aspects including housing and family property, agriculture, industry, business and road. Specifically, the asset loss contains residential housing loss, family property loss, agricultural loss, industrial asset loss, business asset loss, highway loss and railway loss. The output loss contains industrial output loss and business main income loss.
For the asset loss calculation, the method proposed by Penning-Rowsell is used [37]. The asset loss is computed by where V ij is the assets in category i under depth j, η ij is the loss rate in category i under depth j, D A i is the asset loss in category i, and D A is the total asset loss. The output loss is assumed to be related to the submerged duration, which can be calculated by å åå where L ij is the annual output value in category i in the area with submerged duration j, T ij is the submerged duration of L ij , Y is the time of one year, D O i is the output loss in category i, and D O is the total output loss. Consequently, the flood loss D is computed by Figure 4 depicts the data flow of the flood loss assessment, where A represents agricultural output, R represents residential property, I represents industry assets, and B represents business assets.

Design flood hydrograph for the overflow scenarios
The design flood hydrographs of Kaqun canal, which contains 5-, 10-, 20-, 50-, 100-year flood hydrographs, are used as upstream boundary conditions for the overflow scenarios. In this study, the design peak discharges and

Model calibration and validation
The Yiganqi hydrological station, which is located between the head of the Kaqun canal and the head of the canal at the middle reaches, is selected as the reference station. The model is calibrated by the flood discharge and water level data in the years 1971 and 2010 at Yiganqi hydrological station, and then is validated by the data in the year 2012. Three parameters are adjusted to reduce the gap between the observed and simulated discharges and water levels at the reference station during the process of calibration, namely the synthesis loss, Mannings roughness coefficient and drying/flooding/wetting depth.
The observed discharge data at the Kaqun hydrological station (upstream) and the Yiganqi hydrological station (downstream) in 1971, 2010, 2012 is selected to calculate the synthesis loss, and the result is 140 mm/d. The channel roughness in the MIKE 11 model is set between 0.025 and 0.04 in the calculated sections, and the roughness in the MIKE 21 model is determined by technical rules and the Hydraulic Manual [34] (table 5). The drying depth, flooding depth and wetting depth are set as 0.05 m, 0.08 m, and 0.1 m, respectively. After calibration, the water level and discharge data of a typical flood in 2012 are used for validation. The independent flood events for each year for calibration and validation. The temporal resolution of the model simulation was at a daily scale, allowing us to capture the daily variations in streamflow. The spatial resolution of the model simulation was determined by the grid size, with a maximum grid area not exceeding 0.05 km 2 . This resolution was chosen to capture the local variations in topography and channel characteristics within the study area. Figure 6 shows the results of calibration and validation. Table 6 demonstrates a high level of agreement between the observed and simulated discharges and water levels for both calibration and validation of the MIKE11 model. The results of the validation errors show that the absolute value of the maximum water level error at the Yiganqi station is 0.03 m, and the absolute value of the maximum discharge relative error is 2.84%, which are very satisfactory.  The inundation areas for different return periods are compared in figure 8. The inundation area rises with the increase of the return period, but the growth rate of the inundation area declines gradually. In the overflow   Considering the distribution of flood water depth for a 100-year return period, two isolated islands are formed in the flood depth map. To verify the rationality of the results, the inundation layers are superimposed in Google Earth. A topographic profile map is extracted to analyze the topographic relationship between the noninundated locations and the surrounding inundated area, as shown in figure 9. The elevation of noninundated locations is higher than that of the nearby inundated area, which verifies the 2D calculation results.

Flood risk mapping
The flood risk maps for the overflow scenarios are presented in figure 10  The proportion of land with high and very high risks also shows an upward trend. It can also be observed from figure 10 that low-risk areas gradually become medium-risk or high-risk areas when the return period rises.
The flood risk maps for the dike-break scenarios are presented in figure 11 and the comparison of areas rated at different risk are shown in table 8. The total risk areas from eastern-and-western-dikes breaking, the easterndike breaking and the western-dike breaking are 119.64 km 2 , 77.99 km 2 and 42.29 km 2 , respectively. The risk  areas caused by DB_01 almost equals the sum of those in DB_02 and DB_03. In scenario DB_01, the area with high risk accounts for 19.24% of the total, and the area with very high risk is 19.24%. DB_02 and DB_03 almost have the same proportion of areas related to different risks as that of DB_01. It can be seen in figure 11 that the areas with high and very high risks of three dike-break scenarios are all located along the river.

Flood loss assessment
The flood loss rate is determined by the recommended data from China Institute of Water Resources and Hydropower Research and statistics data for historical floods in the surrounding area of the Yarkant River. Details on the relationship between the flood loss rate and flood depth can be found in table 9. The comparison of flood losses for different scenarios is shown in figure 12. In the overflow scenario, the total loss shows an upward trending with the increase of return period. The loss caused by the flood happened in 1999 lies between the losses caused by the 20-year flood and the 50-year flood. In the scenarios for floods due to dike failure, the loss caused by both eastern and western dikes breaking accounts the most of the total loss. The loss in DB_01 approximately equals to the sum of the losses in scenarios DB_02 and DB_03. Among all the scenarios, the maximum total loss is 177.6 million yuan of scenario OF_05. The residential housing loss is 34.78 million yuan, which accounts for the largest proportion (23.6 %) of the total loss. For OF_01, OF_02, OF_03 and OF_1999 scenarios, the agricultural loss makes up the largest proportion. For OF_04, OF_05 and the dike-break scenarios, the residential housing loss constitutes the largest proportion. It can be concluded that agricultural loss is the most substantial loss among all losses caused by small overflow floods, while housing loss is the most severe loss of all losses generated by large-magnitude overflow floods and dike-break floods.

Discussion and conclusion
The present study aims to compile flood risk maps to identify areas prone to flooding and to evaluate losses in the GLOF-dominated areas with intense human impacts. According to on-site investigations, the flood risk patterns vary depending on the magnitude of floods due to the unique flooding and surface conditions in this area. To address this issue, two typical scenarios, namely overflow and dike-break scenarios, were adopted to evaluate flood risks and losses. The MIKE FLOOD model is used to compute risk indicators such as the inundation extent and flood depth for different scenarios in the reach between the head of the Kaqun canal and the head of the canal at the middle reaches in the Yarkant River basin. The method based on water balance theory is adopted to  a The data in parentheses represents the percentage of the area of a certain risk level to the total area of all risks. compute synthesis loss during the flood propagation simulation. The simulation results show that the method performs satisfactorily incalculating the evaporation and infiltration loss during the inundation process. The simulated flood risk indicators can be used for flood risk mapping and socioeconomic loss assessment. For the overflow scenarios, the flood risk and economic loss rise with the increase of the return period. For the dike-break scenarios, the flood risk and economic loss caused by the simultaneous burst of eastern and western dikes are almost the same as the sum of the flood risk and economic loss caused by the respective burst of the eastern dike and the western dike. Through the analysis of various scenarios, the study area is capable of controlling floods only with return periods less than 5 years. Thus, structural and nonstructural measures are required in this area for flood protection. It is observed that the loss is proportional to the flood magnitude and that the residential housing loss and agricultural loss are usually large compared to other losses.
Finally, the study shows the potential for responsible authorities and decision makers to use flood risk maps and flood loss assessment to reduce the negative impacts of floods. Although this study conducted in a small area of the Yarkant River basin with data scarcity, it can provide ideas and preliminary approaches for the future studies in many other similar regions.