A method for simulating karst runoff by coupling SWAT model with antecedent precipitation index

Abstract

The application relates to the field of karst basin runoff simulation calculation, and particularly discloses a karst SWAT model runoff simulation method coupled with an early-stage precipitation index, which comprises the following steps: step one, establishing a SWAT model with the characteristics of a karst area, generating a land use map with the characteristics of the karst area and hydrological response units (HRUs) with the distribution characteristics of the karst and non-karst; step two, modifying the source code of the SWAT model to construct a K-SWAT model with the function of karst runoff generation and concentration; step three, embedding a normalized early-stage precipitation index (MIWAPI) in the K-SWAT model to construct a K-W-D-SWAT model with the function of identifying the effect of dry and wet events on the karst runoff generation and concentration process; and step four, optimizing the parameters of the SWAT model, the K-SWAT model and the K-W-D-SWAT model by using the SUFI-2 algorithm in the SWAT-CUP software, and evaluating the runoff simulation level of each model in the karst area. The karst SWAT model runoff simulation method coupled with the early-stage precipitation index is suitable for the simulation of the runoff process in the karst area with complex climate change, and can improve the simulation accuracy and rationality of the hydrological process in the karst area.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of karst basin runoff simulation calculation, and particularly relates to a karst SWAT model runoff simulation method coupled with a previous precipitation index. BACKGROUND

[0002] About 25% of the world's population uses water from karst groundwater. With the intensification of climate change and the enhancement of human activities, as well as the unique hydrogeological structure and frequent alternation of dry and wet events in karst areas, the regional hydrological process is becoming more and more complex, leading to frequent, expanding and intensifying flood disasters in karst areas, which seriously threatens water safety and the lives and property of local residents. Therefore, accurately identifying the effects of frequent alternation of dry and wet events on karst hydrological processes and achieving accurate simulation of runoff in dry and wet seasons throughout the year are crucial for improving flood warning capabilities and water resource management levels in karst areas, and for mitigating flood disasters and hydrological droughts. This is also the focus of research by scholars in recent years.

[0003] Due to the difficulty in obtaining comprehensive data on the hydrogeological and geometric characteristics of karst systems, most previous karst hydrological studies have not fully described the overall characteristics and spatial distribution differences of karst hydrological systems. In particular, in highly complex karst areas, neglecting the overall characteristics of karst can lead to unreasonable deviations in the water yield contribution of the runoff generation system. Secondly, due to the large spatial variability of rainfall and karst distribution characteristics, the frequent alternation of dry and wet events may become a key factor affecting model performance. In seasonal rivers, the model often overestimates low flow and underestimates high flow, reflecting the sensitivity of the runoff generation system to seasonal changes and its inadequate dynamic response capacity. Although previous studies have improved model accuracy by separating and recombining simulations at the dry and wet seasonal scales, they still face the problems of parameter staticization and time discontinuity of simulated elements. In addition, the meteorological and hydrological characteristics of strongly developed karst areas are significantly different from those of karst areas, especially in terms of frequent alternation of extreme dry and wet events. Therefore, when studying karst hydrological processes, the combined effects of karst system structure and climate change on hydrological processes must be considered.

[0004] Runoff simulation is a key technical means for regional flood warning and water resource management. However, for karst areas with frequent alternation of dry and wet events, most models have not fully considered the effects of climate conditions (such as precipitation characteristics) on the variation process of runoff generation and concentration capacity, and the static nature of model parameters limits the dynamic response capacity of the runoff generation and concentration process, restricting the further development of hydrological models. Therefore, developing a karst hydrological model that can accurately identify dry and wet events to enhance the dynamic response capacity of the model to spatial characteristics and climate change in the basin, and achieve accurate simulation of karst basin hydrological processes throughout the dry and wet seasons, is a key problem that needs to be solved in the field of hydrological simulation. SUMMARY

[0005] The present application aims to at least solve one of the above-mentioned technical problems, provide a karst SWAT model runoff simulation method coupled with the antecedent precipitation index, which is suitable for runoff process simulation in complex karst areas under climate change. The method breaks through the time and space scale limitation of fixed parameters in traditional hydrological models, and improves the simulation accuracy and rationality of hydrological processes in karst areas.

[0006] In order to achieve the above-mentioned purpose, the technical scheme adopted by the present application is as follows: a karst SWAT model runoff simulation method coupled with the antecedent precipitation index, comprising the following steps:

[0007] Step one, establish a SWAT model with characteristics of karst area, determine the surface and underground boundaries of the basin according to the basin boundary vector diagram, hydrogeological map and terrain elevation data, use GIS technology to superimpose hydrogeological map, land use map and depression distribution map, generate land use map with karst characteristics, and combine soil data and slope data to generate hydrological response unit HRUs with karst and non-karst distribution characteristics;

[0008] Step two, modify the source code of SWAT model, build K-SWAT model with karst runoff function, K-SWAT model includes surface karst zone, karst matrix and karst pipeline three systems;

[0009] Step three, on the basis of K-SWAT model, embed normalized antecedent precipitation index MIWAPI, build karst K-W-D-SWAT model with dry and wet effect identification mechanism, to accurately identify the effect of dry and wet events on karst runoff process, realize the simulation of basin response in dry and wet seasons all year round;

[0010] Step four, use SUFI-2 algorithm in SWAT-CUP software to optimize the parameters of SWAT model, K-SWAT model and K-W-D-SWAT model, and evaluate the runoff simulation level of each model in karst area through coefficient of determination R 2 , Nash efficiency coefficient NSE, percentage bias PBIAS and Kling-Gupta efficiency coefficient KGE.

[0011] Preferably, the water balance equation of the surface karst zone system is:

[0012]

[0013] The water balance equation of the karst matrix system is:

[0014]

[0015] The water balance equation of the karst pipeline system is:

[0016]

[0017] wherein, E H is the water level of epikarst zone; P r，i is the recharge of epikarst zone on the i-th day; Q sec is the triggered secondary spring flow of epikarst zone; Q EH is the delayed outflow; Q sepbtm,i is the recharge of epikarst zone to matrix on the i-th day; Q F,i is the recharge of epikarst zone to karst conduit on the i-th day; E E,T is the evaporation of epikarst zone system; M is the water level of matrix system; Q s，i is the diffusion recharge of epikarst zone to karst matrix system on the i-th day; Q sc is the exchange water amount of conduit to matrix system; Q SM is the delayed outflow of matrix system; E T，1 is the evaporation of matrix system; Q FC is the rapid outflow of conduit; E T，2 is the evaporation of conduit system.

[0018] Preferably, the formula for describing the dry-wet event is as follows:

[0019] Based on the weighted average value WAPI, IWAPI of the previous precipitation index to identify the regional dry-wet event, the calculation formula of WAPI, IWAPI is as follows:

[0020]

[0021]

[0022] The above formula is normalized to scale IWAPI to [-1, 1] to show the intensity of drought and flood:

[0023]

[0024] Based on MIWAPI, the description parameter of runoff yield capacity is adjusted, and the specific adjustment method is shown in the following formula:

[0025]

[0026] wherein, P a is the annual average rainfall; ω = 1 if the model parameters have a promoting effect on the runoff yield process, otherwise ω = -1, Kopt is the number of days affected by the previous precipitation; N opt is the decay coefficient; i is the simulation time; j represents the HRU number; Parm origin is the original parameter allocated by SUFI-2 algorithm

[0027] Preferably, in the step three, the initial interception rate λ in the SCS-CN model for calculating the surface runoff in the SWAT model is dynamically adjusted, and the calculation formula is:

[0028]

[0029] I a = λS (26)

[0030] In the formula, Q surf is the surface runoff; P day is the daily precipitation; I a is the initial loss; S is the maximum interception; and λ is the initial loss rate, which is defined as 0.2 in the SWAT model.

[0031] Preferably, in the step four, the coefficient of determination R 2 , the Nash efficiency coefficient NSE, the percentage bias PBIAS, and the Kling-Gupta efficiency coefficient KGE are used to evaluate the runoff simulation level of each model in the karst area, and the calculation formula is:

[0032]

[0033]

[0034] Wherein Q0 and Q s are the observed runoff and the simulated runoff respectively, indicate the average values of the observed runoff and the simulated runoff respectively, n is the length of the sequence, r indicates the correlation coefficient, α indicates the ratio of the standard deviation of the simulated runoff to the observed runoff, β indicates the ratio of the average simulated runoff to the observed runoff; m is the ranking of the sample in descending order, and P is the frequency of the flow exceeding a certain intensity in the observation period

[0035] Preferably, in the step four, the flow duration curve FDC is further used to evaluate the runoff simulation level of each model in the karst area.

[0036] Beneficial effects are that, compared with the prior art, the karst SWAT model runoff simulation method coupled with the antecedent precipitation index has the following advantages:

[0037] 1. The karst SWAT model runoff simulation method coupled with the antecedent precipitation index breaks through the time and space scale limitation of the fixed parameters in the traditional hydrological model, improves the simulation level and rationality of the hydrological process in the karst area, and is suitable for runoff process simulation in the complex karst area with climate change. By embedding the karst system into the SWAT model, the K-SWAT model is proposed, and the normalized antecedent precipitation index MIWAPI is introduced, and the K-W-D-SWAT model is further proposed. Compared with the traditional SWAT model, the K-SWAT and K-W-D-SWAT models not only significantly improve the accuracy of runoff simulation in the karst area, but also can truly reflect the runoff process in the karst area.

[0038] 2. The K-W-D-SWAT model proposed in the application realizes dynamic updating of parameters in time and space scale by coupling the normalized antecedent precipitation index MIWAPI, so that the runoff process can respond to the conversion of dry and wet effects in real time, and the limitation of single time and space scale of parameters in the traditional hydrological model is broken through.

[0039] 3. The K-W-D-SWAT model proposed in the application significantly improves the runoff simulation accuracy in the wet and dry periods, solves the problem of underestimation of high flow by SWAT and overestimation of low flow by K-SWAT.

[0040] 4. The K-W-D-SWAT model proposed in the application is driven by climate conditions, and is also applicable to research areas with different hydrological conditions in theory. BRIEF DESCRIPTION OF DRAWINGS

[0041] The specific embodiments of the application will be further described in detail below with reference to the accompanying drawings, in which:

[0042] Figure 1 The flow chart of the karst SWAT model runoff simulation method coupled with the antecedent precipitation index of the application;

[0043] Figure 2 The Diaojing River Basin profile schematic diagram in the specific embodiment of the application;

[0044] Figure 3 The multi-year monthly average rainfall and runoff process of the two catchment areas of Hekou station and Malong station;

[0045] Figure 4 The runoff simulation results of the SWAT, K-SWAT and K-W-D-SWAT models at Hekou station and Malong station;

[0046] Figure 5 The FDC results of the runoff simulation of the SWAT, K-SWAT and K-W-D-SWAT models at Hekou station and Malong station. DETAILED DESCRIPTION

[0047] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings of the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments of the present application, all the other embodiments obtained by a person of ordinary skill in the art without creative effort belong to the scope of protection of the present application.

[0048] It should be noted that when a component is referred to as being "fixed" to another component, it can be directly on the other component or intervening components can also be present. When a component is referred to as being "connected" to another component, it can be directly connected to the other component or intervening components can also be present. When a component is referred to as being "disposed" on another component, it can be directly on the other component or intervening components can also be present, and when a component is referred to as being "disposed in the middle", it is not only disposed in the middle position, but also disposed within the range defined by the middle. The terms "vertical", "horizontal", "left", "right", and similar expressions used herein are for illustrative purposes only.

[0049] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used in the description of the application herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. As used herein, the term "and / or" includes any and all combinations of one or more of the associated listed items.

[0050] As shown in Figure 1 The application discloses a karst SWAT model runoff simulation method coupled with a previous precipitation index, and comprises the following steps:

[0051] Step one, establish a SWAT model with karst region characteristics, determine the surface and underground boundaries of the basin according to the basin boundary vector diagram, hydrogeological map and terrain elevation data, generate a land use map with karst characteristics by superimposing the hydrogeological map, land use map and depression distribution map by using GIS technology, and generate hydrologic response units (HRUs) with karst and non-karst distribution characteristics by combining soil data and slope data;

[0052] Step two, modify the source code of the SWAT model, and construct a K-SWAT model with karst runoff function, wherein the K-SWAT model comprises a surface karst zone, a karst matrix and a karst pipeline system;

[0053] Step three, on the basis of K-SWAT model, embed normalized antecedent precipitation index MIWAPI, build karst K-W-D-SWAT model with dry and wet effect identification mechanism, to accurately identify the effect of dry and wet events on karst production and concentration process, realize the response simulation of dry and wet seasons in the whole year;

[0054] Step four, the SUFI-2 algorithm in the SWAT-CUP software is used to optimize the parameters of the SWAT model, the K-SWAT model and the K-W-D-SWAT model, and the determination coefficient R 2 , Nash efficiency coefficient NSE, percentage bias PBIAS and Kling-Gupta efficiency coefficient KGE are used to evaluate the runoff simulation level of each model in the karst area.

[0055] As Figure 2 shown, the present application takes the typical karst basin of Diaojiang in southwest China as an example for specific description, including the following steps:

[0056] Step one, build the SWAT model project of karst distribution characteristics of Diaojiang basin. Based on the basin boundary vector map, hydrogeological map and terrain elevation data, the surface and underground boundaries of the basin are determined. By using GIS technology to superimpose the hydrogeological map, land use map and depression distribution map, the land use map with karst distribution characteristics is generated, and the soil data and slope data (divided into 0-15°, 15-30°, 30-45° and >45° four grades) are superimposed, 579 hydrological response units (HRUs) are generated for Diaojiang basin using a combination threshold of 15%, of which 441 are karst HRUs and 138 are non-karst HRUs;

[0057] Step two, based on Visual studio 2019 compilation software, the SWAT model is developed again, the development contents mainly include integrating karst structure into the source code of SWAT model, and adding the identification and judgment function of karst HRU and non-karst HRU, and finally recompiling the source code after development, and generating an executable file, that is, K-SWAT model with karst production and concentration function, K-SWAT model includes surface karst zone, karst matrix and karst pipeline three systems.

[0058] a) Surface karst zone system

[0059] The recharge of the surface karst zone comes from the direct infiltration recharge of the bottom of the soil profile or bare surface, and its water balance equation can be expressed as:

[0060]

[0061] where E H is the water level in the epikarst zone, mm; P r,i is the water recharge in the epikarst zone on the i-th day, mm; Q sec is the triggered secondary spring discharge in the epikarst zone, mm; Q EH is the delayed discharge in the epikarst zone, mm; Q sepbtm，i is the recharge from the soil profile bottom or bare land surface to the epikarst zone on the i-th day, mm; Q F，i is the recharge from the soil profile bottom or bare land surface to the epikarst zone on the i-th day, mm; E E，T is the evaporation, mm.

[0062] Considering the time lag effect of soil or bare land surface recharge to the epikarst zone, an exponential decay weighting function is used to express:

[0063]

[0064] where P prec，i is the recharge from the soil profile bottom or bare land surface to the epikarst zone on the i-th day, mm; δ is the time required for water to leave the soil profile bottom or bare land surface to reach the epikarst zone, day.

[0065] In the epikarst reservoir, a nonlinear lag transfer function can be used to describe the delayed outflow, and the hydrological connectivity of the epikarst zone is represented by the change in reservoir water level ε eh :

[0066]

[0067] where ε eh is the hydrological connectivity of the epikarst zone; E H，1 is the threshold water depth for the epikarst zone to have delayed outflow, mm; E H，2 is the threshold water depth for the epikarst spring to have secondary flow, mm.

[0068] The nonlinear time-lag outflow Q eh of the epikarst zone is calculated in combination with the hydrological connectivity of the epikarst zone ε EH :

[0069]

[0070] where k eh is the outflow coefficient of the epikarst zone to have delayed outflow; α eh is the exponential coefficient of the epikarst zone delayed outflow.

[0071] When the water level E H reaches a certain threshold water level E H,2 , secondary flow Q sec will occur from the epikarst spring, which can be described as:

[0072]

[0073] where k sec The outflow coefficient of the secondary flow triggered by the epikarst zone.

[0074] For the water recharge process from the epikarst zone to the karst matrix system, the exponential decay weighting function applied to the groundwater response model of the SWAT model is used to describe the lag effect of the karst matrix receiving water recharge from the epikarst zone:

[0075]

[0076] where δ s represents the time required for water to leave the bottom of the epikarst zone and reach the karst aquifer zone, day.

[0077] The process of water infiltration from the epikarst zone to the pipe reservoir is linear and rapid recharge, and only the area of underground river opens this module for calculation:

[0078] Q F,i = k f × (max [E H -E H,1 , 0]

[0079] where k f is the coefficient of water transfer from the epikarst zone to the pipe system.

[0080] b) Karst matrix system and karst pipe system

[0081] For the karst matrix system, the water balance equation is:

[0082]

[0083] where M is the water level of the matrix reservoir, mm; Q s,i is the diffusion recharge amount of the epikarst zone to the karst reservoir on the i-th day, mm; Q sc is the exchange water amount transported by the pipe to the matrix reservoir, mm; Q SM is the delayed outflow of the matrix reservoir, i.e. slow flow, mm; E T，1 is the evaporation amount of the matrix reservoir, mm.

[0084] The delay of water in the epikarst zone and the karst matrix recharge process, the exponential decay weighting function is used to describe the lag effect of the karst matrix receiving water infiltration Q sepbtm，i from the epikarst zone:

[0085]

[0086] where δSM delay coefficient of the epikarst zone, day.

[0087] The outflow of the matrix reservoir is characterized by a linear storage-discharge relationship:

[0088]

[0089] where δ M delay coefficient of the matrix outflow, day.

[0090] Considering the water exchange Q SC between the matrix and conduit reservoirs, the system water level difference function is defined as:

[0091]

[0092] where k sc is the exchange coefficient between the matrix and conduit; and α SC is the exchange exponent between the matrix and conduit.

[0093] C is the water level of the conduit reservoir, mm.

[0094] For the karst conduit system, the water balance is shown as follows:

[0095]

[0096] where Q FC is the rapid outflow of the conduit, mm; and E T，2 is the evaporation of the conduit reservoir, mm.

[0097] Unlike the matrix system, the rapid outflow process of the conduit system is represented by an exponential function:

[0098]

[0099] where k FC is the outflow coefficient of the conduit flow; α FC is the specific flow coefficient of the conduit flow; and C H is the threshold water depth of the conduit, mm.

[0100] Step three, on the basis of K-SWAT model, embed normalized MIWAPI of antecedent precipitation index into K-SWAT model source code to integrate dry and wet effect feedback mechanism, use Visual studio 2019 software to compile the developed source code, regenerate the executable file, that is, build K-W-D-SWAT model with dry and wet effect recognition mechanism to accurately identify the effect of dry and wet events on karst runoff concentration process, realize the simulation of basin response in dry and wet seasons all year round, specifically, the weighted average value WAPI and IWAPI of antecedent precipitation index are used to identify regional dry and wet events, and the calculation formula of WAPI and IWAPI is as follows:

[0101]

[0102] The formula is normalized to make IWAPI shrink to [-1, 1] to show the intensity of drought and flood:

[0103]

[0104] Based on MIWAPI, the runoff yield capacity description parameter is adjusted, and the adjustment method is shown in the following formula:

[0105]

[0106] In the formula, P a is the annual average rainfall, IWAPI ranges from-1 to 1, if the model parameter has a promoting effect on runoff yield process, ω=1, otherwise-1, Kopt is the number of antecedent days, day; N opt is the recession coefficient, i represents the simulation time, day; j represents the HRU number; Parm origin is the original parameter allocated by SUFI-2 algorithm.

[0107] In addition, the initial interception rate λ (which has an inhibitory effect on runoff yield process) in the SCS-CN equation used by SWAT model to calculate surface runoff is dynamically adjusted, and the core formula of SCS-CN is as follows:

[0108]

[0109] I a = λS

[0110] In the formula, Q surf is the surface runoff, mm; P day is the daily precipitation, mm; I a is the initial loss, mm; S is the maximum interception, mm; λ is the initial loss rate, which is defined as 0.2 in SWAT model.

[0111] Step four, the SUFI-2 algorithm in the SWAT-CUP software is used to optimize the parameters of the original SWAT model, the K-SWAT model and the K-W-D-SWAT model. The coefficient of determination R 2 , the Nash efficiency coefficient NSE, the percentage bias PBIAS and the Kling-Gupta efficiency coefficient KGE are used to evaluate the simulation performance of the model in simulating different magnitude of flow levels (runoff in wet and dry periods). The correlation calculation formula is as follows:

[0112]

[0113] Wherein, Q0, Q s are the observed runoff and the simulated runoff, respectively, n is the length of the sequence, r represents the correlation coefficient, a represents the ratio of the standard deviation of the simulated runoff to the observed runoff, b represents the ratio of the average simulated runoff to the observed runoff; m is the ranking of the sample in descending order, and P is the frequency of the flow exceeding a certain intensity in the observation period.

[0114] To verify the applicability of the method, the present application takes a typical karst basin in southwest China, the Diaojiang Basin, as an example, uses the rainfall stations, weather stations and underlying surface terrain data in the basin to build a model, and selects the runoff observation data of the Hekou Station and the Malong Station as the verification object of the model. The time setting of the three versions of the SWAT model is: 2007 is the preheating period, 2008-2015 is the calibration period, and 2016-2020 is the verification period. The SUFI-2 algorithm is used to calibrate the sensitive parameters of the three models, and the coefficient of determination R 2 , the Nash efficiency coefficient NSE, the percentage bias PBIAS and the Kling-Gupta efficiency coefficient KGE are used to evaluate the model accuracy. At the same time, the simulation period is divided into hydrological wet and dry periods, and the performances of the indicators of the simulation results of the models in different hydrological stages and the flow duration curve FDC results are compared.

[0115] The Diaojiang River is a first-order tributary of the Hongshui River in the Xijiang River system of the Pearl River Basin, with a total length of 229 km, an elevation range of 80-1148 m, a large terrain fluctuation of 200-500 m, and a basin control area of about 3601 km 2 . The region belongs to a subtropical monsoon climate, with an average annual temperature of 15.6-22.9°C. Influenced by tropical cyclones, the Diaojiang Basin has abundant rainfall, but due to the complex terrain and the development of karst landforms, the temporal and spatial distribution of rainfall is extremely uneven. The rainfall from May to September each year accounts for 70% of the total annual rainfall, which easily leads to flood disasters, and the rainfall in the remaining months accounts for only 30%, which easily leads to drought and water shortage problems. The average annual precipitation in the basin is 1400-1700 mm, floods occur frequently, and the average annual runoff is as high as 2.43 billion m3 The general situation of the Diaojiang River Basin is shown in FIG. 1. Figure 2

[0116] The basic data for constructing the SWAT model project in the present application include DEM topographic elevation, land use type, soil type, hydrogeological map, precipitation, maximum and minimum air temperature, relative humidity, average wind speed, sunshine duration, and measured runoff data, and the data-related information is shown in Table 1.

[0117] Table 1 Data properties used in the present application

[0118]

[0119] The present application adopts the daily precipitation data of 14 rainfall stations in the Diaojiang River Basin, and the daily maximum, minimum temperature, relative humidity, sunshine duration, and average wind speed data of 6 meteorological stations, with a time span of 2007-2020, for driving the three hydrological models of SWAT, K-SWAT, and K-W-D-SWAT to simulate runoff. At the same time, the daily runoff data of the estuary station and Malong station from 2008 to 2020 are used to calibrate and verify the three models of SWAT, K-SWAT, and K-W-D-SWAT. The data of the rainfall stations and the hydrological stations are derived from the China Hydrology Yearbook, and the meteorological station data is obtained from the China Meteorological Data Network. The relevant data details are shown in Table 2.

[0120] Table 2 Hydro-meteorological station information of the study area

[0121]

[0122] The parameters are crucial to the runoff simulation results of the model, so parameter calibration is needed. In this study, the SUFI-2 algorithm is used to calibrate the parameters of the three models of SWAT, K-SWAT, and K-W-D-SWAT. Table 3 lists the initial ranges of the sensitive parameters of the three models.

[0123] Table 3 Description and range of calibration parameters of the model

[0124]

[0125] (R_) represents the relative change of the parameter, where the current value is multiplied by 1 plus a factor; (*) represents a positive effect on runoff generation of the confluence, i.e. w = 1, (**) represents w = -1.

[0126] ​The optimal values and sensitivity ranking of parameters for SWAT, K-SWAT, K-W-D-SWAT at Hekou and Malong stations are listed in Table 4. For SWAT, the results of parameter calibration at both stations show that CH_N2 is the most sensitive parameter, followed by CN2, CH_K2, ALPHA_BNK, GWQMN and GWQMN. The main sensitive parameters of K-SWAT are a_sc, h_(M,0), d_s, GW_DELAY, E_(H,2) and GW_REVAP, highlighting the significant influence of karst structure on hydrological processes. For K-W-D-SWAT, CN2 is one of the main sensitive parameters at Hekou station, indicating that there is a certain proportion of non-karst area in the catchment area controlled by this station, and the surface structure still plays an important role in the process of runoff generation and concentration. N_opt as the most sensitive parameter at Malong station further confirms the significance of dry-wet events in karst areas.

[0127] Table 4 Optimal parameter values of three models in the catchment area controlled by two hydrological stations

[0128]

[0129]

[0130] Table 5 and Figure 4 The runoff simulation results of SWAT, K-SWAT, K-W-D-SWAT at Hekou and Malong stations are shown. The original SWAT model has low accuracy in runoff simulation at both stations, especially at Malong station, with R 2 and NSE less than 0.60. The simulation performance of K-SWAT, which considers the structure of karst system, is significantly improved: R 2 and NSE are higher than 0.70, increasing by 0.10-0.27 and 0.11-0.18 respectively compared with the original SWAT. However, the results of PBIAS and KGE show that K-SWAT model has a serious overestimation of runoff, especially at Malong station with rich karst distribution (PBIAS: -45.1% in calibration period and -49.2% in validation period; KGE: 0.54 in calibration period and 0.50 in validation period). This bias may be due to the frequent alternation of dry-wet events in karst areas, making it difficult for K-SWAT model to coordinate the runoff simulation in wet and dry seasons. In contrast, K-W-D-SWAT model, which incorporates MIWAPI index, significantly improves the runoff simulation performance at both stations, outperforming SWAT and K-SWAT models in both calibration and validation periods, and the reduction of PBIAS indicates that K-W-D-SWAT effectively reduces the overestimation of low flow by K-SWAT. In addition, the improvement of PBIAS and KGE at Malong station is particularly significant, indicating that dry-wet events in karst areas are more frequent than in non-karst areas.Figure 4 The results of the runoff simulation show that K-SWAT and K-W-D-SWAT have better ability to capture flood events and reflect the rapid recharge process of underground pipes in karst areas, and the simulation accuracy is significantly better than that of the original SWAT. Most importantly, K-W-D-SWAT can better reflect the hydrological process simulation capability of both drought and flood events by considering the influence of dry and wet events on runoff production and convergence processes.

[0131] Table 5 Comparison of runoff calibration and validation performance of three models in the Diaojing River Basin

[0132]

[0133]

[0134] Based on Figure 3 The multi-year monthly average rainfall runoff change process is shown, with May-September as the wet season and October-April of the next year as the dry season. Table 6 shows the runoff simulation results of the three models, SWAT, K-SWAT, and K-W-D-SWAT, in the wet and dry seasons. The model performance comparison is as follows: in the wet season, the original SWAT performs poorly and does not meet the basic requirements of simulation accuracy, while K-SWAT shows significant improvement, with R 2 and NSE are higher than 0.70, but the PBIAS and KGE of Malong Station still need to be improved. K-W-D-SWAT model performs better than the first two in all aspects, with the best KGE, all exceeding 0.70, and R 2 , NSE, and PBIAS have improved to varying degrees, effectively alleviating the overestimation of high flows in the flood season by SWAT and K-SWAT. In the dry season, the original SWAT does not perform well, especially in the PBIAS of the Hekou Station verification period (-32.18%), which overestimates low flows. Similarly, K-SWAT performs worse in the dry season than in the wet season, with poor indicators in all aspects. In contrast, the performance of K-W-D-SWAT model has improved, with KGE higher than 0.60 and PBIAS lower than ±20%, showing good simulation capability. At the same time, the runoff simulation performance of extreme dry and wet events is evaluated based on the flow duration curve (FDC) results of the three models at Hekou Station and Malong Station, as shown in Figure 5 Overall, the K-W-D-SWAT model results best fit the FDC of the measured runoff during the entire simulation period, and K-W-D-SWAT shows the best ability in dealing with high flow (0-0.1 flow exceedance probability) and low flow events (0.9-1.0 flow exceedance probability). Conversely, from Figure 5It is obvious that SWAT and K-SWAT have defects in simulating different flow levels at the two stations, indicating that the coordination ability of the model in runoff simulation during the whole period is weak, and the applicability of the model in the region is poor. In high flow events, the performance of the model is ranked as K-W-D-SWAT > K-SWAT > SWAT, and the simulation level of the peak flow of SWAT is lower than that of K-SWAT and K-W-D-SWAT, and with the increase of the proportion of karst area, the gap is further expanded. In low flow events, K-W-D-SWAT also performs best, while K-SWAT performs relatively poor in areas with stronger karst development (Malong Station). Overall, although K-SWAT model considers the karst structure, it still has deficiencies in simulating different flow levels in karst areas with frequent alternation of dry and wet events, reflecting the high dynamic nature and complex variability of the runoff generation and concentration process in karst areas. The results show that it is difficult to fully capture the hydrological characteristics of karst areas by parameterizing the karst structure alone, especially in the case of significant alternation of dry and wet events, and K-W-D-SWAT fully considers these factors, making it perform best in the simulation of hydrological processes.

[0135] Table 6 Comparison of runoff simulation performance of three models in dry and wet event period of Diaojiang River Basin

[0136]

[0137] In conclusion, the runoff simulation method of the karst SWAT model coupled with the antecedent precipitation index proposed in the present application greatly improves the runoff simulation accuracy of the traditional hydrological model. The present application takes into account the effects of karst structure and dry-wet events on hydrological processes, and can meet the high dynamic nature and variability of the runoff generation and concentration process in karst areas. It has great practical significance for the study of regional runoff simulation under climate change and complex underlying surface, and has good application prospect.

[0138] The above examples are only used to illustrate the technical solutions of the present application and not to limit it, any modification or equivalent replacement within the spirit and scope of the present application should be covered in the scope of the technical solutions of the present application.

Claims

1. A method for simulating karst SWAT model runoff by coupling antecedent precipitation index, characterized in that, Comprise the following steps: Step one, establish a SWAT model with karst area characteristics, according to the basin boundary vector diagram, hydrogeological map and topographic elevation data, determine the surface and underground boundary of the basin, use GIS technology to superimpose hydrogeological map, land use map and depression distribution map, generate land use map with karst characteristics, and combine soil data and slope data, generate hydrological response unit HRUs with karst and non-karst distribution characteristics; Step two, modify the source code of SWAT model, build K-SWAT model with karst runoff function, K-SWAT model includes surface karst zone, karst matrix and karst pipeline three systems; Step three, on the basis of K-SWAT model, embed normalized MIWAPI, build karst K-W-D-SWAT model with dry and wet effect identification mechanism, to accurately identify the effect of dry and wet events on karst runoff process, realize the simulation of basin response in dry and wet seasons all year round, the description formula of the dry and wet event is as follows: The weighted average value WAPI, IWAPI based on the antecedent precipitation index is used to identify regional dry and wet events, the calculation formula of WAPI, IWAPI is as follows: ， ， The above formula is normalized, IWAPI is scaled to [-1, 1], to show the intensity of drought and flood: ， Based on MIWAPI, the description parameters of runoff capacity are adjusted, the specific adjustment method is shown in the following formula: ， where, is the regional parameter, , is the average annual rainfall; = 1 if the model parameter has a promoting effect on the runoff production and sink process, otherwise = -1, is the antecedent precipitation impact days; is the recession coefficient; i is the simulation time; j represents the HRU serial number; is the multi-year average value of the jth HRU is the original parameter assigned by the SUFI-2 algorithm;​​ Step four, use SUFI-2 algorithm in SWAT-CUP software to optimize the parameters of SWAT model, K-SWAT model and K-W-D-SWAT model, and evaluate the runoff simulation level of each model in karst area through the coefficient of determination R², Nash efficiency coefficient NSE, percentage bias PBIAS and Kling-Gupta efficiency coefficient KGE.

2. The karst SWAT model runoff simulation method of coupling antecedent precipitation index according to claim 1, characterized in that, The water balance equation of the surface karst zone system is: ， The water balance equation of the karst matrix system is: ， The water balance equation of the karst pipeline system is: ， wherein, is the water level of epikarst zone; is the time scale; is the rate of change of water level of epikarst zone per unit time; is the recharge of epikarst zone on the ith day; is the triggered secondary spring discharge of epikarst zone; is the delayed discharge; is the recharge of epikarst zone to matrix on the ith day; is the recharge of epikarst zone to karst conduit on the ith day; is the evaporation of epikarst zone system; is the water level of matrix system; is the rate of change of water level of matrix system per unit time; is the diffusive recharge of epikarst zone to karst matrix system on the ith day; is the delayed discharge of matrix system; is the evaporation of matrix system; is the water level of matrix system; is the rate of change of water level of epikarst zone per unit time; is the exchange water of conduit to matrix system; is the rapid discharge of conduit; is the evaporation of conduit system.

3. The karst SWAT model runoff simulation method of coupling antecedent precipitation index according to claim 1, characterized in that, In the step three, the initial interception rate in the SCS-CN equation for calculating the surface runoff in the SWAT model Dynamic adjustment is made, and the calculation formula is: ， ， where, is the surface runoff; is the daily precipitation; is the initial loss; is the maximum interception; is the initial loss rate, defined as 0.2 by the SWAT model.

4. The karst SWAT model runoff simulation method of coupling antecedent precipitation index according to claim 1, characterized in that, In step four, the coefficient of determination R², Nash efficiency coefficient NSE, percentage bias PBIAS and Kling-Gupta efficiency coefficient KGE are used to evaluate the runoff simulation level of each model in karst area, the calculation formula is as follows: ， ， ， ， ， wherein , observed and simulated runoff, respectively, , denote the mean of observed and simulated runoff, respectively, n is the length of the series, r denotes the correlation coefficient, a denotes the ratio of the standard deviation of simulated and observed runoff, b denotes the ratio of the mean simulated and observed runoff; m is the rank of the sample in descending order, P is the frequency of the flow exceeding a certain intensity during the observation period.

5. The karst SWAT model runoff simulation method of coupling antecedent precipitation index according to claim 4, characterized in that, In step four, the flow duration curve FDC is used to further evaluate the runoff simulation level of each model in karst area.

Images