Revealing the impact of future climate change on the characteristics and evolutionary patterns of meteorological and hydrological droughts and exploring the joint distribution characteristics of their drought characteristics are essential for drought early warning in the basin. In this study, we considered the Jinghe River Basin in the Loess Plateau as the research object. The standardized precipitation index (SPI) and standardized runoff index (SRI) series were used to represent meteorological drought and hydrological drought with monthly runoff generated by the SWAT model. In addition, the evolution laws of the JRB in the future based on Copula functions are discussed. The results showed that: (
Keywords: climate change; meteorological drought; hydrological drought; spatial-temporal evolution; the JRB
Global warming and the increased frequency of extreme climate events have led to frequent water shortages in many areas in recent years caused by large-scale drought events [[
Different types of drought characteristics can be obtained using "Run theory," which is widely used by researchers at home and abroad [[
Global circulation models (GCMs) can predict future climate characteristics at different emission levels, and these models also have a universal applicability [[
The SWAT model is a semi-distributed hydrological model developed by the American Center for Agricultural Research. The model is widely used to simulate hydrological processes under the influence of climate change and human activities due to its relatively complete consideration of physical mechanisms [[
The main purpose of this study was to evaluate the spatial and temporal evolution laws of the meteorological and hydrological drought characteristics in the JRB under climate change in the future. The research contents of this study include (i) to obtain the future spatial and temporal evolution of precipitation and runoff; (ii) to identify the meteorological and hydrological drought spatiotemporal evolution characteristics of the JRB under different future emission scenarios; (iii) to reveal the joint distribution of the meteorological and hydrological drought characteristics under different periods; (iv) and to explore the probabilities of different types of drought propagation events in the future for the JRB.
The Jinghe River Basin (north latitude 34°46′~37°19′, east longitude 106°14′~108°42′) is located in the middle of China's Loess Plateau (Figure 1), with a total area of 45,412 km
We obtained the daily rainfall, daily maximum temperature, daily minimum temperature, daily average wind speed, and sunshine hours from 14 weather stations in the Jinghe River Basin from 1990 to 2017. The input data used to drive the SWAT model included a 90 m resolution of digital elevation DEM data, a resolution of 1 km of land use data, and 1 km resolution of soil type data.
The monthly runoff observation data of the Zhangjiashan Hydrological Station from 1990 to 2017 were obtained from the National Earth System Science Data Center. The observed monthly runoff data from 1991 to 2005 was used to calibrate the SWAT model parameters, and data from 2006 to 2017 were used to verify the SWAT model parameters. This study used the GFDL-CM3 global climate model data in CMIP5 to analyze climate change in the future period (2021–2060) and considered two representative emission scenarios named RCP4.5 and RCP8.5. The SWAT model input data and hydrometeorological data are summarized in Table 2. The GFDL-CM3 climate model data were simulated by the Geophysical hydrodynamics Laboratory of the National Oceanic and Atmospheric Administration, and the climate data fit well. Climate predictions were corrected for bias using the quantile mapping method (QM) [[
The Mann–Kendall (MK) method can effectively distinguish whether a process displays natural fluctuations or has a certain change trend [[
The MK test statistic Z is generally used to identify the degree to which a trend is consistently decreasing or increasing [[
The SDSM is a mixed method comprising multiple linear regression and a random weather generator [[
(
where Y is the local predictor and x
(
where ϕ is a constant cumulative distribution function and F is an empirical distribution function of
An important sub-model of the SWAT model is the hydrologic process module. We used this sub-model to simulate the spatiotemporal evolution of the JRB runoff (RCP4.5, RCP8.5) in the period of 1990–2017 and a future period (2021–2060).
The constructed SWAT model divided the basin into 43 sub-basins with 1226 hydrological response units. The SUFI-2 algorithm was used in this study to rate determine the model parameters, and the rate-determined parameter results are shown in Table 1. The monthly runoff data of the Zhangjiashan Hydrological Station of the JRB Exit Hydrological Station were used for verification.
In order to reflect the main research content, we have drawn a flowchart (Figure 2), which mainly illustrates the input datasets, methods, and output results of this study.
The standardized precipitation index (SPI) and the standardized runoff index (SRI) are used to describe the occurrence of meteorological and hydrological drought in the JRB. The SPI was proposed by McKee et al. It has been applied to meteorological and hydrological drought assessments globally [[
In this study, the run theory is applied to identify drought and derive drought characteristics based on the SPI/SRI with multiple time scales. The drought severity, duration, and intensity [[
According to the dryness levels of the SPI/SRI, −0.5 (R
Drought events have many characteristic properties such as duration, severity, and peak severity. However, it is difficult to comprehensively describe the characteristic properties of drought events only using single features. The Copula function can construct the joint distribution function for different drought characteristics. For the two-dimensional marginal distribution functions of F
(
In this study, we considered the duration and severity of meteorological and hydrological droughts; therefore, the drought univariate feature recurrence period can be expressed as:
(
(
(
(
where
(
(
where
In order to evaluate the SWAT model in this study, the root mean square error (RMSE), Nash–Sutcliffe efficiency coefficient (NSE), and the coefficient of determination (R
(
(
(
where N represents the number of samples;
Different CMIP5 models shows different simulation performances for different climatic elements in different regions; therefore, five commonly used CMIP5 models were selected in this study for comparing and analyzing the simulation effects of different climate models under two RCPs on precipitation in the JRB (Figure 4). the reasonableness of the model selection is comprehensively examined in terms of the correlation coefficient, ratio of standard deviation, and centralization root mean square error by using Taylor diagrams [[
The simulation shows that the GFDL-CM3 climate models are better than other climate models, with spatial correlation coefficients reaching nearly 0.8. In general, the simulation of precipitation under the RCP8.5 emission scenario is better than the RCP4.5 scenario.
The performance of the statistical downscaling model (SDSM) was evaluated by three elements: daily average precipitation, daily average maximum temperature, and daily average minimum temperature from the 14 weather stations. Accordingly, the performance of the downscaling model was evaluated using standard statistical methods (R
The SWAT model parameters were determined and verified according to the observed monthly flow data of Zhangjiashan Hydrological Station from 1990 to 2017. The year 1990 was the initial preheating period of the model operation. The Zhangjiashan Hydrological Station is the main outlet hydrographic station of the JRB and controls more than 95% of the total area of the basin. To reduce the impact of the initial conditions, the model considered the period from 1991 to 2005, and the model verification period was from 2006 to 2017. The uncertainty of the SWAT model parameters was analyzed by using the SUFI-2 method. R
From the relationship between the precipitation data of 14 weather stations and the simulated runoff in Zhangjiashan, the simulated runoff was higher in the months with more precipitation, indicating that the runoff is correspondingly large under greater precipitation, and the two are positively related. In terms of the correspondence between precipitation and runoff, the 14 meteorological stations selected during the modeling process had a high spatial correspondence with the Zhangjiashan hydrological station (Table 4). According to the results, the simulated runoff series and the observed runoff series were well-fitted, which indicated that the SWAT model constructed in this study accurately simulated the runoff process and could be adapted to the JRB and used in the next step.
Figure 6 illustrates the changes in and trends of the standardized precipitation index (SPI) in the JRB under different scenario modes (RCP4.5 and RCP8.5) and different time scales. The analysis indicated that the SPI value of the JRB will be in a state of alternating drought and flooding in the future, with more complex periodic changes. Overall, the RCP4.5 scenario mode predicted more frequent drought at the monthly, seasonal, and yearly scales than the RCP8.5 scenario mode.
The changing trend of meteorological drought in different scenarios was clearly different, and the frequency of alternating drought and flooding at the monthly, seasonal, and annual scale decreased successively. On the monthly scale, the drought trend in the RCP8.5 scenario was more intense than that under the RCP4.5 scenario, and the drought trend reached a 0.05 significance level in 2024. It is interesting that the seasonal scale was similar to the monthly scale situation. However, on the annual scale, the RCP4.5 scenario was more arid than the RCP8.5 scenario. Especially after 2033, the overall drought trend under the RCP8.5 scenario would be eased, which basically showed a wetting trend. In contrast, the RCP4.5 scenario showed characteristic drought and flood alternation patterns, and the drought severity was more serious.
Combined with the changes in the M-K mutation curve, the different time scales of the JRB under the RCP8.5 scenario mode were basically from drought aggravation to drought mitigation, while the opposite was found under the RCP4.5 scenario. The mutation points were concentrated in 2030, 2032, and 2033. The simulation prediction results showed that the overall annual precipitation of JRB in the river basin increased after 2032 because the SPI value was directly affected by the amount of precipitation. Therefore, the trend of JRB wetting is consistent with the trend of precipitation changes. In addition, the spatiotemporal distribution of the SPI of the JRB varied in different time scales, and thus the drought trend mutation points were also different.
In order to reveal the spatial evolution characteristics of meteorological drought at different time scales under two different future scenarios (RCP4.5 and RCP8.5), Figure 7a–f and Figure 8a–f revealed the spatial evolution pattern of meteorological drought duration and drought severity in the future period (2021–2060). On the monthly scale, the meteorological drought duration of the JRB under RCP4.5 lasted between 2.78 and 3.35 months, with a drought severity of between 2.12 and 2.60. The duration and severity of meteorological drought under RCP8.5 are similar to that of RCP4.5. However, the area with a long drought duration and large severity under RCP4.5 is more extensive, and more obvious in the middle reaches of JRB, while under the RCP8.5 emission scenario, JRB is more prominent in the upstream Yanchi station.
On the seasonal scale, the JRB meteorological drought lasted between 4.52 and 5.63 in RCP4.5, and the drought severity was between 3.86 and 4.80, while the meteorological drought duration and severity were reduced in the RCP8.5 emission scenario. The difference is that the meteorological drought duration and severity of the JRB in the RCP8.5 emission scenario decreased northwest to southeast, while the meteorological drought duration and severity of the JRB under the RCP4.5 emission scenario are larger in the middle reaches.
On the annual scale, the JRB lasted between 9.79 and 19.14, and the drought severity was between 8.74 and 15.49. The meteorological drought under the RCP8.5 emission scenario lasted between 12.72 and 16.90 months, with a drought severity of between 9.62 and 13.26. However, under the RCP4.5 emission scenario, the overall drought duration and severity of the whole JRB river basin were large, mainly concentrated in the middle and upper reaches of the JRB. Interestingly, under the RCP8.5 emission scenario, the drought duration distribution was similar to that in the RCP4.5 emission scenario, but the large values of the drought severity were concentrated at the Yanchi meteorological station point upstream of the JRB.
The SWAT model was used to drive the GCMs downscale atmospheric dataset to simulate the JRB emissions in the future period under different runoff sequences to obtain the standardized runoff index (SRI) time series (Figure 9). The future SRI was predicted to show "drought_flood_drought_flood" periodic changes, and with the increase in the time scale, the drought duration and severity gradually increased.
Combined with the results of the M-K trend test, the results showed that the JRB had two mutation years under the RCP4.5 scenario, and the hydrological drought mutation years were 2021, 2049, 2022, 2049, 2023, and 2050, where the drought trends were similar. For example, at the monthly scale, the JRB showed a significant wetting trend in the overall watershed during 2021–2049. After 2049, the JRB began to show a drying trend. Interestingly, the JRB hydrological drought trend on the monthly and seasonal scales has never exceeded the significance level
Under the RCP8.5 scenario, the JRB has only one mutation point on the monthly, seasonal, and annual scales in 2024, 2026, and 2029, respectively; a remarkable wetting trend continued after the mutation year until 2060. Meanwhile, the JRB hydrological drought trend did not exceed the significance level
The statistical downscale atmospheric dataset of GCMs was input into the SWAT model to simulate the spatial and temporal change distribution of the JRB runoff under different scenario modes in the future (2021–2060). Therefore, we used run theory to reveal the spatial and temporal distributions in the JRB, as shown in Figure 10 and Figure 11.
Overall, the drought duration and severity of hydrological drought were prolonged and increased compared with the meteorological drought in the JRB. On the monthly time scale, the JRB hydrological drought duration lasted between 4.40 and 9.00 months under the RCP4.5 emission scenario, and the drought severity was between 3.82 and 7.53. Taking Huanxian-Yaoxian as the boundary, the hydrological drought in the west of the boundary lasted for a long time; the drought severity was also large. However, the JRB experienced the strongest hydrological drought duration and severity in Xifeng and Pingliang areas under the RCP8.5 emission scenario, lasting for 13.4 months and with a severity that reached 9.6.
From a seasonal perspective, under the RCP4.5 emission scenario, the hydrological drought duration of the JRB was between 6.82 and 15.46 months, and the drought severity was from 5.99 to 13.47 months, being more intense in Xifeng and Changwu than in other regions. In contrast, under the RCP8.5 emission scenario, the drought duration was longer than under the RCP4.5 emission scenario, lasting from 6.43 to 20.64 months, and the drought severity increased slightly. The duration and severity in the west of the middle reaches were more pronounced.
On the annual scale, the overall drought duration of the basin was longer under the JRB emission scenario of RCP4.5; the drought lasted for a long time, reaching 30 months west of the line of Huanxian-Xifeng-Changwu. The severity of drought ranged from 16.81 to 30.91 and reached 30.91 in the Changwu area. Under the RCP8.5 emission scenario, the JRB hydrological drought duration distribution was similar to the severity distribution. The hydrological drought duration ranged from 21.76 to 62.34 months, and the drought severity ranged from 16.06 to 48.47 months. It is worth noting that in the east of the Huanxian-Xifeng Changwu boundary line, there was little difference in the hydrological drought duration or severity between RCP4.5 and RCP8.5 scenarios. However, west of the boundary line, the hydrological drought duration and severity under the RCP8.5 scenario were much greater than under the RCP4.5 scenario.
The Copula function was adopted to optimize the marginal distribution functions of different drought types and characteristics in historical and future periods. The analysis successively used the K-S, AIC, BIC, and other test methods to optimize the joint distribution models of different marginal functions [[
We selected the marginal distribution functions that were most suitable for the drought duration and severity from the six probability distribution functions of normal, exponential, log-normal, gamma, Weibull, and generalized extremal distribution, as shown in Table 5. The results of the K-S test indicated that the gamma distribution, Weibull distribution, and log-normal distribution functions could effectively fit the JRB drought eigenfunctions. Additionally, corresponding to Table 6, the Gumbel and Frank functions could be used to combine the marginal distribution characteristic functions of both drought duration and drought severity.
Based on the Copula joint distribution function under different scenario combinations, the "AND" and "OR" drought characteristics of MD and HD in the historical period (1990–2017) and future period (2021–2060) are shown in Figure 12 and Figure 13.
As can be seen from Figure 9, under the same recurrence period, the MD severity of both the RCP4.5 and RCP8.5 scenarios in the future period was significantly increased compared with the historical period. However, the MD duration under the RCP4.5 scenario was comparable to that of the historical period. It is worth noting that the duration of MD was longer under the future RCP8.5 scenario compared to the historical period. For example, in the "AND" joint distribution model, when the JRB encountered a 20-year return period MD, the drought duration was 11.64, 11.3, and 13.26 months in the historical period, future RCP4.5, and future RCP8.5, respectively, and the drought severity was 8.70, 11.52, and 11.98, respectively. Accordingly, there were similar characteristics in the "OR" joint distribution model.
It can be seen from Figure 10 that under the same return period, the HD severity under the scenarios of RCP4.5 and RCP8.5 would decrease in the future, and the HD situation is alleviated. Meanwhile, the duration of HD under the RCP4.5 scenario is significantly extended, but it is significantly shortened under the RCP8.5 scenario. For example, when the JRB encounters an HD with a 20-year return period in the "AND" joint distribution model, the duration of HD in the historical period, future RCP4.5, and future RCP8.5 scenarios are 32.40 months, 55.25 months, and 22.65 months, respectively, and the drought severity is 46.95, 19.40, and 19.45, respectively. Accordingly, there are similar characteristics in the "OR" joint distribution model.
This study explored the spatiotemporal evolution characteristics of MD and HD in the JRB under different climate emission models. We also studied the correlations between different drought characteristics. In addition, we should also be clear about how the MD propagates the HD and what types of drought propagation events would arise in the future. These studies are also very important and cannot be ignored. Different types of droughts, MD, and HD propagate under certain conditions. However, the process of drought propagation has diverse effects, including merger, attenuation, extension, and hysteresis. Therefore, there are many forms and characteristics of drought propagation. The present study divided drought transmission into the following situations.
S0: There is no correlation between MD events and HD events, i.e., when MD events occur, hydrological drought events do not occur. In this case, the duration of MD is likely to be short, and the severity is slight. After the MD event occurs, the HD event has been relieved before the underlying surface system receives a response.
S1: One MD event corresponds to one HD event because an MD event with a long duration and high severity is "reduced" by the land surface system, resulting in attenuation and triggering a hydrological drought event with a relatively short duration and low severity.
S2: Multiple MD events correspond to one HD event. After multiple MD events have occurred, an HD event with a long duration and high severity is generated due to combination, lag, or other effects. After the previous MD event, with the lag effects, the HD has not ended but has meanwhile ushered in a new MD event until the MD event no longer occurs or has been alleviated. At this time, an HD event formed by these conditions lasts for a long time and has a high severity with adverse effects.
S3: Multiple MD events correspond to multiple HD events. This situation is usually rare, and it is often caused by complex meteorological fluctuations or human activities. After this type of drought propagation event occurs, it is more likely to have serious consequences for the local residents. According to the drought propagation characteristics of the JRB, we matched the MD and HD events under different scenarios in the JRB historical period and the future. The drought propagation types are shown in Figure 14.
According to Figure 14, compared with the duration period, the probability of an S0 event occurring in the scenario of RCP4.5 of JRB increases, the probability of an S1 and S3 event occurring is basically unchanged, and the probability of an S2 event occurring is basically unchanged. Under RCP8.5, the probability of S0 events is almost unchanged, the probability of S1 events decreases, and the probability of S2 and S3 events increases. Therefore, more attention should be paid to drought propagation events under an RCP8.5 scenario in the future. After the occurrence of MD in this scenario, it is easy to form a long-duration and high-severity HD event that is difficult to recover from. Furthermore, although the impact of the drought propagation events of the JRB under an RCP4.5 scenario is relatively small, its MD events are frequent, which also requires us to retain our attention.
In this study, the meteorological data from the GFDL-CM3 model were first used to predict the pattern of meteorological drought characteristics for the JRB over the next 40 years (2021–2060s), and then their future climate model data were driven into the SWAT model. Even though the average annual precipitation of the JRB increases in the future, but its spatial and temporal distribution of precipitation becomes more heterogeneous, which often leads to significant differences in the spatial and temporal characteristics of meteorological droughts and an increase in the frequency of extreme weather events [[
The JRB experienced a slight increase in meteorological drought on a 3-month scale, but the JRB experienced a slowdown in hydrological drought, which is consistent with Sun's findings [[
Due to the limited data collected, this study only explored the meteorological and hydrological drought characteristics of the JRB under natural conditions. However, the occurrence of hydrological drought is not only influenced by natural meteorological elements, but also closely related to local underlying surface conditions and anthropogenic disturbances [[
Furthermore, different GCM models have different simulations for precipitation and temperature in different regions [[
Many studies indicated that there are some improvements from the CMIP5 to CMIP6 GCMs in simulating the mean and extreme temperature and precipitation globally and in China as a whole [[
Guo et al. evaluated and compared the performance of CMIP6 and CMIP5 models in simulating the runoff on global-scale and eight large-scale basins and found that CMIP6 models have less uncertainty on the global scale when compared with CMIP5 models, but it has not made outstanding progress on the basin scale [[
This study explored the temporal and spatial evolution of meteorological drought and hydrological drought in the JRB at different time scales by driving the SWAT model and downscale climate datasets for two representative future emission scenarios. Moreover, the Copula joint distribution function was used to reveal the relationship between different drought characteristic values and the joint distribution characteristic law. The main conclusions are as follows:
(
(
(
It is worth noting that the combined return period of the JRB drought duration and severity of the MD as well as HD in the future were slightly larger than those of the co-occurrence return period, indicating that when one drought characteristic variable occurred, the probability of another drought variable occurring simultaneously was higher. The numerous and short-duration MD events causing long-duration HD events would increase. In that case, the HD caused by this situation would last relatively longer, and the drought severity would be high, making the recovery thereof difficult. Although the area affected by HD under the RCP8.5 scenario was smaller than that under the RCP4.5 scenario, some of its local areas were vulnerable to long-duration and high-severity HD events would have unfavorable effects after the occurrence.
Graph: Figure 1 Study area and its location.
Graph: Figure 2 Flowchart including the input dataset, method, and output result.
DIAGRAM: Figure 3 Diagram of drought identification using the run theory.
Graph: Figure 4 Taylor diagrams of precipitation under different RCPs scenario: (a) RCP4.5 and (b) RCP8.5.
Graph: Figure 5 Comparison between simulated runoff and observed runoff in the JRB.
Graph: Figure 6 Evolution trends of different time scales of the SPI.
Graph: Figure 7 Distribution of spatiotemporal scale of meteorological drought characteristics under the RCP4.5 scenario: (a) Drought Duration of SPI−1; (b) Drought Duration of SPI−3; (c) Drought Duration of SPI−12; (d) Drought Severity of SPI−1; (e) Drought Severity of SPI−3; (f) Drought Severity of SPI−12.
Graph: Figure 8 Distribution of the spatiotemporal scale of meteorological drought characteristics under the RCP8.5 scenario: (a) Drought Duration of SPI−1; (b) Drought Duration of SPI−3; (c) Drought Duration of SPI−12; (d) Drought Severity of SPI−1; (e) Drought Severity of SPI−3; (f) Drought Severity of SPI−12.
Graph: Figure 9 Evolution trends of different time scales of the SRI.
Graph: Figure 10 Distribution of spatiotemporal scale of hydrological drought characteristics under the RCP4.5 scenario: (a) Drought Duration of SRI−1; (b) Drought Duration of SRI−3; (c) Drought Duration of SRI−12; (d) Drought Severity of SRI−1; (e) Drought Severity of SRI−3; (f) Drought Severity of SRI−12.
Graph: Figure 11 Distribution of spatiotemporal scale of hydrological drought characteristics under the RCP8.5 scenario: (a) Drought Duration of SRI−1; (b) Drought Duration of SRI−3; (c) Drought Duration of SRI−12; (d) Drought Severity of SRI−1; (e) Drought Severity of SRI−3; (f) Drought Severity of SRI−12.
MAP: Figure 12 Contour map of "AND" and "OR" type MD return periods of JRB under different scenarios (a) "AND" type in the history; (b) "AND" type RCP4.5 in the future; (c) "AND" type RCP8.5 in the future; (d) "OR" type in the history; (e) "OR" type RCP4.5 in the future; (f) "OR" type RCP8.5 in the future.
MAP: Figure 13 Contour map of "AND" and "OR" type HD return periods of JRB under different scenarios (a) "AND" type in the history; (b) "AND" type RCP4.5 in the future; (c) "AND" type RCP8.5 in the future; (d) "OR" type in the history; (e) "OR" type RCP4.5 in the future; (f) "OR" type RCP8.5 in the future.
Graph: Figure 14 Characteristics of drought propagation under different scenarios.
Table 1 A statistical analysis of the annual average precipitation and temperature.
Climate Factors Mean Median Std Cv Annual precipitation 493.12 476.71 81.98 0.17 Annual temperature 9.63 9.62 0.62 0.06
Table 2 SWAT model input data and hydrometeorological data.
Data Type Spatial/Temporal Resolution Source DEM 90 m The geospatial data cloud; Soil type 1 km World Soil Database (HWSD); Land use 1 km Resource and Environmental Sciences and Data Centre; Meteorological station 1990–2017 China Meteorological Data Network; CMIP5 2021–2060 World Climate Research Programme; Observed runoff 1990–2017 National Earth System Science Data Center;
Table 3 Optimization parameters of SWAT model in the JRB.
Parameter_Name Definition of Parameters Fitted_Value Min_Value Max_Value 1:R__CN2.mgt Runoff curve number −0.5203 −0.833 −0.197 2:V__ALPHA_BF.gw Baseflow coefficient 0.5506 0.0834 0.5529 3:V__GW_DELAY.gw Groundwater delay days 612.15 337.97 638.71 4:V__GWQMN.gw Limiting of shallow aquifers 376.24 −13.65 428.57 5:R__SOL_AWC(..).sol Soil water availability 0.044 −0.1878 0.2652 6:R__SOL_K(..).sol Saturated hydraulic conductivity 0.1472 −0.2225 0.4059 7:R__SOL_BD(..).sol Moist bulk density −0.3232 −0.8726 −0.261 8:V__REVAPMN.gw Guaranteed to shallow groundwater for re-evaporation 276.26 81.38 286.88 9:V__ESCO.hru Soil evaporation compensation factor 0.507 0.2847 0.7309 10:R__HRU_SLP.hru Average slope steepness −0.0957 −0.2323 0.2584 11:V__SLSUBBSN.hru Average slope length 11.61 10 14 12:V__CH_K2.rte Effective hydraulic conductivity of the mainstream 13.09 12.44 31 13:V__EPCO.hru Plant evaporation compensation factor 0.0848 −0.6598 0.1366 14:V__RCHRG_DP.gw Permeability coefficient of deep aquifers 0.277 −0.2745 0.3217 15:V__SURLAG.bsn Surface runoff delay factor 3.3509 0.75 22.7291 16:V__SFTMP.bsn Snowfall temperature parameters 4.3803 −1.8696 9.7401 17:V__SMTMP.bsn Snowmelt minimum temperature −5.187 −13.3622 −2.1377 18:V__SMFMN.bsn Minimum snow melt day factor 1.5257 −0.8357 7.1689 19:V__TRNSRCH.bsn Proportion of main channel propagation losses into deep aquifers 0.2142 0 0.5 20:V__ESCO.hru Soil evaporation compensation factor 0.9305 0.7 1 21:V__CANMX.hru Maximum canopy interception 3.8433 3 5
Table 4 Model evaluation of the calibration and validation periods at the hydrological station.
Hydrological Station Calibration (1991–2005) Validation (2006–2017) R2 NSE RMSE R2 NSE RMSE Zhangjiashan 0.75 0.72 9.45 0.83 0.79 5.56
Table 5 MD and HD duration and severity fitting distribution function and goodness test under different scenarios.
Periods Drought Type Drought Characteristics Marginal Distribution Parameter K-S Test a b D History MD Duration Gamma 2.9 1.64 0.1143 0.7544 Severity Gamma 1.27 2.03 0.0754 0.9869 HD Duration Gamma 1.94 6.21 0.1216 0.9695 Severity LN 1.63 1.19 0.1347 0.9317 RCP4.5 MD Duration Gamma 2.32 1.82 0.1174 0.5998 Severity Gamma 1.14 2.89 0.0902 0.8254 HD Duration Gamma 1.19 7.23 0.1571 0.5451 Severity Weibull 3.13 13.52 0.2022 0.7399 RCP8.5 MD Duration Gamma 1.87 2.44 0.0845 0.8699 Severity Gamma 1.36 2.67 0.0701 0.9657 HD Duration Gamma 1.54 4.92 0.1399 0.7547 Severity Weibull 0.57 7.58 0.0873 0.9927
Table 6 Best-fit Copula function test between duration and severity of MD and HD under different scenarios.
Periods Drought Type Copulas Correlation Parameters AIC Value BIC Value History MD Frank 0.69 9.53 −34.71 −33.23 HD Gumbel 0.71 3.85 −20.73 −20.09 RCP4.5 MD Gumbel 0.72 3.33 −63.5 −61.72 HD Gumbel 0.7 3.31 −24.1 −15.52 RCP8.5 MD Frank 0.75 12.55 −69.22 −67.39 HD Gumbel 0.79 4.72 −40.93 −39.89
Conceptualization, T.H. and Z.W. (Zhiyong Wu); methodology, T.H. and Z.S.; software, T.H.; validation, T.H. and Y.L.; formal analysis, T.H., Z.W. (Zhiyong Wu) and J.W.; investigation, Z.W. (Zhiyong Wu); resources, T.H. and Z.W. (Zhiyong Wu); data curation, T.H.; writing—original draft preparation, T.H. and Z.W. (Zhiyong Wu); writing—review and editing, T.H., Z.W. (Zhiyong Wu) and Z.S.; visualization, Z.W. (Zhihui Wang); supervision, TH and Z.W. (Zhiyong Wu); project administration, Z.W. (Zhihui Wang) and P.X.; funding acquisition, Z.W. (Zhihui Wang) and P.X. All authors have read and agreed to the published version of the manuscript.
Not applicable.
The authors declare no conflict of interest.
By Tingting Huang; Zhiyong Wu; Peiqing Xiao; Zhaomin Sun; Yu Liu; Jingshu Wang and Zhihui Wang
Reported by Author; Author; Author; Author; Author; Author; Author