Evaluation of Widespread Flooding of the Karkheh Basin in Iran using SWAT Model and GLDAS Database

DOI: https://doi.org/10.21203/rs.3.rs-1788047/v1

Abstract

Karkheh basin is a flood-prone region in Iran that was severely affected by devastating river floods in 2019. This study addressed some of the factors that affected on this event and were emphasized in the government inquiry committee1. These include strategic questions such as the effect of prior precipitation in the basin and how the dams could mitigate the floods peak and volume. These questions as well as deficits in the rainfall data led to the application of the Soil and Water Assessment Tool (SWAT) model and the global land data assimilation system (GLDAS) rainfall data to address the research questions. The results showed the initial managements of the dams prior to the occurrence of these floods was definitely affected by a decade continuous drought in the basin and the concerns about its continuation in 2019. The events occurred during March and April 2019, however, the precipitation occurred prior on October 2018 along with the resulted snowpack and soil saturation played a significant role in intensifying the floods. Although there were some limitations for the full operation of Seymareh Dam, the decision regarding its full operation could reduce the peak inflow to the Karkheh Dam from 8529 to 5447 m3/s. Finally, it is crucial to provide more accurate prediction systems, undertaking rapid and flexible responses and do not be misled by continuous droughts

1https://nfr.ut.ac.ir/en

1. Introduction

Karkheh basin often experiences devastating river floods. The important factors that affect these events in the basin include the meteorological conditions, snowpack, dam storage levels, vegetation cover and soil moisture at the time of the event. The 21 March to 4 April 2019 flood events (herein called “the 2019 flood event”) was those in which several factors combined to create some of the most catastrophic flooding recorded in the basin. This event inundated five provinces in southwestern Iran and inflicted many casualties and much loss. It is clear that a hydrological evaluation of such an exceptional event is essential for flood risk management in order to implement necessary measures to mitigate the negative effects of such an event in the future.

In a hydrological evaluation, conceptual hydrological models are helpful. They make it possible to simulate and distinguish between the different factors that form the characteristics of flood events. To this end, reconstruction and simulation of flood events have been considered in different studies. For example, Stucki et al. (2018) applied the semi-distributed rainfall-runoff model PREVAH to reconstruct a historical extreme flood event in the Lago Maggiore catchment in 1868. Dasallas et al. (2019) simulated the 2002 Baeksan flood in Korea by implementing the combined 1D–2D hydraulic computation of the HEC-RAS model. The roles of the factors affecting flood events include the different climate and management scenarios that can be introduced to a hydrological model to investigate their effects on such flooding.

Hussein et al. (2019) studied the effect of land-use changes on flood events occurring on the east coast of the UAE using the HEC-HMS. In their study, the curve numbers (CN) of sub-watersheds were estimated using 1996, 2006 and 2016 remotely sensed images. They concluded that urbanization has increased the peak discharge, runoff volume and extent of the flooded area. However, their approach was less sensitive to extreme events. Similarly, Hu and Shrestha (2020) investigated the effect of land-use change on flood peaks using the same model a watershed in the Midwestern United States, but by applying different CNs. They predicted increases in the urban areas and dry land of 17.47% and 14.05%, respectively, by 2028. Accordingly, increases in flood peaks of 2.9–3.5% for small scale floods and 0.4–1.1% for a big scale flood were anticipated.

It was reported that the peaks and volumes of the 2019 Karkheh basin flood events were significant. However, it was emphasized by the Special Reporting Committee on Iran Floods 2019 (SRCIF) (SRCIF, 2019a) that the damage was mainly due to the flood volumes rather than the peaks. Having in mind these features as well as the scale of the basin and its vast agricultural activity, we decided to use the soil and water assessment tool (SWAT) conceptual model (Arnold et al., 1998). This model is continuous and capable of simulating different climate and management scenarios. It is also semi-distributed in that it divides each sub-basin into smaller hydrological response units (HRUs) based on soil type, crop patterns and management practices (Neitsch et al., 2011). The model incorporates many modules, including crop growth, groundwater, dams and river routing to accommodate the required simulations. Although the original time-scale of the model is daily, it has been applied for flood studies as well.

Schilling et al. (2014) applied the SWAT model to assess changes in downstream flood risks under different types of land use for a large rural basin in Iowa. Their study focused on increasing the cultivation of perennials on the landscape and found that this type of management could decrease the likelihood of flood events and reduce the frequency of severe floods, but that their durations would not be substantially affected. Lee et al. (2017) applied SWAT to evaluate the effects of two upstream dams on the frequency of downstream floods in the Han River basin in South Korea. They also applied this model to individually simulate regulated and unregulated daily streamflows entering the third dam, which is located at the outlet of the basin. The estimated daily flood peaks were used for frequency analysis and showed that the two upstream dams were able to reduce downstream floods by approximately 31%. The model also was applied by Rajib et al. (2020) in conjunction with LISFLOOD-FP for simulation of floods and their respective inundation maps along the Ohio River basin in the United States. The results showed 70–80% consistency in the areas inundated using what had been captured from remotely sensed images.

In addition to hydrological modeling, preparing a relevant set of rainfall data that can illustrate the spatial and temporal variation of a storm is crucial for flood simulation. In reality, meteorological stations are not usually evenly distributed in basins and some of them may even be inaccessible during flood events, as we experienced in 2019 (SRCIF, 2019b). Satellite-based grid precipitation products and reanalysis datasets can be a relevant solution to tackling this issue. For the area presently under study, databases such as the global land data assimilation system (GLDAS) (Delavar et al., 2019), TRMM (Xu et al., 2017), CMORPH (Sinta et al., 2022), CHIRPS (Gummadi et al., 2022) and PERSIANN-CDR (Zhang et al., 2022) were evaluated. Delavar et al. (2019) also evaluated a number of these databases and reported that GLDAS performed the best from among them. For this reason, it was selected for use in the present study.

The current study is in line with numerous efforts to enhance flood risk management downstream of Karkheh Dam, which features a large population and substantial agricultural activity. The research concentrated on inflows to the dam and addresses questions that specifically arose after the 2019 floods. It sought to identify the ability of the combined SWAT model and GLDAS data for flood simulation in the basin as well as for future flood forecasting, understand the role of the precipitation prior to the flood events in the basin on the specification of 2019 flood events and examined the role of upstream dams on mitigating the inflow to the Karkheh Dam.

2 Material And Method

2.1 Study area

The Karkheh River originates in the Zagros Mountains and culminates at the Hoor-al-Azim swamp on the Iran–Iraq border. It is one of the most flood-prone basins in Iran and frequently sustains huge losses from flooding. The river results from the joining of the Seymareh and Kashkan Rivers. The area of research was focused upstream of Karkheh Dam (Fig. 1), which encompasses a total area of 42,600 km2 (82.9% of the entire basin). The elevation ranges from 172 to 3638 m above sea level and the average annual precipitation varies from 300 mm at the dam site to 800 mm in the mountains to the northwest.

To mitigate flooding of the basin as well as to facilitate agricultural development, the Karkheh and Seymareh Dams were constructed and have operated since August 2002 and September 2011, respectively. The Mashureh Dam currently is in the study phase with plans for construction on the Kashkan River. Some of the specifications of the dams are shown in Table 1. The mean daily discharge of the river at the inlets of the Karkheh, Seymareh and Mashureh Dams (study phase) are 115.9, 63.6 and 25.5 m3/s, respectively. The Karkheh Dam itself also is a carryover dam that protects six cities and 537 villages with a total population of 208,710 (Statistical Center of Iran, 2016). A number of improved irrigation and drainage networks with an approximate area of 250,000 ha are fed by this dam.

Table 1

Specifications of the main dams of the study area

Year of operation

Reservoir volume at operating level (MCM)

Height of foundation(m)

Type

River

Location

Dam

2011

2800

180

Doul-arched concrete

Seymareh

40 km northwest of DareShahr city

Seymareh

2002

5600

127

Earthen with clay core

Karkheh

22 km northwest of Andimeshk

Karkheh

-

1639

137

Doul-arched concrete

Kashkan

500 meters below the intersection of Kaka Reza and Cham Zakaria branches

Mashureh

2.2 Data

2.2.1 Ground stations

About 60 meteorological stations measure meteorological variables throughout the basin. Of these, only 21 stations (Fig. 1) were able to record the rainfall amounts of the 2019 flood events. There are five hydrometric stations located along the main river channel and its tributaries (Fig. 1). Table 2 provides some of the specifications of these stations. For the purposes of the current study, the record length for 2002 to 2014 has been applied for calibration as well as further evaluation of the 2019 flood events.

Table 2

Specifications of hydrometric stations of the study area

raw

Station Name

River

Longitude

Latitude

Elevation(m)

1

Qurbaghestan

Qaresu

47.25

34.23

1300

2

Pole Chehr

Gamasiab

47.43

34.34

1280

3

Nazarabad

Seymareh

47.43

33.17

559

4

Pole Dokhtar

Kashkan

47.71

33.16

650

5

Payeh Pol

Karkheh

48.15

32.42

90

2.2.2 GLDAS database

The ground stations in the study area are not well distributed over the basin; thus, not all of them could record the 2019 rainfall events. The lack of data over the west and southwest of the basin was especially considerable. To rectify this deficiency, the GLDAS global database was applied. This database applies the land information system (Kumar et al., 2006) and the land data assimilation system, which consists of multiple land surface models and integrates the observed ground data to generate the land surface state and flux parameters. The present study used GLDAS, version 3, with spatial and temporal resolutions of 0.25° and 3-hourly data (ftp://hydro1.sci.gsfc.nasa.gov/data/s4pa/GLDAS_V1/). Furthermore, 3-hourly rainfall data were retrieved from the database and converted into a daily time base for the model simulations. Figure 1 also shows the GLDAS gird over the study area.

2.2.3 Other data/information

In addition to hydro-meteorological data, the SWAT model requires data/information such as from a digital elevation model (DEM), soil map, land use map and agricultural management information. Table 3 lists the datasets and sources applied in this study.

Table 3. Datasets and respective sources for model parameterization

2.3 Catastrophic 2019 floods

As stated, Karkheh basin is one of the most flood-prone basins in the country. Evaluation of the records from 1955 to 2019 from Karkheh station shows that approximately 76% of the recorded floods were less than 2000 m3/s, 17% were 2000 to 3000 m3/s and 4% were 3000 to 4000 m3/s. The 2019 flood event was the highest recorded events with a 5909 m3/s peak.

Different factors combined to create such a huge flood. Those that have received the most attention are the extensive spatial area affected and the prolonged duration of the event (SRCIF, 2019b). During this event, two rainfall systems affected the basin. They occurred during 4–6 April and 10–13 April in 2019 and released a total of 195 mm of rain over the basin. These temporal and spatial patterns made the management of the events (including dam operation) more complicated. Ultimately, five provinces, 200 cities and 4,300 villages were affected and nearly 80 people lost their lives (SRCIF, 2019b).

2.4 Conceptual modeling framework

The methodology followed for this study is presented in Fig. 2. Its four main components are as follows:

  • Rainfall-runoff simulation using the SWAT model

  • Model parameterization, including calibration and sensitivity analysis

  • Creation of weather datasets for the 2019 flood event

  • Evaluation of the weather and management scenarios of the event

2.4.1 SWAT model

SWAT is a conceptual, semi-distributed model which can simulate the effect of management activities on hydrological processes at the catchment level (Arnold et al., 1998). The model first divides a basin into several sub-basins and then–depending on the type of land use, soil cover, and slope–divides each sub-basin into HRUs. Its basic equation is the hydrological cycle balance, which can be estimated for each HRU as follows (Neitsch et al., 2011):

$$S{W_t}=S{W_0}+\sum\nolimits_{{i=1}}^{t} {({R_{day}} - {Q_{surf}} - {E_a} - {W_{seep}} - {Q_{gw}})}$$
1

where SWt is the amount of water in the soil at time t; SW0 is the initial amount of water in the soil, Rday is the rainfall, Qsurf is the surface runoff, Ea is the evapotranspiration, Wseep is infiltration into the unsaturated water layer and Qgware is the return water from groundwater at time i.

This model uses the water balance simulations for a dam or floodgate as follows (Neitsch et al., 2011):

V (t+1) = Vstored(t) + Vflowin(t) – Vflowout(t) + Vpcp(t) – Vevap(t) – Vseep(t) (2)

where V is the volume of stored water, Vstored is the volume of water stored in the reservoir, Vflowin is the volume of water entering the reservoir, Vflowout is the volume of water flowing out of the reservoir, Vpcp is the volume of precipitation falling on the reservoir, Vevap is the volume of water removed from the reservoir by evaporation, and Vseep is the volume of water lost from the reservoir by seepage (all in m3) as well as (t), which is the Julian day number.

To calculate Vflowout, SWAT applies one of the following methods: measured daily outflow; measured monthly outflow; average annual release rate for uncontrolled reservoir; controlled outflow with target release. In this study, the first option was used. Eq. (2) requires the surface area of the reservoir to calculate Vpcp, Vevap and Vseep, which are updated daily as:

$$SA={\beta }_{sa}\times {V}^{{exp}sa}$$
3

where SA is the surface area of the water body (ha), \({\beta _{sa}}\) is a coefficient, V is the volume of water in the impoundment (\({\text{m}}^{3}{\text{H}}_{2}\text{O}\)), and expsa is an exponent. More details about these equations are available in (Neitsch et al., 2011).

2.4.2 Model parameterization

Using the DEM, land use maps (1993, 2000, 2005, 2010, and 2015) and soil maps, Kerkheh Dam basin was divided into 153 sub-basins and 1658 HRUs. A complete agricultural management schedule for dry farming and irrigated land was applied to the model to compile the HRUs more accurately. Ultimately, ten elevation bands were included in the model for better snowmelt simulation.

2.4.3 Sensitivity analysis

For sensitivity analysis, the calibration, uncertainty assessment and parameterization steps as well as the sequential uncertainty fitting algorithm (SUFI-2) were adapted from the SWAT-CUP software (Abbaspour, 2013). The most sensitive parameters were initially found. The final ranges then were the average of all the discharge stations (Table 4). Accordingly, CN, CH-N2, Sol-AWC, ESCO and GWQMN all showed significant effects on the output.

Table 4

The most important parameters used in calibration of river flow in SWAT

rank

Parameter

unit

Description

Initial range

Final range

Method

1

CN2.mgt

-

SCS runoff curve number

35–98

-0.3-0.1

r_

2

CH_N2.rte

-

Manning's ‘n’ value for the main channel

-0.01-0.3

6–14

r_

3

SOL_AWC.sol

\(\frac{mm {H}_{2}O}{mm Soil}\)

Available water capacity of the soil layer

0–1

-0.42-0.2

r_

4

ESCO.hru

-

Soil evaporation compensation coefficient

0–1

0.8-1

V_

5

GWQMN.gw

mm

Threshold depth of water in the shallow aquifer required for return flow to occur

0-5000

0.9–2.1

r_

6

SFTMP.bsn

ᵒC

Snowfall temperature

-20-20

0.1–0.3

r_

7

TLAPS.sub

ᵒC

Temperature lapse rate

-10-10

-8- -5.5

V_

8

LAT_TTIME.hru

day

Lateral flow travel time

0-180

0.48–1.12

r_

9

SOL_K.sol

\(\frac{mm}{hr}\)

Saturated hydraulic conductivity

0-2000

1.2–2.8

r_

10

ALPHA_BF.gw

\({day}^{-1}\)

Baseflow alpha factor

0–1

0.3–0.7

V_

11

OV_N.hru

-

Manning's n value for overland flow

0.01-30

-0.7- -0.3

r_

12

SURLAG.hru

day

Surface runoff lag time

0.05-24

0.1–0.2

V_

13

SLSUBBSN.hru

-

Average slope length

10–150

-0.3- -0.1

r_

2.4.4 Model calibration

The runoff was calibrated at the five discharge stations of Pole Chehr, Ghurbaghestan, Nazarabad, Pole Dokhtar (the most closest station to Mashureh Dam) and Payeh Pol (inflow to Mashureh Dam) (Fig. 1). The degree of conformity of the simulated discharge results in each step was achieved using the R2 and NS values. The model calibration and validation periods for most stations were from 2002 to 2014 and 1996 to 2001, respectively, and the results are shown in Table 5. It can be seen that the results were acceptable. For the calibration period, R2 was 0.79 to 0.86 and NS was 0.73 to 0.86. For the validation period they were 0.69 to 0.82 and 0.56 to 0.76, respectively. These meet the criteria recommended by Moriasi et al. (2015).

Table 5

Results of daily discharge simulation at the selected stations using statistical indicators

Calibration and Validation

Station

Period

(NS)

(R2)

(NS)

(R2)

1996–2014

0.84

0.84

0.76

0.77

Pole Chehr

1996–2014

0.81

0.82

0.69

0.76

Qurbaghestan

1996–2012

0.85

0.86

0.56

0.69

Nazarabad

1996–2001

0.73

0.81

0.75

0.80

Pole Dokhtar

1996–2000

0.79

0.79

0.81

0.82

Payeh Pol

3 Results And Discussion

3.1 Evaluation of GLDAS for 2019 storms

Previous studies have revealed the GLDAS database as performing the best of those tested (Delavar et al., 2019). However, because of the importance of the rainfall records during the 2019 flood event, additional pre-assessments were done before using the GLDAS data as input to the calibrated hydrological model. For this aim, the performance of this database for the storm period (21 March to 4 April 2019) was evaluated in terms of rainy-day detection by the FAR (false alarm ratio) and POD (probability of detection) indices (Behrangi et al., 2011). The accuracy of the estimated daily values are presented below.

3.1.1 FAR and POD

FAR is the ratio of false alarms to the total number of false and correct (hit) alarms. This index reveals which parts of the GLDAS database estimates have not been recorded at rain gauge stations. The best-case value is 0 and the worst is 1. The index was observed to be desirable at most stations with a maximum value of 0.37. The POD index is a binary indicator of the correct detection of rainy days in the databases. This index is the ratio of hit alarms to the total hit-and-missed detections by GLDAS at gauging stations. The POD value ranges from 0 to 1, where 0 is the worst and 1 is the best performance. It was observed that the POD values for all stations were appropriate with a minimum value of 0.78. Figure 3 shows maps of the FAR and POD values across the study area.

3.1.2 Coefficient of determination (R2) and Nash-Sutcliffe (NS)

Both R2 and NS are well-known criteria for model performance (i.e. observed vs. estimated values). The best performance value is 1 and the worst is 0 for R2 and the best performance value is 1 for NS and the worst is -\(\infty\); however, NS < 0 indicates that the observed mean is a better predictor than the model (Krause et al., 2005; Moriasi et al., 2015). Figure 3 shows the spatial performance of GLDAS compared with the observed data. The results were generally acceptable as all values were greater than 0.6.

3.1.3 Storm isohyetal map

The isohyetal map of the storm over the basin was produced using the GLDAS dataset. As stated, two systems affected the region during the 2019 flood event. Figure 4 shows the maps that were produced for 21–29 March and 30 March to 4 April 2019 using the observed data and GLDAS dataset from the 21 basin stations and 70 grid cells shown in Fig. 1, respectively.

A comparison revealed the reasonable performance of the GLDAS datasets for the study area. GLDAS properly detected the cores of the first storm over the eastern part of the basin (Kashkan sub-basin) and the second storm over the central part of the basin.

3.2 Simulation of 2019 floods using rainfall datasets

At this stage, different combinations of the observational rainfall data and the GLDAS database were examined to determine the best dataset for simulation of the 2019 floods. They are as follows:

  • Available recorded rainfall data (blue circles in Fig. 1)

  • Precipitation cell data from GLDAS database (P-GridGLDAS grids in Fig. 1)

  • GLDAS data for the closest its grid to the ground rainfall stations with no data (P-GLDAS: black and blue circles in Fig. 1)

  • Combined observed data and GLDAS database (Combined: Observation and P-GLDAS)

Next, the SWAT model was run using these datasets to simulate the flood event at the main gauging stations of Ghorbaghistan, Pole Chehr, Pole Dokhtar, Seymareh Dam and Karkheh Dam (Fig. 1) for the period of 9 March to 14 April 2019. The performance of the different datasets were visually evaluated from the simulated hydrographs for base flow, ascending and descending limbs and peak flow. Figure 5 shows the simulated daily flows for the 2019 flood event at the aforementioned stations and reveals the generally acceptable performance of the datasets. Although the periods for calibration and validation of the SWAT model were from 1996 to 2014, the calibrated model was directly used for the simulation of the 2019 flood event and demonstrated robust calibration.

Despite the acceptable performance of the model, it can be seen that the simulated peaks from the observed rainfall only (P-Observation dataset) were commonly underestimated compared to the observed hydrographs while the combined observation data and GLDAS database mode (Combined (Obs.+GLDAS)) better simulated the hydrographs and its peaks. This supports the decision to apply the GLDAS database to fill the gaps in the data in the upper catchments and fill in the missing rainfall records during the 2019 storms. To improve the accuracy of the results, they were also evaluated for the entire course of the event (9 March to 14 April 2019) using R2, NS and the normalized root mean square error (NRMSE). Figure 6 shows the best performances of these as 1, 1 and 0, respectively. This evaluation also confirms the better performance of the Combined dataset. One priority of this selection was to preserve the observed data in the simulations as well; thus, this dataset was used as the basis for further analyses.

3.3 Evaluation of 2019 flood event

For this step, the calibrated SWAT model and the selected rainfall dataset (Combined) have been implemented to address the research questions. A brief explanation of the status of the study area prior to the 2019 flood event has been provided below.

3.3.1 Initial conditions of earlier rainfall and water stored behind dams

The status of the rainfall before the 2019 floods is relevant to the event itself. The Iranian water year begins in October and ends in September, meaning that the flood event occurred over the 2018–2019 water year. It is important to note that the 2017–2018 water year was the conclusion of a very low annual rainfall condition. The total national annual rainfalls from 2005 to 2018 were continuously below normal and experienced rainfall shortages of up to -45%.

Similar conditions were experienced in the basin as well. Figure 7 shows that this drought continued until October 2018, but started to change from November onward, such that the monthly accumulation of rainfall reached 865.5 mm in April 2019. This is notable because this amount is more than twice the long-term average. It is noteworthy that most of the 3- and 6-month rainfall forecasts did not predict such massive rainfall (SRCIF, 2019a).

Similarly, and considering the previous continuous droughts, the Karkheh Dam started the new water year (2018-19) at its minimum level in October 2018 (2528.55 MCM). The dam managers were eagerly attempting to store inflow for irrigation of the autumn crops beginning in October 2018. However, from November 2018 onward, it was realized that the dam inflow was increasing in an irregular manner (Fig. 7). It then was decided to release more water in order to control the amount of stored water. There were limitations to releasing an amount of water that exceeded the bankfull discharge of the river. As it was harvest time, any submergence due to overtopping of the river bed would cause severe crop loss for farmers. Nevertheless, the amount of water became an issue and inevitably the discharge exceeded the bankfull discharge limit.

Seymareh Dam was another player for the management of flooding. The dam storage is 2875 MCM and it had stored 1374.76 MCM of water on Oct 2018. However, it was not fully under operation in 2019, because part of the land that was supposed to be submerged by the dam had not yet been purchased by the government. At that time, about 2000 MCM could be stored; however, as the situation became critical, it was decided to make the dam fully operational and reimburse farmers for the losses incurred by submergence of the land (SRCIF, 2019c).

3.3.2 Role of rainfall before 2018–2019 water year

How the rainfall status before the 2019 floods affected the magnitude of the floods was an additional issue (Fig. 7). To address this, the event was simulated assuming the rainfall prior to March 2019 was the same as during March 2018 (2018 scenario) and the SWAT model was run with the Combine dataset. Figure 8 shows the simulated hydrographs and their accumulative volumes for the initial weather conditions. The results at the inlet of Karkheh Dam showed that the difference between the peaks was not significant (5190 and 5545 m3/s). These volumes were 4045.5 and 6087.7 MCM under the 2018 and 2019 conditions, respectively, which shows that the hydrograph volume was a major contributor to the 2019 flood event.

3.3.3 Role of Seymareh Dam on inflows to Karkheh Dam during 2019 flood event

The Seymareh dam is located upstream of Karkheh Dam (Fig. 1). This dam became fully operational during this event, which was a critical decision during the 2019 flood event (sect. 3.3.1). The model was run for the following scenarios of Seymareh Dam operation: storage of zero water (i.e. Seymareh Dam did not exist) and partial operation of the dam (maximum storage of 2000 MCM) in comparison with full operation of the dam (i.e. actual operation and storage of up to 2875 MCM of water). Figure 9 shows the effect of the existence of the dam, which was able to reduce the peak discharge of Karkheh Dam inflow from 8250 to 5545 m3/s. It also reduced the volume of the floods from 6755.7 to 6087.7 MCM.

3.3.4 Effect of full operation of Seymareh Dam on inflow to Karkheh Dam

A SWAT model was executed for the maximum dam storage of 2000 MCM (partial operation due to land ownership issues) and 2875 MCM (real operation) in order to evaluate the full operation of the dam. The initial storage capacity of Seymareh Dam for both scenarios was set at 1600 MCM on 1 January 2019 as per the recorded data. Figure 10 shows the effect of this decision on the inflow to Kakheh Dam. It can be seen that this decision had no effect for the first peak, but had a positive effect on the second peak. When the peak for full operation of the dam was 5447 m3/s, partial operation increased the peak to 8529 m3/s. Figure 10 includes the time series of the daily cumulative dam storage of Karkheh Dam, which demonstrates how the storage of the Seymareh Dam changed under the two operational scenarios and that the differences between the storage levels started to grow after the second peak.

3.3.5 Effect of inflow to Kashkan River on inflow to Karkheh Dam

The Kashkan River is another important tributary that contributes to the Karkheh River (Fig. 1). As stated, construction of the Mashureh Dam is planned for this river and it is in the study phase. To evaluate how this dam could mitigate the inflows to the Karkheh Dam, the SWAT model was run with the assumption of this dam being operational. Figure 11 shows the peak and volume of the inflows to the Karkheh Dam decreased from 5545 to 4103 m3/s and 6087.7 to 5094.6 MCM for this event, respectively.

4 Conclusion

This research attempted to address some of the critical questions about the causes and management of the 2019 flood event in the Karkheh basin. The responses to these questions as well as the modeling system that was developed to address them can enhance the ability of basin management to cope with future events.

The 2019 flood event occurred after about ten consecutive years of drought in the basin, which can be a strong sign for the effects of climate change (SRCIF, 2019b). The 3- to 6-month rainfall forecasting systems failed to provide sufficient warning of possible flooding (both local and global databases). Thus, the initial management of the dams prior to the occurrence of these floods was definitely affected by the extended previous droughts and lack of forecasting systems. Considering the these uncertainties, It is crucial to facilitate the water authorities with more adaptive management by undertaking rapid and flexible responses as well as not to be misled by long-term droughts.

The combination of observed and GLDAS rainfall data as input to the SWAT provided a useful framework for simulating the considered hydroclimate and management scenarios. This compensates for the lack of access to the gauging stations during the flood event. Published reports of this event emphasized the resulted damages were more affected by the flood volumes rather than their peaks. The framework could handle both characteristics and considering that Karkheh is a large basin and that a daily time-scale could be sufficient in this regard.

Although the flood event occurred during March and April 2019, the rainfall and snowpack that occurred prior to this event started on October 2018 played a significant role in intensifying the flood. Simulation of this event by assuming the prior conditions, such as the rainfall during the 2017-18 water year, reduced the peak flow and volume of the 2019 hydrograph from 5545 to 5190 m3/s and its volume from 6087.7 to 4045.5 MCM, respectively. This confirms reports of the exceptional volume of this flood.

Full operation of the Seymareh Dam was a political decision made during the flood event. The evaluation of the partial (storage capacity of 2000 and 2875 MCM, respectively) and full operation of the dam revealed that this decision reduced the peak inflow to the Karkheh Dam from 8529 to 5447 m3/s. This provided more time to the operators of Karkheh Dam to mitigate flooding downstream of the dam.

The Mashureh Dam on the Kashkan River is still under study. Our simulations showed that, assuming it had become operational during this event, it could have reduced the peak and volume of the inflow to the Karkheh Dam from 5545 to 4103 m3/s and 6087.7 to 5094.6 MCM.

Finally, integrated drought and flood management is the most important strategy to save this basin from such devastating floods. This study only addressed one aspect of such a management and many other issues must be considered to increase resiliency of the basin to flooding. Fortunately, positive steps have been taken, but there is still a long way to go in this regard.

References

  1. Abbaspour KC 2013. Swat-cup 2012.SWAT calibration and uncertainty program—A user manual
  2. Arnold JG, Srinivasan R, Muttiah RS, Williams JR (1998) Large area hydrologic modeling and assessment part I: model development 1. JAWRA J Am Water Resour Association 34(1):73–89
  3. Behrangi A, Khakbaz B, Jaw TC, AghaKouchak A, Hsu K, Sorooshian S (2011) Hydrologic evaluation of satellite precipitation products over a mid-size basin. J Hydrol 397(3–4):225–237
  4. Dasallas L, Kim Y, An H (2019) Case study of HEC-RAS 1D–2D coupling simulation: 2002 Baeksan flood event in Korea. Water, 11(10), p.2048
  5. Delavar M, Morid S, Hajihosseini HR, Ahmadi H, Shokri Kuchak V (2019) Model preparation Preparation of a comprehensive model of hydrological simulation of Karkheh basin to evaluate the effects of environmental changes on discharge upstream of Karkheh dam and development of long-term flow forecasting system in it. Khuzestan Water and Power Authority, Ahvaz, Iran
  6. Gummadi S, Dinku T, Shirsath PB, Kadiyala MDM (2022) Evaluation of multiple satellite precipitation products for rainfed maize production systems over Vietnam. Sci Rep 12(1):1–18
  7. Hu S, Shrestha P (2020) Examine the impact of land use and land cover changes on peak discharges of a watershed in the midwestern United States using the HEC-HMS model. Papers in Applied Geography 6(2):101–118
  8. Hussein K, Alkaabi K, Ghebreyesus D, Liaqat MU, Sharif HO (2020) Land use/land cover change along the Eastern Coast of the UAE and its impact on flooding risk. Geomatics Nat Hazards Risk 11(1):112–130
  9. Hussein K, Alkaabi K, Ghebreyesus D, Liaqat MU, Sharif HO (2020) Land use/land cover change along the Eastern Coast of the UAE and its impact on flooding risk. Geomatics Nat Hazards Risk 11(1):112–130
  10. Krause P, Boyle DP, Bäse F (2005) Comparison of different efficiency criteria for hydrological model assessment. Adv Geosci 5:89–97
  11. Kumar SV, Peters-Lidard CD, Tian Y, Houser PR, Geiger J, Olden S, Lighty L, Eastman JL, Doty B, Dirmeyer P, Adams J (2006) Land information system: An interoperable framework for high resolution land surface modeling. Environ Model Softw 21(10):1402–1415
  12. Lee JE, Heo JH, Lee J, Kim NW (2017) Assessment of flood frequency alteration by dam construction via SWAT simulation. Water, 9(4), p.264
  13. Moriasi DN, Gitau MW, Pai N, Daggupati P (2015) Hydrologic and water quality models: Performance measures and evaluation criteria. Trans ASABE 58(6):1763–1785
  14. Neitsch SL, Arnold JG, Kiniry JR, Williams JR (2011) Soil and water assessment tool theoretical documentation version 2009. Texas Water Resources Institute
  15. Rajib A, Liu Z, Merwade V, Tavakoly AA, Follum ML (2020) Towards a large-scale locally relevant flood inundation modeling framework using SWAT and LISFLOOD-FP. Journal of Hydrology, 581, p.124406
  16. Schilling KE, Gassman PW, Kling CL, Campbell T, Jha MK, Wolter CF, Arnold JG (2014) The potential for agricultural land use change to reduce flood risk in a large watershed. Hydrol Process 28(8):3314–3325
  17. Sinta NS, Mohammed AK, Ahmed Z, Dambul R (2022) Evaluation of Satellite Precipitation Estimates Over Omo–Gibe River Basin in Ethiopia. Earth Systems and Environment, pp 1–18
  18. SPCIF (2019a) Report of Hydrology and Water Resources Working group, Special Reporting Committee on Iran Floods 2019. URL: https://nfr.ut.ac.ir/
  19. SRCIF (2019b) Report of Meteorological Working group, Special Reporting Committee on Iran Floods 2019. URL: https://nfr.ut.ac.ir/
  20. SRCIF (2019c) Report of Education and Human Resources Working group, Special Reporting Committee on Iran Floods 2019. URL: https://nfr.ut.ac.ir/
  21. Stucki P, Bandhauer M, Heikkilä U, Rössler O, Zappa M, Pfister L, Salvisberg M, Froidevaux P, Martius O, Panziera L, Brönnimann S (2018) Reconstruction and simulation of an extreme flood event in the Lago Maggiore catchment in 1868. Nat Hazards Earth Syst Sci 18(10):2717–2739
  22. Xu R, Tian F, Yang L, Hu H, Lu H, Hou A (2017) Ground validation of GPM IMERG and TRMM 3B42V7 rainfall products over southern Tibetan Plateau based on a high-density rain gauge network. J Geophys Research: Atmos 122(2):910–924
  23. Zhang Y, Wu C, Yeh PJF, Li J, Hu BX, Feng P, Jun C (2022) Evaluation and comparison of precipitation estimates and hydrologic utility of CHIRPS, TRMM 3B42 V7 and PERSIANN-CDR products in various climate regimes. Atmos Res 265:105881