Impacts of land use changes and climate variability on transboundary Hirmand River using SWAT

Many river basins are facing a reduction of ﬂ ows which might be attributed to changes in climate and human activities. This issue is very important in transboundary river basins, where already existing con ﬂ icts about shared water resources between riparian countries can easily escalate. The decrease of stream ﬂ ow in the transboundary Hirmand (Helmand) River is one of the main challenges for water resources management in Iran and Afghanistan. This research aims to quantify the causes of this problem which has a direct impact on the dryness of the Hamoun wetlands being an international Ramsar site. To achieve this, the land use changes in the Middle Helmand Basin (MHB) in Afghanistan were evaluated for three time periods between 1990 and 2011 using remote sensing data and the Soil and Water Assessment Tool (SWAT) Model for understanding watershed response to environmental changes. It was concluded that the total irrigated area in the region has increased from 103,000 ha in 1990 to 122,000 ha in 2001 and 167,000 ha in 2011 (62% increase). According to the results, the average annual discharge when adapting the land use during the simulations was 4,787 million cubic meters (MCM)/year and while employing the land use of 1990 from the beginning of the simulations, the average annual discharge was 5,133 MCM/year. Therefore, the agricultural developments in the Helmand basin decreased the discharge with about 346 MCM/year accompanying an increase of 64,000 ha in an irrigated area in MHB after 1990. Notably, the impact of land use change increases signi ﬁ cantly for more recent periods and causes a reduction of 810 MCM in annual stream ﬂ ow for the MHB. The amount of water depletion (i.e. actual evapotranspiration) per hectare has increased from 5,690 in 1985 to 7,320 m 3 in 2012. The applied methodology of this study is useful to cope with such a data scarcity region. It can help quantify the impact of land use change on the region and formulates strategies that can improve the situation between Iran and Afghanistan.


INTRODUCTION
Many river basins are facing a reduction in flows which might be attributed to climate change and human activities.
The distinction between these two factors has received widespread attention, particularly in recent literature (Zhi et al.

The Hirmand (Helmand) River basin in Iran and
Afghanistan is an example of a transboundary river basin facing a decline in streamflow in recent decades. The river is the most important water resource shared between Iran and Afghanistan and plays an important role in the society and economy of this region. However, fast agricultural developments (UNODC ) and changes in climatic variables (Vining &  On the other hand, SWAT and similar models need long-term time series of climate and hydrological data, while data collection and exchange are usually a serious obstacle in many transboundary basins due to political issues (Bitew & Gebremichael ). Moreover, Afghanistan has remained largely undeveloped due to the unstable situation in recent decades. Thus, its infrastructure is still poor and the available data are also very limited (Hajihoseini et al. ). For such a situation, the application of global climate databases and remote sensing data to create land use maps and produce the required inputs of the hydrological models can be a useful solution (Mango et al. ).
The global database of the Climate Research Unit (CRU) has been often considered in this regard, as it has the advantage of using a considerable number of ground observing stations (Harris et al. ) and has a long record length.
Notably, the combination of SWAT and CRU has been applied for a number of hydrological modeling studies in large basins (e.g. Abbaspour et al. ; Xu et al. ; Khoi & Hang ). Therefore, it can be a relevant option for the present study. This paper aims to study the impacts of climate variability and land use changes on streamflow of the Hirmand River during the period 1985 till 2012. Such evaluations are essential to identify the necessary bilateral policies between Iran and Afghanistan to manage the basin more cooperatively and especially, for the restoration of the international Hamoun wetlands (Ramsar site). Moreover, this period is a distinctive period in Afghanistan including many political events that have affected its development and consequently its water demand. One of the novelties of this paper is the development of a methodology to respond to the research questions of a transboundary river basin in a data-scarce region, while using data-intensive models like SWAT. Moreover, the results of this research provide information that decision-makers need in order to manage the basin and especially, save the Hamoun international wetlands.

Study area
The transboundary Helmand basin is shared between Iran and Afghanistan. It includes a few principal tributaries originating from the south and westerly slopes of the Hindu Kush Mountains near Kabul and flows approximately 1,100 km from its source to the international Hamoun wetlands at the Afghan-Iranian border. Lake Hamoun is a seasonal lake and water is generally only present during the spring melt season (Mojtahed-Zadeh ; USGS 2016: https://earthshots.usgs.gov/earthshots/node/59). The average annual precipitation in the Helmand basin varies from 100 to 400 mm (NCDC ).
The focus of this study is on the Transboundary Hir- There is another tributary, the Arghandab River that joins the main stream at Qala-Bust in MHB. However, its importance for the total inflow to MHB is small due to irrigation consumption before reaching the Hirmand River (Helmand River Delta Commission ; Williams-Sether ). Figure 1 shows the discretization of the basin in this study.

Data used
For hydrological modeling of the Helmand basin, different data were needed including (1) observed hydrological data, (2) global climate databases, (3) land use and land cover information (4) operation rules of the Kajakai dam, (5) a Digital Elevation Model (DEM) and (6) a soil map and soil properties (e.g. particle size distribution, bulk density, organic carbon content and available water capacity). Table 1 shows the data used in this study and related references. Part of the required information needed some prepreprocessing that is explained in the next sections.

CRU database
The CRU data were used in this study. The applied version is TS v3.21 (2013) (http://www.cru.uea.ac.uk/), which contains several climate variables. The monthly CRU data were converted to a daily time stepnecessary for SWAT inputusing the daily weather generator algorithm 'dGen' (Geng et al. ). This algorithm has been applied in similar studies as well (e.g. Schuol & Abbaspour ; Schuol et al. ). The CRU precipitation data were evaluated using the observed data. For this, the mean annual precipitation map of the entire basin (for the period 1961-1976) was produced using the CRU dataset and the available recorded data of meteorological stations. Over 250 CRU grid cells and 15 stations in Afghanistan, 2 in Iran and 1 in Pakistan were used to construct the maps using the Kriging interpolation method that performed better than other spatial interpolation methods. None of the meteorological stations is located in MHB.  Table 1). To aid the analysis of images, complimentary information was collected including (1) 1973,1985,1990,1993,1998,2001,2009,2011

Climate data analysis
The resulting maps of annual precipitation are shown in For the temperature, due to the limited number of available records and also less spatial variation compared with precipitation, the assessment was solely done for Panjab station, which is located in UHB (Figure 1). It is also reported that CRU temperatures usually fit better with observations than other datasets (Döll et al. ). Figure 5 shows the average mean, maximum and minimum monthly observed and CRU temperature.

Simulation of Kajakai dam operation
As shown in Figure 1, the main inflow to MHB is the release of  Simulation of streamflow

SWAT model calibration and validation
The model calibration was performed using the available observed discharge data at Char-Burjak station and different global datasets. The land use of 1990 was applied for the calibration period, because it could be assumed that due to the       The difference in outputs between scenarios S2 (S3) and S1 indicates the impact of climate variability (land use change) on streamflow in SWAT. The formula for determining the impact is as follows: where R S1 , R S2 and R S3 are the average runoff from scenarios S1, S2 and S3, respectively; ΔR Climate (ΔR Landuse ) is the difference in SWAT output between scenarios S2 (S3) and S1.

Scenario 2: Response to climate variability
In order to study changes in discharge time series due to climate variability, possible trends in the climate data were evaluated using the Mann-Kendall test. The results showed that there is no significant trend in climate variables except for the minimum temperature (Table 2).
Furthermore, the Pettitt test was applied and detected meaningful changes in 1976. The time series of minimum temperature is detrended using the method suggested by Hamlet & Lettenmaier () and the SWAT input files were modified accordingly. While the recorded average annual minimum temperature was 1.5 C (1941-2012), its detrended value reduces to 0.9 C. In Figure 9, we compare the minimum temperature of the CRU grid at one of the CRU grid (a 254).
Applying the detrended time series of monthly minimum temperature, the calibrated SWAT model was used to simulate discharges at Char-Burjak station. The discharge time series based on detrended climate data and the discharge time series based on observed climate data (without detrending) are shown in Figure 10. It can be seen that the differences are not very significant. For instance, the annual mean discharges are 4,733 MCM and 4,787 MCM, respectively.

Scenario 3: Response to land use change
The results in the accuracy assessment of the land use classification showed a Kappa coefficient value between 72% and 93%, while the overall classification accuracy was between  82% and 90%. Congalton & Green () suggest that the values above 80% are acceptable. Using the processed images, the changes in land use from 1990 to 2011 were determined and the results can be found in Table 3.
To describe the changes, three main agricultural regions in MHB are discussed below.
Nawa Barakzai area (West Lashgargah city) As Figure 11 indicates, the majority of marshland (swamp and wetland), located at the downstream end of the Boughra Creek, has dried up after 1990 and converted into irrigated lands of which their total area has increased from 11,909 ha in 1990 to 24,454 ha in 2011.

Nowzad area
Comparing the results in Table 3, it can be seen that the rainfed area has severely increased from 1990 to 2011, mostly in Musa Qala and Nowzad highlands. In this region, a sharp increase in the cultivated area is observed (Figure 12).

Nad Ali area (West Boughra Creek)
Increasing cultivation area (irrigated and rainfed) has also occurred in the upstream part of Boughra Creek. Groundwater (qanats and wells) is a source of water supply for the increased agricultural lands (Favre & Kamal ) (see Figure 13).
Due to the above-mentioned changes, the total area of agricultural lands (irrigated and rainfed) was estimated to be 200,000 ha in 2011. Comparing the results of 1990 with 2001 shows that dense land had changed to bare land. Comparing the results of the classification of different periods (Table 3), it can be seen that the irrigated area has increased from 103,000 ha in 1990 to 122,000 ha in 2001 and then to 167,000 ha in 2011.        Figure 15 shows the simulated streamflow time series at Char-Burjak station. The reduction of streamflow is considerable in recent years and is discussed in the following section.

CONFLICTS OF INTEREST STATEMENT
None declared.