Impact of basin-wide dry climate conditions and non-climatic drivers: an isolation approach

The imbalance between available water and growing water demand has become a source of major concern and a heavy burden facing water managers and decision-makers in shared river watersheds, where competition between upper and lower riparian countries to meet water demand targets is considerable. Drought spells emerging from climate change coupled with upstream non-climatic drivers (e.g. construction and management of reservoirs, water diversion structures, inter-basin water transfer schemes, water withdrawals and land use changes) in the transboundary river watersheds are becoming key challenges to sustainably managing the water resources of lower riparian states. Previous research (Tharme 2003; Orr & Carling 2006; Keskinen et al. 2010; Abushandi & Merkel 2011; Naik & Jay 2011) has indicated the impacts of climate change and river regulation on runoff alteration. Al-Faraj & Scholz (2014b) have explored a three-scalar framework, which formulated the transboundary watershed management difficulties in a three-level system. The three-level system encompasses (1) the national interests of the upstream country, (2) the transboundary (basin-wide) scale, and (3) the national interests of the downstream country.

Alemaw & Kileshye-Onema (2014) have stated that effects of droughts and correlated resilience options should be considered when developing basin-wide sustainable water resources planning and management strategies. Naumann et al. (2014) has pointed out the importance of understanding the drought vulnerability of the impacted areas, and the need to develop an appropriate drought early warning system to lessen forthcoming droughts.

Various drought indices (IES 2008) have been developed worldwide. These indices range from single index parameters such as streamflow drought index (SDI) and standardized precipitation index (SPI) to multi-parameter indices such as the Palmer drought severity index. In the present study, the characteristics of the hydrological droughts were assessed using the SDI and the monthly-based variable truncation level (i.e. drought threshold), while the SPI was used for characterizing the meteorological droughts. The SDI was developed by Nalbantis (2008), and has been broadly used in many recent studies (Shukla & Wood 2008; Nalbantis & Tsakiris 2009; Yang 2010; Tigkas et al. 2012; Tabari et al. 2013; Haslinger et al. 2014).

The truncation level (or drought threshold (Byzedi et al. 2012)) method is based on the theory of runs, which defines droughts as periods during which imbalance between water availability and water demand occurs. Byzedi et al. (2012) highlighted the simplification of the proposed method by adopting a constant threshold level (constant demand). Hence, droughts are defined as periods during which the runoff is less than the threshold level. Recent literature demonstrates that the threshold levels are flows that are likely to be equalled or exceeded at 70–95% of the time (Hisdal et al. 2002, 2004; Engeland et al. 2004; Andreadis et al. 2005; Fleig et al. 2006; Byzedi & Saghafian 2009; Tallaksen et al. 2009; Wong et al. 2011; Van Loon & Van Lanen 2012). Byzedi (2011) and Kwak et al. (2012) have used the method of truncation level at 70%, which has been adopted in this article.

The characterizations of the meteorological droughts were explored using the precipitation-based drought index SPI (McKee et al. 1993), which has frequently been utilized in research, particularly in recent articles (Vincente-Serrano et al. 2004; Wilhite et al. 2005; Khadr et al. 2009; Rasheed 2010; Karavitis et al. 2011; Al-Timimi & Al-Jiboori 2013; Palchaudhuri & Biswas 2013; UNESCO 2014). Following the Lincoln Declaration on Drought Indices, the experts of the World Meteorological Organization (WMO 2009) have reported that the SPI should be used to characterize meteorological droughts by all National Meteorological and Hydrological Services around the world.

In this study, the SDI and SPI computations were performed using the DrinC software version 1.5.71 (Tigkas et al. 2013). The monthly time series of the streamflow and the precipitation were fitted to a two-parameter Gamma distribution function in the computation process to obtain the SDI and SPI indices. The monthly threshold drought levels (truncation levels at 70%) were computed using the HEC Data Storage System Visual Utility Engine (HEC-DSSVue) version 2.01 and the Statistical Software Package (HEC-SSP) version 2.0 of the US Army Corps of Engineers Hydrologic Engineering Centre (USACE).

The study aims to: (1) assess the hydrological drought characteristics under the natural flow regime; (2) investigate the collective impacts of basin-wide dry climate conditions and upstream water use practices on available water in a downstream state; (3) assess the meteorological drought characteristics; (4) isolate the proportional influence of upstream man-made activities from the mutual impacts of human interventions and basin-wide dry climate conditions (which cause droughts); and (5) identify the monthly-based proportionate impact of upstream man-made activities on water availability for the downstream nation.

Findings will help water resources decision-makers, particularly lower riparian states, to deepen and broaden understanding of the combined consequences of the upstream artificial river regulation schemes and dry climate environments at transboundary scale on the sustainable use of water resources. In addition, it highlights the risk confronting water management in downstream states in the absence of proper practical transactions to address the problem.

The study targets the Diyala (Sīrvān) river watershed, shared between Iraq and Iran, which lies between 33.216 ° and 35.833 °N, and 44.500 ° and 46.833 °E. The basin occupies a total drainage area of about 32,600 km², of which about 43% lies in Iraq and 57% in Iran (MWR 1982). Mountains represent nearly 62% of the basin, while the plateau and plain areas embody approximately 38%. The total length of the river is 386 km from its source in Iran to its confluence with the Tigris river in Iraq, south of Baghdad. This basin supports important agricultural activities in the upper and lower riparian countries. The watershed has complex and heterogeneous hydrological features. The basin is characterized by a Mediterranean climate; cool and wet winters, and hot and dry summers. Nearly 90% of the annual rainfall occurs between November and April, most of it in the winter months from December to March. The remaining 6 months are often dry, particularly the non-rainy months between June and August. Precipitation is irregular in time, quantity and locations. The total annual precipitation falls between about 110 mm at the lowest part of the basin and approximately 970 mm in the highlands of the watershed.

Thirteen synoptic meteorological stations, within and in proximity to the studied basin, were investigated. The monthly observed precipitation data between the hydrologic years 1981 and 2010 were analysed. The data were made available by the Ministry of Agriculture and Water Resources in the Kurdistan region of Iraq (MA & WR-KRG) and the Iranian Meteorological Organization. Some other meteorological data in Iraq were obtained from FAO (2004). Figure 1 shows the location of the examined meteorological stations.

A new generic statistical methodology has been introduced to isolate the proportional impact of upstream man-made river regulation schemes from the mutual influences of upstream anthropogenic pressures and basin-wide dry climate conditions. The following steps describe the application of the new proposed methodology to the Diyala watershed:

1. For the pre-regulation monthly mean extended streamflow time series (1925–1982):

   • (a) The flow duration curve analysis (FDCA) was conducted for the 58-year unimpaired monthly mean flow time series observed at the DHS to determine the monthly-based truncation levels at 70%. The HEC-DSSVue and the HEC-SSP were applied for the FDCA. Twelve monthly truncation levels were extracted (i.e. one threshold value per month).

   • (b) For each month, the monthly mean streamflows, which are less than the corresponding monthly-based truncation level (calculated in step 1a), were marked.

   • (c) The ratios between the marked streamflow values in step 1b and their corresponding monthly-based truncation level as obtained in step 1a were calculated. For each month, the standard deviation of the calculated ratios was determined.

2. For the post-regulation monthly mean streamflow time series (1983–2013):

   • (a) The same process as described in step 1b was followed.

   • (b) The process as outlined in step 1c was repeated.

   • (c) The ratios between the standard deviation values computed in step 1c and the ones determined in step 2b were estimated to isolate the proportion of impact of the upstream artificial regulation from the mutual impacts of upstream anthropogenic pressures and basin-wide dry climate conditions.

   • (d) Finally, the monthly proportion of the impact of the upstream artificial regulation equals the value obtained by the following simple calculation: 1–(ratio calculated in item 2c) ×100.

The SPI was used to assess the characterization of the meteorological droughts and as a useful complement to support the findings of the hydrological droughts in the pre-regulated flow condition. For each meteorological station, four reference overlapping SPI time series were generated exemplifying four time spans, 3, 6, 9 and 12 months, respectively. Quantities and descriptive states of the SDI and the SPI indices documented in literature (Bonsal & Regier 2007; Tigkas et al. 2012; Tabari et al. 2013) were considered in this paper for drought interpretations: extremely wet when (SDI and SPI) ≥ 2.00, and extremely dry, when the (SDI and SPI) ≤ –2.00. The severity classifications are very wet (1.50 ≤ SDI and SPI) < 2.00), moderately wet (1.00 ≤ (SDI and SPI) < 1.50), near normal (–1.00 < (SDI and SPI) < 1.00), moderately dry (–1.50 < (SDI and SPI)≤ –1.00), severely dry (–2.00 < (SDI and SPI) ≤ –1.50).

The statistical description of the streamflow time series is given in Table 1. Results show that the long-term mean streamflow observed at DHS of the unimpaired and impaired streamflow paradigms were about 173 m³/s and 131 m³/s, respectively. This suggests that the artificially-influenced streamflow period had seen a flow reduction of approximately 25%. The 25th, 50th, and the 75th percentiles were dropped by about 47, 33 and 19%, respectively. The streamflow reductions can be attributed to the collective impacts of anthropogenic pressures on the river basin in the upstream riparian state and basin-wide dry climate conditions.

The period between 1925 and 1998 shows a significant temporal disparity in SDI magnitudes punctuated by non-drought spells. This can be explained by the typical characteristics of the hydrologic regime dominated by a rainy period between October and May, and a non-rainy period between June and September. The dry period may span over one complete hydrologic year or several water years such as 1948, 1958 and 1960. Dealing with the four reference time intervals, prolonged droughts of mild to moderate severity were detected over 5 successive years between 1928 and 1932, with SDI ranges from –0.05 to –1.12. Between 1933 and 1946, only 2 years of mild to moderate drought spells were observed, in 1935 and 1944, with SDI ranges between –0.43 and –1.08. Between 1947 and 1957, mild to moderate droughts were recorded in 1947, 1948, 1951, 1955 and 1956. The SDI values fell between –0.03 and –1.66. Moderate to mild drought episodes were sustained for the period from 1958 to 1962, with some signs of short-term extreme drought spells. The SDI values ranged from –0.34 to –2.49. Short-term droughts were registered for the hydrologic years 1963–1990, followed by a mild drought in 1991. A moderate to mild drought was also recorded in 1997, with SDI ranges between –0.53 and –1.18.

Findings reveal that droughts were frequently detected between 1999 and 2013. Droughts of extreme severity were acknowledged for the hydrologic years 2000, 2001, 2008 and 2009. No tangible artificial regulations and hydraulic water diversions were implemented and commissioned between 1925 and 1982 in the upstream riparian state. Likewise, limited hydraulic regulation works and water exploitation schemes in the upper part of the basin were observed between 1983 and 2004 (Al-Faraj & Scholz 2014a, 2014b). Hence, the droughts can be assigned mainly to the abnormality of precipitation. This is supported by the findings of the SPI values (Table 2) for four time intervals (i.e. 3, 6, 9 and 12 months). A close examination of Table 2 reveals that the water years of 1984, 1989, 1991, 1996, 1997 and 1999–2001 had noticeable drought spells for various time intervals over the examined meteorological stations. Moreover, results indicate that the drought spells were not limited to a number of stations, yet droughts were observed in all examined meteorological stations. This suggests that the entire watershed was frequently prone to drought episodes, in particular from 1999 onwards.

Table 2

SPI analysis for four reference periods for all examined meteorological stations

View Large

Outstanding river regulation arrangements were put in place between 2004 and 2013, twinned with extreme basin-wide dry climate conditions, which span the years 2008–2009. The SDI peaked at –2.72 in 2009. According to SDI magnitudes, the successive drought periods 1999–2001 and 2008–2009 were the most influential to agriculture and water resources of the downstream country. Between 2010 and 2013, the severity of the drought ranged from severe to mild. The disparity in severity and duration of drought spells, which extended for the period 2004–2013, is attributed to collective impacts including precipitation anomalies, the volume of water abstraction, diversion across the main river corridor and its tributaries, and the operational practices and management strategies of the upstream reservoirs.

Extreme drought spell characteristics were recognized for the years 1948 and 1960, with SDI ranges from –2.09 to –2.53. Mild and moderate drought spells lasted for about 7% and 39%, respectively, of the observed time period. In comparison, the maximum time during which extreme and severe drought spells were observed was about 7%. The severity and duration of droughts between 1931 and 1961 were due to precipitation deviations compared to normal conditions.

Concerning the upper and middle combined reaches of the watershed, the SDI time series between 1925 and 1961 at DDS for 1-month time intervals and four reference periods are shown in Figures 5 and 6, respectively. Findings from the non-overlapping SDI of 1-month time intervals reveal that drought spells were perceived at about 54% for the pre-river regulation period (1925–1961). No non-drought spells were detected for the hydrologic years 1931, 1932, 1935 and 1948. Extreme droughts were observed in February 1948, February–April 1960, July 1960 and October 1961. The 1-month SDI peaked at –2.80 in March 1960.

The SDI and the SPI values have clearly indicated that the basin has been recurrently suffering from meteorological and hydrological droughts. Since 2004, the hydrological drought observed in the lower riparian country has been considerably governed by basin-wide dry conditions and anthropogenic-induced river modifications in the upper riparian state.

The increase in monthly standard deviations from the pre- to the post-regulation condition can be explained by upstream intense damming arrangements and significant consumption of water and diversion measures. Findings disclosed that the upstream river regulation schemes considerably reduced the water available for the downstream country. The relative impact ranged from a minimum value of 5% in February to the highest value of 54% in July. The average relative impacts between April and October and between November and March were about 46% and 17%, respectively.

A new generic statistical methodology has been proposed to detect the relative impact of man-made river regulation practiced by an upstream country (Iran), which influences adversely the downstream country (Iraq). This was accomplished by isolating the relative impact of upstream human pressures from the collective influence of basin-wide dry climate conditions and upstream anthropogenic activities. The proposed simple statistical approach supports water managers in unbiased, timely and spatially relevant decision-making processes.

For the upstream reach of the examined basin, findings from the non-overlapping SDI of 1-month time interval reveal that drought episodes were obtained for 44% of the unimpaired period (1925–1982), 40% of the transition time frame (from the pre- to the post-impact period) for the hydrologic years 1983–2003, 92% of the time for the regulated period between 1999 and 2013, and 93% of the time between 2004 and 2013, during which intense damming and significant consumption of water and diversion were put in place. For the four overlapping reference periods, results indicate that droughts due to the combined influences of human activities and dry climate conditions were common between 1999 and 2013. Drought events between 1925 and 1982 can be directly linked to precipitation abnormalities. According to SDI magnitudes, the 1999–2001 and 2008–2009 drought periods impacted mostly on the agriculture and water resources of the downstream country. Between 2010 and 2013, the severity of droughts ranged from severe to mild. The disparity in the severity and duration of drought spells, which span the period between 2004 and 2013, is attributed to various impacts including precipitation deviations, the volume of water abstraction across the main river and its tributaries, and operational practices of upstream reservoirs.

Concerning the upstream and middle combined reaches, findings from the non-overlapping SDI of 1-month time interval show that drought spells were observed for about 54% regarding the pre-river regulation period (1925–1961). Drought spells dominated the hydrologic years 1931, 1932, 1935 and 1948. Extreme droughts were noticed in February 1948, February–April 1960, July 1960 and October 1961. The 1-month SDI peaked at –2.80 in March 1960. Results also show that the relative impact of the upstream river regulation ranged from 5% in February to the highest value of 54% in July. The average relative impact between April and October was 46%. In contrast, the corresponding value was 17% between November and March.

The authors recommend further research on improving estimation of the proportion of impact of upstream river regulation within shared river basins on the sustainable management of water resources in the state or states located downstream. Furthermore, lower riparian states should increase their water usage efficiency and update the rule curves of reservoirs to allow for a better allocation of the available water in time and space.

Flow data used in this study were made available by the General Directorate of Dams and Reservoirs, the Directorate for Operation of Derbandikhan Dam, and the Directorate for Operation of Dokan Dam (Ministry of Agriculture and Water Resources-Kurdistan Regional Government in Iraq).