Assessing Climate Change Impacts on the Lower Zab River Watershed: A Predictive Study

Climate change is identified as one of the most significant environmental challenges confronting the globe today, as mentioned in the Sixth Assessment Report (AR6) released by the Intergovernmental Panel on Climate Change (IPCC) [1]. The report stated that the surface temperature of the globe rose by 1.1℃ by 2020 relative to the levels from 1850 to 1900. This increasing pattern is anticipated to continue until the beginning of the 22nd century, reaching 1.5℃ [2, 3]. The impacts of climate change have been observed in numerous regions worldwide, including elevated temperatures, sea-level rise, ice melting, and degradation of air quality [4]. These effects are acknowledged as the most significant threat to human existence and livelihood on Earth [1, 2, 5]. The leading cause of climate change is the emission of greenhouse gases (GHG) due to the utilization of fossil fuels and alterations in land use/ land cover. The consequences of global warming are expected to worsen in the future, exerting a significant impact on hydrological processes [6].

The Middle East has been significantly impacted by global climate change because of its arid and semi-arid environment [7]. Iraq is one of the countries in the Middle East that has been most severely affected by climate change. Accordingly, Iraq will encounter significant environmental issues, including sudden floods, water shortages, sandstorms, and heatwaves [8]. Abdulsahib et al. [2], Hassan and Hashim [9], Mohammed and Fallah [10] revealed that extreme events are anticipated to increase in Middle Eastern countries due to the impacts of climate change.

Scientists frequently employ General Circulation Models (GCMs) to simulate and project climatic conditions under different emission scenarios [11]. The GCMs are numerical models with outputs of inferior spatial resolution. Thus, the GCMs' results cannot be employed directly to examine the effects of climate change. Consequently, downscaling methods must be applied to enhance the spatial resolution of the data [12, 13]. Downscaling techniques are classified into two categories: dynamic and statistical. An overview of downscaling methods was provided by Trzaska and Schnarr [14]. The GCM outcomes generated by the CMIP, an abbreviation for the Coupled Model Intercomparison Project, are the most widely applied [15-17]. The Working Group on Coupled Modelling (WGCM) has introduced the latest CMIP6 scenarios, which involve the most recent set of emission scenarios referred to as shared socio-economic pathways (SSPs), based on various socio-economic assumptions [2]. The Long Ashton Research Stations Weather Generator (LARS-WG) software is one of the most common Statistical Downscaling (SD) methods that reliably identifies complex patterns in spatial-temporal data prediction. The LARS-WG model requires fewer computational resources and provides a user-friendly interface compared to dynamical downscaling methods [1, 2].

The LARS-WG model has been effectively applied in various parts of the globe [11, 18-21]. Several investigations in this regard were carried out in Iraq. Mohammed et al. [22] applied LARS-WG to examine the impacts of climate change on ten weather stations around the Lower Zab River Basin in northeastern Iraq. Mohammed and Hassan [1] utilized LARS-WG 6 to predict changes in temperature and precipitation in southern Iraq under RCP4.5 and RCP8.5 over the three periods from 2021 to 2100. Abdulsahib et al. [2] used LARS-WG 8 to predict climate variables from 2021 to 2040 in northern Iraq by two scenarios (SSP2-4.5 and SSP5-8.5). The results of these studies revealed a continuous rise in minimum and maximum temperatures, while the rainfall trend exhibited noticeable fluctuation depending on the region, scenario, and period.

The Little Zab River (LZR), also known as the Lesser or Lower Zab, represents the Tigris River's second major tributary within Iraqi lands [23]. The length of the river is about 375 km. The LZR has a catchment area of 20000 km² located in northeastern Iraq and northwestern Iran [24-26]. In addition to supplying the Tigris River, the Little Zab River Watershed (LZRW) is a significant area for agricultural and natural resources [25]. Throughout the 20th century, the LZRW has been significantly affected by conflicts, population growth, and climate change [27].

Although many researchers have conducted studies on climate change and forecasting climate factors in Iraq, including [28-30], there is a lack of thorough studies showing the impact of climate change and its repercussions on the watershed scale. Thus, the current study aims to project precipitation, minimum, and maximum temperature in the LZRW for three future periods: 2041-2060, 2061-2080, and 2081-2100. The LARS-WG 8 model was applied to project the climatic factors for 10 weather stations adjacent to the LZRW using five GCMs and two emission scenarios (SSP5-8.5 and SSP2-4.5). The Kriging tool in ArcMap 10.2 was used to convert the point data source at the weather station into a spatial map covering the LZRW. The research methodology employed will provide a precise vision of the extent to which each part of the study area is affected by climate change impacts. The results of this study will provide policymakers with a strong tool for fulfilling the requirement of sustainable natural and water resources within the basin.

3.1 Calibration and validation of the LARS-WG

The daily rainfall and temperature data from the ten weather stations associated with the LZRW, spanning from 1985 to 2015, were used for model calibration and validation. The model's efficiency in downscaling the GCMs' outputs in the study region was evaluated by statistical tests. The K-S test was used to verify the similarity of the daily weather data distribution derived from observed and simulated data. Furthermore, this research employed a p-value test to determine whether the hypothesis that both data sets (observed and projected) may have the same distribution. The simulated data is considered identical to the measured data if the p-value is high and the K-S value is low [2, 44]. The statistical indices calculated for the Halabcha station, as an example, are shown in Table 4.

Table 4. The K-S test and p-value for Halabcha station

According to Semenov and Barrow [45], the fit is poor if the p-value is less than 0.4. For 0.4 ≤ p-value < 0.7 and 0.7 ≤ p-value < 1.0, the fit is considered good and very good, respectively, while a perfect fit is achieved for p-value equal to 1.0. The evaluation results displayed in Table 4 indicate that the model's performance in predicting the distribution of daily Tmax, Tmin, and rainfall is perfect (P), with occasional instances of very good (VG) results. As a result, the downscaling technique was employed with more confidence in this study.

Additional statistical indicators were used to assess the performance of the LARS-WG model in Halabcha station. The indicators applied are the root mean square error (RMSE), coefficient of determination (R²), and mean bias error (MBE) [1, 2].

RMSE is a model evaluation statistic that works as an error index. The nearer the RMSE is to zero, the higher the model's efficiency:

$RMSE=\sqrt{\frac{\mathop{\sum }_{1}^{n}{{\left( {{G}_{i}}-{{O}_{i}} \right)}^{2}}}{n}}$           (1)

R² signifies a relativity coefficient that spans from zero to unity, with the optimal value being 1:

${{R}^{2}}=\frac{\mathop{\sum }_{1}^{n}{{G}_{i}}{{O}_{i}}}{\sqrt{\mathop{\sum }_{1}^{n}G_{i}^{2}\mathop{\sum }_{1}^{n}O_{i}^{2}}}$         (2)

MBE signifies the error in the model estimation. It may possess either a negative or a positive value. Nevertheless, the optimal value is zero:

$MBE=~\frac{\mathop{\sum }_{1}^{n}\left( {{G}_{i}}-{{O}_{i}} \right)}{n}$             (3)

where $G_i$ is the generated value of the weather variable, $O_i$ is the observed value, and $n$ is the data number.

The results (Table 5) revealed that the LARS-WG model can effectively predict weather variables. The statistical criteria demonstrated a good agreement between the observed and generated data, as the indicators determined were within acceptable limits [46, 47].

Table 5. Statistical indicators result for model calibration and validation during the baseline period

3.2 Projection results

The calibration and validation results obtained from the LARS-WG model indicated that its effectiveness was sufficient for downscaling and predicting daily precipitation and temperature (minimum and maximum) at the ten weather stations. Consequently, the calibrated model was used to forecast the daily climatic variables for each station over the three intervals (i.e., 2041-2060, 2061-2080, and 2081-2100) using five GCMs and two scenarios.

A prevalent method in climate change effect research is to evaluate the phenomenon at a specific location separately [1, 13, 30, 48]. Abdulsahib et al. [2], Noor et al. [49] displayed the phenomenon spatially using the Inverse Distance Weighted (IDW). In this study, the Kriging tool in ArcGIS software was employed to create the spatial-temporal maps of the climatic variables. Kriging is a geostatistical prediction technique that generates weights from adjacent known values to estimate values at unmeasured sites [50]. Amelia et al. [51] and Shi et al. [52] stated that Kriging is superior to IDW, as it exhibits the highest interpolation accuracy and effectively captures the changing trends of precipitation across extensive areas. Additionally, Abdulsahib et al. [2] stated that the bull's-eye effect is one of the IDW's defects which can be overcome using the Kriging tool.

According to the author's cartographic skills and previous works [2, 53], the following points were considered while preparing the maps: (1) the weather variables were divided into five classes, choosing smaller and larger number of classes makes the maps uninformative or confusing, respectively, (2) the lowest and highest possible values of the weather variable of all scenarios and future periods were determined to fix the classes across all maps, enabling the reader to compare the results by simply seeing the maps, and (3) the classes' coverage percentage (%) relative to the entire area was calculated and written for each class.

3.2.1 Projection of minimum temperature

The spatial maps of the annual average minimum temperature (Tmin) for the LZRW are displayed in Figures 3 to 5. For all the figures, the minimum temperatures were equally divided into five classes, ranging from class 6-9℃ (the lowest) to class 18-21°C (the highest). Figure 3 shows the mean Tmin maps during the first future period (2041-2060) using the two scenarios (SSP 2-4.5 and SSP 5-8.5) compared to the observed map (1985-2015). The observed map involved the first three classes. The eastern parts of the watershed exhibited the lowest temperatures (class 6-9℃), followed by the central and western parts. The projected maps indicated an increase in the minimum temperature as a new class (15-18℃) appeared and covered the western areas. In the SSP 2-4.5 map, the extent of class 6-9℃ reduced from 38.32% in the observed map to 4.87%, while class 15-18℃ increased from 0% to 26.74%. It can be observed that the coldest class (6-9℃) disappeared in the SSP5-8.5 map, and the map was approximately similar to the observed map, but with one class shifting. In other words, the area of class 6-9℃ was covered by the higher class (9-12℃) and so on.

tu_pian_1.png

Figure 3. Results of minimum temperatures for the observed and projected maps (2041-2060)

tu_pian_2.png

Figure 4. Results of minimum temperatures for the observed and projected maps (2061-2080)

The comparison between the Tmin of the baseline period and the projection results for the second future period is shown in Figure 4. The figure indicated a continuous rise in Tmin from 2061 to 2080 across the LZRW. The most significant increase can be observed in the SSP5-8.5 map, where the hottest class (18-21℃) covered 20.72% of the basin. Furthermore, one can notice that the SSP2-4.5 map is approximately similar to that of the SSP5-8.5 in the first period.

Figure 5 depicts the effect of climate change on the minimum temperature at the LZRW during the third period (2081-2100). The same increasing trend can be observed within the basin. The predominant class in the central part, for instance, of the basin during the baseline period was class 9-12℃, while it shifted to class 12-15℃ and 15-18℃ in the maps of the SSP2-4.5 and SSP5-8.5, respectively.

tu_pian_3.png

Figure 5. Results of minimum temperatures for the observed and projected maps (2081-2100)

Figure 6 summarizes the predicted effects of climate change on the minimum temperature within the LZRW. The figure displays the percentage change in the mean minimum temperature (Tmin) of the basin during the future periods compared to that of the baseline period. The SSP2-4.5 predicted that the mean minimum temperature will increase by 19.14, 28.13, and 33.97%, while the SSP5-8.5 projected higher percentages of 29.38, 42.49, and 58.09% during the periods 2041-2060, 2061-2080, and 2081-2100, respectively. The SSP5-8.5 projected a greater rise in Tmin than the SSP2-4.5 due to the unrestrained greenhouse gases [2].

tu_pian_4.png

Figure 6. Percentage change of minimum temperature for the projection periods

3.2.2 Projection of maximum temperature

This section displays and discusses the projected mean maximum temperatures (Tmax) in the LZRW by the two scenarios and for the three future periods. The temperatures were divided into five equal classes. The lowest Tmax values were represented by the class 20-23℃, while the highest values were categorized by the class 32-35℃.

Figure 7 shows the mean Tmax for the observed and the first projection period (2041-2060). The majority of the LZRW (44.17%) was occupied by the coldest class (20-23℃). On the other hand, the projected maps revealed a significant increase in Tmax, with the warmer class (29-32℃) appearing in the western parts of the basin. The extent of class 29-32℃ was 25.98% in the SSP2-4.5 map and 30.35% in the SSP5-8.5 map. At the same time, the coldest class (20-23℃) decreased by 42.81 and 98.32% in the SSP2-4.5 and SSP5-8.5 maps, respectively.

tu_pian_5.png

Figure 7. Results of maximum temperatures for the observed and projected maps (2041-2060)

A greater increase in the maximum temperature can be observed during the second future period, as shown in Figure 8. The SSP5-8.5 exhibited the most significant rise, with the spatial extent of the warmest class (32-35℃) increasing to 21.73%, accompanied by the disappearance of class 20-23℃.

tu_pian_6.png

Figure 8. Results of maximum temperatures for the observed and projected maps (2061-2080)

The projection outcomes for the third period are displayed in Figure 9. Compared to the observed map, the class 6-9℃ vanished totally in the SSP2-4.5. Thus, the predominant class in the eastern, central, and western parts was changed from 20-23, 23-26, and 26-29℃ to 23-26, 26-29, and 29-32℃, respectively. A higher increase in Tmax can be observed in the SSP5-8.5 map, where the warmest class covered 23.43% of the basin.

tu_pian_7.png

Figure 9. Results of maximum temperatures for the observed and projected maps (2081-2100)

Figure 10 explains the percentage change of the projected Tmax compared to the observed data. The increasing pattern is similar to that of Tmin, but with a slower rate of increase. The figure illustrates that the percentage change in the SSP2-4.5 scenario is always smaller than that of the SSP5-8.5. In the third future period, the Tmax increased by 15.68 and 26.65% using the SSP2-4.5 and SSP5-8.5 scenarios, respectively, compared to the baseline's Tmax.

tu_pian_8.png

Figure 10. Percentage change of maximum temperature for the projection periods

One can notice that the percentage change of Tmin is greater than that of Tmax. This could be attributed to the calculation of the percentage change. For example, Tmin during the baseline period and during the third future period (2081-2100) by the SSP5-8.5 was 10.45 and 16.52℃, respectively. This means that the Tmin increased by 6.07℃. During the same period, the Tmax increased from 24.05 to 30.46℃ (i.e., increased by 6.41℃). To determine the percentage change, the increase in Tmin (6.07℃) is divided by the initial Tmin value (10.45℃), resulting in a percentage change of 58.09%. On the other hand, the increase in Tmax (6.41℃) is divided by 24.05℃, resulting in a smaller percentage change (26.65%).

3.2.3 Projection of precipitation

The observed and projected rainfall maps are displayed in Figures 11 to 13. The annual accumulated rainfall over the LZRW was divided into five classes, ranging from 375-475 mm (the lowest) to 775-875 mm (the highest).

tu_pian_9.png

Figure 11. Rainfall results for the observed and projected maps (2041-2060)

tu_pian_10.png

Figure 12. Rainfall results for the observed and projected maps (2061-2080)

tu_pian_11.png

Figure 13. Rainfall results for the observed and projected maps (2081-2100)

Figure 11 shows the comparison between the observed rainfall map and the projected maps for the period 2041-2060. The observed map shows the first four classes. The class 375-475 mm covers 27.23% of the basin, and it extends mainly in the western parts of the watershed. The class 475-575 mm covers the majority of the LZRW (40.22%), with a spread across the central and eastern parts. The class 575-675 mm represented 29.29% of the catchment area. The highest rainfall occurred in the northern region, where the class 675-775 mm was extended. Both of the projected maps revealed increasing rainfall; the higher rise was observed in the SSP2-4.5 map. The spatial extent of the lowest class (375-475 mm) decreased by 21.00 and 18.66% in the SSP2-4.5 and SSP5-8.5 maps, respectively, compared to the observed map. Meanwhile, the spatial coverage of the class 775-875 mm increased from a non-existent state in the observed map to 3.38 and 0.48% in the SSP2-4.5 and SSP5-8.5, respectively.

Figure 12 depicts the observed map and the results of the second projection period. The SSP2-4.5 exhibited increasing rainfall. The lowest rainfall class reduced by 28.76% in the projected map compared to the observed map, while the highest class rose from zero in the observed map to 0.82% in the projected map. In contrast, the SSP5-8.5 and the observed maps had approximately similar rainfall distribution.

The comparison between the observed map and the projected maps of the third future period is displayed in Figure 13. A significant increase is observable in the projected maps. Most of the rainfall increment occurred in the northern parts of the LZRW, where the class 675-775 mm increased from 3.35% in the observed map to 5.83% in the SSP2-4.5 map, and similarly, the class 775-875 mm rose from zero to 0.48%.

Figure 14 illustrates the precipitation trend during the future periods. One can observe that the projection scenarios for all periods exhibited rainfall increment. The SSP2-4.5 predicted the most significant rise of 9.35% during the first period (2041-2060). This increment decreased in the subsequent periods. On the other hand, the SSP5-8.5 showed a fluctuating increase trend; the percentage changes during the first, second, and third periods were 3.18, 0.02, and 2.59%, respectively.

tu_pian_12.png

Figure 14. Percentage change of rainfall for the projection periods

The results obtained align with studies on climate change undertaken in Iraq and neighboring nations [1, 2, 6]. Global warming, induced by increasing GHG emissions, will ultimately lead to elevated temperatures in the oceans and rising levels of water, increasing water evaporation and boosting global humidity and precipitation [54]. Nonetheless, it must be acknowledged that the predicted rise in rainfall may also pose some challenges, including an elevated risk of flooding and soil erosion, depending on the intensity, duration, and season of the rainfall [55].

The authors express gratitude for the assistance provided by their institutions. The governmental, Corporate, and non-profit organizations did not provide distinct funding for this study.