3.1 Results of implementation and calibration of SWAT model
Table 5 presents the values obtained from the model performance evaluation index in runoff simulation for all studied stations. As can be seen, the Nash-Sutcliffe coefficient and the coefficient of determination are close to one in all hydrometric stations except the five stations of Sarab Seyed Ali, Pulchehr, Noorabad, Hamidiyeh and Pai-e-Pol, in the calibration and validation periods. In other words, the coefficients obtained in the other 6 stations are more than 0.5, which indicates that the model has an acceptable ability to simulate runoff. In Sarab Seyed Ali, Polchehr and Noorabad stations, the SWAT model could not simulate the runoff well, which could be due to the location of these stations in the high areas of the basin and tributaries, and as a result, it is snow-covered. Lack of proper distribution of meteorological stations in these areas makes the model unable to properly simulate snow runoff. In Hamidiyeh and Pai-e-Pol stations, the SWAT model could not establish an acceptable relationship between the observed and simulated runoff due to the impact of the construction of Karkheh Dam on the hydraulic process of the river flow. The results of calibration and validation of the SWAT model (observational data, simulated and 95% uncertainty estimation band) as a sample for Afrineh hydrometric station are shown in Figs. 2 and 3, respectively.
Table 5
Statistical comparison of monthly simulated and observational discharges in the calibration period
Hydrometric station | Statistical period length (years) | Nash - Sutcliffe | The coefficient of determination |
Validation | Calibration | Validation | Calibration | Validation | Calibration |
Hamidiyeh | 9 | 20 | -0.04 | -0.19 | 0.22 | 0.02 |
Pai-e-Pol | 9 | 20 | 0.18 | 0.02 | 0.29 | 0.47 |
Pol-e-Zal | 9 | 20 | 0.68 | 0.56 | 0.8 | 0.87 |
Pol-e-Dokhtar | 9 | 20 | 0.75 | 0.73 | 0.75 | 0.86 |
Afrineh | 9 | 20 | 0.7 | 0.74 | 0.73 | 0.85 |
Cham Anjir | 9 | 20 | 0.7 | 0.76 | 0.75 | 0.88 |
Sarab Seyed Ali | 9 | 20 | 0.38 | 0.25 | 0.52 | 0.42 |
Noorabad | 9 | 20 | 0.4 | 0.22 | 0.4 | 0.45 |
Doab Merk | 9 | 20 | 0.61 | 0.68 | 0.68 | 0.84 |
Kakarza | 9 | 20 | 0.65 | 0.73 | 0.68 | 0.86 |
Polchehr | 9 | 20 | 0.42 | 0.11 | 0.42 | 0.22 |
3.2 Results of validation of simulated data by CCT implementation
Table 6 shows the validation results of the models used to microscale the GCM models for precipitation data. Tables similar to this table were created for maximum temperature and minimum temperature. These tables are based on the average accuracy of the models used in each of the 17 meteorological stations. Based on the statistical indicators shown in these tables, the model with the highest R2 and the lowest MAE and RMSE was chosen as the best and most reliable model. Hence, as shown in Table 6, the GFDL-ESM2 had the best performance in simulating precipitation in the Karkheh river basin under both RCP8.5 and RCP 2.6 scenarios. Based on the maximum temperature table, the HadGEM2-ES model is more suitable for simulating this data than the other models in both scenarios. The minimum temperature table also shows that the MIROC model is the most efficient in both scenarios for simulating this data. In all three tables, the other models, although having acceptable values of R2, are not recommended due to the relatively high degree of calibration characteristics.
Table 6
Validation of the CCT simulated precipitation in Karkheh River Bains
Scenario | Model | RMSE | MAE | R2 |
RCP2.6 RCP8.5 | GFDL-ESM2 | 3.2 4.1 | 1.1 1.5 | 0.9989 0.9971 |
RCP2.6 RCP8.5 | HadGEM2-ES | 4.5 5.8 | 2.3 3.5 | 0.9906 0.9908 |
RCP2.6 RCP8.5 | IPSL-CM5A-LR | 7.2 6.7 | 5.6 4.2 | 0.9901 0.9902 |
RCP2.6 RCP8.5 | MIROC | 5.6 8.1 | 3.2 5.4 | 0.8946 0.8934 |
RCP2.6 RCP8.5 | NoerESM1-M | 5.3 4.6 | 3.4 2.6 | 0.7936 0.8961 |
3.3 Results of the CCT uncertainty evaluation
Figure 4 shows the average weight of five GCM models with two scenarios, RCP2.6 and RCP8.5, to estimate future precipitation changes. Similar shapes were created for the maximum temperature and minimum temperature in the future. As given in these tables, the total weight of the variables in different models and scenarios is equal to one per month. The results of weighted GCMs in precipitation prediction show that the weight of all GCMs is less than 0.9. Except for the three summer months, the GFDL-ESM2 model had the highest weight under both scenarios. Based on the maximum temperature results, GCMs weigh less than 0.7. In all months, especially in the three summer months, HadGEM2-ES had the highest weight in estimating the maximum temperature under both scenarios. The minimum temperature results show that GCMs weigh less than 0.45, and the MIROC model, with a slight difference, ranked as the best predictor of minimum temperature.
In general, changes in the weight of GCM models showed that these models differed significantly in estimating maximum temperature and precipitation than minimum temperature. In other words, GCM models have a more similar potency in estimating minimum temperature than maximum temperature and precipitation. Precipitation, maximum temperature and minimum temperature produced using the best-performed models and scenarios were analyzed to investigate the future effect of climate change on temperature and precipitation.
3.4 Precipitation changes
Precipitation was generated using the GFDL-ESM2 model for the near (2071 − 2043) and distant future (2100 − 2072) under two scenarios of RCP2.6 and RCP8.5. Figure 5 shows the changes in the mean monthly precipitation of the next two periods compared to the base period. To calculate these changes, the predicted precipitation for the near and distant future periods was subtracted from the mean of these precipitations in the base period. The vertical lines added to the graphs indicate the standard error rate of mean precipitation at different meteorological stations. The results showed that the highest increase in precipitation in the distant future is projected to occur under the RCP2.6 scenario in January (4.3 mm) which will be about 3.6 mm in the near future in January. The largest precipitation increase under the RCP8.5 scenario was projected to be 3.6 mm for the near future and 1.7 mm for the distant future both in January. As shown in the figure, the amount of precipitation will decrease in agriculturally-important months of May, June, July, August and September during the next two periods as compared to the base period. In general, precipitation is projected to decrease from late spring to early autumn and increase in winter and early spring during both near and distant futures.
To better understand how precipitation changes, seasonal and annual variations of precipitation are shown in Table 7. The results of seasonal precipitation variations show that winter will experience the highest precipitation increase of 2.31 and 2.3 mm in the near future and 2.99 and 0.99 mm in the distant future under RCP2.6 and RCP8.5 scenarios, respectively. Summer is expected to have the highest precipitation reduction of 0.99 and 1.1 mm in the near future and 1.04 and 1.05 mm in the distant future under RCP2.6 and RCP8.5 scenarios, respectively. Annual changes in precipitation also indicate an increase in precipitation between 0.73 to 0.59 mm in the near future and 0.89 to 0.26 mm in the distant future under RCP2.6 and RCP8.5 scenarios, respectively.
Table 7
Seasonal and annual precipitation changes based on GFDL-ESM2 model under RCP2.6 and RCP8.5 emission scenarios
Emission scenario | Period | Winter | Spring | Summer | Autumn | Annual |
RCP2.6 | Near future | 2.31 | 0.54 | -0.99 | 1.07 | 0.73 |
Distant future | 2.99 | 0.85 | -1.04 | 0.74 | 0.89 |
RCP8.5 | Near future | 2.3 | 0.42 | -1.1 | 0.74 | 0.59 |
Distant future | 0.99 | 0.47 | -1.05 | 0.65 | 0.26 |
The temperature changes were calculated similarly to the methodology employed for measuring changes in precipitation. The results showed that the highest maximum temperature increase of 1.63 ° C is projected to occur in July of the near future under the RCP2.6 scenario. This amount in July of the distant future is expected to be 1.6°C. Under the RCP8.5 emission scenario, the highest maximum temperature increase was projected to occur in the July of the distant future by 1.76 ° C. In the near future, this amount is expected to be 1.6°C in July. The maximum temperature will also increase under both scenarios as well as in both future periods from March to October. The highest minimum temperature increase of 0.92 ° C will occur in July of the distant future period under the RCP2.6 scenario. This increase in July of the near future is expected to be 0.91° C. The highest minimum temperature increase of 1.1 ° C was found to be in July of the distant future under the RCP8.5 emission scenario. This value July of the near future is projected to be 0.95 ° C. Generally, the minimum temperature increase is projected to occur in April to October of both future periods compared to the base period. Both minimum and maximum temperatures will increase from early spring to early autumn and decrease from late autumn to late winter in both future periods.
The results of seasonal changes indicate that the highest maximum temperature increase of 1.55 and 1.57 ° C in the near future and 1.53 and 1.68 ° C in the distant future, under the RCP2.6 and RCP8.5 scenarios, respectively, will occur in summer. Moreover, winter will experience the highest maximum temperature drop which is projected to be 0.41 and 0.39 ° C in the near future and 0.38 and 0.28 ° C in the distant future under the RCP2.6 and RCP8.5 scenarios, respectively. Annual changes also indicate an increase in maximum temperature between 0.58 to 0.6 ° C in the near future and 0.57 to 0.72 ° C in the distant future under the RCP2.6 and RCP8.5 scenarios, respectively.
Similar to the maximum temperature, summer will experience the highest minimum temperature increase of 0.84 and 0.87 ° C in the near future and 0.84 and 0.98 ° C in the distant future under the RCP2.6 and RCP8.5 scenarios, respectively and winter is expected to have the highest minimum temperature drop of 0.42 ° C in the near future and 0.42 (RCP2.6 scenario) and 0.33 ° C (RCP8.5 scenario) in the distant future. Annual changes also indicate an increase in minimum temperature between 0.21 and 0.22 ° C in the near future and 0.22 to 0.32 ° C in the distant future under the RCP2.6 and RCP8.5 scenarios, respectively.
3.5 Results of the study of the relationship between meteorological and hydrological drought in the base period
In this study, SPEI and SRI indices were used to investigate the relationship between meteorological and hydrological droughts in Karkheh watershed. These indices were calculated for 3 meteorological stations and 11 hydrometric stations in 3, 6 and 12 month periods. Afterward, the drought of each meteorological station together with those of adjacent hydrometric stations was statistically analyzed. Based on these results, the SPEI index for each of the 17 meteorological stations on a quarterly scale showed more positive values and with increasing the length of the period, the number of dry periods increased and this index showed more negative values. The longer the time scale, the lower was the SPEI fluctuations and changes. The SPEI diagrams of all meteorological stations exhibited the most severe drought periods during 1992 − 1991, 2007 − 2006, 2012, 2015 and 2018.
Based on the SRI index results of the Hamidiyeh, Pol-e-Zal, Pai-e-Pol, Afrineh and Pol-e-Dokhtar hydrometric stations, the number of wet periods was higher than the dry periods in 3 and 6-month scales. As the length of periods increased, the SRI index showed negative values and periods of drought such that since 2007, the index was negative for most 12-month periods. Based on this index in other hydrometric stations including Cham Anjir, Sarab Seyed Ali, Kakarza, Doab Merk, Noorabad and Polchehr in 3 and 6-month scales, the number of wet periods was larger than dry periods. As the length of periods increased, the SRI index showed negative values and drought periods from 2007 onwards, while in 2013 and 2017, the index showed positive values and wet periods.
In general, these results indicate that the severity and duration of hydrological drought in the region were more than meteorological drought. The duration of the meteorological drought in most stations in the region was less than one to two years, while the hydrological drought began in 2007 and continued until the end of the statistical period under study. Figure 6 shows an example of changes in the SPEI index relative to SRI at Ahvaz and Hamidiyeh stations.
3.6 The status of meteorological and hydrological drought across the Karkheh watershed
To investigate the meteorological and hydrological drought situation in the study area, the annual weighted mean of SPEI and SRI indices of all studied stations was determined over the 29-year statistical period based on the area affected by each station and using the Theissen method (Fig. 7). According to the results of the drought situation in the region, the SPEI index showed more negative values than the SRI index at the beginning of the statistical period, while the SRI index followed a decreasing trend in the last years of the statistical period.
Based on the mean value of drought indices, the average number of meteorological droughts in the region was 14 years, the most severe of which occurred in 1991. The number of periods of hydrological droughts was 10 years with the most severe one being occurred in 2009. Based on the results of changes in the indices, it can be concluded that at the beginning of the study period (2006–2009), meteorological drought had minimal effect on hydrological droughts and the SRI index showed positive values in most years. In the following years, although the severity of meteorological droughts was not higher than those of the previous years, the severity of hydrological drought increased with negative SRI values in most years.
3.7 The results of the relationship between meteorological and hydrological drought in the futureReferences
In the next period, as in the baseline period, meteorological and hydrological drought indices of Karkheh watershed in 3, 6 and 12 months periods were calculated. For example, the comparative graphs of these indicators are presented in Fig. 8. These figures were obtained for all 5 climate models and both RCP2.6 and RCP8.5 scenarios in both the near and distant future periods. Based on these results, in the next two periods, the SPEI index on a quarterly scale shows more negative values and as the length of the period increases, the number of dry periods decreases and this index shows more positive values. The longer the time scale, the less the SPEI fluctuations and changes. For the near future, in the SPEI diagrams of all models and scenarios, drought periods with more intensity during the years 2046 − 2045, 2051 − 2050, 2065, 2069 and 2071 will be observed. For the distant future, this situation will occur in the late 70s, 80s and 90s.
In the case of the SRI index for the next two periods on a three- and six-month scale, the number of wet periods is greater than the number of dry periods. As the length of periods increases, this index shows negative values and periods of drought; as in most twelve-month periods, this index has been negative. For the near future, the SRI diagrams of all models and scenarios will show more severe droughts in the late 1940s, 1950s, and 1960s. For the distant future, this situation will occur in the early 70s, 80s and 90s.
In general, these results show that the severity and duration of hydrological drought in the region is more than meteorological drought, so that the duration of meteorological drought will not be more than one to two years.
3.8 Investigation of the correlation between meteorological and hydrological drought
The statistical relationship between drought meteorological index and hydrology for periods of 3, 6 and 12 months was calculated. The results of Pearson correlation between SPEI and SRI values of the studied stations for the highest correlation values in the base period are shown in Table 8.
Table 8
Pearson correlation between SPEI index of meteorological stations and SRI index of their corresponding hydrometric stations
Station name | SPEI-3 | SPEI-6 | SPEI-12 |
Ahvaz- Hamidiyeh | SRI-3 **0.32 | SRI-6 **0.31 | SRI-3 **0.2778 |
Bostan- Hamidiyeh | SRI-3 **0.451 | SRI-12 **0.41 | SRI-12 **0.266 |
Safiabad-Pol Zal | SRI-6 **0.174 | SRI-6 **0.21 | SRI-6 **0.36 |
Mazo-Pai Pol | SRI-12 **0.26 | SRI-3 **0.177 | SRI-12 **0.274 |
Dezful-Pai Pol | SRI-3 **0.27 | SRI-3 **0.158 | SRI-12 **0.36 |
Khorramabad-Afrineh | SRI-12 **0.22 | SRI-3 **0.44 | SRI-12 **0.51 |
Kuhdasht-Poldakhtar | SRI-3 **0.17 | SRI-3 **0.42 | SRI-12 **0.51 |
Aleshtar-Cham Anjir | SRI-6 **0.156 | SRI-6 **0.29 | SRI-12 **0.61 |
Boroujerd-Sarab Seyed Ali | SRI-6 **0.249 | SRI-3 **0.633 | SRI-12 **0.65 |
Poldakhtar-Kakarza | SRI-3 **0.21 | SRI-12 **0.17 | SRI-3 **0.167 |
Ilam-Do Ab-e-Merk | SRI-12 **0.369 | SRI-12 **0.127 | SRI-12 **0.38 |
Malayer-Noorabad | SRI-12 **0.45 | SRI-3 **0.152 | SRI-12 **0.47 |
Kermanshah-Noorabad | SRI-12 **0.66 | SRI-3 **0.51 | SRI-12 **0.75 |
Noorabad-Noorabad | SRI-6 **0.31 | SRI-12 **0.498 | SRI-6 **0.51 |
Kangavar-Polchehr | SRI-3 **0.52 | SRI-12 **0.49 | SRI-12 **0.466 |
Hamedan-Noorabad | SRI-12 **0.31 | SRI-3 **0.51 | SRI-6 **0.52 |
Ravansar-Pulchehr | SRI-6 **0.174 | SRI-12 **0.31 | SRI-12 **0.53 |
** It is significant at the level of 1%. Bold numbers indicate a higher degree of correlation. |
A similar table was created for the near future and the distant future. Based on the results obtained in the base period, the correlation between meteorological drought index and hydrological drought index in different stations studied did not have the same trend. Of the studied stations, 10 meteorological stations and their corresponding hydrometric stations (Mazo-Pai Pol, Dezful-Pai Pol, Khorramabad-Afrineh, Kuhdasht-Poldakhtar, Aleshtar-Cham Anjir, Boroujerd-Sarab Seyed Ali, Ilam-Do Ab-e-Merk, Malayer-Noorabad, Kermanshah-Noorabad and Ravansar-Pulchehr) had the highest significance between hydrological and meteorological drought indices for the 12 month-period; while the highest significance between meteorological and hydrological drought indices was quarterly for four stations of Ahvaz-Hamidiyeh, Bostan-Hamidiyeh, Poldakhtar-Kakarza and Kangavar-Polchehr. The greatest significance between the 12-month meteorological drought index and the six-month hydrological drought index was observed for three stations of Safiabad-Pol Zal, Noorabad-Noorabad and Hamedan-Noorabad.
For the next two periods, from the total of climate models and their scenarios, there was the greatest significance between the hydrological and meteorological drought indices of 12 months.
3.9 Relationship between SPEI of each period and with one, two, three and four months lag time SRI
To analyze the relationship between meteorological drought and the time of hydrological drought between the SPEI index of each period with the SRI index of the same period and with a lag of one, two, three and four months, the Pearson correlation was obtained and their correlation was determined.
The results of the study of the correlation between meteorological drought index and hydrological drought index with delay for the highest correlation values in the base period are presented in Table 9.
Table 9
Pearson correlation of SPEI and SRI indices with lags of 1, 2, 3 and 4 months
Station name | SPEI-3 | SPEI-6 | SPEI-12 |
Ahvaz- Hamidiyeh | 3 months delay **0.35 | The same period **0.31 | 1 month delay **0.178 |
Bostan- Hamidiyeh | 3 months delay **0.52 | The same period **0.41 | The same period **0.266 |
Safiabad-Pol Zal | The same period **0.174 | 2 months delay **0.25 | 3 months delay **0.4 |
Mazo-Pai Pol | 3 months delay **0.3 | The same period **0.177 | The same period **0.274 |
Dezful-Pai Pol | The same period **0.27 | The same period **0.158 | 3 months delay **0.41 |
Khorramabad-Afrineh | 4 months delay **0.25 | 1 months delay **0.64 | 4 months delay **0.71 |
Kuhdasht-Poldakhtar | The same period **0.17 | The same period **0.42 | 3 months delay **0.53 |
Aleshtar-Cham Anjir | The same period **0.156 | The same period **0.29 | 4 months delay **0.76 |
Boroujerd-Sarab Seyed Ali | The same period **0.249 | The same period **0.633 | 3 months delay **0.7 |
Poldakhtar-Kakarza | 4 months delay **0.321 | The same period **0.17 | The same period **0.167 |
Ilam-Do Ab-e-Merk | The same period **0.369 | The same period **0.127 | The same period **0.49 |
Malayer-Noorabad | The same period **0.45 | The same period **0.152 | 4 months delay **0.61 |
Kermanshah-Noorabad | 3 months delay **0.791 | 1 months delay **0.62 | The same period **0.75 |
Noorabad-Noorabad | The same period **0.31 | The same period **0.498 | 3 months delay **0.59 |
Kangavar-Polchehr | 4 months delay **0.653 | The same period **0.49 | The same period **0.466 |
Hamedan-Noorabad | The same period **0.31 | 2 months delay **0.68 | 4 months delay **0.81 |
Ravansar-Pulchehr | The same period **0.174 | 1 months delay **0.35 | 4 months delay **0.687 |
**: It is significant at the level of one percent. Bold expressions and numbers indicate a greater degree of correlation. |
According to this table, meteorological and hydrological drought occurred simultaneously in Ilam meteorological station and there was no time delay in this hydrometric station corresponding to this hydrological drought station. There was a high correlation between 12-month SPEI at 10 stations and 12-month with 1, 2, 3 and 4-month lag SRI. There was a high correlation between 3-month SPEI at 6 stations and 3-month SRI with 3 and 4-month lag time. The highest correlation among all stations and all periods was related to the 12-month SPEI of Hamedan meteorological station with 12-month SRI of Noorabad hydrometric station with 4-month lag time and a correlation coefficient of 0.81.
For the near future, meteorological and hydrological droughts have not occurred simultaneously for any of the climate models and scenarios. There is a high correlation between SPEI and 12-month SRI with a time delay of 4 months. After that, the highest correlations between these two indicators are related to delays of 3 and 2 months, respectively. The highest correlation among all models and scenarios is related to IPSL-CM5A-LR model under RCP2.6, with a delay of 4 months and a correlation coefficient of 0.8.
For the distant future as for the near future, for none of the models and scenarios did meteorological and hydrological drought occur simultaneously. There is a high correlation between SPEI and 12-month SRI with a time delay of 3 months. After that, the highest correlations between the two drought indicators are related to the 4-month delay. The highest correlation among all models and scenarios is related to IPSL-CM5A-LR model under scenario 8.5, with 3 months delay and correlation coefficient of 0.83.