4 DISCUSSION

4.1 Literature comparison of the main attribution results

The main attribution of changes in AET and TWSA in the EIB were semi-quantified based on the water balance in the closed basin. The main contribution to the decrease in the TWSA was the increase in the AET. The incremental AET consumption of other water sources and the precipitation increase contributed approximately 70% and 30% to the increased AET, respectively. Although few studies have focused on the entire EIB, there have been many studies focusing on the attribution to changes in TWSA at the basin scale; these results were compared with the results from this study. For example, the Caspian Sea level has been gradually decreasing over the past 20 y. Some studies have attributed this decrease to meteorological factors (van Dijk, Renzullo, Wada, & Tregoning, 2014), while others suggest that evaporation from the sea is the main impact factor (Chen, et al., 2017). Other studies have suggested that it is these two factors combined with agricultural irrigation diversions (Rodell et al., 2018). According to the water balance of the closed basins, this study demonstrates that the main factor causing the decreasing sea level was the increase in AET. The increase mainly includes an increase in water surface evaporation and agricultural irrigation diversion. For example, the Volga River delivers roughly 80% of its runoff to the Caspian Sea; however, there are 11 dams located in the basin that ensures a steady water supply for crop irrigation (Avakyan, 1998; Rodell et al., 2018). There was a negligible correlation between crop production and precipitation, suggesting that irrigation effectively mitigates the impact of drought. Based on the long-term observation data from 1961 to 2020, there was a clear dry period over the past 20 y (Figure 6). Therefore, the runoff into the Caspian Sea has decreased due to the inevitable increase in agricultural irrigation diversion throughout this dry period. In addition, the reservoir capacity of dams in the CSB exceeds 75% of the total inflow, and the total capacity of reservoirs is 223 km3; these dams mostly occur in the Volga River basin (Akbari et al., 2020). Although precipitation in the headwaters of the Volga River is increasing, the impact of increased precipitation on runoff into the sea may be weakened as a result of water storage regulation of reservoirs during the dry seasons. Therefore, the precipitation increase in the headwaters has a limited impact on the Caspian Sea level.
The extent to which the increase in AET has contributed to the TWSA decline in the ASB from 2002 to 2020 was greater than the decreasing precipitation; this is consistent with the conclusions of previous studies (Yang et al., 2020). The increase in the AET consumption of other water sources contributed >90% to the increase in AET, while the precipitation increase contributed <10%. The increase in AET consumption of other water sources largely involves the increase in water surface evaporation caused by higher evaporation rates and the increase in AET from irrigation diversion. Although the impact of agricultural irrigation diversion on the water storage of the basin has been gradually weakening post-2005 (Wang et al., 2020), agricultural irrigation diversion remains the most important water resource problem in this region. There is a still a need to improve the irrigation efficiency, crop use efficiency, and water resource management in this region. The precipitation decrease was the main impact factor driving the decrease of the TWSA in the IIRB; this is consistent with previous studies (Khaki et al., 2018).
The significant increasing trend in the TWSA for the QB and QPB was consistent with previous studies (Bibi et al., 2019; Meng et al., 2019; Liu et al., 2019); however, the main causes attributed to this trend differ from those identified in previous studies. The increasing TWSA in the QB is mainly caused by the precipitation increase, which contributes >90 % to the increase in AET. In the QPB, the increasing TWSA has largely been attributed to the decrease in AET.
Similar to the ASB results, the increase in AET was also the main cause for the decrease in the TWSA for the TaRB; this differs from the results of previous studies (Yang et al., 2017; Xu et al., 2019). The extent to which the increase in AET consumption of other water sources contributed to the increasing AET (>90%) was much higher than that of precipitation (<10%). The increase in the AET consumption of other water sources was mainly due to the elevated consumption of water resources by human activities and the increase in melt water from glacier retreat.
There were non-significant decreasing trends in the TWSA of the BLB and ISB for 2002–2020; this is contrary to the rising water levels in Balkhash Lake and Issyk-Kul Lake reported in previous studies (Alifujiang et al., 2017; Duan et al., 2020) and may be due to the differences in the study period. These two study periods for the previous works ended in 2012 and 2013, respectively, while this study period ended in 2020. The precipitation decrease and the increase in the AET consumption of other water sources from elevated evaporation rates accounted for approximately half of the AET changes in the BLB and ISB.
Based on the results of five GRACE TWSA products, there was a significant decreasing trend in the TWSA for the GHC for 2002–2020. This differed from the results obtained at other time periods when only one product was used (Cao et al., 2018; Wang et al., 2020). Similar to the EIB and most of its closed basins, the decreasing TWSA of the GHC was mainly due to the increase in AET; >60% of the precipitation increase contributed to the AET increase.

4.2 GRACE TWSA products and uncertainties

It was difficult to determine which product was more suitable for the EIB because of the lack of observational data for validation. However, the area of the Caspian Sea is sufficiently large and thus, may be considered a typical region to evaluate TWSA products. Chen et al. (2017) found that the observed level of the Caspian Sea decreased by -67.2 cm/10a for 1996–2015. This was more consistent with the two mascon products (JPL-v2 and GFSC-v1), compared to the three spherical harmonic coefficients products (CSR-v3, GFZ-v3, and JPL-v3). This was because the TWSA of the Caspian Sea from the two mascon products decreased by >-60 cm/10a for 2002–2020, as consistent with the observed decrease. The TWSA of the three other products was <-33 cm/10a, which is much lower than the observed value; this indicates that the mascon products in the Caspian Sea region outperforms the spherical harmonic coefficients products. Some studies have also highlighted the many advantages of the mascon products for hydrologic studies (Scanlon et al., 2016; Rodell et al., 2018). Overall, the five TWSA products showed high consistency in most other regions of the EIB.
The uncertainties associated with the AET simulation were mainly due to uncertainties in the input data and during simulation processes. As the input precipitation and TWSA data were grid data, the actual basin boundary did not coincide with the grid boundary. In this study, the area weighting of the basin in the boundary grids was used to reduce the uncertainty caused by the lack of alignment with the two boundaries. Many studies have shown that GPM precipitation products have high accuracy at regional scales (Tang, Ma, Long, Zhong, & Hong, 2016; Le, Lakshmi, Bolten, & Bui, 2020; Islam, Yu, & Cartwright, 2020); this was also the case in this study, in which the GPM precipitation products were highly consistent with the station observations in the EIB. Although the five TWSA products were generally highly consistent in the EIB, there were still some differences (e.g., Figures 8(c) and 8(d)). This study used the mean value series of five TWSA products to simulate the monthly AET series, causing uncertainties in the simulated AET. In addition, we also compared the median value series of the five products, finding marginal differences between the mean and median series. Monthly AET series simulated by the two series were also highly consistent. The accuracy of GRACE TWSA data directly determined the accuracy of the related applications. As such, further research on GRACE TWSA inversion is required for hydrology and water resource applications.
In the simulation of monthly AET, ΔS represents the change in the water storage over a month (i.e., the difference between the water storage at the end and beginning of the month). However, the GRACE TWSA data represents the mean water storage in the month. In this study, the difference between the TWSA of the previous month and the next month was used to represent the ΔS in the basin; this may also have caused uncertainties. This study simulated the monthly AET series for a closed basin, in which the runoff process within the basin was ignored; however, for a finer temporal scale or an exorheic basin, the runoff process within the basin must be considered.