A constructed long-term (1950-2014) runoff dataset in Balkhash Lake basin

We collected gauge runoff data from 13 hydrological stations in all five rivers in the BLB from the hydrological yearbook and literature. Seven stations are located in the Ili River: the Ushjarma (1939–1985), Kapchagay (1930–2000), Dobyn (1929–2014), and Yamate (1949–1986) stations in the mainstream, and three stations including the Tekes (1956–1979), Kunes (1960–1979), and Kashi (1953–2015) stations in the mountainous reaches (the Tekes, Kunes, and Kashi Rivers). For the East Rivers, two stations are located in the Karatal River: the downstream Ushtobe station (1939–1985) and the upstream Tekeli station (1940–1985). Two stations are located in the Lepsy River: the downstream Lepsy station (1936–1985) and the upstream Lepsinsk station (1936–1985). Runoff data were acquired from the upstream Chann station (1936–1981) in the Aksu River and from the downstream Ayaguz station (1949–1986) in the Ayaguz River. To construct a long-term annual runoff series for all 13 sub-basins, we first developed linear regression models relating observed gauge runoff (R) to the corresponding yearly aggregated gridded runoff from both GLDAS-2.0 (RGLDAS) and ERA5-Land (RERA5-Land) within their common period of overlap with the gauge data, as formulated in Equation 1 (where a₁, a₂, and R₀ are regression coefficients). The regression equations were then used to infill missing gauge-recorded runoff values based on RGLDAS and RERA5-Land inputs, thereby reconstructing a continuous runoff series. We determined that combining both GLDAS-2.0 and ERA5-Land products yielded more accurate runoff reconstructions than using either product alone. Since the overlapping period of GLDAS-2.0 and ERA5-Land spans from 1950 to 2014, we reconstructed runoff over this interval to maximize reliability.