**2. Study Area**

Qingtu Lake, the terminal lake of the Shiyang River, is located on the northeast edge of Minqin County, Gansu Province, between 39◦04 and 39◦09 N latitude and 103◦36 and 103◦39 E longitude (Figure 1). The area has a temperate continental arid desert climate. The mean annual temperature is 8.87 ◦C; the maximum temperature is 37.8 ◦C and the minimum temperature is −29.5 ◦C. The mean annual precipitation is 110 mm, with 73% of this occurring in June, July, and August. The mean annual evaporation is 2644 mm [20]. The main vegetation species include *Phragmites australis*, *Nitraria tangutorum*, *Suaeda glauca*, *Haloxylon ammodendron*, and *Kalidium foliatum*. Accompanying shrubs include *Lycium ruthenicum*, *Artemisia sphaerocephala*, and *Kalidium foliatum*. Herbaceous plants include those such as *Salsola ruthenica*, *Zygophyllum fabago*, *Cynanchum sibiricum*, *Salsola collina*, and *Agriophyllum squarrosum*. They are all typical desert vegetation types [21]. The water of Qingtu Lake is mainly used to maintain the ecological system balance around the wetland, especially for improving the vegetation ecosystem [20,22]. The water in Qingtu Lake is weakly alkaline and Na–SO4 (Cl) type, with high salinity and total dissolved solids (TDS) [23]. The lithology is shown in the cross section (Figure 2). The sediments in this area consist of mainly sand, silt, and loam [24].

**Figure 1.** Location of the study area (**a**) and monitoring locations in the Qingtu Lake wetland (**b**).

**Figure 2.** Simplified geological cross section from V01 to V06 (locations are shown in Figure 1b).

During the Western Han Dynasty, Qingtu Lake covered an area of 4000 km2. Due to the increase of water consumption in the upper reaches of Qingtu Lake, the surface area of the water body of Qingtu Lake began to gradually diminish after 1924. In the 1950s, the construction of the Hongyashan Reservoir caused the Shiyang River to dry quickly [25]. Since the 1960s, there have been a series of ecological problems in the lower reaches of Shiyang River, such as the decline of the groundwater level, vegetation degradation, and soil salinization [23]. In order to restore the regional ecological environment, the local governmen<sup>t</sup> decided to begin transferring ecological water from Hongyashan Reservoir to Qingtu Lake in September 2010. The water is transferred through irrigation channels from September to October of each year. It is a typical ecological water transportation system, carried out during the non-growing season.

#### **3. Data and Methods**

#### *3.1. Data Sources and Preprocessing*

Ten monitoring sites were constructed in July and August 2018. Instrumentation at the V01–V10 locations was implemented using 5TE probes developed by the Decagon Company, with a precision of 3% for monitoring the soil water content at different depths along the soil profile (Figure 1b). The G01–G10 wells with screens at depths between 1 and 3.6 m were installed with Canadian Solinst3001 titanium gold edge water-level loggers with a precision of 0.05% for recording the groundwater level, water salinity, and temperature (Figure 1b). The soil water content and parameters in the groundwater were recorded every 30 min. Meteorological data in the study area (temperature, precipitation, air pressure) were downloaded from the China Meteorological Data Sharing Network (http://data.cma.cn).

Considering the timing of water transfer from Hongyashan Reservoir to Qingtu Lake and vegetation evolution, the remote sensing data used in this study were acquired from Landsat series satellite data, including Landsat 8 OLI, Landsat 7 OFF, and Landsat 4-5 TM from 2009 to 2018, with a resolution of 30 m, a synthetic image resolution of 15 m, and a temporal resolution of 16 d. The remote sensing data from after May 2013 were from the Landsat 8 OLI series, while data from before May 2013 were from the Landsat 7 OFF or Landsat 4-5 TM series. The data from the Landsat 7 OFF series were damaged; the images showed data overlap and about 25% of the data was lost. All of the satellite data of the Landsat 7 OFF series used in this study were repaired by the strip removal method [26]. The available Landsat data from before May 2013 were limited, and there were nine images from 2009 to 2018.

All the image data were preprocessed with ENVI 5.2 for radiometric calibration, layer stacking, atmospheric correction, geometric correction, and cropping [27].
