Influences of climate change on area variation of Qinghai Lake on Qinghai-Tibetan Plateau since 1980s

Qinghai-Tibetan Plateau is the most sensitive region to global warming on Earth. Qinghai Lake, the largest lake on the plateau, has experienced evident area variation during the past several decades. To quantify the area changes of Qinghai Lake, a satellite-based survey based on Landsat images from the 1980s to 2010s has been performed. In addition, meteorological data from all the seven available stations on Qinghai-Tibetan Plateau has been analyzed. Area of Qinghai Lake shrank ~2% during 1987–2005, and then increased ~3% from 2005–2016. Meanwhile, the average annual temperature increased 0.319 °C/10 y in the past 50 years, where the value is 0.415 °C/10 y from 2005–2016. The structural equation modeling (SEM) shows that precipitation is the primary factor influencing the area of Qinghai Lake. Moreover, temperature might be tightly correlated with precipitation, snow line, and evaporation, thereby indirectly causes alternations of the lake area. This study elucidated the significant area variation of water body on the Qinghai-Tibetan Plateau under global warming since 1980s.

anthropogenic changes of the lakes. Therefore, Qinghai Lake is an ideal candidate to explore the responses of waterbody alternations to climate variations.
The aim of this study is to investigate the correlations among climate change, average altitude of snow line, and the area of Qinghai Lake based on spatial analysis of remotely sensed imagery and meteorological data. The remote sensing images of Qinghai Lake during 1987-2016 were collected to calculate its area. Meanwhile, the meteorological data of Qinghai-Tibetan Plateau during 1970-2016 was collected and analyzed.

Results
Climate fluctuation. Between 1970 and, annual precipitation for the seven meteorological stations in Qinghai-Tibetan Plateau increased by 10.7 mm per decade. Compared with temperature, precipitation variation on Qinghai-Tibetan Plateau had evidently spatial heterogeneity. Annual precipitation variedamong the seven stations, e.g., the annual precipitation changes of Dachaidan, Dulan, Gangcha, Golmud, Xining, Yushu and Lhasa were 3.295, 15 Evaporation is an important factor related to water budget on Qinghai-Tibetan Plateau, which is affected directly by climate change 3 . Annual pan evaporation of Dachaidan, Dulan, Gangcha, Golmud, and Xining has decreased by 6.255, 11.059, 1.343, 16.21 and 19.433 mm/y respectively during the period of 1970-2003 (Fig. 4).      (Fig. 6). Xining station shows the lowest increase of temperature in the past 50 years, probably due to that it is on the edge of the plateau.
Dynamic changes of Qinghai Lake area and SEM results. Remote sensing images in the period of mid-July to mid-August were further selected to study the changes among years from 1987 to 2016. In Table 1 The rationality test of SEM includes coefficient rationality, significance test, and goodness of fit test. The overall fitness of the model is expressed by normed fit index (NFI), Goodness of fit index (GFI), Comparative fit index (CFI), Incremental fit index (IFI), and Root mean square error of approximation (RMSEA). Fitting results of the model are given directly by AMOS software (Table 2). It indicates the model fitted well, and the relationships among the lake area, temperature, precipitation, evaporation, and snow line can be well reflected.
The maximum likelihood method was applied to evaluate the model. As shown in Fig. 7, the lake area is affected by precipitation, snow line, evaporation, and temperature. It confirmed a strong positive correlation

Discussion
Change of lake area is tightly related to climate and environment evolution 16,17 . The alternations of the Qinghai Lake area, ideally, mostly reflect the natural processes 17 . Precipitation, evaporation, temperature, and snow line are common climatic factors, which affect the area of Qinghai Lake more directly than other natural factors in a short term 17,18 , i.e., since 1980s in this study. Climatic characteristics of the Qinghai-Tibetan Plateau are heterogeneous based on seven meteorological stations, excluding temperature. This confirms the sensitivity Qinghai-Tibetan Plateau to global warming.
The area of Qinghai Lake had shrunk slowly in the period of 1987 to 2005. In the downward trend of lake area from 1987-2005, precipitation increased slightly, along with lower snow line and quite higher evaporation. Then, a rapid increase of lake area was observed from 2005-2016. Within this period, it also had increasing precipitation, but with elevated snow line and reduced evaporation. In addition, there is a certain fluctuation of lake area from 2014 to 2016, which is consistent with the precipitation changes (also with evident fluctuation) in Gangcha station (the nearest meteorological station to Qinghai Lake). Therefore, the change of the lake area on the plateau should be attributed to combined climatic variables. As an important response area of global climate change, the temperature was rising in Qinghai-Tibetan Plateau, especially in recent years, which has also been revealed in the previous studies 5, [19][20][21][22] . Moreover, although temperature has no significantly direct influence on the area of Qinghai Lake, it drives the changes of snow line, thereby affecting the lake area. Furthermore, evaporation usually contributes to outflow of lakes, but it has the low magnitude on the plateau. It is also possible that the vegetation and wind, in addition to water bodies, could influence evaporation. Therefore, evaporation has limited effect on the area of Qinghai Lake based on our results. As the SEM model shows that precipitation is the primary factor of Qinghai Lake area variation. It is consistent with the previous study 23 , which reconstructed the fluctuation of lake level in the last 600 years and the results showed that there was a tight relationship between lake level and precipitation. Rivers, whose runoff due to glaciers and precipitation, plays important roles in the supply of natural water of Qinghai Lake. Although surface runoff accounts for 46.59% of the water in Qinghai Lake, D-O water isotope experiments had demonstrated that the primary source of runoff was precipitation, not glacial meltwater 18 . Therefore, although evaporation, snow line, and temperature have certain influences on Qinghai Lake area, the primary driver of the change in Qinghai Lake area is the increase of precipitation. Meanwhile, all the available data of remote sensing images and meteorological information related to the research were collected and analyzed. More available data, including both remote sensing and climatic data, would promote our further understanding of the global climate changes on the Qinghai-Tibetan Plateau.

Methods
Data collection. Meteorological data of all available bases of China Meteorological Network (http://data. cma.cn/) in Qinghai-Tibetan Plateau was collected, i.e., Dachaidan, Dulan, Gangcha, Golmud, Xining, Yushu, and Lhasa. The locations of the seven stations are shown in Fig. 1 and Table 3. Annual mean temperature and annual precipitation data were collected from 1970 to 2015 (all the available data). In addition, evaporation data from 1970 to 2016 were also analyzed. From 1970 to 2003, annual pan evaporation was provided. After 2003, only daily evaporation was available and different evaporator were applied. Thus, average daily evaporation from June to August after 2003 was selected to match the following GIS data.
Landsat TM/ETM+/OLI images covering the studied area from 1987 to 2015 were selected, which were downloaded from USGS website (http://landsat.usgs.gov/). Data of several years cannot be obtained due to cloud interference, which is shown in Table 1. Snow line of Gangshika peak was analyzed by GIS images during July-August, when Qinghai has the highest temperature. Two Gaofen-1 WFV4 images were selected due to the lack of appropriate images of the lake (year 2013 and 2016), which can be gained from RS Cloud Mart (http:// www.rscloudmart.com/en). Additionally, remote sensing images of all months were selected to study the changes of Qinghai Lake for reference and calibration. Data analysis. EVNI (Version 5.0) and ArcGIS software (Version 10.2 for Desktop) were applied to process the Landsat TM/ETM+/OLI images (L1T) for lake area calculating. Modification of the normalized difference water index (MNDWI) 24 and reclassification were selected to calculate the area of Qinghai Lake by Landsat TM/ETM+/OLI images. Gaofen-1 WFV4 images require pretreatment, including radiometric calibration, ortho-rectification, and atmospheric correction. Band ratio (band 2/band 4) was applied to calculate the area via GF-1 WFV4 images.
Bands 5, 4, and 3 (for R, G and B respectively) of Landsat TM/ETM+ were selected for false color composites (bands 6, 4, and 3 for Landsat8 OLI), so that snow and glaciers could be differentiated. More than 20 points were selected along the snow line of Lenglong mountain range, with equal points distributed on the south and north slope. In addition, distances between the two adjacent points on the mountain range were relatively equal. Altitudes of the points were extracted from ASTER GDEM V2 with the resolution of 30 m.
Structural equation modeling (SEM) integrates factor analysis and path analysis 25 . It has been widely applied to ecological and environmental studies 19 . The model was constructed by IBM SPSS AMOS 21 software. The model of this study was established by the potential relationships between the five possible variables, i.e., area of Qinghai Lake, temperature, precipitation, evaporation, and snow line. The data of precipitation, evaporation, and temperature were collected from Gangcha meteorological station, which is the nearest meteorological station to Qinghai Lake. The snow line data was collected from Gangshika peak. Then, the stochastic relationship between variables was tested by model reasonableness test.  Table 3. Coordinates of the seven available meteorological stations on Qinghai-Tibetan Plateau.