The middle and upper reaches of the Yarlung Zangbo River, including the banks of the Yarlung Zangbo River and the main tributaries of the Maquan River, Nianchu River, and Lhasa River, the Himalayas in the South and Gangdise in the north of the basin, have complex landform types and a significant altitude range (Li et al., 2011). The upper reaches are plateau include wide valley landforms with elevations of about 4500-4800 m; while the midstream river valleys are broad and narrow, with elevations of about 3500-4400 m (La et al., 2014). The climate is cold and dry, the heat level in the basin is not high and the precipitation is rare, so it belongs to the cold temperate climate. The annual average temperature ranges from -1.5 ℃ to 8.5 ℃, the average temperature in the hottest month is lower than 15 ℃, the daily extreme low temperature is -44.6 ℃, and the daily maximum temperature is 30.3 ℃ (Ge et al., 2013). The annual average precipitation in the study area is 200-600 mm, and the precipitation in the rainy season (May to October) accounts for more than 80% of the annual rainfall. The ratio of evaporation to precipitation is 7.7-12.2 (Jia et al., 2008). There are various types of soils in the region, mainly alpine soils, including black clay, straw felt soil, and calcareous soil (He et al., 2005). The types of vegetation from the upper reaches to the middle are alpine desert, desert steppe, alpine steppe, alpine meadow, and shrub grassland (Zhang et al., 2008; Wang et al., 2019). Alpine shrubs are mainly distributed in the Xigaze section and along the Nianchu river of the Yarlung Zangbo River, usually including Sophora moorcroftiana, Potentilla parvifolia, Caragana versicolor, and others; while the steppe plants are mainly represented by Trike raiahookeri, Orinus thorodii, Stipa purpurea, Pennisetum centrasiaticum; and Kobresia pygmaea and Kobresia littledale are the dominant populations of the meadow type (Wang et al., 2019).

A traditional ecological survey method combining transect and sampling plots was used (Fig. 1). The study sites included Horba, Phayam, Chungpa, Saga, and Sangsang in the upper reaches of the Yarlung Zangbo River; Angren, Lazi, and Xigaze in the middle reaches; Bailang, Gyangze and Kangma in the Nianchu River Basin; and Mozhugongka in the Lhasa River Basin. Forty-four plots were set on the 1500 km transect, and 220 sample plots were surveyed (Wang et al., 2019). Plots of 20 m×30 m were set by comprehensively considering the plant type, community structure and topography, landform, and elevation. A 5 m×5 m shrub plot was set in each plot, and a 1 m×1 m herb plot was set in the center of each shrub plot. Five replicates of the shrub and herbaceous plots were included. The survey recorded the plant species, average height, coverage, and the number of plants in each plot. A GPS unit was used to measure and record necessary information, such as altitude, latitude and longitude, slope, and aspect.

According to the classification principle of “Chinese Vegetation”, plant communities are divided based on habitat characteristics, community structure, species dominance and species composition. Statistical analysis used SPSS software (SPSS Inc., 2002) sub-module of hierarchical cluster, and squared Euclidean distance.

Plant diversity index and related parameters were calculated by:

In the formula: D is Simpson index, H is Shannon-Wiener index; S is the number of plant species in the sample; P_i= N_i/N, N is the sum of the coverage of all species in the sample, and N_i is the coverage of species i in the sample plot.

Environmental data acquisition and analysis: Based on the recorded data from the meteorological stations in the middle and upper reaches of the Yarlung Zangbo River, Tibet, for the past 30 years (1985-2015), a series of climate data graphs for Mean Annual Temperature (MAT) and Mean Annual Precipitation (MAP) were obtained, and the MAT and MAP values for each survey point were extracted.

Combined with the environmental factors such as altitude, slope and slope position obtained during the field survey, a 44 (sample plot) × 5 (environmental variable) dimension environmental data matrix was obtained. The partial redundancy analysis (Partial RDA) in the R-Vegan software package (R version 3.6.3) was used to reveal the rate of contribution of the environment to the plant diversity index. The level of significant differences was set at P<0.05. Verification of the diversity index was obtained through the semi-variance function model of ArcGIS 10.4 software. The plant diversity prediction map was drawn through the ordinary Kriging interpolation method of ArcGIS 10.4 software, and Sigmaplot 12.0 software was used for the drawing.

According to the importance values of plant species, the cluster analysis method was used to divide the association types. Because the distributions of species with importance values less than 0.01 and numbers of occurrence less than 2 on each gradient have great uncertainty, we excluded these species from the cluster analysis. The importance values of 68 common sandy plants in the 44 plots from Horba to Mozhugongka were calculated, and the plants in the study area were divided into seven main communities as shown in Fig. 2.

The structural characteristics of each association are shown in Table 1 and described in detail here.

I Artemisia minor + Stipa purpurea association (five sampling sites): This association is mainly distributed near the rocky beaches and flowing sand dunes at altitudes of 4750 m to 5000 m, where the average temperature of the coldest month is -12.9 ℃ to -11.6 ℃, and the average annual temperature is 3.4 ℃ to 4.7 ℃. The dominant species are Artemisia minor, Stipa purpurea, Potentilla saundersiana, Saussurea tibetica and Lasiocaryum densiflorum. The number of species is 5-12, the average height is 5-20 cm, and the total coverage is less than 15%. This association has strong zonality, and the slope has little effect on the association distribution. It is often called a desert or desert grassland ecosystem.

II Artemisia wellbyi + Festuca ovina association (four sampling sites): This association is mainly distributed on the rocky desert beaches or sand dunes at altitudes of 4600 m to 4800 m, where the average temperature of the coldest month is -9.6 ℃ to -9.1 ℃, and the annual average temperature is about 6.5 ℃. The dominant populations are Artemisia wellbyiand Festuca ovina, and the number of species is generally not more than 10. The average height of this association is 10-20 cm. It is often called a desert grassland ecosystem.

Ⅲ Potentilla fruticose + Orinus thoroldii association (seven sampling sites): This association is distributed in the range of 4500 m to 4700 m, where the average temperature of the coldest month is 6.3 ℃ to 7.0 ℃, and the annual average temperature is 7.5 ℃ to 7.7 ℃. Potentilla fruticosa dominates the main part of the dune, while Orinus thoroldii and Carex moorcroftiiare distributed on the slightly flat sand around the dune. The number of species is generally 3-7, with an average height of 20-60 cm and a total coverage of less than 20%. It is a typical desert grassland or grassland ecosystem.

Ⅳ Trikeraia hookeri + Artemisia frigida association (four sampling sites): This association is mainly distributed on hillsides with altitudes of 4300 m to 4500 m, where the average cold month temperature is -8.0 ℃ to -6.1 ℃, and the average annual temperature is about 7.5 ℃. It mainly consists of Trikeraia hookeri, Artemisia frigida, Elymusnutans and Anaphalisxylorhiza. The number of species is generally about six, the average height of the association is 20-50 cm, and the total coverage is often less than 20%. It is a typical desert grassland or grassland ecosystem.

Ⅴ Kobresia pygmaea association (eight sampling sites): This association is mainly distributed between 4000 m to 4300 m of elevation, where the average temperature of the coldest month is -7.8 ℃ to -0.9 ℃, and the annual average temperature is 7.4 ℃ to 10.6 ℃. It mainly consists of Kobresia pygmaea, Potentilla saundersiana, Astragalus arnoldii and Poa annua. The number of species is generally about 20, the average height of the community is 5-10 cm, and the total coverage is often more than 30%. It is a common alpine meadow ecosystem.

Ⅵ Sophora moorcroftiana + Artemisia hedinii association (eight sampling sites): This association is present at altitude ranges from 3700 m to 4200 m, where the average temperature of the coldest month is -8.0 ℃ to -6.1 ℃, and the average annual temperature is about 7.5 ℃. The dominant species are Sophora moorcroftiana, Artemisia hedinii, Poa tibetica, Carex moorcroftii and Kobresia tibetica. This association is mainly distributed in the Nianchu River and Lhasa River. The number of species is generally 10-20, and it is the most widely distributed and most common community type in the study area. The average height of the community is 30-60 cm, and the total vegetation coverage is 40%-55%. It is a shrub or shrub grassland ecosystem.

Ⅶ Sophora moorcroftiana + Pennisetum centrasiaticum association (eight sampling sites): This association is widely distributed in the elevation range of 3700 m to 4100 m. The established populations are Sophora moorcroftiana, and Pennisetum centrasiaticum. The number of species is 8-15. The average height is 40-65 cm, and the total vegetation coverage is 40%-60%. It is a shrub grassland ecosystem.

3.2.1 Precipitation and temperature distribution patterns

In order to facilitate the analysis and comparisons, the study area is divided into four parts: source area, upstream section (Angren-Saga section), Nianchu River Basin, and Lhasa River Basin. The precipitation and temperature ranges in each of the different areas are shown in Fig. 3. The annual average precipitation shows a distribution of 200 mm (150-290 mm) in the source area, less than 310 mm (280-350 mm) in the upstream section, less than 390 mm (320-450 mm) in the Nianchu River Basin, and less than 430 mm (350-520 mm) in the Lhasa River Basin. The annual average temperature shows a distribution of 3.3 ℃ (2.0 ℃-4.5 ℃) in the source area, less than 5.1 ℃ (4.2 ℃-6.0 ℃) in the upstream section, less than 6.5 ℃ (5.8 ℃-7.2 ℃) in the Nianchu River Basin, and less than 8.2 ℃ (5.0 ℃-9.2 ℃) in the Lhasa River Basin. Overall, the trends of water and heat conditions in the study area are consistent, and the pattern of the changes are that the average annual precipitation and temperature both gradually increase from northwest to southeast.

3.2.2 Variation of plant diversity along the hydrothermal gradient

In general, with the increases of precipitation and temperature, the plant diversity also increases, which is a universal law recorded in most studies (Yu et al., 2013; La et al., 2014; Zhu et al., 2017). In the upper valley, the Simpson index and Shannon-Wiener index also showed similar trends (Fig. 4). When the average annual precipitation is in the range of 300-400 mm, the plant diversity index has the greatest variation, where the Simpson index varies from 0.27 to 0.89 and the Shannon-Wiener index varies from 0.52 to 1.58. This may be due to the wide range of 300-400 mm for the precipitation in the area, and factors such as temperature, topography, altitude, and soil also vary greatly, resulting in differences in plant diversity. In addition, land desertification and overgrazing are also important factors affecting biodiversity.

A fitting analysis using the annual average temperature in the study area and the plant diversity index (Fig. 5) shows that the Simpson index has a linear relationship with the average annual temperature (P<0.001). With the increase of the average annual temperature, the Simpson index increases from 0.28 to 0.77. The change in the Shannon index with annual average temperature is fitted in the form of a quadratic function (P<0.01). As the temperature rises, the Shannon index keeps increasing. When the average annual temperature is in the range of 3-6.3 ℃, the Shannon index grows more rapidly and the slope of the fitted curve is greater, reaching a maximum value (1.52) when the average annual temperature is 6.7 ℃. At higher temperatures, the increasing trend slows down and the slope of the fitted curve becomes progressively less steep.

3.2.3 The relationship between plant diversity and altitude gradient

Overall, the plant diversity index in the study area gradually decreases with increasing altitude, as shown in Fig. 6. As the altitude rises from 3500 m to 5000 m, the Simpson index decreases from 0.81 to 0.3, and the Shannon index decreases from 1.67 to 0.54; and the fitting of the two indexes and the altitude gradient both conform to the form of a quadratic function, with R² values of 0.4921 and 0.4786, respectively.

Partial RDA was used to measure the explanatory effects of environmental variables on plant community heterogeneity. The five environmental factors are divided into three components: water, energy and habitat heterogeneity, which explain 55.1% (a+d+e+g), 38.9% (b+d+f+g) and 15.3% (c+e+f) of the plant diversity in the middle and upper reaches of the Yarlung Zangbo River, respectively (Fig. 7). Water is the primary factor that affects the pattern of plant diversity, with an individual interpretation rate of 19.3% (a). Habitat heterogeneity has the smallest impact on plant diversity, at only 2.3% (c). Water and energy together explain 60.8% (a+b+d+e+f+g) of plant diversity variation, with a common interpretation rate of 9.2% (g). In addition, there is a remaining 36.3% due to unexplained factors, which may include the chemical properties of the soil, species interactions between communities, and interference factors.

The plant diversity semi-variation function model and ANOVA results (Table 2) show that the two plant diversity indexes in the study area conform to the spherical model.

The C₀ value represents the spatial heterogeneity caused by sampling errors and here it is smaller than the random portions on the sampling scale. A larger value of C₀+C₁ indicates a higher total spatial heterogeneity. The ratio of C₀/(C₀+C₁) indicates the spatial correlation caused by system variables. A ratio of less than 0.25 indicates a strong spatial correlation, 0.25-0.75 indicates a moderate degree of spatial correlation, and higher than 0.75 indicates a weak spatial correlation (Ma et al., 2008; Wang et al., 2016). Here, the two plant diversity indexes are within the 0.25-0.75 range, so the plant diversity in the middle and upper reaches of the Yarlung Zangbo River has a moderate spatial correlation.

Based on the analysis of the semi-variance model, the ordinary Kriging interpolation method of ArcGIS 10.4 was used to obtain the plant diversity prediction maps of the middle and upper reaches of the Yarlung Zangbo River (Fig. 8). Plant diversity has obvious vertical and latitudinal changes, rising from west to east and from south to north. The Zhongba area has less precipitation, lower heat, and severe land desertification, so its plant diversity is generally low. The Shigatse region is also affected by wind and sand erosion, but has comparatively more precipitation and higher temperatures, and the plant diversity is better than that of Zhongba region. The Lhasa River Basin, with the best water and heat conditions, shows higher plant diversity.

4.1.1 Distribution pattern of plant community types

The middle and upper reaches of the Yarlung Zangbo River are located on the Qinghai-Tibet Plateau, which is characterized by extremely harsh natural conditions. The plants in the upstream area have remained basically unchanged or only slightly changed, while the changes in plant coverage in the middle reaches have been relatively dramatic. The DCA ranking of species shows a strong spatial correlation between the composition of plant community species and changes in environmental factors (Wang et al., 2019). In field measurements, with the increase of longitude, the vegetation type gradually evolved from desert steppe to alpine steppe, and then it developed into shrub steppe. The upstream area is dominated by Artemisia, Stipa, Gramineae, Trichosanthes, Pennisetum, and Kobresia (Li et al., 2013), and has harsh soil and water conditions and a relatively simple community composition, which is typical of desert grassland communities. In addition to the above species, the midstream area also includes a large number of shrubs, such as members of the families Rosaceae and Fabaceae, and it is dominated by alpine grassland communities and shrub communities (Zhang et al., 2008; Wang et al., 2019).

4.1.2 Water and temperature are the primary factors affecting the distribution of plant diversity

The distribution pattern of plant diversity is the result of multiple environmental factors including precipitation, temperature, and terrain, but these factors will exert different effects in the different study areas. At the regional scale, temperature and precipitation are the decisive factors that determine the distribution of vegetation. At the local scale, non-zonal environmental factors, including terrain and soil, have an important impact on the distribution pattern of vegetation by adjusting local climatic conditions (Yu et al., 2013). This study shows that plant diversity in the middle and upper reaches of the Yarlung Zangbo River is mainly controlled by the water and heat conditions. As affected by altitude and longitude, a gradient of increasing moisture and heat is formed from west to east, so plant diversity gradually increases along this gradient. In some local areas, the diversity fluctuates greatly, indicating that localized interference has a substantial impact on plant diversity. The interference factors in the study area mainly include land desertification, overgrazing, and construction activity. Among them, land desertification has the most significant impact. The results of different studies on the diverse distribution patterns of plants in the Qinghai-Tibet Plateau are consistent (Tang and Fang, 2004; Liao et al., 2013; Yu et al., 2018). However, some researchers believe that the distribution pattern of plant diversity in Tibet should be unimodal. As the altitude increases, the species diversity index shows a trend of increasing first and then decreasing, with the highest levels in the mid-altitude regions (Duan et al., 2011). The results of this study indicate that plant diversity is monotonically increasing, which may be due to the fact that the middle and upper reaches of the Yarlung Zangbo River include a wide range of areas. Due to the topography and geomorphology, the elevation range in some areas is significant, which weakens the impact of altitude on plant diversity, allowing moisture and temperature to become more important factors.

The results of the semivariogram model and analysis of variance indicate that the spatial heterogeneity of plant diversity in the study area is very strong, with moderate spatial correlation and a high nugget value. The Simpson index and Shannon index predictions obtained by ordinary Kriging interpolation also show that water and heat conditions have significant effects on plant diversity, which indicates that the diversity exhibits an upward trend with the increases of precipitation and temperature. The influence of changes in longitude is more significant than that of changes in latitude.

The unique geographical environment and climatic conditions in Tibet make this area a fragile ecosystem with great significance (Wang et al., 2019). The above conclusions can guide the selection of suitable native plant species and other exotic plants during plant restoration projects on the desertified land of the Yarlung Zangbo River to a certain extent, which is of great significance to the protection and improvement of the ecological environment of the Yarlung Zangbo River region.

The results of this study on the distribution pattern of plant diversity and the factors which influence it in the middle and upper reaches of the Yarlung Zangbo River show that the plant communities can be divided into seven distinct associations: Artemisia minor + Stipa purpurea association, Artemisia wellbyi + Festuca ovina association, Potentilla fruticosa + Orinus thoroldii association, Trikeraia hookeri + Artemisia frigida association, Kobresia pygmaea association, Sophora moorcroftiana + Artemisia hedinii association and Sophora moorcroftiana + Pennisetum centrasiaticum association. The plant diversity in the study area gradually increased from west to east. As the longitude increased and the latitude decreased, the Simpson index increased from 0.18 to 0.89 and the Shannon index increased from 0.32 to 1.96. For every 300 m increase in elevation, the number of plant species decreases by two on average. When the average annual precipitation is between 300 mm and 400 mm, there is a large variation in the plant diversity index. The distribution pattern of plant diversity in the study area is mainly affected by precipitation and temperature, with obvious zonal characteristics. Water and energy factors together explained 60.8% of the spatial variability of plant diversity, with individual explanation rates of 19.3% and 5.7%, respectively, and the coupling effect between the two was obvious. Water and temperature are the primary factors affecting the distribution of plant diversity. At the same time, plant diversity in some areas has decreased due to land desertification. Reasonable restoration measures should be implemented in these areas to restore the natural vegetation as soon as possible to increase ecosystem stability and gene flow in the different areas (Ma et al., 2008). There are many important sandy plant resources in this area. These species have to be protected and developed properly.