2 Methods

2.1 Study sites

The site of this study was located in the middle reaches of Yarlung Zangbo River basin in Naidong district of Shannan City, Tibet of China (91.585 °E, 29.258 °N, 3560 m average above sea level). This region belongs to the semi-arid plateau temperate monsoon climate area. The temperature is low with unclear seasons. The annual daily mean temperature is approximately 8.2 ℃ (from -18.3 ℃ in January to 30.1 ℃ in July). During the last decade, this region has the annual precipitation of approximately 382.3 mm, mostly ranged from June to September. The soil is mainly aeolian sandy soils with low contents of soil organic matter and total nitrogen, which is highly susceptible to erosion (Liao et al., 2020a).

2.2 Experiment design and data collection

The experiment was initiated in 2017 at the natural shrubland ofS. moorcroftiana population within a little other plant species, such as Oxytropis sericopetala , Stipa bungeana Trin.,Orinus thoroldii , et al., which had undergone over decade of natural restoration. The degradation had happened in this region before 2000 due to the overgrazing and firewood cutting (Shen et al., 2012), however there was visible improvement for vegetation coverage to 2017 during decade of natural restoration (Fig. 1). Three topographies (micro-topographies) measuring 10 ha (100 × 100 m) each were selected separately along the streamway in the same alluvial fan. The margin of alluvial fan was a fixed sand land with flat area (T1) (Fig. 1b and Table 1). The middle of alluvial fan was a semi-fixed sand land at the west slope of approximately 25° (T2) (Fig. 1c and Table 1). The top of alluvial fan was also a semi-fixed sand land at the northwest slope of approximately 35° (T3) (Fig. 1d and Table 1). Three of 5 × 5 m small quadrats were set along a diagonal line within each topography for field survey and sampling. For each quadrat, the location, species, vegetation coverage, height, wide (WC) and narrow (NC) crown diameters, elevation and slope were recorded. The crown projection area (CPA) was simply calculated as the area of an ellipse (Zhang et al., 2016). The soil samples were collected and detected as the background characteristics at the beginning of the experiment (Table 2).
In each topography, 3D point cloud data was acquired by a RIEGL VZ-400i TLS on June 30, 2017. The VZ-400i TLS has an ± 5 mm accuracy at a range up to 800 m and capture the point cloud unit at a speed up to 500,000 points per second, with a field view of 360° horizontal (H) × 100° vertical (V) (http://www.riegl.com/). Depending on complexity of topography, we used different configurations for plot-level scans to maximum acquisition of the ground surface and vegetation parament (Fig. 2). The TLS was placed at the center of T1 and eight surrounding locations approximating the cardinal directions, with the distances (approximately 30 m) from the center point, due to the uniform distribution could comprehensively capture point clouds in the flat and open terrain (Fig. 2a). While for the T2 and T3, four of six scans were conducted on the ridge to take advantages of locations (Fig. 2b).

2.3 LiDAR data processing

2.3.1 Plot cutting and pre-processing

For all scans in each topography, each point cloud was first merged by RiScan Pro software to form a comprehensive point cloud of plot based on the location and reference targets of each scan, such as the house and surveyor’s poles of plot (Li et al., 2019). And then, one big plot of 50 × 50 m was cut from each point cloud of plot, using LiDAR360 software. We removed the noise and discriminated ground vs. non-ground points through this procedure. The digital elevation model (DEM) and digital surface model (DSM) were generated by the triangulated irregular network (TIN) interpolation with a spatial resolution of 0.05 m.

2.3.2 Vegetation and micro-topography metrics

Based on the DEM and DSM, the canopy height model (CHM) was derived with a 0.05 m resolution (Li et al., 2019; Liao et al., 2020b). The location, height and CPA of individual shrub within the CHM were detected, computed and exported by “rLiDAR” package from R (R x64 3.6.1.lnk) statistical software (R Core Team, 2014). The slope and aspect was derived from the 0.05 m DEM using the ‘terrain’ tool in LiDAR360 software. The micro-topography parameters (including elevation, slope and aspect), matched with the location of individual shrub, were extracted by R.

2.4 Data analysis

2.4.1 Population structure

The plant height, regarded as one of the morphological characteristics, was directly and accurately extracted from LiDAR data (Luo et al., 2015). Therefore, height classes were used to explain the structures of S. moorcroftiana population, which was an undershrub species and does not have notable annual rings. All S. moorcroftiana individuals, captured by single tree segmentation of LiDAR, were classified into different classes based on the vegetation height (Class 1 was less than 30 cm; 30 cm were added gradually from Class 2 to 4 groups; Class 5 was more than 120 cm) which were consulted from previous researches (Yang et al., 2011). The relationship between CPA and height was fitted by a linear model in each topography. A normal distribution curve was used to analyze the plant height distribution structure in each topography.

2.4.2 Spatial pattern analysis

The relationship between spatial distribution of S. moorcroftianapopulation and spatial covariate (elevation, slope and aspect) was estimated by nonparametric estimation of the dependence of a spatial point process on spatial covariates in this study. The LiDAR data was captured as finite set spatial points, with the locations of individuals (y). Additionally, the values Z(u) of a spatial covariate (elevation, slope and aspect) at every spatial location y were extracted from micro-topography metrics. The y as a realization of a spatial distribution of individuals Y with density function λ(u) depending on Z(u) was modeled as follow,
\(\lambda\left(u\right)=\rho(Z(u))\) (1)
where ρ is a resource selection function reflecting preference for particular micro-topographic conditions in ecological applications where the points are the locations of individual organisms (Baddeley et al., 2012).