The study was conducted at the Shayar site on the upper (U), at the Yingbazar site on the middle (M), and at the Arghan site on the lower (L) reaches of the Tarim River (40.57°N–40.08°N, 82.22°E–88.21°E, ASL 1000–800 m) within the northern edges of Taklamakan Desert, northwest China (Figure 1). This area is characterized by a typical continental temperate arid desert climate (Kottek et al., 2006). The mean annual temperature, precipitation, and evaporation range from 9 to 11°C, 50 to 80 mm, and 2500 to 3000 mm, respectively (Aishan et al., 2016; Keyimu et al., 2017a). There is low precipitation, and the main water source is the Tarim River which is derived from Tianshan and Karakorum Mountain’s snow meltwater and rainfalls. According to the Food and Agriculture Organization (FAO) classification of the world’s soils, Fluvisols, Gleysols, Solonchak, and Arenosols occur as the soil types in the study area (Thevs et al., 2008). The Tarim River is fringed with riparian forests, and the growing season is from April to October. The main vegetation species are P. euphratica, Tamarix spp., Phragmites australis, Alhagi sparsifolia, Karelinia caspia, Lycium ruthenicum, Halimodendron halodendron, Apocynum venetum, Hexinia polydichotoma (Kurban et al., 2019), among these, the P. euphratica is the dominant tree species (Zeng et al., 2009; Thomas et al., 2016).

A total of 18 sample plots were established in the upper, middle, and lower transects of the Tarim River. At each transect, six P. euphratica plots (U1–U6, M1–M6, L1–L6) with a quadrat area of 50 m × 50 m were set up maintaining a distance of 20, 200, 400, 600, 800, and 1000 m from the river channel, respectively (Table 1). Considering the P. euphratica forest distribution area, and the river line changes complexity, the plots cannot be arranged following one line, particularly at the middle reaches. But we ensured that the straight distance of each plot from the river channel had an equidistantly increased pattern (Figure 1). In each plot, P. euphratica is the only tree species. Considering that seedlings below 1.5 m were difficult to distinguish from shrubs, affecting the segmentation of individual trees, and thus only the trees taller than 1.5 m were identified using TLS. The biometric parameters (TH, tree height; DBH, diameter at breast height; CD, crown diameter) were subsequently determined (Figure 2). TLS data was collected in May 2020 under wind speeds lower than 3 m/s (Yusup et al., 2020). The average groundwater depth was obtained from groundwater monitoring wells set up in each section by the Tarim River Basin Administration Bureau. Soil moisture, salinity, and pH values were measured in each plot at 50 cm soil depth using Hydra soil moisture sensor (Stevens, Co. Ltd., Horn, Germany).

Populus euphratica forests in the study area are mainly distributed in the flat terrain areas in the riparian zone (Table 1) and grow under the same exposure condition. Previous research results indicated that groundwater condition (Cao et al., 2020; Thomas and Lang, 2020) and soil physicochemical properties (Han et al., 2016; Zhang et al., 2019) were the main environmental factors affecting the distribution of desert riparian forests. Therefore, we selected groundwater depth (GD), soil moisture (SM), soil conductance (SC), and pH values as abiotic factors correlated with the spatial distribution pattern index by the Pearson correlation analysis method. Stata (V13.0, Stata, Co. Ltd. Texas, TX, United States) software was adopted for calculation.

The numbers of total and young trees, the SDI, TH, and CPA values significantly reduce from the upper to the lower reaches of the Tarim River (p < 0.05). In contrast, the number of old trees and DBH and CD values peaked in the middle transect. Mature and near mature tree numbers were the highest in the lower reaches (Table 2).

The population structure of P. euphratica trees in the upper reaches exhibits a reversed–J distribution (logistic fit, R² = 0.90) (Figure 4A). The near-mature trees occupy the highest ratio (43%), followed by young trees (29%), mature trees (23%), and then old trees (5%) (Table 2). The number of young trees (n = 197) is almost six times that of old trees (n = 34). The population structure of the middle and lower reaches follows a Gaussian fit (R² = 0.95 and 0.97) (Figures 4B,C), where near-mature trees occupy the highest ratio (45 and 48%), followed by mature trees (25 and 29%), young trees (18 and 14%), and old trees (11 and 9%) (Table 2). Young and near mature trees occupy a higher ratio in the middle reaches compared to the lower reaches (29%, 23%).

Individual P. euphratica tree locations were recorded and acquired kernel intensity map for each plot based on the TLS data (Figure 5). Trees are more evenly distributed at plots U1–U5 (upper reaches) and M2–M3 (middle reaches), while the trees of the remaining plots in the middle and lower reaches are unevenly distributed. Young and near mature trees are concentrated in plots were close to the river channel (Figure 5); mature trees are evenly distributed in each plot; and old trees are generally located at distances far from the river channel (e.g., plots M5–M6 and L5–L6).

The Kernel density of the spatial distribution patterns reveals a random distribution for the trees in the upper reaches; in the middle reaches, the trees are typically distributed in aggregated patterns inside the plot areas, with random patterns outside the aggregated area; and the majority of trees in the lower reaches exhibit an aggregated distribution, with young and near mature trees distributed around the mature trees (Figure 5).

Variations are observed in the distribution patterns (Figure 6). The point pattern analysis based on the g(r) function under the CSR null model reveals upper reaches plots U1–U5 [g(r) ≈ 1] to exhibit a random pattern, while plot U6 (located 1000 m away from the river channel) is associated with an aggregated pattern. In the middle reaches, the distribution presents a random pattern [g(r) ≈ 1] for plots M1 and M2, which are located close to the river channel, and an aggregated pattern [g(r) > 1] for plots M3–M6. The distribution of the lower reaches generally exhibits an aggregated pattern [g(r) > 1], except for plot L3. The highest degree of aggregation is observed for plots L2 and L6 [g(r) = 9.7, 3.7] in the lower reaches, followed by plots M3 and M6 [g(r) = 3.1, 2.2] in the middle reaches, and plot U6 [g(r) = 1.8] in the upper reaches (Figure 6). The g(r) function average value exhibited the following order: lower reaches [g(r) = 1.3] > middle reaches [g(r) = 1.1] > upper reaches [g(r) = 1.0], indicating that the main distribution pattern of trees from the upper to lower reaches has changed from random to cluster distribution pattern.

Table 3 showed that the ANN index values are high at the upper reaches ranged from 0.95 to 1.18, except the plot U4 and U6, other plots indicating a generally random distribution; values of the middle reaches range from 0.72 to 1.06, indicating a distribution that changes from random to aggregated pattern; and the lower reaches values range from 0.54 to 0.94, revealing an aggregated pattern, except plot L3 which showed random pattern. The DMI, HIS and I_δ index results also same as the ANN index, at the 0.05 significant levels. These observations agree with g(r) function result.

The correlation between the different abiotic factors of the plots, including the groundwater depth (GD), soil moisture (SM), soil conductance (SC), soil pH value and distance from the river channel (DR), and ANN index was assessed (Table 4). The GD, SM, and SC correlate with the ANN index.

The average GD and SM values exhibit a decreasing trend from the upper to lower reaches (Figure 7). Correlation analysis reveals the ANN index to be significantly positively correlated with GD and SM (R = 0.67 and 0.56, respectively, P < 0.05).

Various traditional field measurement methods have been used for analyzing the spatial distribution pattern of riparian forests (Han et al., 2014; Xu et al., 2016; Miao et al., 2020). However, traditional methods are hard to obtain detailed structural parameters, and spatial distribution patterns of the forest stand in the same period, which hindering our deeper understanding of forest structure. As an active remote sensing method, LiDAR has obvious advantages for determining forest structural parameters (Guo et al., 2018; Calders et al., 2020). In this study, we used TLS data for the first time on riparian forests and demonstrated that the LiDAR techniques have high advantages for measuring the desert sparse forest type (Figure 3). TLS can quickly and accurately determine forest structure attributes in different stand density plots. We comparatively analyzed the stand structure parameters, population dynamics, and spatial distribution pattern of P. euphratica forest in different transects of the Tarim River.

The number of trees, SDI, CPA, and TH decreased significantly in our study sites from the upper to lower reaches of the Tarim River (Table 2). This indicated that the structural characteristics of P. euphratica forest have changed in the different sections of the Tarim River. For example, the upper transect exhibited the highest average TH and lowest average DBH, while the CPA values exceeded 80%, demonstrating the high vitality of the forests (Han et al., 2017). The TH in the middle reaches less than the upper reaches, while the average DBH was the largest, with CAP values higher than 60%, indicating a relatively good forest growth trend. In the lower reaches, the average TH and DBH were lower than the middle reaches, and the CPA was smaller than 40%, revealing that the forests were degraded. Previous studies have reported the individual growth of P. euphratica to adapt to deteriorated environments by adjusting the size of the CD and reducing the TH (Yang et al., 2014; Aishan et al., 2015). The growth of P. euphratica highly depends on the groundwater and soil moisture (Halik et al., 2009; Keyimu et al., 2017a). Under extreme drought stress, morphological characteristics of trees shift the form to extend their survival period to wait for a favorable change (Miao et al., 2020). Monda et al. (2008) demonstrated near mature P. euphratica trees at the Heihe River to exhibit damage at the top part of the crown, while the dieback phenomenon was observed, suppressing their height. Trees can survive increasingly dry environments by abandoning assimilation organs, thereby suppressing height growth as a price for survival (Xu et al., 2016). The higher canopies above thick trunks and wider leaves of P. euphratica become disadvantageous under severe drought stress (Li et al., 2017). In addition, the increased intensity of inter-species competition reduces the density and survival rate of the P. euphratica population, significantly changing structural indicators such as TH, DBH, and CD (Shi et al., 2021). Therefore, the change of stand structural attributes and growth status of the P. euphratica indicated the adaptation to different environmental conditions.

In harsh dry conditions, the growth potential of the P. euphratica decreases while the mortality rate increases (Chen et al., 2014; Keram et al., 2021). Thus, the population structure evolves from a growing to a declining population, and the forest stand gradually changes from a juvenile to an over mature forest (Han et al., 2017). Our result indicated distinct population structure characteristics across the upper, middle, and lower reaches of the Tarim River (Figure 4). The plots of the upper reaches exhibited a DBH class structure that followed a logistic fit, demonstrating a growing population trend, while the middle and lower reaches plots followed a Gaussian fit, with a stable population trend. However, the scarcity of young trees leads to a weak anti-disturbance ability of the forest population in the lower reaches, and thus the renewal and rejuvenation of forests are difficult under natural conditions. Han et al. (2014) found the young tree ratio in the upper reaches of the Tarim River to be high, corresponding to the Deevey I type survival curve (Deevey, 1948). This indicated an expanding population trend. Under suitable environmental conditions, shallow groundwater depth during the flood season, which coincides with the dispersal period of P. euphratica seeds, enables the production of more seedlings in the forest. The DBH class exhibited a normal distribution in the middle reaches, with relatively few young trees and more mature trees, demonstrating a stable population trend (Xu et al., 2016; Han et al., 2017). In the lower reaches, Keyimu et al. (2017b) and Zhou et al. (2018) found the mature trees to occupy a large population ratio. Therefore, the living conditions must be improved to enhance the survival rate and the number of young P. euphratica trees in the middle and lower reaches of the Tarim River. The water resources in the three reaches of the Tarim River are unevenly distributed across time and space, and the gradually declining groundwater level from the upper to lower reaches consequently resulted in different population structures in the respective transects.

Our results reveal that the P. euphratica forests exhibit varying spatial patterns in the three transects of the Tarim River (Figure 5). In the upper reaches, the spatial pattern index g(r) function value was close to 1 and fluctuated within the CSR null model confidence interval, representing a random distribution pattern (Figure 6). This was also true for the ANN, DMI, and I_δ indexes (Table 3). These observations agree with previous work findings reported by Zhang et al. (2019) and Miao et al. (2020), which showed that in the upper reaches of the Tarim River, the P. euphratica live trees were typically randomly distributed. The trees in the middle reaches were unevenly distributed, except for two plots (M1 and M2) closest to the river (Figure 5), with the g(r) function producing high values above the CSR null model confidence interval. Moreover, the ANN index demonstrated a decreasing trend, implying that the spatial distribution changed from random to aggregated patterns with increasing distance to the river channel. Wang et al. (2020) found that the old tree ratio increased with the distance from the river channel, and the distribution pattern changed. Zeng et al. (2019) and Zhang et al. (2019) reported the aggregation intensity of the Morisita index to increase from the upper to the middle reaches of the Tarim River. Yu et al. (2021) revealed that the P. euphratica trees in the lower reaches of the Heihe River to be randomly distributed within a 100 m distance from the river channel and aggregation intensity was increase as increasing the distances from the river. These results are consistent with the current study (Table 3). The plots (L1∼L6) in lower reaches exhibited g(r) function values above the CSR confidence interval, and ANN index values less than 1 indicated the aggregated pattern. Moreover, the Kernel intensity showed strong cluster intensity in a partial area within the plots (Figure 5). These results are consistent with previous work by Zhou et al. (2018) and Zhang et al. (2019). P. euphratica forest aggregation intensity generally increased from the upper to lower reaches, and the spatial distribution changed from random to aggregated pattern.

Spatial pattern differences result from the interaction between plants and abiotic factors, particularly the biological plant characteristics, the competition between species, and the habitat environment (Li and Zhang, 2015; Podlaski, 2019). In arid regions, water from rivers infiltrates into the riparian groundwater (Westermann et al., 2008). River diversion and flooding are the main factors influencing the dominance of the Tarim River basin in global P. euphratica forest locations (Zhang et al., 2019). P. euphratica trees are phreatophytes and thus rely on groundwater to survive (Thomas and Lang, 2020). The spatial heterogeneity of the water environment induces changes in the morphological and physiological characteristics of P. euphratica (Shi et al., 2021). Ling et al. (2015) determined the minimum groundwater depths for young, near-mature, mature, and old trees as 4.0, 5.0–5.4, 6.9, and 7.8 m, respectively. Moreover, the groundwater level exhibits a descending gradient from the upper to lower reaches (Chen et al., 2006; Zeng et al., 2019). The soil moisture has also been observed to decrease gradually from the upper to lower reaches of the Tarim River (Han et al., 2016), thus resulting in differences in the growth and distribution of P. euphratica (Gries et al., 2003; Xu et al., 2016; Zhang et al., 2019). The survival and growth rate of P. euphratica decreased with the groundwater level (Aishan et al., 2015, 2018; Ling et al., 2015, 2017; Yu et al., 2021). Our results revealed a significant positive correlation between the spatial pattern index ANN and values of GD and SM (Figure 7). Furthermore, reductions in the groundwater level and soil moisture content were associated with an increase in the aggregation intensity of spatial distribution patterns. In the upper reaches plots, the shallow groundwater level and relatively abundant soil moisture conditions contributed to maintaining a high forest stand density (Table 3) and growing population structure (Figure 4), and thus the spatial distribution was generally formed random pattern. Miao et al. (2020) found the riparian forest growing in an area of the Daliyaboyi Oasis with a groundwater depth of 3–5 m to present a random distribution. In the middle and lower reaches of the Tarim River, the GD and SM decreased with increasing distance from the river (Figure 7), and consequently, the ANN index values also exhibited a decreasing trend.