Study on spatiotemporal changes and influencing factors of frozen soil moisture during freeze-thaw period under different ecological construction measures in the loess plateau of China

The climatic conditions and soil characteristics of the Loess Plateau in China make it have a unique freeze–thaw process and the distribution of soil unfrozen water. In order to elucidate the spatio-temporal variation of soil unfrozen water during freeze–thaw period and its influencing factors under different ecological construction measures, The spatial and temporal variation of soil unfrozen moisture and its main influencing factors in five different ecological plots in the Loess Plateau during freeze–thaw period were studied by combining field monitoring and indoor calculation. The results showed that: (1) the spatiotemporal changes of unfrozen water content in soil were different under different vegetation restoration methods. The variation of unfrozen water content in fallow land was the largest at 20cm depth, while the variation in surface 10cm was the largest in the other four plots. (2) The average unfrozen water content of soil in the representative soil layer can be estimated more accurately, and the best representative soil layer in the five sample plots is concentrated in 20–30 cm. (3) During seasonal freeze–thaw period, the main controlling environmental factors of unfrozen water content are air temperature and surface temperature. The results can provide reference for soil protection and water resources management in northwest China, especially in the Loess Plateau.


Introduction
Soil freezing and thawing, as a significant natural phenomenon, exerts direct influence on the spatial variability of soil moisture and plays a crucial role in various domains including agriculture, water resources, environmental systems, and infrastructure construction (Yi et al 2014, Musa et al 2016).Understanding the intricacies of water and heat transport during soil freezing and thawing processes holds paramount importance in the assessment of surface and subsurface water resources, utilization of soil water and heat reservoirs, determination of technical parameters for agricultural irrigation, engineering construction, environmental preservation, and prevention of soil salinization, among others.(Liang et al 2019, Chen et al 2018, Hayashi 2013).The northwestern region of China is characterized by limited precipitation, prominent evaporation rates, and vulnerable ecological environments.Consequently, investigating the distribution of soil moisture and delineating the factors influencing its dynamics during the freeze-thaw process in this area assumes heightened ecological significance (Lei et al 2015, Zhang et al 2019, Wu et al 2022).
During the process of soil freezing, soil moisture tends to migrate from unfrozen regions to frozen regions (Nagare et al 2012), leading to an increase in soil water content in the frozen soil (Zhang and Sun 2011).In the subsequent thawing phase, as temperatures rise, the upper soil layer thaws first, followed by the deeper layers.Consequently, in the initial stages of thawing, the soil water content in the uppermost layer exhibits a noticeable increase (Chen et al 2013).In contrast, the lower soil layers experience negligible changes in water content as they are still frozen, creating limited opportunities for water infiltration (Bing et al 2015).Simultaneously, variations in air temperature during the thawing process contribute to the freeze-thaw cycles within the soil profile, thereby influencing the distribution of soil water content (Luo et al 2003).Previous studies have primarily focused on examining the migration patterns of water, heat, and salt during freeze-thaw processes (Hafsteinsdóttir et al 2011, Song et al 2017, Zhang et al 2021).Currently, the Chinese government has implemented large-scale vegetation restoration efforts to improve the ecological environment and prevent soil erosion (Liu et al 2023, Liu et al 2023,Wang et al 2023).Some studies have indicated that different vegetation restoration methods can impact the spatiotemporal variation of soil moisture and temperature during freezethaw processes (Zhu et al 2016, Xiao et al 2020, Bo et al 2021).However, research specifically focused on northwest China remains limited (Zhu et al 2020).Additionally, the majority of studies are predominantly based on laboratory experiments and model simulations, lacking field data validation and analysis.
This article focuses on five different ecological construction models in the hilly area of the Loess Plateau.Through comprehensive on-site monitoring and rigorous data analysis, the spatiotemporal distribution and influencing factors of frozen soil moisture under different ecological construction conditions during the freezethaw period are studied.The main purpose of this study is to reveal the complex distribution patterns of unfrozen soil moisture during the freeze-thaw period under different ecological construction scenarios.To provide substantive references and insights for effective hydrological management measures in the northwest region of China, and to provide valuable guidance for sustainable land management practices in these areas.

Research methodology
2.1.Overview of the study area Xindian Yulin Gully (figure 1) is located on the left bank of the middle reaches of the Wuding River in the east of Suide County, Yulin City, Shaanxi Province, China.It has undulating ridges and gullies, and serious surface erosion.At an altitude of 800∼1100 m, the whole watershed is cut by 31 branch gullies of over 200 m long.In addition to a loose soil texture and a thin arable layer (20∼30 cm), this area has a fragmented topography formed by 16 interlacing gullies, with a gully density of 7.26 km/km 2 .Located at 110.25°E and 37.52°N, the study runoff plot was constructed in 1987, with a runoff plot slope of 25°, a slope direction of 14°N by W, a horizontal length of 20 m, an inclined length of 22.53 m, and an average width of 5 m.
In this study, five field runoff plots were arranged on a typical loess slope in Xindian Yulin Gully, Suide County.The soil type is mainly yellow spongy soil .The specific information of the runoff plots is shown in figure 2.

Field monitoring
For real-time dynamic monitoring of soil moisture and temperature, online soil moisture temperature monitoring instruments (Cloud Intelligent Soil Moisture Temperature Observer, ET100-5AH) were placed in the middle of the slope of five runoff communities: agricultural land, fallow land, shrubland, shrub and grass mixed land (hereinafter referred to as shrub-grass land, and woodland (arbor).The slope of the runoff plot is 25°, the aspect of the slope is 14°to the west of the north, the horizontal length is 20 m, the inclined length is 22.53 m, the average width is 5 m.The data were collected at an interval of 1 h, the monitoring depth was 0∼100 cm.Moisture and temperature sensors were placed every 10 cm along the depth direction to obtain the realtime water content and temperature data of different soil layers.ET100-5AH uses frequency domain reflectometry to measure soil moisture content.The technique uses the relationship between soil dielectric constant and soil moisture content to calculate the proportion of soil moisture content by emitting high frequency signals into the soil and measuring the phase and amplitude changes of the reflected signals.In addition, ET100-5AH has built-in a set of temperature sensors to monitor real-time changes in soil temperature.

Meteorological data
The meteorological data of 2019-2020 (including precipitation, temperature, evapotranspiration, relative humidity, wind speed, etc) were obtained from xindiangou meteorological station, the meteorological station automatically observes a set of meteorological data every 1H, which is consistent with the monitoring time of soil moisture and temperature on-line monitor.

Relative difference method
The relative difference mean and relative difference standard deviation of each point can describe the time stability characteristics of soil unfrozen water (Vachaud et al 1985).The relative difference of soil unfrozen water RD i at observation time j in the observation layer i is given by: ̅ ̅ ( ) where q ij is the observed value of unfrozen water content of i at time j; ̅ q j is the average of the unfrozen water content in all soil layers at time j.
Therefore, the relative difference mean(MRD i ) and the relative difference standard deviation (SDRD i ) at soil i are calculated by: where M is the number of observation dates.
It is generally believed that the closer the MRD i is to 0, the more the unfrozen water content of the soil layer can represent the average soil unfrozen water content of the sample land (Bo et al 2021).A smaller SDRD i indicates a higher the temporal stability of the soil layer (Grayson and Western 1998).Thus, the average unfrozen water content of soil in the study area is usually estimated based on soil layers with a MRD i close to 0 and a small SDRD .i

Change in water content
To further analyze the variation characteristics of soil unfrozen water content in the whole seasonal freeze-thaw period under different vegetation restoration methods, the change amplitude of unfrozen water content (Dθ) is introduced, and it can be calculated by: where q D is the variation amplitude of soil unfrozen water content; q max and q min are the maximum and minimum unfrozen water contents during the seasonal freeze-thaw period, respectively.

Data processing
The data were preprocessed before statistical analysis, the daily average data were obtained in the development tool module of Excel 2010 through self-written VB program code.Use statistical analysis software to calculate and analyze the unfrozen soil moisture content, and use EXCEL and other software to draw charts.The temporal stability of soil moisture was analyzed through the relative difference method.Redundancy analysis (RDA) was performed by R language data programming.

Seasonal variation characteristics of soil unfrozen water content
Vegetation restoration methods will affect the micrometeorological and micro-topographic conditions of the soil, and further change the freeze-thaw conditions inside the soil, which in turn will impact the spatial and temporal distribution of the unfrozen water content of the soil (Lu et al 2021, Cao et al 2022).
Overall, among the five samples, the unfrozen water content of the shallow soil layer (10∼50 cm) changed drastically during the monitoring period.Compared with shallow layers, the unfrozen water content of deep soil layers (60∼100 cm) was stable.As shown in figure 3, the unfrozen water content at 10∼100 cm in soil of shrubland had the highest unfrozen water content, yet woodland had the lowest.
Table 1 presents the statistical results of the variation amplitude ( q D ) of unfrozen water content at different soil depths in the five plots during the freeze-thaw period.The variation range of unfrozen water content at 20∼40 cm soil depth showed a descending trend of shrubland > shrub-grass land > fallow land > agricultural land > woodland.The variation amplitude of unfrozen water content at 50∼60 cm soil depth showed was all mixed shrubs > fallow land > agricultural land > shrubs and grasses mixed > trees.The variation amplitude of unfrozen water content in soil at depths of 70 cm and 90 cm showed shrubland > fallow land > agricultural land > woodland > shrub-grass land.Among them, agricultural land, shrubland, shrub-grass land, and woodland all had the largest change in soil unfrozen water content at the surface layer (10 cm), while the largest change in fallow land was at 20 cm.

Time stability analysis of unfrozen soil water content
The MRD of unfrozen water content at different soil depths is shown in figure 4, where the error bars indicate the size of SDRD.In general, the variation in unfrozen water content in the soil of the five samples showed a trend of first decreasing and then increasing during the whole seasonal freeze-thaw period.The MRD of unfrozen water content in agricultural land was close to 0 and relatively stable, but it fluctuated wildly in the other four samples.The SDRD value of soil unfrozen water content was the highest in the 10 cm soil layer of the five sample plots, indicating the largest dispersion.This is mainly because shallow soil is more affected by the environment, resulting in poor stability.
To study the temporal stability of soil water in the sample plots, it is necessary to determine the representative soil layers with the mean unfrozen water content of the soil during the seasonal freeze-thaw period (Corbari et al 2010).According to the abovementioned conditions of MRD i and SDRD , i the representative soil layers of unfrozen water content in the five sample plots (agricultural land, fallow land, shrubland, shrub-grass land, and woodland) are estimated to be 40, 30, 60, 50 and 60 cm, respectively.2. The variation coefficient is used to measure the degree of variation in soil moisture.Herein, Cv 10% means weak variation, 10% < Cv < 100% indicates medium variation, and Cv 100% shows strong variation (Nielsen and Bouma 1985).
In general, the variation coefficient of unfrozen water content of the five sample plots decreased with the rising soil depth.The soil unfrozen water content at 10∼100 cm showed weak variation and medium variation.The 10 cm soil layer had the largest variation coefficient (medium variation), and its variation degree was in the following correlation: shrub-grass land > fallow land> agricultural land> shrubland > woodland.The unfrozen water content of agricultural land, fallow land, and shrubland reached weak variation at 70∼100 cm soil layer, shrub-grass land at 60∼100 cm soil layer, and woodland at 30∼100 cm soil layer.According to the above analysis, the soil layer in weak variation means it has not entered the freezing period.Thus, it can be concluded that the unfrozen water content of woodland soil was much more stable throughout the seasonal freeze-thaw period.

Impact factor analysis
The 10∼100 cm soil unfrozen water content and environmental factors of five plots are used for redundancy analysis, so as to reveal the impact of environmental factors on the unfrozen water content in soil of different vegetation types (Vieira et al 2018).The data of the nine environmental factors (P, ET, CSR, H, AP, T, WV, WD, and TD) in the study area are collected.They were precipitation, evapotranspiration, solar radiation intensity, relative humidity, atmospheric pressure, air temperature, wind speed, wind direction, and surface temperature.R language data programming software is applied for the redundancy analysis of unfrozen water content and environmental factors of the five sample plots, and the sorting analysis diagram is shown in figure 5. Specifically, the blue rays represent the unfrozen water content at 10∼100 cm soil layers, and the red rays represent the environmental factors.
It can be clearly seen from the figure that the environmental factors of the five plots had a great influence on the unfrozen water content at shallow soil layers (10∼50 cm), and such impact varied with the environmental factors.In deep soil layers (60∼100 cm), the correlation between unfrozen water content and environmental factors weakened with the increasing soil depth.The unfrozen water content of the surface soil layer (10∼20 cm) in agricultural land, fallow land, shrubland, and shrub and grass mixed land had a strong correlation with ET, CSR, AP, T and TD, but a weak correlation with P, H, WV and WD.The surface soil (10∼20 cm) in woodland had a strong correlation with all the nine environmental factors.

Spatial heterogeneity of soil unfrozen water content
During the freezing period, the temperature decline triggers the downward freezing of the topsoil and consequently reduces the liquid water content of the soil (Xu et al 2022).Conversely, during the thawing period, rising temperatures initiate the melting process from the surface layer downwards, as well as from the maximum freezing depth upwards, resulting in an increase in the liquid water content within the soil (Xu et al 2018).Consequently, the unfrozen water content within the shallow soil exhibits a pattern of initial decline followed by an increase throughout the freeze-thaw period.In contrast, deep soil layers are less susceptible to external temperature variations, leading to a lower degree of freezing and a smaller solid water content, contributing to their relative stability.Research indicates that environmental and meteorological factors predominantly influence soil moisture dynamics during freeze-thaw periods, while vegetation types also play a role in moisture changes (Chu et al 2013, Lin et al 2023).These findings highlight the complex interplay between temperature dynamics, soil properties, and vegetation characteristics in shaping soil moisture patterns during the freezethaw period.

Rationality analysis of representative soil layers of unfrozen water content
To assess the reasonableness of the representative soil layers, regression analysis was performed on the unfrozen water content and average soil water content of each layer.The resulting regression models and related accuracy parameters for the representative soil layers in the five sample plots, along with the corresponding soil average unfrozen water contents, are summarized in table 3. The determination coefficient (R 2 ) values ranged from 0.73  to 0.94 for agricultural land and from 0.86 to 0.94 for fallow land, indicating a high level of accuracy in estimating the average unfrozen water content in the soil using the representative soil layer of 30 cm.However, when applying the relative difference method to the representative measurement points, the regression analysis revealed R 2 values below 0.5 for shrubland, shrub-grass land, and woodland plots.This suggests that the relative difference method is not appropriate for these three types of land cover.These findings underscore the importance of selecting suitable measurement methods based on the specific characteristics of different land cover types during the analysis of soil water content dynamics.Consequently, regression analysis was conducted to examine the relationship between the unfrozen water content and the mean values of shallow soil layers (20∼40 cm) in the aforementioned shrubland, shrub-grass land, and woodland.The results revealed determination coefficient (R 2 ) ranges of 0.81 to 0.90 for both shrubland and shrub-grass land, and a range of 0.65 to 0.87 for woodland.These values indicate a high degree of accuracy in estimating the average unfrozen water content in the soil using the representative soil layers of 20 cm, 30 cm, and 20 cm, respectively.This highlights the applicability of shallow soil layers in predicting the soil water content dynamics in different land cover types such as shrubland, shrub-grass land, and woodland.Proper consideration of the representative soil layers corresponding to each land cover type is crucial for accurately estimating soil moisture conditions during the freeze-thaw period.
Based on the aforementioned analysis, it can be deduced that the representative soil layers for the five plots predominantly ranged from 20 to 30 cm.Therefore, it can be inferred that the relative difference method is suitable for predicting the average unfrozen water content in the representative soil layers of agricultural land and fallow land.Conversely, this method is not deemed appropriate for estimating the unfrozen water content in soil for shrubland, shrub-grass land, and woodland (arbor).Moreover, regression analysis of the shallow unfrozen water content and average soil unfrozen water content has demonstrated its potential in determining the representative soil layers.In particular, the prediction of unfrozen water content in soil for shrub-grass land exhibits a higher level of accuracy.These findings emphasize the significance of employing suitable measurement techniques and considering the specific characteristics of various land cover types when investigating soil water dynamics during the freeze-thaw period.

Influencing factors of soil unfrozen water content
Tables 4-8 provide a comprehensive overview of the explanations and contributions of environmental factors within each sample plot to the unfrozen soil moisture content, as well as the results of Anova statistical testing of the variables.Through the analysis of the table, it is evident that evapotranspiration and temperature have a significant impact, contributing over 10% of the changes in unfrozen soil moisture content in farmland.Under the three ecological construction conditions of fallow land, shrubbery, and mixed grassland, evapotranspiration, air temperature, and surface temperature all contribute over 10% of the unfrozen water content changes.In addition, rainfall, temperature, and surface temperature have been identified as key factors in forest soil, resulting in changes in unfrozen water content exceeding 10%.These findings emphasize the importance of considering these environmental factors when examining and predicting soil moisture changes during freezethaw periods for different land cover types.The degree of contribution serves as an indicator of the influence exerted by environmental factors on soil unfrozen water content (Cheng et al 2021).Specifically, evapotranspiration and air temperature emerged as the primary contributors to variations in unfrozen moisture content.In the case of fallow land and shrubland, surface temperature was identified as the dominant environmental factor influencing the unfrozen moisture content in soil.For shrub and grass mixed land, the largest contributions to unfrozen water content were observed with evapotranspiration and surface temperature.In the context of woodland, rainfall and surface temperature were found to have the most substantial contributions to unfrozen moisture content.These findings provide valuable insights into the specific environmental factors that have the most significant impact on soil moisture dynamics during the freeze-thaw period in different land cover types.
During seasonal freezing and thawing periods, the unfrozen soil water content of agricultural land is primarily influenced by evapotranspiration and air temperature.This can be attributed to the increased exposure of agricultural land, leading to higher surface evapotranspiration rates (Miao et al 2017, Zhao et al 2019).In the case of fallow land and shrubland, surface temperature emerges as the dominant environmental factor controlling soil unfrozen water content.This can be attributed to the fact that surface temperature plays a crucial role in regulating the freezing and thawing processes in these land types.For shrub-grass land, which encompasses a greater presence of plant leaves, the unfrozen water content is predominantly affected by evapotranspiration and surface temperature.This can be attributed to the stronger transpiration rates associated with the presence of shrubs and grass, resulting in higher evapotranspiration rates (Man et al 2019).Finally, in woodland soil, the primary environmental factors influencing unfrozen water content are rainfall and surface temperature.This can be attributed to the significant role of rainfall in providing supplemental moisture to the soil, coupled with the influence of surface temperature on the freezing and thawing dynamics in woodlands.These findings provide valuable insights into the specific environmental factors governing the dynamics of unfrozen soil water content during the freeze-thaw period in different land cover types.

Conclusions
This study elucidates the spatiotemporal distribution characteristics of thawed soil moisture across various vegetation restoration modes and identifies the environmental factors controlling soil moisture dynamics during seasonal freeze-thaw periods under different vegetation restoration methods.The findings of this research hold significant theoretical and practical implications for regional soil erosion control and hydrological management.The primary conclusions derived from this study are as follows: (1) The temporal and spatial variation of unfrozen water content is found to exhibit distinct patterns in relation to different vegetation restoration methods.Across the five sample plots, the unfrozen water content in the 10 to 100 cm soil layer generally demonstrates weak to moderate variation, with the greatest degree of variation observed in the 10 cm soil layer.However, regardless of land type, the unfrozen water content consistently decreases with increasing soil depth after reaching the maximum range of change.
(2) The soil depth ranging from 0 to 10 cm exhibits the highest degree of variability and relatively low stability in terms of unfrozen water content.However, accurate estimation of the average soil unfrozen water content is feasible in representative soil layers, with the five sample plots highlighting the concentration of representative soil layers within the 20 to 30 cm depth range.These observations provide valuable insights into the spatial distribution and stability of unfrozen water content within different soil layers, contributing to a sound understanding of soil moisture dynamics during the freeze-thaw period.
(3) During the seasonal freeze-thaw period, evapotranspiration and air temperature emerge as the primary environmental factors governing the unfrozen soil moisture content in agricultural land.The unfrozen soil water content of both fallow land and shrubland is predominantly influenced by surface temperature.Moreover, in shrub-grass land, evapotranspiration and surface temperature play crucial roles in shaping the unfrozen water content dynamics.Additionally, rainfall and surface temperature exert the most substantial influence on the unfrozen soil water content in woodland.These findings contribute to a deeper understanding of the complexities involved in soil moisture dynamics during the freeze-thaw period and have significant implications for agricultural and ecological management strategies.
Although this study has achieved certain results, only the natural freeze-thaw soil water and heat changes have been studied, which cannot solidify many external factors.The detailed impact of specific individual factors on soil water and heat changes during the freeze-thaw period has not been solved, and further research will need to rely on indoor freeze-thaw experiments in the future.

Figure 1 .
Figure 1.Location of the Loess Plateau in China (a); Location of the study site on the Loess Plateau (b); Xindiangou Watershed Digital Elevation Model and location of the study site (c).

Figure 3 .
Figure 3. Seasonal variation of unfrozen water content in soil restored by different vegetation methods.(a) ∼ (j) represents the seasonal changes of soil unfrozen water content in different depths of 10 ∼ 100cm vegetation restoration.

3. 3 .
Analysis of difference of soil unfrozen water content and its influencing factors 3.3.1.Analysis of statistical parameters of soil unfrozen water content To better show the moisture content in different land use soils, this paper analyzes the difference of unfrozen water content of soil at a depth of 10∼100 cm, and the results are given in table

Figure 4 .
Figure 4. Average relative deviation and standard deviation of soil moisture content.

Table 1 .
Variation range of unfrozen water content in different soil depths (%).

Table 2 .
Difference analysis of unfrozen water content in seasonal freeze-thaw loess.

Table 3 .
Accuracy parameter of representative soil layer to estimate soil unfrozen water.

Table 4 .
Redundancy analysis results of agricultural land.

Table 5 .
Redundancy analysis results of fallow land.

Table 6 .
Redundancy analysis results of shrubland.

Table 7 .
Redundancy analysis results of shrub-grass land.

Table 8 .
Redundancy analysis results of woodland.