Multi-Scale Periodic Variations in Soil Moisture in the Desert Steppe Environment of Inner Mongolia, China

Abstract

Uncovering the complex periodic variations in soil moisture can provide a significant reference for climate prediction and hydrological process simulation. We used wavelet analysis to quantify and identify the multi-scale periodic variations of soil moisture in the desert steppe of Inner Mongolia from 2009 to 2019 and analyzed the differences between the multi-scale periodic changes in soil moisture at the bottom (BS) and upper slope (US). The results show that the interannual variability in soil moisture at the BS and US has a significant upward trend. Moreover, the amount and volatility decrease with the increase in soil depth in the range of 0–30 cm, and the soil moisture at the BS is 26.4% higher than the US. The soil moisture has periodic changes of 0.5a, 1a, 2a, 3a and 3.5a in the desert steppe environment of Inner Mongolia. The periodic structure and intensity of different slope positions are greatly different. Soil moisture at the US has more complex multi-scale periodic changes, and the periodic oscillations of 3.5a, 3a, and 1a are dominant. The periodic oscillations of 0.5a and 1a are dominant at the BS. At the BS, the intensity of periodic oscillation of 1a after January 2015 has weakened. The weakening of soil moisture by temperature, rainfall and soil temperature caused the change in the multiple time-scale periodic oscillation of soil moisture.

1. Introduction

Soil moisture is an important parameter and plays an essential role in characterizing land–atmosphere coupling and surface hydrology [1]. On the one hand, soil moisture becomes a critical medium that affect the water cycle and atmospheric changes through the water vapor transfer method of evaporation and transpiration [2,3]. In surface hydrology, soil moisture controls the redistribution of surface water. For example, low soil moisture leads to more surface infiltration and smaller surface runoff after rainfall. On the other hand, soil moisture affects atmospheric changes by changing land surface characteristics such as soil heat capacity, vegetation coverage, and surface albedo [4,5]. Meanwhile, under the joint action of climate factors and surface characteristics, soil moisture has a complex multi-scale periodic variation. In addition, soil moisture is an important factor restricting the growth of vegetation [6,7]. Soil moisture deficit leads to a significant decline in vegetation productivity [8,9], especially in arid and semi-arid desert steppe areas, and the growth and development of vegetation are more sensitive to the change in soil moisture [10,11,12]. Soil moisture becomes an important factor affecting the structure and function of grassland ecosystems [13]. The study of soil moisture change characteristics has important scientific value for soil moisture monitoring, drought risk warning, climate change prediction, and vegetation ecological management.

The desert steppe of Inner Mongolia, located in the mid-latitudes of the northern hemisphere, is a transitional area from a typical steppe to the desert. As a typical ground–atmosphere coupling region in Eurasia, this area is affected by the Siberian winter monsoon and the Pacific East Asian monsoon in winter and summer, respectively. Accompanied by climate change, the frequency and uncertainty of extreme weather events have become increasingly higher [14,15,16]. The climate in the desert steppe of Inner Mongolia has shown a trend of warming and drying [17,18], and the increasing frequency and intensity of droughts significantly impact the land–atmosphere coupling [19,20,21], which has an impact on the dynamic process of soil moisture in the desert steppe and its multi-scale period. The implementation of the “enclosure and grazing prohibition” and the “two-screens, three-belts” ecological security strategy have dramatically changed the vegetation coverage and community structure of the desert steppe in recent years [22,23]. This change increases the vegetation’s demand for soil moisture, thereby increasing the variability and uncertainty of the dynamic changes in soil moisture [24,25] and breaking the original periodic variations in soil moisture. Therefore, it is necessary to clarify the multi-scale periodic variations in soil moisture in the desert steppe, which will help us to understand the region’s meteorological changes and water balance with strong land–atmosphere coupling represented by the desert steppe and provide a theoretical basis for people to formulate a drought prevention and animal husbandry development model under climate change.

The dynamic change and multi-scale period of soil moisture has always been an important basic research area and focus for studying the interaction between soil and the near-surface atmosphere. Deng et al. [26] found that global soil moisture has shown a significant downward trend in recent decades, and global soil will continue to be dominated by aridification. Cheng and Huang [27] proposed that significant soil drying firstly occurred in the dry–wet transitional zone, including East Asia. In China, soil moisture has shown a significant downward trend in the past 30 years [28]. In terms of multi-scale periodic variation studies, Ma et al. [29] used 11-year soil moisture data from 98 observatories in China and found that soil moisture change has a cycle of 3–4 years. Jia et al. [30] found that soil moisture in plain areas has a time cycle of 3–5 years and 1.5–2.5 years. At the same time, topographic differences affect soil moisture changes through the redistribution of rainfall, and the difference in slope position is an important factor reflecting the impact of topography on soil moisture. Su et al. [31] showed that slope position has a significant impact on soil moisture and its change within a certain range of soil depth, and Meng et al. [32] found that soil moisture decreases with the increase in slope position in Maowusu sandland. Zhang et al. [33] showed that the difference in slope position affected the decay rate of soil moisture. Recent studies [29,34,35,36,37] have explored the spatiotemporal evolution characteristics, multi-scale periodic variations, and response to slope position of soil moisture in some typical ground–atmosphere coupling regions in China. However, most researchers only consider the difference in soil moisture and its dynamic changes at different slope positions, and there are relatively few studies on the difference in soil moisture periodic variation between different slope positions. Wavelet analysis, known as the mathematical microscope, can better explore the period frequency and local features of time series. Wavelet analysis was widely used to analyze long-term climate change [38,39], surface runoff characteristics [40,41,42,43] and other fields, and it can effectively identify and explore the multi-scale periodic variations in soil moisture. Scientific identification and analysis of the periodic variation in soil moisture in the desert steppe can provide an essential reference for hydrological process simulation and climate prediction in this area.

Therefore, the aims of this study were to (1) characterize the time series trend characteristics of soil moisture based on the ground high-time-resolution automatic soil moisture observation data; (2) study soil moisture’s multi-level time-scale structure and periodic change characteristics using wavelet analysis to determine the main periodic and oscillating characteristics; and (3) reveal the multi-scale periodic variations in soil moisture time series at different slope locations and the differences in influencing factors in different periods to raise awareness of the differences in soil moisture caused by different slope positions. It provides a theoretical basis for water cycle simulation and meteorological prediction of the coupling system of soil and air in the desert steppe. In practice, it provides management and a decision-making basis for rational allocation of water resources and vegetation restoration in the steppe.

2. Materials and Methods

2.1. Study Area

The study area (41°20′ N–42°40′ N, 109°16′ E–111°25′ E) is located on the northern edge of the Yinshan Mountains, in the transition zone from the Yinshan Mountains to the Inner Mongolia Plateau. It belongs to Darhan Muminggan United Banner, Baotou, Inner Mongolia. The topography of the study area is low in the north and high in the south, with an average elevation of 1367 m. It belongs to the mid-temperate semi-arid continental monsoon climate, with an arid and windy spring and autumn, abundant rainfall in summer, and a dry and cold winter. The mean annual precipitation is 284 mm, mainly concentrated in July to September, accounting for 76–80% of the annual precipitation. The mean annual evaporation is 2305 mm, and the moisture coefficient ranges from 0.13 to 0.31. The mean annual temperature is 2.5 °C, the annual accumulated temperature ranges from 1985 °C to 2800 °C, the annual average sunshine duration is 3100–3300 h, and the area experiences a frost-free period of about 83 d. The average wind speed is 4.5 m/s, and the maximum speed is 27.0 m/s. The main wind directions throughout the year are north and northwest.

The study area is located in the small Shangdong River watershed, a tributary of the Tabu River. Shangdong River is a seasonal river, and 3–5 floods flow into the Tabu River in summer under the influence of rainfall. The thickness of the aquifer is generally about 3–8 m, the buried depth of the roof is less than 10 m, the buried depth of the floor is 6–20 m, and the water level is 3–6 m. Plants depend almost entirely on natural precipitation for their water needs. The soil type is millet, and the soil texture is mainly sandy loam and light loam. The effective soil layer thickness is about 40 cm. The average soil porosity is 69.13%, and there were significant differences between BS and US in soil porosity (p < 0.05,

Table 1. Statistical characteristics of soil physical properties in different slope positions.

	Soil Bulk Density /g·cm−3	Soil Porosity/%	Clay Sand/% (<0.05 mm)	Fine Sand/% (0.05–0.1 mm)	Coarse Sand/% (0.1–2 mm)
US	1.346 ± 0.122	65.711 ± 2.338	36.238 ± 4.95	20.905 ± 4.113	42.857 ± 6.068
BS	1.34 ± 0.143	72.549 ± 2.388	41.053 ± 9.036	18.082 ± 5.56	40.865 ± 7.672
Mean	1.343 ± 0.127	69.13 ± 4.222	38.646 ± 7.387	19.493 ± 4.891	41.861 ± 6.676

). The zonal vegetation in the study area is mainly shallow root grass and grasses. Stipa krylovii Roshev is the founding vegetation group. The dominant vegetation species are Artemisia frigida, Cleistogenes squarrosa, Convolvulus ammannii Desr, Heteropappus altaicus, Agropyron cristatum, and Leymus chinensis. The vegetation height is 30–50 cm, and the coverage is 25–45%. The plant roots are mainly distributed in the 0–30 cm soil layer.

2.2. Data

Soil moisture and meteorological data come from the National Field Scientific Observation and Research Station of the Eco-hydrology of the Desert Steppe on the Southern Edge of the Inner Mongolia Plateau. In the study area, we set up two different slope positions on a typical slope (the slope is 3°, the slope direction is northeast-southwest): the upper slope (US, 41°21′10″ N, 111°12′34″ E, the altitude is 1610 m) and the bottom slope (BS, 4°20′55″ N, 111°12′22″ E, the altitude is 1600 m) (Figure 1). Soil moisture observation stations were established at two slope positions to monitor soil moisture for a long time (began in 2008). The two observation stations are 541 m apart. The observation instrument is an AZ-DT soil moisture monitoring station (produced by IMKO company, Germany, and the data collector is DT-80 produced in Australia) to collect soil volume water content. The soil moisture sensors were placed at soil depths of 5 cm, 15 cm, and 25 cm, respectively, representing the soil moisture of 0–10 cm, 10–20 cm, and 20–30 cm. This study selected soil moisture data from 25 May 2009 to 16 August 2019, with a time resolution of 30 min.

Aberrant values of soil moisture that were outlierswere eliminated, and then the average value of the data in the adjacent period was calculated for interpolation. Missing and abnormal data were less than 5% of the total data volume, and their impact was negligible. To ensure the reliability and consistency of the data, we used the soil moisture observed by the UGT at a depth of 5 cm to verify the soil moisture observed by the AZ-DT. The two soil moisture data sets were highly correlated and showed good consistency.

Meteorological data come from the UGT Automatic Meteorological Station (produced by UGT Germany, 41°21′13″ N, 111°12′27″ E, the altitude is 1600 m, Figure 1) in this study area. It is located on the same slope and has similar site conditions to the soil moisture monitoring station. Meteorological data include air temperature, rainfall, soil temperature, wind speed, and solar radiation. The air temperature sensor is 1.5 m from the ground with a measurement accuracy of 0.1 °C, the rain gauge is a non-heated type with an accuracy of 0.1 mm, and the buried depth of the soil temperature sensor is 5 cm. The measurement range of wind speed is 0.5–40 m/s, and the measurement range of solar radiation data is 0–1400 W/m² with an accuracy of 1 W/m². The meteorological data’s research period and time resolution are consistent with the soil moisture data.

2.3. Data Analysis

Firstly, we used the original soil moisture data to calculate the daily average and monthly average values, and the Z-score method was used to standardize the soil moisture time series:

$$X_{st}=\frac{\left(X-\overline{x}\right)}{\sigma }$$

(1)

where $X_{st}$ is the value after standardization, $X$ is the value to be standardized in the data, and $\overline{x}$ and $\sigma$ are the mean and standard deviation of the time series respectively.

Then, we analyzed intermonthly periodic variation characteristics of soil moisture using the wavelet analysis method, which can simultaneously realize time- and frequency-domain analysis to reveal multiple cycles’ changes hidden in time series. Compared with traditional time series analysis methods, it can characterize the local characteristics of time series at a different time and frequency for accurate frequency positioning for non-stationary time series affected by multiple factors. The soil moisture time series can be decomposed into discrete signals. The basic principle of wavelet analysis is to use a cluster of wavelet functions to represent or approximate the signal. Therefore, the key to wavelet analysis is the wavelet function. The wavelet function refers to a type of function that is oscillating and can quickly decay to zero, that is, the wavelet function $\psi \left(t\right)\in L^2\left(R\right)$ and satisfies:

$${\int }_{-\mathrm{\infty }}^{+\mathrm{\infty }}\psi \left(t\right)dt=0$$

(2)

where $\psi \left(t\right)$ is the fundamental wavelet function, which can form a cluster of functions through scale expansion and translation on the time axis:

$${\psi }_{a,b}\left(t\right)={\left|a\right|}^{-\frac{1}{2}}\psi \left(\frac{t-b}{a}\right) a,b\in R,a\ne 0$$

(3)

where ${\psi }_{a,b}\left(t\right)$ is a sub-wavelet, $a$ is a scale factor, which reflects the period length of the wavelet, and $b$ is the translation factor, which reflects the amount of translation in time. If ${\psi }_{a,b}\left(t\right)$ is the sub-wavelet given by Formula (3), then for a given finite signal $f\left(t\right)\in L^2\left(R\right)$, the continuous wavelet transform equation is:

$$W_f\left(a,b\right)={\left|a\right|}^{-\frac{1}{2}}{\int }_R^{}f\left(t\right)\overline{\psi }\left(\frac{t-b}{a}\right)dt$$

(4)

where $W_f\left(a,b\right)$ is the wavelet transform coefficient, $f\left(t\right)$ is a signal or square integrable function, $a$ is the scaling scale, $b$ is a translation parameter, and $\overline{\psi }\left(\frac{t-b}{a}\right)$ is the complex conjugate function of $\psi \left(\frac{t-b}{a}\right)$. In actual situations, the time series are usually discrete. Set the function $f\left(k_{\Delta \mathrm{t}}\right)$, $(k=\mathrm{1,2},\dots ,N; \Delta \mathrm{t} \mathrm{i}\mathrm{s} \mathrm{t}\mathrm{h}\mathrm{e} \mathrm{s}\mathrm{a}\mathrm{m}\mathrm{p}\mathrm{l}\mathrm{i}\mathrm{n}\mathrm{g} \mathrm{i}\mathrm{n}\mathrm{t}\mathrm{e}\mathrm{r}\mathrm{v}\mathrm{a}\mathrm{l})$, then the discrete form of the above formula is:

$$W_f\left(a,b\right)={\left|a\right|}^{-\frac{1}{2}}\Delta t{\sum }_{k=1}^{N}f\left(k\Delta t\right)\overline{\psi }\left(\frac{k\Delta t-b}{a}\right)$$

(5)

The wavelet basis function selected in this study is a Morlet continuous complex wavelet function, which can satisfy the multiple time-scale characteristics of soil moisture in this experiment. Moreover, the phase difference between the real part and the imaginary part of the complex wavelet function is π/2, which can eliminate the false oscillation generated by using the real wavelet coefficient as the judgment basis. This study used Morlet wavelet analysis in the Matlab wavelet toolbox to carry out continuous wavelet transformation for soil moisture time series and calculate the wavelet coefficients under every $a$ and $b$ value. The real part and modulus of the wavelet coefficients were used to draw the contour map of the real part and wavelet coefficient modulus, respectively. The wavelet real-part contour map represents the distribution characteristics of signals at different times and frequencies. Positive and negative wavelet coefficients indicate relatively high and low periods of soil moisture, respectively, and a wavelet coefficient of 0 indicates a sudden change point. The wavelet coefficient modulus represents the periodic oscillation intensity of the corresponding period and scale, which is used to verify the wavelet coefficient’s real part distribution.

Calculating the wavelet variance can further reflect the distribution of time series fluctuation energy on each time scale and is used to determine the primary cycle in the process of soil moisture change. The $x$ value corresponding to the maximum peak value of the curve in the wavelet variance map is primary cycle, and each primary cycle has a corresponding real-part process line of the wavelet coefficient, which can identify the fluctuation characteristics of soil moisture under the primary cycle. Integrate the square value of the wavelet coefficient in the $b$ domain to get the wavelet variance:

$$Var\left(a\right)={\int }_{-\mathrm{\infty }}^{+\mathrm{\infty }}{\left|W_f(a,b)\right|}^{2}db$$

(6)

IBM SPSS^® Statistics 25 software was used to conduct outlier tests and eliminate abnormal data, while standardizing time series. Wavelet analysis was calculated by Matlab 2016 and Office Excel software 2021. Correlation analysis was done by undertaken with IBM SPSS statistics 25 software. Images were drafted in Origin 2018 software.

3. Results

3.1. Time Series of Soil Moisture

Figure 2 shows the interannual variation trend of soil moisture in each soil layer of 0–30 cm at BS and US from May 2009 to August 2019, displaying a similar and consistent fluctuation trend between the slope positions. Soil moisture in each soil layer showed a significant upward trend (p < 0.05). From the upper to the lower layer, soil moisture at the BS was consistently higher than at the US, averaging 20.22%, 12.94%, and 11.16%, in contrast to 15.03%, 10.43%, and 9.59%, respectively. Soil moisture gradually decreases with the deeper soil. The mean soil moisture in the 0–30 cm soil layer of BS was 1.26 times that of US, the standard deviation of BS was 4.97, the standard deviation of US was 2.97, and there were significant differences in soil moisture between different slope positions (p < 0.05). Surface soil moisture at the BS and US fluctuates wildly, and the fluctuation range and standard deviation decreases with deeper soil (

Table 2. Statistical characteristics of soil moisture and correlation between monthly variation (MV) and precipitation in each layer.

	BS 0–10/%	BS 10–20/%	BS 20–30/%	US 0–10/%	US 10–20/%	US 20–30/%
Average variation	3.57	3.04	2.45	3.18	1.76	1.52
Standard deviation	5.46	4.86	4.54	3.81	2.6	2.47
Correlation between MV and precipitation	0.265 **	0.267 **	0.299 **	0.086	0.204 *	0.207 *

Note(s): * Significant correlation at the 0.05 level; ** Significant correlation at the 0.01 level.

). The correlation between rainfall and the monthly variation in soil moisture (Table 2) shows that rainfall’s uncertainty and pulse characteristics make the soil surface moisture more volatile, especially at the BS.

3.2. Periodicity of Soil Moisture Characterized by Wavelet Analysis

Wavelet real-part contour maps of the six time series (0–10 cm, 10–20 cm, 20–30 cm at BS, and 0–10 cm, 10–20 cm, 20–30 cm at the US) show the multi-time-scale periodic distribution characteristics of the soil moisture (Figure 3).

Overall, there are multiple time-scale characteristics of the change process of soil moisture from May 2009 to August 2019. There are cyclical changes in soil moisture on scales of 6–11 months, 12–24 months, 25–42 months, and 51–64 months at the BS. On the time scale of 51–64 months, the cyclical change in soil moisture experienced a global alternation of four high-value and three low-value periods, and on the time scale of 25–42 months, the cyclical change in soil moisture experienced a global alternation of six high-value and six low-value periods, but the intensity of cyclical oscillations on these two time scales was weak. As time increases, the periodic oscillation of soil moisture on this time scale gradually increases, and the phenomenon of increased periodic oscillation is more obvious at 10–30 cm (Figure 3c,e). On the time scale of 6–11 months and 12–24 months, the intensity of cyclical oscillations of soil moisture was high, and on the time scale of 12–24 months, soil moisture underwent five “high–low” oscillation alternations between January 2010 and June 2015.

There are cyclical changes in soil moisture on the scales of 6 to 24 months and 30 to 64 months at the US. Compared with the BS, the periodic distribution on each time scale is complex at the US, and the periodic oscillations on each time scale transform with each other. At 0–30 cm, the periodic oscillations on the 30- to 54-month time scale were obvious, and became weaker after January 2014. The periodic oscillations on the 48- to 64-month time scale were enhanced, and the periodic oscillations on the 30- to 54-month time scale shifted to 48 to 64 months. Meanwhile, compared with the BS, the cyclical change in soil moisture experienced a global alternation on the time scale of 6 to 24 months (Figure 3b,d,f).

Figure 4 shows the contour map of the wavelet coefficient modulus of the six soil moisture time series. The modulus variation is consistent with the strong periodic oscillation distribution in the real-part contour map. The periodic intensity on the 25- to 42-month time scale is global during the study period at the BS; however, as the soil deepens, the cyclical changes in soil moisture on this time scale slightly weaken, and the weakening trend is more evident before January 2015. Meanwhile, January 2015 was also the time node when the modulus disappeared on the 9-month and 18-month time scales. The periodic intensity on the 36- to 54-month and 48- to 64-month time scales was apparent during the study period at the US. In particular, with the deepening of the soil, the 30- to 54-month periodic oscillation shifted to the 48- to 64-month periodic oscillation more obviously. The periodic change on other time scales was not apparent.

The soil moisture at the BS has four primary cycles: 9-month, 18-month, and 34 to 36-month, respectively. The 18-month cycle is the strongest, followed by the 9-month, and 34- to 36-month. The first primary cycle of soil moisture at the US is 63 months for all layers, but the second and third primary cycles are different by layer (Figure 5). The second primary cycles of 0–10 cm, 10–20 cm and 20–30 cm were 9-month, 19-month and 52-month, and the third primary cycles of 0–10 cm, 10–20 cm and 20–30 cm were 19-month, 11-month and 19-month, respectively. Compared with the BS, the distribution of primary cycles in the soil layers of the US is inconsistent.

The real-part process line of wavelet coefficients (Figure 6) shows that on the 18-, 9-, 34 to 36-month time scale, the average cycle of soil moisture change at the BS is about 12 (1a), 6 (0.5a), and 24 months (2a), respectively. For the 18-month time scale, the intensity of cyclical oscillations was relatively small before July 2011 and after January 2015. The periodic fluctuation of soil moisture in 10–20 cm and 20–30 cm is relatively weak compared to 0–10 cm on this 34- to 36-month time scale in the early period of this study. At the US, the average cycle of soil moisture change is about 42 (3.5a), 36 (3a), 12 (1a), and 6 (0.5a) months on the 63-, 52-, 19-, 9-month time scales, respectively. On the 63-month time scale, the intensity of cyclical oscillations increases gradually over time, and cyclical oscillations of soil moisture on the 19- and 9-month time scales are similar to that of BS. It is worth noting that the periodic oscillation of the 52-month time scale only exists in 20–30 cm of US, and the periodic variation waveform of soil moisture is better (Figure 7).

3.3. The Relationship between Soil Moisture and Influencing Factors

It is worth noting that the soil moisture changes for the period of 1a on the 18-month time scale, and the intensity of periodic oscillation reflects the intensity of the seasonal variation in soil moisture. As mentioned previously, January 2015 is an important time node for periodic oscillation of soil moisture. At the BS, the periodic oscillation on the 18-month time scale suddenly disappeared in this time node (Figure 4 and Figure 6). Taking January 2015 as the boundary, two time periods were divided: May 2009–January 2015 (pre-period) and February 2015–August 2019 (post-period) to examine the abrupt changes in soil water oscillation intensity.

There were no significant differences in soil moisture (SM), air temperature (AT), rainfall (R), wind speed (WS), soil temperature (ST), or solar radiation (SR) between the two periods (p > 0.05) (

Table 3. Statistical characteristics of BS soil moisture and its influencing factors in two time periods.

		SM 0–10/%	SM 10–20/%	SM 20–30/%	AT/°C	R/mm	WS/m·s−1	ST/°C	SR/W·m−2
Total (n = 124)	Mean	20.22	12.94	11.16	3.29	20.15	2.52	7.03	189.88
	SD	5.49	4.88	4.55	12.98	24.63	0.7	12.38	62.44
	CV	0.27	0.38	0.41	3.94	1.22	0.28	1.76	0.33
Pre-period (n = 69)	Mean	19.99	12.38	10.94	2.82	21.66	2.62	6.51	186.6
	SD	6.05	5.29	4.94	13.21	25.52	0.78	12.2	61.73
	CV	0.3	0.41	0.45	4.68	1.18	0.3	1.87	0.33
Post-period (n = 55)	Mean	20.51	13.08	11.44	3.88	18.27	2.39	7.68	194
	SD	4.71	4.36	4.05	12.76	23.56	0.58	12.69	63.64
	CV	0.23	0.33	0.35	3.29	1.29	0.24	1.65	0.33

). As the primary factor affecting soil moisture change, the change in air temperature, rainfall, and soil temperature in the study area showed noticeable seasonal change (Figure 8) In the pre-period, soil temperature, air temperature, and rainfall were significantly positively correlated with soil moisture (p > 0.05), while in the post-period, soil moisture was not significantly correlated with any influencing factor (

Table 4. Correlation between BS soil moisture and its influencing factors in two time periods.

	Soil Layer/cm	AT/°C	R/mm	WS/m·s−1	ST/°C	SR/W·m−2
Total (n = 124)	0–10	0.293 **	0.258 **	−0.083	0.291 **	0.175
	10–20	0.17	0.142	−0.077	0.162	0.072
	20–30	0.291 **	0.231 **	−0.099	0.283 **	0.163
Pre-period (n = 69)	0–10	0.321 **	0.301 *	−0.21	0.331 **	0.16
	10–20	0.19	0.19	−0.19	0.2	0.04
	20–30	0.316 **	0.305 *	−0.21	0.326 **	0.13
Post-period (n = 55)	0–10	0.25	0.2	0.22	0.23	0.2
	10–20	0.13	0.07	0.16	0.1	0.11
	20–30	0.25	0.12	0.15	0.22	0.21

Note(s): * Significant correlation at the 0.05 level; ** Significant correlation at the 0.01 level.

). Under the influence of seasonal meteorological factors, the periodic oscillation on the 63-month time scale was weakened in the pre-period. The periodic oscillation on the 18-month time scale was prominent in the pre-period. As the correlation became insignificant in the post-period, the periodic oscillation on the 18-month time scale also disappeared in January 2015. Therefore, the variation in soil moisture oscillation intensity at the BS might not result from the mutation of soil moisture, but from the influence of temperature and rainfall after 2015 nonlinearly.

4. Discussion

4.1. The Time-Series Trend Characteristics of Soil Moisture and Influencing Factors

Soil moisture in the desert steppe in the study area increased significantly. Wang et al. [44] also found that soil moisture showed an overall upward trend in the Mongolian Plateau but an insignificant one, probably due to the large spatial scale. During the study period, there were significant differences in soil moisture between different slope positions and different depths, and the total amount and volatility of soil moisture gradually decreased with the deepening of soil depth, consistent with Fang et al. [45] and Zhang et al. [46]. Compared with deep soil, the surface soil is more sensitive to the supplement of rainfall and the surface evaporation dominated by temperature change, which leads to the fluctuation in surface soil moisture. The correlation analysis of this study also proves this point. The correlation between surface soil moisture and temperature and precipitation is more significant. The correlation between rainfall and monthly changes in soil moisture (Table 2) shows similar results to those reported by Zou et al. [47].

In arid and semi-arid areas, rainfall is the main source of soil moisture, the uncertainty and pulse characteristics of rainfall signal itself are transmitted from the soil surface to the lower layer, and the rainfall signal is attenuated in the downward transmission process, resulting in the decrease in the value and fluctuation of soil moisture with the deepening of soil, which is more volatile in BS 0–10 cm soil moisture. Consistently with the results of this study, Sun et al. [35] also believed that the amplitude of soil moisture signal will decrease with the increasing soil depth.

The mean value and standard deviation of soil moisture between BS and US were different, which was consistent with the results found by Meng et al. [32] and Zhao et al. [48] In addition to the supplement of precipitation, the soil moisture at the lower slope position is also supplemented by the surface runoff and interflow of other high terrains. When a rainfall event of >25 cm occurs, the soil moisture content at the BS increases or even reaches supersaturation to form surface confluence [49]. Furthermore, surface soil moisture at the BS is more volatile than at the US due to surface runoff and soil flow from other high terrains, increasing the soil moisture content and uncertainty [45]. As a static influencing factor of soil moisture, although there is little difference in the physical properties of soil at the BS and US, soil porosity and clay content at the BS are higher (Table 1), which have an important impact on soil water storage and holding capacity [13].

4.2. Multi-Scale Periodic Variation Characteristics of Soil Moisture

The periodic variations in soil moisture are complex, and the smaller periodic changes are nested within the larger periodic changes. The real-part process line of the wavelet coefficient (Figure 7 and Figure 8) shows that the average cycle of soil moisture change at 9 months, 18 months, 34 to 36 months, 52 months and 63 months were 0.5a, 1a, 2a, 3a and 3.5a, respectively. This means that the soil moisture has alternating dry and wet cycles of 0.5a, 1a, 2a, 3a and 3.5a. Huang and Ding [50] found that soil surface moisture had periodic fluctuations of 5.5 years around the research area, consistent with the 63-month first primary cycle of soil moisture obtained in this study. On a larger spatial scale, Ma et al. [29] found that soil moisture has a 3- to 4-year cycle in the mid-latitude region, which is basically consistent with the periodic variation (3a) of soil moisture in this study. Some scholars have also found that atmospheric factors have similar periodic fluctuations with soil moisture in the region. Sun [51] showed that temperature changed periodically for three years, consistent with the periodic change (3a) of soil moisture at the US in this research. Jiang et al. [52] found that the periodic fluctuation in precipitation within 4 to 5 years was larger than the fluctuation period of soil moisture in this study, which may be caused by the difference in rainfall in the study area.

Soil moisture was positively correlated with temperature and precipitation. Influenced by seasonal atmospheric factors (Figure 8), soil moisture also changed in a seasonal period of 1a. Compared with deep soils, surface soil moisture is more susceptible to periodic atmospheric factors, radiation fluxes and vegetation communities, especially in temperate regions with alternating seasons [49,53]. The 2a periodic change in soil moisture mainly shows the interannual alternation between the high-value period and the low-value period, and a low-value year and a high-value year constitute a 2a variation cycle of soil moisture. It is worth noting that the periodic oscillation on the 63-month time scale at the US gradually increases with time, and it is speculated that this time scale will remain for the main cycle and the periodic oscillation will continue to increase in the next few years. For the 9- and 18-month time scales, the strength of the periodic oscillations was relatively small before July 2011 and after January 2015. The periodic oscillation of soil moisture reflects the replacement of dry and wet soil in this region, while in the BS, the periodic oscillation waveform is poor and the periodic oscillation is relatively weak on all time scales except the 18- and 9-month ones (Figure 7).

4.3. The Difference in Soil Moisture Periodic Variations at Different Slope Positions

The characteristics of periodic variation in soil moisture on different slope positions are obviously different, including the differences in the primary cycle of soil moisture and the intensity of periodic oscillation. The periodic oscillation of soil moisture, dominated by the 18-month and 9-month time scales at the BS is more robust than the 9- to 19-month one at the US; however, the periodic oscillation of soil moisture, dominated by the 63-month and 52-month time scales at the US, is more robust than the same time scales at the BS. The periodic variation in soil moisture is the result of the comprehensive action of climate, soil, vegetation and topography, etc. As a dynamic influencing factor of soil moisture, the dynamic change in climate factors at various scales is the main influencing factor for the periodic oscillation of soil moisture. The precipitation redistribution caused by slope position resulted in differences in vegetation (root distribution, water holding capacity of plants, etc.), soil properties (bulk density, soil permeability, etc.), and surface heat characteristics (soil water–heat flux, etc.), which weakened or strengthened the effect of atmospheric factors on the periodic variation in soil moisture. However, with the increase in the research scale, this effect would gradually decrease. Temperature (air temperature and soil temperature) is the critical factor causing soil moisture fluctuation in the dry–wet climate transition zone [27]. Studies have shown that the positive feedback effect of temperature change on soil moisture in global climate change is greater than the supplementary effect of rainfall on soil moisture. Under the long-term disturbance of climate change and SST anomalies, the coupling mechanism between land and air in the dry–wet climate transition zone will become more complex and uncertain [54].

This study provides a scientific basis for quantitative monitoring of drought and flood disasters in desert steppe areas in the future. However, a complete understanding of drought evolution still needs to combine meteorological drought and socioeconomic indicators. As an essential link in the water cycle, under the influence of the atmospheric circulation, soil properties, human activities, soil moisture reserves and periodicity show large variability. In the larger space and water system, soil moisture and variability influence factors will become more complex. The future impact on the soil water cycle should be focused on the whole regional water cycle system to examine periodic characteristics of soil moisture through model simulation.

5. Conclusions

This paper analyzed the periodic variation in soil temperature in the desert steppe from 2009 to 2019 by applying a wavelet analysis method. The inherent multi-scale period in the soil moisture time series was determined, and the difference in the structure and intensity of soil moisture periodic variations at different slope positions was analyzed. These results provide a new understanding of soil moisture evolution in the desert steppe environment.

The interannual variation in soil moisture has a noticeable upward trend. Moreover, the volatility decreases with the increase in soil depth. Soil moisture was significantly different at different slope positions. The soil moisture of BS is higher than that of US, and the periodic structure and intensity of different slope positions are greatly different. The soil moisture has periodic changes of 0.5a, 1a, 2a, 3a and 3.5a in the desert steppe environment of Inner Mongolia. The soil moisture at the BS mainly has three types of periodic variations: 0.5a, 1a, and 2a. The periodic oscillations of 0.5a and 1a are dominant and stable. Compared with the BS, the soil moisture at the US has more complex multi-scale periodic changes, and the soil moisture of the three soil layers mainly has four types of periodic variations: 3.5a, 3a, 1a, and 0.5a. The periodic oscillations of 3.5a, 3a, and 1a are dominant and stable. On the periodic variations of 1a at the BS, the oscillation intensity was relatively weakened after 2015. The weakening effect of temperature, rainfall, and soil temperature on soil moisture is the main reason for changing soil moisture multi-time scale cycle oscillation. Rainfall and soil moisture in this region have similar periodic fluctuations, and the fluctuation in rainfall is still the main influencing factor of soil moisture in the desert steppe region.