Fluctuations of continuous soil moisture evaporation under different rainfall conditions during the growing period of the non-monsoon season, the eastern Loess Plateau

Study area

The study area was located in an open farmland with the implementation of the policy of returning farmland (112°66′E, 37°74′N) in the ECLP, 772 m above sea level (Fig. 1). Mountains surrounded the experimental area on three sides, with terrain high in the north and low in the south (Li et al., 2022b). The climate pattern of the study area belongs to the continental temperate monsoon climate, and the annual average temperature is 9.5 °C, with large interannual and seasonal temperature fluctuations (the extreme maximum temperature is 37 °C, and the minimum temperature is −22.8 °C). Precipitation in the study area was relatively rare and concentrated, with an average annual precipitation of 468.4 mm, and was mostly concentrated in summer. The average annual potential evaporation in the study area was 1,774.5 mm (Sun et al., 2020a). Under the strong influence of the East Asian monsoon, the rainy season begins in the study area at the end of June. During the rapid growth period of plants before the rainy season, there is scarce precipitation, strong soil water evaporation, and large plant water demand. This is the most prominent contradiction between water resource supply and demand in this region, especially for agricultural systems (Xiang et al., 2020). Therefore, the non-monsoon growth period from April 1 to July 1 was selected as the study period. The average temperature in this period was 12 °C, the average precipitation was only 57.5 mm (13–15% of the year (Sun et al., 2020b)), and the potential evaporation was 517.4 mm (29% of the year). The soil type in the study area is mainly brown soil with loose soil quality and a deep soil layer (Guo et al., 2019). In addition, the study area is a typical dryland agricultural area, and crop planting has a large demand for water resources. The main crops include winter wheat, spring wheat, and spring corn (Liu et al., 2022).

Information on sampling and measurement

Soil water samples were collected from the test site of the Ecological Environment Research Center of the Middle Reaches of the Yellow River at Shanxi Normal University. In this study, soil samples were collected daily continuously in the study area from April 1, 2023, to July 1, 2023, and sampling activities were conducted daily from 18:00 to 19:00. Three parallel quadrats, 1 × 1 m, were selected in each sampling activity for further study. Soil samples were obtained using a 38-mm diameter handheld soil drill. To avoid the influence of air on the soil samples, the 1–2 cm soil surface was removed using a shovel before each sampling. Soil samples were collected from a depth of 0–100 cm at 20 cm intervals. The different samples from the same layer of three parallel quadrats were mixed to create a composite sample. After sampling, part of the soil was loaded into a 10 mL glass bottle, quickly sealed with a Parafilm membrane, and brought back to the laboratory for soil water extraction. The other part was layered and packed into a sealed bag to measure soil moisture content (SWC). Atmospheric precipitation samples were collected during precipitation events. A rain bucket was placed near the sample point to collect the precipitation. Precipitation samples were sealed immediately after precipitation (stored in a 10 mL sampling tube and sealed with parafilm), and the corresponding meteorological elements (precipitation time, real-time precipitation, temperature, relative humidity, and wind speed) were simultaneously recorded. During the sampling period (April 1, 2023–July 1, 2023), 460 soil samples and 16 precipitation samples were collected.

During the sampling period, meteorological data of the study area (air temperature (T), precipitation amount (P), relatively humidity (RH), wind speed (WS)) were downloaded from the https://cds.climate.copernicus.eu/datasets/reanalysis-era5-land?tab=download (Taiyuan airport), and the ET0 data were calculated by using the Penman-Monteith formula calculator.

The stable isotopes of the water samples were measured using a Liquid Water Isotope Analyzer (Los Gatos Research, Inc., San Jose, CA, USA, Type: GLA431-LWIA) at the Ecological Environment Research Center in the middle reaches of the Yellow River at Shanxi Normal University. The measurement precision was 0.2‰ and 0.03‰ for δ²H and δ¹⁸O, respectively. The Vienna Standard Mean Ocean Water (V-SMOW2) was used to revise the measured results (δ-values, shown as Eq. (1)). To reduce cross-contamination, the number of samples measured in each batch was controlled at approximately 35. Before each batch of samples was measured, a standard sample was prepared, the syringe was cleaned, and the diaphragm was replaced. Each sample was retested six times, and the LWIA Post Analysis software (v. 3.1.0.9) was used to reanalyze the measured data. The average of the last four needles of the measured data was selected as the output value (Sun et al., 2020c).

(1) $$\mathrm{\delta }\left(‰\right)=\left({\displaystyle \frac{R_{\mathrm{s}\mathrm{a}\mathrm{m}\mathrm{p}\mathrm{l}\mathrm{e}}-R_{\mathrm{s}\mathrm{t}\mathrm{a}\mathrm{n}\mathrm{d}\mathrm{a}\mathrm{r}\mathrm{d}}}{R_{\mathrm{s}\mathrm{t}\mathrm{a}\mathrm{n}\mathrm{d}\mathrm{a}\mathrm{r}\mathrm{d}}}}\right)\times 1000\mathrm{\%}$$where R_sample is the ratio of ²H or ¹⁸O in the measured sample, and R_standard is the ratio of ²H or ¹⁸O to the Vienna standard mean ocean water.

Lc-excess

In this study, the intensity of soil water evaporation relative to local precipitation was characterized by calculating the Lc-excess for each soil water sample. The linear relationship of δ²H and δ¹⁸O is defined as the local atmospheric waterline (LMWL) and soil waterline (SWL) in precipitation and soil water, respectively. In contrast, the deviation between LMWL and SWL can be defined by Lc-excess, which can be used to determine the evaporative fractionation process of different water bodies (Hasselquist et al., 2018; Hervé-Fernández et al., 2016).

(2) $$Lc-excess={\delta }^{2}H-a{\delta }^{18}O-b$$where a and b represent the slope and intercept of LMWL, respectively, and δ²H and δ¹⁸O values represent the hydrogen and oxygen isotope values in the sample, respectively. Generally, the annual average value of lc-excess in precipitation is 0‰, whereas the water body lc-excess is less than 0‰ (and is below LMWL), indicating that the water body is affected by strong evaporation. However, lc-excess showed a positive value in the water samples, indicating that the water samples may be affected by water sources other than precipitation (Landwehr, Coplen & Stewart, 2014).

Quantifying soil moisture evaporation using the C-G model

Soil evaporation loss f under continuous evaporation conditions from April 1, 2023, to July 1, 2023, was quantified using the nonstationary model in the Craig–Gordon model (non-steady-state model) (assuming that the stable isotope composition of soil water is altered by isotopic fractionation during evaporation) (Skrzypek et al., 2015; Li et al., 2023; Craig, Gordon & Horibe, 1963). The equation is as follows:

(3) $$f=1-{\left[{\displaystyle \frac{{\delta }_{L(\mathrm{s}\mathrm{o}\mathrm{i}\mathrm{l})}-{\delta }^{\ast }}{{\delta }_{P(\mathrm{s}\mathrm{o}\mathrm{i}\mathrm{l})}-{\delta }^{\ast }}}\right]}^{{\displaystyle \frac{1}{\mathrm{m}}}}$$where δ_P(soil) is the initial value of the soil-water sample oxygen isotope composition (δ¹⁸O_P(soil)), δ_L(soil) is the final value of the soil-water sample oxygen isotope composition, and δ* is the limiting factor for oxygen isotope enrichment (Gat, 1978). The calculation method is as follows:

(4) $${\delta }^{\ast }={\displaystyle \frac{h\times \delta A+\epsilon }{h-\frac{\epsilon }{1000}}}$$where h is the average relative humidity (fraction), ε is the total isotope fractionation (‰), and δA is the stable isotopic composition of water in ambient air (‰). δA is generally determined based on the isotopic composition of precipitation samples (Gibson & Reid, 2014). The equation is as follows:

(5) $$\delta A={\displaystyle \frac{{\delta }_{rain}-{\epsilon }^{+}}{{\alpha }^{+}}}$$where δ_rain is the measured value of the precipitation isotope at the beginning of each continuous evaporation, and ε⁺ and α⁺ are the equilibrium enrichment factor and equilibrium fractionation factor that change with temperature, respectively. This can be calculated using the following equation:

(6) $${\epsilon }^{+}=\left({\alpha }^{+}-1\right)\times 1000$$

(7) $${{\alpha }_{{(}^{18}O)}}^{+}=e^{\left[-7.685\ast {10}^{-3}+6.7123\ast T^{-1}-1.6664\ast {10}^3\ast T^{-2}+0.35041\ast {10}^6\ast T^{-3}\right]}$$where temperature (T) is expressed in Kelvin.

(8) $$\epsilon ={\displaystyle \frac{{\epsilon }^{+}}{{\alpha }^{+}}+{\epsilon }_k}$$where ε is the total isotopic fractionation factor (‰) and εk is the kinetic isotopic fractionation factor (‰).

(9) $${\epsilon }_k=(1-h)\times n\theta C_D$$where n is a constant, which is related to the correlation between the molecular diffusion resistance and molecular diffusion coefficient and is usually considered to be 1 for non-mobile air layers (e.g., soil moisture evaporation and plant transpiration), θ is the ratio of the molecular diffusion coefficient to the total diffusion fractionation coefficient and is generally considered to be 1 for soil moisture evaporation. CD is a parameter that describes the diffusion efficiency of the molecule and has a value of 25.1‰ and 28.5‰ for hydrogen and oxygen, respectively (Crusius & Thomson, 2000).

(10) $$\mathrm{m}={\displaystyle \frac{h-\frac{\epsilon }{1000}}{1-h+\frac{{\epsilon }_k}{1000}}}$$where m represents the correlation between the isotopic compositions of the local evaporated water vapor and liquid water, also known as the enrichment slope of the local evaporated water line (Wendong et al., 2023).

The redundancy analysis and the multiple linear regression analysis

The redundancy analysis (RDA) was used to explore the relationship between soil moisture evaporation loss and the main influencing factors using RStudio4.3.2 software (Capblancq et al., 2018; Du et al., 2021). In this study, the response matrix was the evaporation loss matrix at different soil depths and the explanation matrix was a matrix composed of environmental factors. Based on Monte Carlo random permutation tests, each ranking axis was tested individually.

Multiple linear regression was used to further fit the relationship between the continuous soil moisture evaporation loss and environmental factors (Sixtine et al., 2021; Nejatian et al., 2023). In this study, the multiple linear regression functions of the explained variable y (soil moisture evaporation loss) and the variables X_2, X_3, …, X_k (the main influencing factors) were as follows:

(11) $$y=\alpha +{\beta }_1{x}_1+{\beta }_2{x}_2+{\beta }_3{x}_3+...+{\beta }_k{x}_k$$where β_j (j = 1, 2,…, k) is the parameter of the model; k is the number of explanatory variables. Below, R² was used to test the goodness of fit, and a t-test was used to test the significance of the regression parameters.

Information for regional air temperature, humidity, potential evaporation, and precipitation

During the study period, the daily air temperature (T) ranged from 4.7 °C to 35 °C, with an average value 22 °C. The lowest and highest air temperatures occurred on April 11 and June 22, respectively. Air temperature showed an overall upward trend, with a rapidly increasing trend appearing between May and June. During the study period, the daily average relative humidity (RH) varied between 11% and 86%, with a mean value of 34%. Relatively lower RH values were observed in June, and relatively higher values were observed in May (Fig. 2). The daily potential evapotranspiration (ET₀) during the study period ranged from 2.2 to 21.4 mm, with a mean value of 5.95 mm, and the total potential evapotranspiration was 549 mm. A relatively stronger ET₀ was observed in June.

Regional precipitation was relatively rare (only 161.1 mm) and less than the regional ET₀ level during the study period (Fig. 2). To facilitate the calculation of continuous soil water evaporation losses, this study divided 10 continuous evaporation periods according to the continuity of the 16 precipitation events. The soil moisture content gradually decreased during the study period. The soil moisture content fluctuated between 1.90% and 30.98%. The soil water content often showed a significant upward trend after regional precipitation events, especially after continuous larger precipitation events.

Temporal variability on regional soil moisture δ¹⁸O

The fluctuation range of the stable hydrogen and oxygen isotopes in the soil moisture was smaller than that of regional precipitation (−15.60% to 5.71% and −104.96% to 30.81%) (Table 1). In this study, the fluctuations in hydrogen and oxygen stable isotopes of soil moisture gradually decreased with increasing soil depth. The δ¹⁸O and δ²H values of surface layer soil moisture had a relatively larger amplitude, which indicated that the surface layer soil moisture was more significantly disturbed by the external environment. The δ¹⁸O and δ²H of soil moisture in the 40–60 cm soil depth layer showed relatively depletion values, whereas, in the surface layer, they showed relatively enriched values (Table 1).

Owing to the relatively similar temporal variability that was observed in the hydrogen and oxygen stable isotopes of regional soil moisture, only δ¹⁸O was used to discuss the temporal variation of soil moisture in this study. In the early part of the study period (2023.4.1–2023.5.6), a smaller fluctuation of the soil moisture δ¹⁸O value appeared in this period, with an average value of −5.65 ± 1.89‰, and the corresponding SWC also showed a stable trend (average value: 22.44%) (Fig. 2). From 2023.5.6 to 2023.5.14, soil moisture δ¹⁸O rapidly decreased at the beginning of this period and remained at a relatively low δ¹⁸O level, especially in the surface layer (Fig. 2). The corresponding meteorological elements also showed significant variations, and regional air temperature and ET₀ showed a significant upward trend during this period (2023.5.6–2023.5.20). The relatively depleted soil moisture δ¹⁸O in this period may be related to the melting and evaporation of the seasonal permafrost in the deep soil due to the influence of surface temperature and strong ET₀. The evaporated soil moisture of the deep soil layer (the depleted δ¹⁸O value) accumulated in the surface soil layer and further resulted in the dilution of soil moisture δ¹⁸O (Li et al., 2020; Song et al., 2017). Subsequently, soil moisture δ¹⁸O values gently increased (mean value: −8.54 ± 2.59‰) after week 6, while the SWC showed a slow fluctuating decline trend (average value 21.96%). From weeks 9–13 (2023.6.3–2023.7.1), soil moisture δ¹⁸O values in the study area presented notable fluctuations with a larger amplitude of the SWC variation (average value 18.95%), and the maximum soil moisture δ¹⁸O value was observed during the study period. Compared with the previous period, the average δ¹⁸O value of soil moisture in this period was more enriched (average value is −6.94 ± 2.23‰). Obviously temporal heterogeneity occurred in the soil moisture δ¹⁸O values in the study period, which was similar to the temporal variation trend of soil moisture δ¹⁸O in other regions (central and western regions) of the Chinese Loess Plateau (gradually enriched from April to June) (Wang et al., 2019; Zhang et al., 2024). In addition, the response of soil moisture δ¹⁸O variability to precipitation events during the 0–100 cm soil depth layers presented inconsistency, especially in April (Fig. 2A). After the precipitation event on April 4, the soil moisture δ¹⁸O in the 0–20 cm soil depth layer was significantly depleted the following day; the depleted soil moisture δ¹⁸O occurred 2 days later in the 20–100 cm soil depth layer. A similar phenomenon occurred after the precipitation event on April 23.

Spatial variability on regional soil moisture δ¹⁸O

Figure 3 shows the vertical spatial variation in the average values of soil moisture δ¹⁸O and δ²H values at different soil depth layers during 10 consecutive evaporation periods after precipitation events (the P1-P10). Overall, the soil moisture’s δ¹⁸O values had a “C” shaped vertical space distribution. Similar vertical spatial distribution characteristics of soil moisture’s δ¹⁸O values in the 0–100 cm depth soil layers were also observed in the central region of the Chinese Loess Plateau (Wang et al., 2019). The soil moisture δ¹⁸O values in the middle soil depth layers were relatively depleted, while relatively enriched soil moisture δ¹⁸O values appeared in the surface and bottom soil layers, which was inconsistent with the spatial variation of soil moisture’s δ¹⁸O values in the western and northern regions of the Chinese Loess Plateau (Xiang et al., 2021; Zhang et al., 2024). During the study period, the average soil moisture δ¹⁸O value was most enriched in the 0–20 cm soil depth layer, which gradually decreased with increased soil depth and reached the minimum value at the 40–60 cm soil depth layer (Fig. 3). In addition, the soil moisture δ¹⁸O was gradually enriched after the 40–60 cm soil depth layer with increased soil depth layer (Fig. 3).

According to the accumulated precipitation, 10 consecutive evaporation periods (the P1–P10) in this study were divided into three main types: accumulated precipitation amounts of 0–5 mm (the P2, P4, P5, P8, and P9), 20–30 mm (the P3 and P6), and 30–40 mm (the P1, P7, and P10). A similar trend of the vertical variability on soil moisture δ¹⁸O was observed during the three evaporation periods of the accumulated precipitation amount of 30–40 mm (the P1, P7, P10), which were also consistent with the vertical variation trend for the entire study period. In the five evaporation periods (the P2, P4, P5, P8, and P9) of the accumulated precipitation amount of 0.2–5 mm, the vertical spatial distribution of soil moisture δ¹⁸O values was also consistent with the entire study period except for the P9. During the P9, a relative enrichment peak value of soil moisture δ¹⁸O was observed in the 40–60 cm soil depth layer. An obviously inconsistency for the vertical spatial distribution of soil moisture δ¹⁸O occurred under 20-30mm precipitation conditions compared with the entire study period (Fig. 3), which showed relatively enrichment δ¹⁸O values at the 20–40 cm soil depth layer, whereas the relatively depletion δ¹⁸O values of the 0–20 cm soil depth layer occurred during this period. The vertical spatial heterogeneity of soil moisture δ¹⁸O was weak during the P1–P5, while the vertical spatial difference of soil moisture δ¹⁸O was gradually increased after the P6.

The relationship between δ¹⁸O and δ²H of soil moisture in different soil layers

During the study period, regional precipitation δ¹⁸O values varied from −15.60‰ to 5.71‰, with an average value of −5.03‰ and a standard deviation of ±5.11. Among them, the relatively higher standard deviation of precipitation δ¹⁸O values occurred in April (6.31‰), which indicated a more complex precipitation process in this month. The larger fluctuation of precipitation δ¹⁸O in April may have been related to regional temperature fluctuations and complex water vapor sources during this period (Craig, 1961). The meteorological water line (LMWL_4-6) during the study period was fitted using δ²H = 5.25δ¹⁸O + 2.39 (R² = 0.78, n = 16). The smaller slope of LMWL_4-6 indicates the regional climate pattern was relatively arid during the study period (Sun et al., 2020c). In this study, the slopes of the soil moisture line (SWL) at different depths of soil layers were smaller than that of the LMWL_4-6, which may be related to the strong evaporation of soil water (Li et al., 2023). The dispersion of the soil moisture sample points in the surface soil layer was greater and gradually decreased with increasing soil depth (Fig. 4). Compared with the other 3 months, the soil moisture samples in May were more dispersed, and significant inconsistencies appeared in the different soil layers during May. Most water moisture samples collected in April were concentrated in the middle of the LMWL_4-6 and were relatively displaced in different soil layers. The locations of the soil moisture sampling points were significantly different from those in the other months (Fig. 4). Most soil moisture samples were on the upper left of the LMWM_4-6 and GMWL, indicating a strong recycling process. The dispersion of the soil moisture sample points in June gradually weakened as the soil layer depth increased and almost coincided with the soil moisture samples in April in the 80–100 cm soil layer. The slope of the SWL in April generally increased with increasing soil depth. The smallest value of the SWL slope appeared at the dispersion of the soil moisture sample points in June, gradually decreased with increased soil layer depth, and almost coincided with the April soil moisture samples in the 80–100 cm soil layer, which indicated that the soil moisture samples were affected by regional evaporation during April. In May, the SWL slope was opposite to that in April, showing a gradually decreasing trend with increasing soil depth. The slope of the SWL in June was generally consistent with that in April.

Lc-excess of soil moisture at different soil depth layers

During the study period, the average lc-excess values in soil moisture were generally less than 0‰ owing to evaporative fractionation of stable isotopes. Mean values of soil moisture lc-excess in the soil layer at the 0–20, 20–40, 40–60, 60–80, and 80–100 cm depth soil layers were −14.88, −11.31, −14.56, −13.05, and −13.26‰, respectively. The average lc-excess value of soil moisture at the 0–20 cm soil depth layer was lower than that of the other soil layers, indicating the soil moisture at the 0–20 cm soil depth layer was more strongly disturbed by regional evaporation. The average lc-excess of soil moisture showed an increasing fluctuation trend with increasing soil depth in the study region, which was consistent with the spatial-temporal trend of lc-excess of soil moisture compared with other regions of the Chinese Loess Plateau (Xiang et al., 2021; Zhang et al., 2024) (Fig. 5).

The soil moisture lc-excess showed a “low-high-low” temporal trend, which was contrary to the temporal trend of soil moisture δ¹⁸O (Fig. 5). During weeks 1–5, all the soil moisture lc-excess was below 0‰ with a slight fluctuation, indicating that the soil moisture at different depths experienced relatively stable intensity evaporation during this period. The increasing soil moisture lc-excess after week 6 (even located above 0‰) indicated the soil moisture was affected by other water sources except for rainfall during this period, which may be related to the condensation of water vapor (evaporated from surrounding vegetation soil and lakes) and being adsorbed by the topsoil at night. This season is the high-incidence period of dew formation in the Loess Plateau. The adsorption of dew (the depleted stable isotopes) by topsoil may be a possible reason for the lower soil moisture δ¹⁸O values and higher soil moisture lc-excess values during this period (Li et al., 2022a). In this period, the depth layer soil moisture also began to be affected by evaporation along with the rapid rise of surface temperature, and the accumulation of the evaporated water vapor (depleted δ¹⁸O) in the surface soil layer may also be an important reason for depleted δ¹⁸O and higher lc-excess during this period (Fig. 5). During the study period, the soil moisture lc-excess gradually decreased in the 0–20 cm soil depth layer, whereas the soil moisture lc-excess in the deeper soil layers was more stable, except in early May (Fig. 5).

Quantitative assessment on evaporation loss of soil moisture

Using the C-G model, we estimated the evaporation loss (f) of the soil layer with a 0–100 cm depth under continuous evaporation during the observation period. Significant spatial and temporal heterogeneity was observed in the soil moisture evaporation loss fraction (f) during the study period (0–23.35%) (Fig. 6). Overall, the evaporation loss of soil moisture was weak at the beginning of the study period, whereas a relatively strong soil moisture evaporation loss occurred between mid-late May and mid-June (Fig. 6). The mean values of soil moisture’s f in April, May, and June were 2.74 ± 1.06%, 11.36 ± 5.70%, and 8.41 ± 3.10%, respectively. Soil water’s evaporation loss in this study area showed inconsistent temporal variation trends compared with other regions of the Chinese Loess Plateau (Yong et al., 2020; Zhang et al., 2024), which posed challenges for the management of soil water resources in the non-monsoon period of the study area. The soil moisture evaporation loss was relatively weak at the end of precipitation, whereas a relatively larger intensity of continuous evaporation loss of soil water usually occurred at longer precipitation intervals. A relatively smaller soil moisture f was observed in the 0–40 cm soil layer, except for the P1 and P2 (Fig. 6). In the 40–80 cm soil layer, the soil moisture f at the P6 and later periods showed significant vertical spatial inconsistency (Fig. 6). At 80–100 cm soil layer, significant soil water evaporation losses occurred only in the P6 and the P8. As for temporal variation, soil moisture evaporation loss showed relatively lower and stable values during the P1 and P2 and then gradually increased. Soil moisture f increased from the P6 and reached an extremely large value on May 20.

The maximum value of soil moisture evaporation loss appeared in the 0–20 cm soil depth layer (8.97%), and the soil evaporation loss fluctuated as the soil depth increased, which was consistent with the lc-excess variation in soil moisture in “Lc-excess of Soil Moisture at Different Soil Depth Layers”. The relatively smaller soil moisture f in April (with a weak vertical spatial difference) indicated weaker regional soil moisture evaporation. A relatively high soil moisture f was observed in May, and the soil moisture evaporation intensity of the surface soil layer was significantly higher than that of the deep soil layer during this month. Significant vertical spatial heterogeneity of soil moisture evaporation loss was observed in late June, which may be related to fluctuations in the main environmental elements (precipitation, temperature, and humidity) during this month.

Determination of the relationship between soil water evaporation loss and influencing factors

Several studies have revealed that the mechanism of soil water migration in the Loess Plateau is complex, and regional soil moisture evaporation loss is evidently affected by the main environmental factors (such as air temperature, precipitation, potential evaporation, soil water content, relative humidity, and vegetation cover) (Dubbert et al., 2013; Li et al., 2020; Song et al., 2017). Because the sampling site of this study area had no evident vegetation coverage or significant groundwater recharge (the altitude of the sampling depth was higher than the regional lake surface), air temperature (T), potential evaporation (ET₀), soil water content (SWC), wind speed (WS), and atmospheric relative humidity (RH) were selected as potential factors for regional soil moisture evaporation loss to conduct the correlation analysis.

The RDA results showed an inconsistency in the main controlling factors of soil moisture evaporation loss at different depths of the soil layers (Fig. 7). Overall, the f of soil moisture in the study area showed a significant correlation with T, ET₀, and SWC. While a relatively weak correlation was observed between f and WS/RH. Among them, a negative correlation was observed between SWC and RH and continuous soil evaporation loss. In the 0–20 cm soil depth layers, regional ET₀, WS, and T were positively correlated with soil moisture f, and the order of the correlation degree was ET₀ > WS > T, indicating that the evaporation process of the surface soil layer was mainly affected by the variability in regional ET₀ and WS. In the 20–60 cm soil depth layer, a positive correlation was observed between regional ET₀ and regional soil moisture evaporation loss, and the influence of regional T variability on soil moisture evaporation loss showed an increasing trend. A weak correlation between ET₀, WS, and T and regional soil moisture f was observed in the 60–100 cm soil depth layer, which indicated that soil layers below 60 cm depth were not disturbed by external environmental elements.

Generally, a correlation existed between the continuous evaporation loss of regional soil moisture and the above five environmental control factors; however, this correlation gradually weakened with increasing soil depth. The variation in regional air temperature had a stronger connection with the f of soil moisture, and increasing air temperature resulted in increased soil moisture f of the surface soil layer. Increased regional relative humidity reduced the evaporation strength of soil moisture by inhibiting the diffusion movement of the evaporated soil moisture, while a greater wind speed accelerated the diffusion of evaporated water vapor in the soil, increasing soil evaporation intensity, which is consistent with previous studies (Lyu & Wang, 2021). In general, SWC variability directly determines the water yield from soil moisture evaporation that can be supplied. Regional ET₀, as an original driver of the regional soil moisture evaporation process, also significantly impacts the strength of the regional soil moisture evaporation loss.

To further reveal the internal relationship between the continuous evaporative loss of soil moisture and environmental elements, a multiple linear regression analysis model was used to construct a driving mechanism model for continuous soil moisture evaporation under different regional precipitation conditions. According to the results in the previous sections, the soil moisture evaporation loss fraction was defined as the independent variable (y) in this study, and five environmental elements were defined as dependent variables (x₁, x₂, x₃, x₄, and x₅), which were analyzed using MLR regression under different precipitation conditions (the 0.2–5, 20–30, and 30–40 mm, respectively). Table 2 shows the results of the evaporation loss driving model of soil moisture at different soil layer depths under different precipitation conditions, with relatively large R² fluctuations varying from 0.14 to 0.95.

At an accumulated precipitation amount of 0.2–5 mm, the R² of the fitted driving models of all soil depth layers did not exceed 0.4 and P > 0.05, which indicated that the fitting effect of the soil moisture evaporation driving model was not significant. A statistically significant environment-driven model for continuous evaporation of the 0–100 cm soil depth layer was fitted under an accumulated precipitation amount of 20–30 mm. Among them, a significant linear relationship between the continuous evaporation of soil moisture and the mainly environmental elements appeared in soil layers of 0–20 and 20–40 cm depths, respectively.

At an accumulated precipitation amount of 20–30 mm, the fitting equation of the soil moisture evaporation driving model for the 0–100 cm soil depth layer was significant (R² = 0.88, P < 0.05). The soil moisture evaporation driving model for the 20–40 cm soil depth layer was particularly significant (R² = 0.95, P < 0.001). With increasing soil depth, the linear relationship between the continuous evaporation of soil moisture and environmental elements gradually weakened. In this study, depth soil evaporation showed a relatively weak response to variation of environmental elements (Fig. 7 and Table 2), which may be related to the following reasons. First, depth layer soil moisture fluctuated less than other soil layers, and the exchange of matter and energy with the surface usually was relatively weak (Du et al., 2021). Second, the influence of air temperature has obvious hysteresis on the deep soil layer, and its influence to the depth layer is weak than that of other soil layer (Lyu & Wang, 2021). Third, precipitation affected the soil moisture f through infiltration, while the depth soil received less surface precipitation than surface soil layers (Piayda et al., 2017). In summary, the environmentally driven model of the continuous evaporation of soil moisture was not statistically significant for relatively small rainfall amounts but was suitable for larger amounts, especially for the surface soil layers. However, this study has some limitations due to the observed data of this study just only collected from a single point during 1 year.