Response of soil water movement to rainfall under different land uses in karst regions

Soil water is a critical factor closely related to hydrological and ecological processes. Owing to the complex surface conditions with heterogeneous soil thickness and abundant underlying fissures, soil water in the karst region has been a complicated issue. In this study, the dynamic changes of soil water in the vertical profile of selected grassland, farmland and bare land on a karst yellow soil hillslope in southwest China were monitored at five depths including 20, 40, 60, 80 and 100 cm with an interval of 15 min. Results showed that (1) there were spatial differences in the response to rainfall of the soil water content at different depths. When the rainfall amount was similar, the soil water replenishment amount and migration depth under the three land uses decreased with the rainfall intensity. In the case of light rainfall, the soil water content at 20 cm was the most sensitive to rainfall, and the response of soil water to rainfall mainly occurred at 0–40 cm. In the case of moderate and heavy rainfall, soil water could migrate down to 100 cm on grassland but less than 100 cm on farmland and bare land under heavy intensity rainfall. (2) The variation in the soil water content had interlayer differences over time. The response of soil water to rainfall in different soil layers showed multipeak fluctuations. In general, when the rainfall intensity was the same, the soil water fluctuation on grassland and farmland at the same depth was larger than that on bare land; however, the peak value of soil water decreased with soil depth. (3) Land use and the antecedent soil water content had important effects on soil water loss during the dry period. Soil water loss was faster at the beginning and before slowing down. The soil water loss rate on grassland and farmland increased with the length of the dry period, but decreased gradually on bare land. These results can support the utilization and protection of soil and water resources on karst yellow soil slopes and help to understand the temporal and spatial dynamics of soil water under natural rainfall.


Introduction
Soil water dynamics is a key factor in understanding hydrological and ecological processes, such as groundwater dynamics , vegetation growth (Bromley et al. 1997;Danelichen et al. 2016;Seghieri et al. 1997) and crop yield (Yang et al. 2016a, b). Soil water directly affects the redistribution of rainfall on the land surface by altering evaporation, plant transpiration (Vereecken et al. 2016), infiltration and runoff processes (Brocca et al. 2009;Koster et al. 2004). Soil water plays a critical role in the water circulation of the terrestrial surface system and has important influence on soil erosion and other earth surface processes (Moreno-de-las-Heras et al. 2020).
However, in karst regions, soil and water resources have formed a two-layer spatial structure due to the long-term carbonate dissolution processes Sohrt et al. 2014), which leads to the soil water movement have complexity, where soil water is a critical environmental factor determining agricultural production and ecological processes (Hartmann et al. 2014). Groundwater in the karst region has an obvious dissolution effect on the carbonate rocks and is more active in hot and rainy summer than in cold winter . Therefore, exploring the response of soil water movement to rainfall is vital for better understanding the hydrological processes in karst ecosystems, and accurate estimation of the movement and loss 50 Page 2 of 17 of soil water can provide helpful information for the water management in these areas. However, soil water has high temporal and spatial variability (Duan et al. 2017;Penna et al. 2013;Vanderlinden et al. 2012), particularly in karst slopes with abundant rock outcrops.
Soil water in karst regions exhibits even higher temporal and spatial variability due to the special heterogeneity in environmental factors (Yan et al. 2019). Special geologic, pedologic and geomorphic conditions result in complex hydrological processes in karst areas (Peng et al. 2020). Thin abundant exposed bedrock makes soil erosion in karst regions more serious and difficult to control (Dai et al. 2017). The soil water content is closely related to rock occurrence direction and characteristics, and the presence of bare rocks changes the surrounding soil water microenvironment. (Li et al. 2014). In addition, soil property, rock fragment content and soil texture are three main soil properties that affect the soil water content significantly . Because of abundant cracks, funnels and sinkholes in karst areas, rainfall is easily transported to groundwater , which reduced the surface runoff coefficient. The runoff coefficients of multiple land use to individual rainfall events are obviously smaller than other regions (Chen et al. 2010).
To understand and quantify the heterogeneity of soil water on a slope, it is necessary to investigate soil water movement and the main related factors. Studies have revealed that soil water is mainly determined by climate factors such as rainfall (Ma et al. 2020a, b) and freeze-thaw cycles (Gao et al. 2020;Niu et al. 2019), elevation (Ibrahim et al. 2021), slope gradient (Mei et al. 2018), vegetation (Fu et al. 2012), soil texture (Zhao et al. 2011) and land use (Bai et al. 2020). Land use, rainfall intensities and antecedent soil water content are the important driving force determining soil water content and evapotranspiration at the slope scale (Chen et al. 2010;Tang et al. 2019).
Although soil water in karst regions has different characteristics from other areas, land use still plays a decisive role within karst regions. Li et al. (2019) found that soil water temporal dynamics were influenced mainly by land use and precipitation by relating soil water to meteorological factors in a karst depression. Based on field observations, soil water had a moderate variation during an 8-month growing season on karst hillslopes, but showed a significant difference among different land uses including economic forestland, native scrubland, abandoned and sloping cropland (Chen et al. 2010), because land-use practices are likely to change the runoff characteristics by intercepting and redistributing rainfall (Bowden et al. 2001). Wang et al. (2016) showed that the soil water content of different vegetation conditions on hillslopes like shrub and mixed grass-shrub was influenced by land use. Soil water would be reduced by 13.1% and 32.1% once the forest was reclaimed as farmland and bare land, respectively, and 70% if the land developed into strong rock desertification (Chen et al. 2009). Rainfall characteristics such as different rainfall amounts and rainfall intensities affect the infiltration and storage of soil water between soil layers (Yl et al. 2018). Soil water movement is strongly dependent on patterns of precipitation. Jia et al. (2016) found that water infiltration into 20-160 cm soil profiles required 1-61 mm of precipitation, respectively. Soil water movement has variability under different antecedent soil water content. Low antecedent soil water content can increase soil water infiltration and reduce soil erosion (Ziadat and Taimeh 2013).
The studies mentioned above have promoted the understanding of soil water dynamics in both horizontal and vertical directions in karst regions, but it is still unclear how soil water movement and soil water content dynamics especially the response characteristics of soil water to individual rainfall events under different land uses. For the limited studies concerning response of soil water to rainfall (Yl et al. 2018;Jia et al. 2016;Sun et al. 2019), the data were all collected from artificial rainfall experiments. Currently, there is a lack of analysis on the response of soil water to natural rainfall under long-term monitoring on karst slopes, which restricts the understanding of soil water dynamics as response to natural rainfall. Thus, it is necessary to study the response of soil water movement and redistribution in a vertical profile to rainfall under different land uses in karst regions by in situ monitoring, which can better determine the soil water dynamics under natural conditions.
In this study, soil water dynamics and their responses to natural rainfall of different land uses were monitored in situ at multiple depths on a slope in the karst region of southwest China. The objectives of this study are to: (1) evaluate the spatial migration depth of soil water corresponding to natural rainfall under the three land uses; (2) quantify the variation in soil water content in profiles under different rainfall amounts and rainfall intensities; and (3) identify the characteristics of soil water loss for each layer of the profile under different antecedent soil water contents during the dry period.

Study site
The study was conducted at the Zunyi soil and water conservation monitoring station in the Huyangshui watershed of Guizhou Province in southwest China (Fig. 1), which has a typical subtropical humid monsoon climate with abundant and concentrated rainfall. The mean annual precipitation is 1024 mm, of which 70-80% is concentrated from May to August. The mean annual temperature is 14.6 ℃. The main vegetation type is evergreen trees and deciduous mixed coniferous broad-leaved forest. The lithology is mainly limestone and crystalline limestone which is exposed frequently on the ground surface owing to serious soil loss. The dominant soil type in the region is yellow soil, which is also the typical agricultural soil of Guizhou Province (Ma et al. 2020b). According to US soil taxonomy, the soil is classified as Udic Ultisols, which are known as Udic Ferrisol in the Chinese soil taxonomy. The silt and clay content of the soil is significantly higher than that of sand. The detailed soil physical properties of the three selected monitoring sites are shown in Table 1.

Field monitoring and instruments
Three sets of EM50 series data loggers were installed on grassland, farmland and bare land slopes. The distance between each land use is 20 m. Farmland and bare land are plowed once a year during the annual cultivation season. Grassland is natural grass grassland without plowing and the bare land is weeded every 10 days on average to keep the vegetation coverage of below 5%. The monitoring sample points are in standard runoff plots, with a length of 20 m and a width of 5 m. Each land-use plot has one monitoring site, and monitoring sensors are buried at the foot of runoff plot. At each monitoring point, five ECH2O sensors (Decagon Corporation, USA) were vertically installed in the profile to measure soil water at 20, 40, 60, 80 and 100 cm. Data were monitored and recorded at an interval of 15 min. In addition, a HOBO U30 high-precision rainfall sensor was installed (Decagon Corporation, USA) near the plots to record precipitation data. The soil water and rainfall data from June 2018 to September 2020 were analyzed for this paper. To relate soil water to a single rainfall event, rainfall data were divided into four grade categories according  to the standard set by the meteorological department (Zhu et al. 2021). The four rainfall amount grades are A, B, C and D. A 24-h rainfall amount of less than 10 mm is a light rainfall event; if 10-25 mm, a moderate rainfall event; if 25-50 mm, a heavy rainfall event; and when more than 50 mm, a storm event. Based on the classification standard above, there were almost no storms of more than 50 mm during the monitoring period, so the rainfall data were grouped into three groups for analysis. Data monitoring began on June 8, 2018, and was done mainly from the local farming period to the autumn harvest period each year. During the studied period, a total of 628 days of soil water and rainfall data were collected. We divided the rainfall events into different types according to the interval of more than 6 h between two rainfall events and then calculated the average rainfall intensity of each rainfall event. To reduce the influence of the antecedent soil water content on the experimental results and to explore the responses of soil water to rainfall under different land uses, 9 typical independent rainfall events from 2018 to 2020 were selected based on the principle of no rainfall events in the previous three days. Three light rainfall events had similar rainfall amounts of 7.8 mm, 9.4 mm and 8.6 mm but different rainfall intensities of 0.78 mm h −1 , 1.5 mm h −1 and 2.81 mm h −1 , respectively. Similarly, three moderate rainfall events had rainfall amounts of 23.8 mm, 22.2 mm and 19.8 mm and rainfall intensities of 0.85 mm h −1 , 1.29 mm h −1 and 3.27 mm h −1 , respectively. Finally, there were three heavy rainfall events with rainfall amounts of 40.8 mm, 33.6 mm and 35.6 mm and rainfall intensities of 1.11 mm h −1 , 3.77 mm h −1 and 7.57 mm h −1 , respectively.

Data analysis
To evaluate the response characteristics of soil water in different profiles to rainfall, we analyzed soil water replenishment and the variation in soil water per layer in each rainfall event. In addition, to identify the loss characteristics of soil water under different antecedent soil water contents, we calculated the soil water loss rate in the dry period. SPSS 26 was used to analyze the rainfall and soil water data. GIS 10.6 was used to map the study area. Origin 2021 was used to draw figures. The following equations were used to calculate the soil water content and related indices: Soil water replenishment in each rainfall event Variation in soil water per layer in each rainfall event Soil water loss rate in the dry period where SWR max is the highest value of soil water content during rainfall, SWR 0 is the antecedent soil water content before a rainfall event, i represents monitoring depth in three kinds of land uses, t indicates the monitoring time and t 0 is the beginning time of the rainfall. Soil water content (%) is expressed as : 1 means the soil water content (%) on the first day after the rainfall, and n indicates the soil water content (%) on day n after the rainfall.

Water movement in the soil profile during rainfall events
During the monitoring period from 2018 to 2020, light rainfall events accounted for 74%, moderate rainfall events accounted for approximately 17%, and heavy and storm rainfall events accounted for only 8% and 1%, respectively. Continuously monitored data of soil water content showed that the soil water under the three land uses was obviously affected by rainfall events from 2018 to 2020.

Soil water movement under light rainfall
Within 6 h after the end of rainfall, the soil water of grassland and farmland hardly changed, while it changed obviously on bare land. The soil water content at 20 cm was the most sensitive to rainfall; however, the response of soil water to rainfall was mainly concentrated at 0-40 cm. The spatial response of soil water to light rainfall is shown in Fig. 2. When the rainfall intensity was 0.78 mm h −1 , the soil water from rainfall replenishment amounts were as follows: grassland 6.4% > farmland 4% > bare land 1.6%. Soil water from rainfall could reach 80 cm in grassland, 60 cm in farmland and 20 cm in bare land. For the rainfall with an intensity of 1.50 mm h −1 , the rainfall replenishment amounts into the soil were as follows: bare land 3% > farmland 2.6% > grassland 2.2%. Soil water could migrate to 40 cm under the three land uses during rainfall. For the rainfall with an intensity of 2.81 mm h −1 , the soil water replenishment amounts from rainfall were as follows: bare land 13.5% > grassland 0% = farmland 0%. Soil water in bare land could migrate up to 40 cm but less than 20 cm in grassland and farmland.
For these three rainfall intensities, the response of the soil water content to rainfall at different depths had distinct interlayer variations among the different periods, as shown in Fig. 3. For light rainfall intensity, the variation in soil water content was grassland > farmland > bare land, and obvious hysteresis occurred at 40 cm. For moderate rainfall intensity, the variation in soil water content was as follows: bare land > farmland > grassland; the highest was at 20 cm for farmland and bare land, while the highest was at 40 cm for grassland. Under the above two kinds of rainfall, the response of soil water to rainfall was faster for grassland and farmland within the rainfall duration, while lagging for bare land, and the response occurred mainly in the late period of rainfall or after the rainfall ended. For heavy rainfall, the variation in soil water content was as follows: bare land > grassland > farmland, which was inconsistent with that under the two previous rainfall intensities due to the low water content in the early stage before rainfall.
In karst areas, the processes of rainfall infiltration on hillslopes are mainly affected by rainfall characteristics (Gregory et al. 2009), vegetation coverage and soil properties (Canton et al. 2016). Rainfall characteristics such as amount, intensity and frequency significantly affect the depth of soil water migration (He et al. 2012;Yaseef et al. 2009). For light rainfall, the depth of soil water movement in grassland and farmland decreased with rainfall intensity, but the opposite was true in bare land. In general, the lower the rainfall intensity is, the more soil water replenishment there is. For grassland and farmland, more rainfall can infiltrate into the soil, surface runoff is reduced, and the infiltration depth is deeper than that on bare land. As a soil water resource in karst regions, light rainfall and moderate rainfall events play an important role in replenishing soil water. He et al. (2012) concluded that rainfall events greater than 15 mm and 20 mm significantly increased the soil water content at depths of 20 cm and 40 cm, respectively, in their study of the response of soil water to rainfall events. However, our study showed that light rainfall events with rainfall intensities less than 1.5 mm h −1 have an important replenishment effect on the surface layer of 0-40 cm soil water.

Soil water movement under moderate rainfall
The spatial response of soil water to moderate rainfall is shown in Fig. 4. When the rainfall intensity was 0.85 mm h −1 , the soil water replenishment amounts from rainfall were as follows: grassland 26.2% > farmland 15.4% > bare land 7.8%. When the rainfall intensity increased to 1.29 mm h −1 , the replenishment amounts from rainfall were as follows: grassland 17.5% > bare land 9.9% > farmland 8.5%. When the rainfall intensity reached 3.27 mm h −1 , the replenishment amounts from rainfall were as follows: grassland 22.2% > farmland 5.3% > bare land 1%. The total soil water replenishment was the highest under light rainfall intensity for the three lands and the lowest on grassland under moderate rainfall intensity but the lowest on farmland and bare land under heavy rainfall intensity.
The response of soil water to rainfall in different soil layers showed multipeak fluctuations. When the rainfall intensity remained the same, the soil water fluctuation on grassland and farmland at the same depth was larger than that on bare land, and the peak value of soil water decreased with the soil depth, as shown in Fig. 5. Under light rainfall intensity, the variation in soil water content was grassland > farmland > bare land, and the response time of grassland was the shortest, while it was the longest for bare land. Under moderate rainfall intensity, the variation in soil water content was grassland > bare land > farmland. Under these two rainfall intensities, infiltration mainly occurred throughout the period of rainfall. Under heavy rainfall intensity, the variation in soil water content was grassland > farmland > bare land, and infiltration mainly occurred in the late period of rainfall. In general, the variation in the soil water content of grassland was higher in the bottom layer while on farmland and bare land it was highest in the surface layer. As a result, the smaller the rainfall intensity is, the higher the variation in the soil water content.
Under moderate rainfall, the soil water infiltration depth of the three land uses was almost the same except for the high intensity. The response depths of soil water to rainfall were 100 cm for grassland, 60 cm for farmland and 20 cm for bare land under heavy intensity rainfall. Under rainfall with light intensity and moderate intensity, soil water could migrate to 100 cm on all three land uses. However, when the rainfall intensity reached 3.27 mm h −1 , the soil water content was the largest on grassland, but only showed a 1% increase on bare land because of the high runoff rate. As the rainfall intensity changed from light to strong, the soil water replenishment on grassland decreased at the beginning before increasing, on farmland it decreased linearly, and on bare land, it was opposite to grassland. In the matter of total soil water replenishment, the smaller the rainfall intensity is, the more soil water is replenished.
A karst slope has poor water holding capacity, and the evaporation of bare land is faster, while vegetation cover can prevent soil water from evaporating on grassland and farmland. Zhou et al. (2019) found that when the soil water content was higher than 39% prior to a rainfall event, the difference in infiltration capacity had less influence on the response time, which was mainly affected by vegetation type. In our study, the average soil water content of bare land prior to a rainfall event was more than 39%, and the infiltration response of farmland and bare land to soil water had an obvious lag effect under heavy rainfall intensity. Owing to the absence of plant roots on bare land, it is difficult to infiltrate and easy to generate runoff. Soil water replenishment plays an important role in surface vegetation patterns (Jian et al. 2015) and the water cycle (Kong et al. 2020); moreover, it can be used to evaluate and predict the risk of drought and soil water dynamics. Our results showed that the soil water replenishment depth increased with rainfall intensity from light to moderate rainfall. When the rainfall is greater than 19.8 mm, the soil water replenishment depth could reach 100 cm. A similar positive correlation between replenishment depth and rainfall was also found by Hao et al. (2008).

Soil water movement under heavy rainfall
As shown in Fig. 6, the response of soil water to heavy rainfall exhibited spatial differences. When the rainfall intensity was 1.11 mm h −1 , the rainfall replenishment amounts were as follows: grassland 79.3% > bare land 39.7% > farmland 38.5%. In the case of rainfall with an intensity of 3.77 mm h −1 , the soil water replenishment amounts were as follows: bare land 26.6% > grassland 18.9% > farmland 11.7%. When the rainfall intensity increased to 7.57 mm h −1 , the soil water replenishment amounts were as follows: grassland 57.5% > farmland 27.3% > bare land 7.6%.
As shown in Fig. 7, the lower the rainfall intensity was, the more obviously the soil water between the soil layers changed. In addition, the greater the rainfall intensity is, the less variable the soil water at the bottom layer. Under light rainfall intensity, the variation in soil water content was grassland > farmland > bare land. The response time was shorter for grassland and farmland and longer for bare land. Under moderate rainfall intensity, the variation in soil water content was in the order of bare land > grassland > farmland, and the response time was the longest for farmland. Under these two rainfall intensities, infiltration mainly occurred throughout the duration of rainfall, and the variation in soil water content showed a decreasing trend from the end of rainfall. In contrast, under heavy rainfall intensity, the variation in soil water content was grassland > farmland > bare land, and water reached to 80 cm on grassland and 40 cm on farmland and bare land.
Compared with moderate rainfall events, the depth of soil water replenishment under heavy rainfall varied obviously with rainfall intensity, although the rainfall amount increased markedly. Rainfall easily infiltrated with light and moderate intensity rainfall and could reach to 100 cm to add soil water. However, the vertical replenishment of soil water decreased under heavy rainfall because of the blocking of soil pores by stronger rain drop hammering. In this case, the migration depth of soil water was 100 cm on grassland, 60 cm on farmland and 40 cm on bare land. The total soil water replenishment was the highest in the case of light rainfall intensity on the three kinds of land but the lowest under moderate rainfall intensity on grassland and farmland and the lowest under heavy rainfall intensity on bare land. The response of soil water to rainfall intensity remained consistent on grassland in contrast to moderate rainfall events, but it fluctuated on farmland and bare land. In addition, soil water did not infiltrate rapidly in a short time under heavy rainfall intensity similar to the situation under moderate rainfall events. The replenishment amount was the lowest on bare land because of more surface runoff formation.

Response of soil water content to rainfall
For a given land use, the response of soil water to rainfall is an important part of the terrestrial surface hydrological cycle (Yin et al. 2020). Precipitation is rich in the karst area of South China, and its characteristics are key factors affecting soil water. Controlled by multiple factors, the soil water content on the slopes of the three kinds of land use has obvious spatiotemporal variability; furthermore, the characteristics of soil water movement are significantly different in different land uses under different rainfall classes. The soil water movement on grassland, farmland and bare land was significantly influenced by rainfall characteristics, such as rainfall frequency, amount and intensity (Figs. 2, 4 and 6). If the rainfall duration was short while the intensity was strong, limited water infiltrated into the soil, and the replenishment amount on bare land was the lowest, probably because of the rapid production of overland flow (Qin et al. 2015). Overall, the response of soil water increased with rainfall amount and intensity. This result was consistent with the results of some previous studies (Albertson and Kiely. 2001;He et al. 2012;Heisler-White et al. 2008), which reported that rainfall with a small amount and low intensity was easily lost via litter interception and canopy. Based on the data in our work, rainfall amount and intensity were major factors affecting the dynamics in soil water content at different soil layers on the plot scale. If the rainfall intensity is smaller, the replenishment amount is higher. Bare land has a high replenishment amount under moderate rainfall intensity, while grassland and farmland have a high replenishment amount under light and heavy rainfall intensity.

Water loss in the soil profile under different water contents
In order to identify the variation characteristics of water loss in different soil layers, according to the average soil water content in the previous 30 days, three dry periods in the growing season were selected. Soil water loss is obvious during dry periods. To maintain similar conditions, the durations of the three dry periods were similar, and the rainfall during the period was in the range of 40-50 mm.

Soil water loss under low water content
The soil water loss process under low water content is shown in Fig. 8. A dry period with a low water content from July 27 to September 3 in 2020 was selected, during which only 3 light and 1 heavy rainfall events with a rainfall of 49.8 mm occurred. The average soil water content on grassland, farmland and bare land in the previous 30 days was 35.1%, 26.7% and 27.2%, respectively. Soil water decreased from the first day of the dry period, but the loss amounts of soil water decreased with increasing soil depth. Due to the interruption of heavy rainfall events, we divided the dry duration into two periods to analyze the soil water decay characteristics.
There were significant differences in soil water loss under the three land uses under low water content conditions (Table 2). At the beginning of the dry period, the soil water loss on grassland was the greatest from the surface layer at 0-60 cm because of the high surface coverage and strong transpiration by plants. As the dry duration continued, the soil water at the bottom 80-100 cm migrated upward to supplement the surface soil water, leading to the most serious loss of soil water at the bottom layer. In the late period of dry duration, evaporation was intense on bare land because of little vegetation cover. As a result, soil water loss was serious in the profile from 0 to 60 cm. However, the soil water loss from the 80 to 100 cm layer was strongest for grassland, a result of plant transpiration. Once heavy rainfall occurred, the rainfall replenishment effect was obvious, but the water infiltration and subsequent soil water change depended on land use. During the whole dry period, the soil water loss rate was as follows: grassland > farmland > bare land, while bare land > grassland > farmland in the late period. Therefore, soil water on bare land can be effectively replenished after rainfall, while the soil water holding capability was the worst during the dry period. Land use was the main factor controlling soil water loss during the dry period.

Soil water loss under moderate water content
The soil water loss process under moderate water content is shown in Fig. 9. The moderate water content during the dry period was selected from July 1, 2018, to August 4, 2018, in which only 1 heavy rainfall event with a rainfall of 43 mm occurred. The average soil water content on grassland, farmland and bare land in the earlier 23 days was 39.1%, 29.3% and 24.3%, respectively. As the rainfall event had a significant impact on the soil water loss rate, we divided the dry duration into two periods to analyze the soil water decay characteristics. The heavy rainfall event within 15 days was the early period, and that after 15 days was the late period. There were significant differences in soil water loss under the three land uses under moderate water content conditions (Table 3). In the early period of dry duration, the soil water  loss rate was the highest on bare land because surface evaporation was strong after rainfall but the lowest on grassland affected by vegetation cover. In the late period of dry duration, the soil water loss rate was the highest on grassland due to plant transpiration and root water absorption, while it was the lowest on bare land. The shorter the dry period was, the more soil water was lost on bare land; in contrast, the more soil water was lost on grassland. The variation pattern of soil water loss in the early period was consistent with that in the whole dry period. The highest soil water loss rate appeared on farmland at 40 cm, but all appeared on bare land at other soil depths. The soil water loss rate was similar to the dry period under low water content, which appeared as grassland > farmland > bare land. In contrast to the whole dry period under low water content, the soil water loss rate on grassland and farmland both increased by 16%, but that on bare land decreased by 15% because the antecedent soil water content in 2018 was lower than that in 2020. The soil water loss rate on grassland and farmland increased with the length of the period, while that on bare land decreased gradually.

Soil water loss under high water content
The soil water loss process under high water content is shown in Fig. 10. A dry period with high water content from July 24, 2019, to August 27, 2019, was selected, in which 11 light and 1 moderate rainfall events with a rainfall of 40.2 mm occurred. The average soil water content of grassland, farmland and bare land in the previous 30 days was 45.0%, 36.3% and 41.1%, respectively. Light rainfall events had no effect on the soil water content while moderate rainfall events had only a weak effect that could be ignored. Due to the random distribution of natural rainfall in dry duration, we defined the first 17 days as the early period and the last 18 days as the late period to analyze the soil water loss characteristics. There were significant differences in soil water loss under the three land uses at high water contents (Table 4). At the beginning of the dry period, soil water loss was serious on farmland, but the opposite was true on bare land. In the late dry period, soil water loss was the greatest on grassland and the lowest on bare land. The variation pattern of the soil water loss rate in the early and wholly dry periods was consistent as farmland > grassland > bare land. In summary, grassland has the highest and lowest soil water loss rates for low and moderate water contents, respectively. In the dry period, when the antecedent soil water content was higher, the soil water loss rate on farmland was greater. The soil water loss rate on bare land was the highest under moderate water content but the lowest under high water content. Although the rainfall amount was approximately equal during the three dry periods, the variation patterns of the soil water loss rate on the three land uses were different and were mainly affected by rainfall characteristics and other combined factors.

Soil water loss characteristics and influencing factors
Karst slopes have shallow soil layers and poorer water holding capacities than non-karst areas. Meteorological factors, vegetation cover types and land use promoting surface evaporation and leading to soil water loss were more obvious. The result of surface soil water loss was more severe than that of the deep layer in the dry period, mainly caused by land use (e.g., Fig. 10), which was consistent with the findings of Zhu and Lin (2011) and Takagi and Lin (2012), who reported that during the growing season, crop and vegetation evapotranspiration exerted obvious effects on soil water loss, especially from 10 to 40 cm. Soil water loss was less than precipitation on three land types during the dry period in karst slopes, while the opposite was true in the semiarid loess hilly area (Jian et al. 2015). Soil water is accompanied by an obvious dry period during the growing season. In the dry period, soil water was low, but its variability was relatively higher, and land use and precipitation were the main factors affecting the variability of surface soil water in karst regions ). For instance, Ge et al. (2020) found that the soil water storage increased and the average evapotranspiration decreased when the land use changed from alfalfa to soybean or other food crops. The soil water content could be rapidly depleted and replenished, and the soil water loss was mainly dominated by rainfall. The influence of rainfall on soil water loss was also identified by Zhou et al. (2019) and Li et al. (2017).
Karst soil has shallow characteristics, with low water holding capacity and high permeability, which promotes rapid movement of soil water to the deep layers (Canton et al. 2016, Yang et al. 2019. Soil water replenishment amount and movement speed decreases with the increasing depth of soil profile (Chen et al. 2017). Thus, the response of soil water movement to rainfall in shallow soil is fast, which accompanied by rapid loss. Karst soil has high content of crushed stone that is different from non-karst regions (Fu et al. 2015a, b), which increases the temperature of surface soil and then accelerates evaporation (Cousin et al. 2003). Meantime, higher gravel content is accompanied by higher infiltration capacity in the profile, which leads to soil water movement is easier to the deep layer. Yang et al. (2016a, b) proposed that hydrological function of soil water movement is affected by the soil properties and its distributions. Interflow is the main type of hydrology due to the blocking of rainfall by outcropping rock and gravel in karst areas, and it is the main form of retaining water (McDaniel et al. 2008). In general, soil water movement are influenced by climate and nearby land uses and its soil erosion condition over a long process.
In summary, our work showed that the soil water loss rate on grassland and farmland increased with the length of the dry period, while that on bare land decreased. Light rainfall events had no effect on soil water during the dry period, while heavy rainfall events had an obvious replenishment effect on soil water. The three dry periods were mainly concentrated in July and August, but there were some differences in meteorological factors between each month of the three years. Furthermore, there were different rainfall events in the dry period, which might affect the accuracy of our experimental analysis. Longer soil water content data series in the future would be helpful to accurately investigate the characteristics of soil water loss in the dry period.

Conclusions
There were spatial differences in the response of the soil water content to rainfall at different depths under different land uses in karst regions. For a similar rainfall amount, the soil water replenishment amount and migration depth decreased with the rainfall intensity. In the case of light rainfall, the soil water content at 20 cm was the most sensitive to rainfall, but the response was limited within 40 cm. Under moderate rainfall and heavy rainfall, soil water could migrate to 100 cm in grassland but less than 100 cm in farmland and bare land under heavy rainfall intensity. The variation in soil water content had interlayer differences over time. When the rainfall intensity remained the same, the soil water fluctuation on grassland and farmland at the same depth was larger than that on bare land, and the peak value of soil water decreased with the soil depth. The response of soil water to rainfall exhibited hysteresis with increasing soil depth. If the rainfall intensity remained the same, soil water on grassland and farmland responded faster to rainfall than that on bare land at the same depth. Land use and the antecedent soil water content had important effects on soil water loss during the dry period. The soil water loss rate was higher at the beginning of the dry period, but slowed gradually. The soil water loss rate on grassland and farmland increased with the length of the dry period, but decreased gradually on bare land.
Grassland has ideal soil water holding capacity and water retention, which is a beneficial land-use types on the karst yellow soil slope. The effects of the three land-use types on soil water provide valuable information for the utilization and protection of soil and water resources. This work improves the understanding of the temporal and spatial dynamics of soil water on karst yellow soil slopes under the influence of natural rainfall and provides a scientific basis for the study of soil water movement and loss in karst regions.