Spatial and Temporal Changes in Infiltration and Aggregate Stability: A Case Study of a Subhumid Irrigated Cropland

Climate change is increasing the occurrence of extreme precipitation events and causing irregular precipitation patterns. This occurs in parallel with the degradation of crop fields, and triggers the occurrence of pluvial floods and droughts on the same field. Consequently, irrigation must be adapted to the changing soil properties. Detailed spatial and temporal measurements of changes in infiltration are required. This study aimed to quantify changes in infiltration for a subhumid irrigated cropland with various soil types (Phaeozem, Solonetz, Chernozem) and field conditions (seedbed and stubble) by simulating rainfall. As the soil structure determines hydrology, the aggregate stability/surface roughness was tested as a proxy of infiltration through photogrammetry. The soil losses caused by precipitation did not exhibit connections to changes in the surface roughness, and lower aggregate stability did not necessarily cause lower infiltration intensities, suggesting that sedimentation could only partly seal drainage pores. The final infiltration intensities varied within a wide range (0.2–28.4 mm h−1). Seedbed preparation did not increase the volume of micropores (<10 μm), which resulted in higher infiltration under stubble. Photogrammetry was found to be a potentially useful tool for measuring aggregate stability, however, further investigations on in situ soil surfaces are required for technical improvement.


Introduction
Intensive horticulture and agriculture that are traditionally based on cultivation can considerably degrade soil structure [1,2]. Poor soil structure results in the rapid breakdown of aggregates, which mitigates the porosity of the uppermost soil layer [3]. In addition, canopy cover is often rare or even missing on crop fields depending on various agrotechniques used [4]. Aggregate breakdown accelerates due to splash erosion and inhibits water infiltration [5]. Moreover, the temporal pattern of rainfall has changed, and the occurrence of extreme precipitation events has increased in most parts of the world [6]. On hummocky landscapes, surface runoff erodes the most fertile part of the soil, forms gullies, fills reservoirs, and transports nutrients and pollutants to surface and subsurface water bodies [7]. Furthermore, runoff from crop fields is responsible for flash floods that often directly damage infrastructure [8]. On non-sloping landscapes, excess water can trigger pluvial flooding and

The Study Site
The studied arable field is located in the middle of the Great Hungarian Plain, where micro-topography is an important soil formation driver. The field is 105 ha at an elevation of 88-92 m above sea level. In the higher parts of the field, loess-derived Chernozems are predominant, while the soils of the lower surfaces mainly contain hydromorphic soils, such as Phaeozems and Solonetz according to the Word Reference Base. The mean annual precipitation and temperature are 540 mm and 9.8 • C, respectively [35]. Intensive cultivation with continuous plowing occurs in the field; hence, the soil structure is highly degraded and the field contains a compact plow pan [36]. As the investigation focuses on irrigation and its result on soil structure, only the uppermost cultivated layer was measured during sampling. Based on the spatial soil heterogeneity of the field [36], three different topsoils were select to measure the in situ water infiltration and estimate the hydrological properties of the field (Figure 1). The measured soil types differ in their texture and chemical properties, hence, their soil type analogies were identified according to the World Reference Base [37] (Table 1). Therefore, the three replicates are (Phaeozem (47.509693 N; 21.271736 E); Solonetz (47.508995 N; 21.273058 E); Chernozem (47.504360 N; 21.277486 E). The first measurement was collected on 8 May 2018 under recently sown sweet corn (Syngenta GSS3071), while soil surface was recently tilled (shallow disking) and bare without any canopy cover. The second measurement was collected on 10 July 2018 after harvest, under stubble conditions, again without any canopy cover. During the growing period, two mechanical weed controls were done (at four and at eight leaves phenophase) with shallow cultivator. As natural precipitation amount exceeded 310 mm in total, no irrigation was used. The studied arable field is located in the middle of the Great Hungarian Plain, where microtopography is an important soil formation driver. The field is 105 ha at an elevation of 88-92 m above sea level. In the higher parts of the field, loess-derived Chernozems are predominant, while the soils of the lower surfaces mainly contain hydromorphic soils, such as Phaeozems and Solonetz according to the Word Reference Base. The mean annual precipitation and temperature are 540 mm and 9.8 °C, respectively [35]. Intensive cultivation with continuous plowing occurs in the field; hence, the soil structure is highly degraded and the field contains a compact plow pan [36]. As the investigation focuses on irrigation and its result on soil structure, only the uppermost cultivated layer was measured during sampling. Based on the spatial soil heterogeneity of the field [36], three different topsoils were select to measure the in situ water infiltration and estimate the hydrological properties of the field (Figure 1). The measured soil types differ in their texture and chemical properties, hence, their soil type analogies were identified according to the World Reference Base [37] (Table 1). Therefore, the three replicates are (Phaeozem (47.509693 N; 21.271736 E); Solonetz (47.508995 N; 21.273058 E); Chernozem (47.504360 N; 21.277486 E). The first measurement was collected on 8 May 2018 under recently sown sweet corn (Syngenta GSS3071), while soil surface was recently tilled (shallow disking) and bare without any canopy cover. The second measurement was collected on 10 July 2018 after harvest, under stubble conditions, again without any canopy cover. During the growing period, two mechanical weed controls were done (at four and at eight leaves phenophase) with shallow cultivator. As natural precipitation amount exceeded 310 mm in total, no irrigation was used.

Measurements on Soil Porosity
The porosity of the uppermost cultivated soil layer was measured by collecting 100-cm 3 undisturbed samples, with three replicates per sample. The initial moisture content, bulk density, and differentiated porosity values were measured following the gravimetric method [38]. The results were tested by conducting a one-way ANOVA with a post-hoc Tukey test and presented as boxplots using IBM SPSS Statistics for Windows, (version 22, IBM Corp., Armonk, NY, USA).

Rainfall Simulation
A Meersmans's rainfall simulator was used to collect the in situ infiltration measurements. The device was developed and constructed at Catholic University of Leuven [39]. We used a Lechler 460.788 full cone nozzle, which should provide a continuous rainfall intensity of 45 mm h −1 under a pressure of 0.2 bar. The formed drop spectrum and kinetic energy agree with those of natural rainfall with a similar intensity [40]. The investigated soil surface is a rectangle of approximately 0.5 m 2 , fenced by aluminum sheets. To avoid border effects, the irrigated area was approximately 2.5 m 2 [19]. The plot locations were randomly selected. The real precipitation amount and intensity were measured using plastic vessels placed next to the plot and analyzed against the theoretical values [19]. The total amount of runoff was continuously measured and collected parallel with time. The sediment yield was determined in the laboratory by evaporating the whole amount of runoff divided to 6-10 intervals per precipitation.

Calculation of Hydrological Properties
The separate runoff volumes were plotted as a function of time. As the precipitation intensity was constant, in the final stage of the measurement, the runoff intensity became constant, which was determined by fitting a linear function to the final 6-10 measured data points. The steepness of the fitted function was equal to the final runoff intensity [5]. As evaporation was set to negligible under the applied circumstances and the surface storage capacity was constant, the final infiltration intensity was calculated as the difference between precipitation intensity and apparent runoff intensity.

Estimating Surface Roughness Changes
As the investigated plots were flat, the surface water storage capacity was taken as a proxy of surface roughness, and was calculated as the difference between the beginning of runoff and appearance of excess water in seconds. A longer period indicates a higher surface roughness. The soil surface roughness was also estimated by photogrammetry. High-resolution (10 MP) photos (25)(26)(27)(28)(29)(30) of the plot surface were captured from 360 • aspects just before and after the precipitation. The images were compiled into a digital elevation model (DEM) using the PhotoScan image processing software (Agisoft, Russia). The horizontal and vertical resolutions of the created DEMs were 3 mm. The elevation histogram of each DEM was analyzed using basic descriptive statistics. The DEMs before and after rainfall were compared and analyzed by applying the topographic range (TR) index, which is the difference in height (mm) between the highest and lowest point of the surface [41]. DEM construction and analysis were conducted using ArcMap 10.3. by Environmental Systems Research Institute, USA.

Results of Porosity Changes
Seedbed preparation created homogeneous and high porosity values (55%-60% v/v) throughout the investigated field, which significantly decreased during the growing season. The volume of the decrease was affected by the soil type. The decrease in the volume of Phaeozem was 10%, while that of Chernozem was only 2%-3%, and significant differences (p < 0.01) were observed between soil types (Figure 2a). The same is true for macropores (>50 µm), although micropores exhibited an opposite trend. The volume of micropores increased during the growing season and was not affected by the soil type. Moreover, the soil type determined the microporosity under both tilled and stubble conditions; Solonetz exhibited the highest amount of micropores, while Chernozem exhibited the lowest ( Figure 2c). Consequently, tillage cannot affect the volume of microporosity, which is consistent with the findings of Wairiu and Lal [42]. Both the Solonetz and Chernozem soils lost a significant part of their moisture content during the growing season (p < 0.01), while the moisture content of Phaeozem did not change (Figure 2e). The volume of micropores was in accordance with the moisture content under seedbed conditions, while it became independent under stubble, indicating that a drought occurred under the latter conditions. by the soil type. Moreover, the soil type determined the microporosity under both tilled and stubble conditions; Solonetz exhibited the highest amount of micropores, while Chernozem exhibited the lowest (Figure 2c). Consequently, tillage cannot affect the volume of microporosity, which is consistent with the findings of Wairiu and Lal [42]. Both the Solonetz and Chernozem soils lost a significant part of their moisture content during the growing season (p < 0.01), while the moisture content of Phaeozem did not change ( Figure 2e). The volume of micropores was in accordance with the moisture content under seedbed conditions, while it became independent under stubble, indicating that a drought occurred under the latter conditions.

Results of Infiltration Measurements
The measured mean rainfall intensities were not uniform due to the occasional winds and pressure fluctuations, but were scattered around 47 mm h −1 ( Table 2), which is slightly higher than the intensity generally applied during sprinkler irrigation using the Linear device on the studied field (42 mm h −1 ). The time taken for excess water to appear on the surface reflects the initial water infiltration capacity of the soil. These values were within one-minute intervals (3-4 min) for recently tilled soil and independent of soil type. In contrast, excess water appearance periods were longer and more diverse under stubble conditions, suggesting that the initial moisture content has a significant role in water infiltration against macroporosity, which was also reported by Ruggenthaler et al. [43]. Macropores created by seedbed preparation can drain and store excess water, which appears on the surface of macro-aggregates due to the higher moisture content in the capillary pores. Moreover, the order of excess water appearance over time is consistent with the micro-porosity values under stubble conditions (compared to Figure 2c).

Results of Infiltration Measurements
The measured mean rainfall intensities were not uniform due to the occasional winds and pressure fluctuations, but were scattered around 47 mm h −1 (Table 2), which is slightly higher than the intensity generally applied during sprinkler irrigation using the Linear device on the studied field (42 mm h −1 ). The time taken for excess water to appear on the surface reflects the initial water infiltration capacity of the soil. These values were within one-minute intervals (3-4 min) for recently tilled soil and independent of soil type. In contrast, excess water appearance periods were longer and more diverse under stubble conditions, suggesting that the initial moisture content has a significant role in water infiltration against macroporosity, which was also reported by Ruggenthaler et al. [43]. Macropores created by seedbed preparation can drain and store excess water, which appears on the surface of macro-aggregates due to the higher moisture content in the capillary pores. Moreover, the order of excess water appearance over time is consistent with the micro-porosity values under stubble conditions (compared to Figure 2c). As a direct indicator of surface runoff, the runoff initiation time revealed that the surface only smoothed over the growing season for Phaeozem ( Figure 3). Vermang et al. [44] observed explicit smoothening due to rainfall simulation, however, they measured runoff from sieved soils under laboratory conditions. Our results suggest that in situ runoff initiation is determined by the interactions of several circumstances beyond surface roughness.  As a direct indicator of surface runoff, the runoff initiation time revealed that the surface only smoothed over the growing season for Phaeozem ( Figure 3). Vermang et al. [44] observed explicit smoothening due to rainfall simulation, however, they measured runoff from sieved soils under laboratory conditions. Our results suggest that in situ runoff initiation is determined by the interactions of several circumstances beyond surface roughness. The final infiltration intensities were lower under seedbed conditions and increased by one order of magnitude under stubble conditions, or almost decreased to zero for Chernozem. Bombino et al. [16] also observed a considerable decrease in the final infiltration due to cultivation on clay soils. Effective infiltration is highly affected by the formation of a surface crust and sealing [45], as the surface is flat and the sediment deposits within a few decimeters. Therefore, the total sediment yield triggers unconsolidated crust generation with no porosity on surface depressions, which mitigates infiltration. This was also speculated by Zhao et al. [46]. Hence, the sediment yield and runoff concentration exhibit an inverse relationship with infiltration on crop fields with no slope. Stubble conditions (surface smoothing) considerably decreases the sediment yield under each soil type (Table  2), emphasizing the role of surface conditions and soil properties on aggregate stability [44]. The final infiltration intensities were lower under seedbed conditions and increased by one order of magnitude under stubble conditions, or almost decreased to zero for Chernozem. Bombino et al. [16] also observed a considerable decrease in the final infiltration due to cultivation on clay soils. Effective infiltration is highly affected by the formation of a surface crust and sealing [45], as the surface is flat and the sediment deposits within a few decimeters. Therefore, the total sediment yield triggers unconsolidated crust generation with no porosity on surface depressions, which mitigates infiltration. This was also speculated by Zhao et al. [46]. Hence, the sediment yield and runoff concentration exhibit an inverse relationship with infiltration on crop fields with no slope. Stubble conditions (surface smoothing) considerably decreases the sediment yield under each soil type (Table 2), emphasizing the role of surface conditions and soil properties on aggregate stability [44].
Chernozem soil has lower Na + and higher organic carbon contents than Solonetz (Table 1). In theory, these circumstances result in Na humate-dominated, less developed, and more vulnerable aggregates for the latter soil. However, the lowest infiltration intensities were detected for Chernozem soil at both times of the year. This suggests that aggregate distribution and stability in intensively cultivated soils is a function of individual tillage operations, rather than physicochemical properties. However, Bottinelli et al. [47] emphasized the role of moisture content and seasonal dynamics. Moreover, these results also indicate the important role of particle size distribution in aggregate stability [48], as the clay content of Chernozem was highest (Table 1).

Changes in Surface Roughness
Altogether, 12 DEMs were built. The before-and after-rainfall DEMs of the Phaeozem soil under seedbed conditions are presented in Figure 4 as an example. Generally, the higher parts of the plots were eroded due to the precipitation that triggered the decrease in the elevation of the concerning clogs. Chernozem soil has lower Na + and higher organic carbon contents than Solonetz (Table 1). In theory, these circumstances result in Na humate-dominated, less developed, and more vulnerable aggregates for the latter soil. However, the lowest infiltration intensities were detected for Chernozem soil at both times of the year. This suggests that aggregate distribution and stability in intensively cultivated soils is a function of individual tillage operations, rather than physicochemical properties. However, Bottinelli et al. [47] emphasized the role of moisture content and seasonal dynamics. Moreover, these results also indicate the important role of particle size distribution in aggregate stability [48], as the clay content of Chernozem was highest (Table 1).

Changes in Surface Roughness
Altogether, 12 DEMs were built. The before-and after-rainfall DEMs of the Phaeozem soil under seedbed conditions are presented in Figure 4 as an example. Generally, the higher parts of the plots were eroded due to the precipitation that triggered the decrease in the elevation of the concerning clogs. The relevant soil losses of the plots were measured, and considerable sedimentation was also recorded at the deepest parts of some plots (Figure 4d). Highly fragmented elevations became uniform and the contours were smoothed, indicating a general leveling of the surface. Jester and Klik [49] presented data on the higher inaccuracy created by using photogrammetry-based DEM to interpolate rough surfaces among macro-aggregates, which were not detected in this study. However, highly contrasting shades and plant residuals triggered local failures (<2% of the area; also lamented by Hwang et al. [50]) that were excluded from further investigation.
The elevation values of each DEM were analyzed to quantify the changes in topography, and the descriptive statistical parameters of the distributions are presented in Table 3. As minimum (0) values were set at the deepest (runoff outlet) points of each plot, no changes were detected, despite The relevant soil losses of the plots were measured, and considerable sedimentation was also recorded at the deepest parts of some plots (Figure 4d). Highly fragmented elevations became uniform and the contours were smoothed, indicating a general leveling of the surface. Jester and Klik [49] presented data on the higher inaccuracy created by using photogrammetry-based DEM to interpolate rough surfaces among macro-aggregates, which were not detected in this study. However, highly contrasting shades and plant residuals triggered local failures (<2% of the area; also lamented by Hwang et al. [50]) that were excluded from further investigation.
The elevation values of each DEM were analyzed to quantify the changes in topography, and the descriptive statistical parameters of the distributions are presented in Table 3. As minimum (0) values were set at the deepest (runoff outlet) points of each plot, no changes were detected, despite the considerable sedimentation that occurred in some sub-basins. Hence, the sedimentation process was only quantified from the mean elevation values calculated following comparative statistical methods. The mean elevation values also decreased due to precipitation under seedbed conditions. In contrast, no general trend was observed under stubble conditions. Phaeozem exhibited the highest reduction I elevation to 55%, while Chernozem did not change and the elevation of Solonetz increased. In theory, the increase in the mean elevation value and sedimentation within the plot should be inversely proportional to the runoff concentration (sediment leaving the plot). However, this is not fully supported by the runoff results (Table 2). Changes in the median values are similar to those reported concerning the means, which may refer to quasi-normal distributions, in line with the skewness and kurtosis values.
The maximum values were equal to the TR and scattered within a wide range before precipitation. The values decreased in all cases due to precipitation, however, the ratio of change differed. The greatest smoothing was recorded for the Phaeozem soil under stubble conditions, where the topographic range decreased to 48%. Similarly, the Phaeozem soil was also the most vulnerable under seedbed conditions, where the topographic range decreased to 71%. The role of Na + in soil structure destabilization has already been reported by several authors [51,52]. In this study, both Solonetz and Phaeozem soils were rich in Na + (Table 1), even though the topographic range decreases of Solonetz and Chernozem were lower and exhibited opposite trends under seedbed and stubble conditions (Table 2). Therefore, the role of Na + in aggregate stability is likely affected by additional drivers, such as the organic matter content and compound, which was also suspected by Rengasamy and Olsson [53]. The lowest change was observed for Chernozem under seedbed conditions (93%). However, the temporal fluctuations during the growing season due to tillage and sealing are not yet fully understood [47]. The changes in TR due to precipitation were analyzed with the final infiltration intensities (Table 4), and they were found to be completely independent.
Runoff initiation (s) (an indicator of surface storage capacity) for rainfall was found to be independent from the surface roughness estimated using TR (Table 2). This supports the results obtained by Appels et al. [54], who reported that infiltration has a greater effect on runoff generation than the surface storage capacity. Our results also demonstrate the importance of the known and still unsolved question of characterizing surface roughness using a single number [49]. Swath analysis [55] provides a generalized cross-section of an area at the landscape scale. Therefore, it could be suitable for measuring surface roughness in the present scale and if the spatial resolution is sufficiently high. Spatial variations in hydrology, erosion, and sedimentation within the plot were only investigated by visual observations. For example, stream capture was observed within a plot (Figure 4b,d) during rainfall, which resulted in hydrological changes. To adequately describe such processes, further GIS applications are available. Theoretically, these databases are suitable for quantifying soil redistribution volumes and comparing them to the measured soil losses due to runoff. For instance, creating elevation difference maps, comparing the DEMs for before and after precipitation, or calculating the water storage capacity and flow direction database of the surface provides such additional information that would be very useful for spatially detailed smart irrigation planning. In this study, the spatial resolution of the DEMs was not sufficiently high to reliably quantify soil redistribution. Moreover, in the absence of measured bulk density values of the original surface (including wheel tracks, and micro and macro aggregates) and the sealed sediments triggered by rainfall, the volumetric prediction would not be sufficiently precise [56]. Nevertheless, our method shows promise, and further tests and calibrations are required for proper application.

Conclusions
The tillage-induced initial topsoil porosity and moisture content differed between the studied conditions. There was no significant difference in the volume of micropores (<10 µm) under seedbed and stubble soil conditions. As a result, the final infiltration intensity was higher under stubble conditions, which occurred alongside the lower sediment concentration in runoff caused by the crusted surface. The final infiltration intensities varied widely (0.2-28.4 mm h −1 ) in both space and time, indicating the importance of precise irrigation. For better water efficiency and reducing soil structure degradation, the irrigation volume must be adapted to the actual infiltration value of the soil in both time and space. Nevertheless, the precipitation-generated sediment losses were not connected to the surface roughness changes, and lower aggregate stability did not necessarily trigger lower infiltration intensities, suggesting that sedimentation could only partly seal the drainage pores. Photogrammetry was found to be a potentially useful tool for measuring aggregate stability, however, further investigations on in situ surfaces are required for technical improvement. Acknowledgments: Gergely Jakab and Balázs Madarász was supported by the Bolyai fellowship of the Hungarian Academy of Sciences.

Conflicts of Interest:
The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.