The Effects of Rainfall Patterns on Runoff, Sediment, and Nutrients Under Various Arti cial Rainfall Experiments

Hanna Mariana Henorman MARA University of Technology: Universiti Teknologi MARA Duratul Ain Tholibon (  duratulaintholibon@gmail.com ) Universiti Teknologi MARA Pahang Kampus https://orcid.org/0000-0002-2729-2694 Masyitah Md Nujid Universiti Teknologi MARA Cawangan Pulau Pinang Hamizah Mokhtar Universiti Teknologi MARA Pahang: Universiti Teknologi MARA Pahang Kampus Jamilah Abd Rahim Universiti Teknologi MARA Pahang: Universiti Teknologi MARA Pahang Kampus Azlinda Saadon University of Selangor Shah Alam Campus: Universiti Selangor


Introduction
Water erosion gives a severe type of soil erosion as it is susceptible to give more impact on the soil surface compared to wind erosion. This is caused by rainfall that is known as the major factor that contributes to water erosion since it caused the soil particle to lose and detached. The mechanism of rainfall erosion i.e. topography, soil properties, and rainfall characteristics had been investigated whether by simulated or natural rainfall. The rainfall that is saturated enough in the soil becomes surface runoff; known as a partial contributor to soil loss. Meanwhile, the energy of raindrops that detach the soil structure is another partly contributor. Merritt et al. (2003) stated that three stages describe the water erosion process are detachment, transport, and deposition of soil particles (refers Fig. 1). Detachment is the rst process that caused soil clods to break into smaller particles and is considered an independent variable and plays an important role as no erosion must happen unless detachment takes place (Sadeghi et al., 2017). Detachment is usually caused by the locally forceful shear stress to the soil surface by raindrop force called rainfall detachment or may be caused by surface runoff when the shear stress on the soil surface exceeds the critical shear stress of soil. The second process of water erosion is transportation and is considered a dependent variable upon detachment. In shallow water conditions especially, raindrops splash provide temporary disturbances that caused static particles to move eventually transported by overland ow (Hajigholizadeh et al., 2018;Kiani-Harchegani et al., 2018). The deposition is the last stage of water erosion and is also considered a dependent variable upon the earlier two stages. Sediment removed in the detachment process is transported by surface runoff. Once detached, sediment particles are transported in the ow. A deposition may take place when sediment transport capacity is lesser than sediment load in the ow (Aksoy et al., 2020).
There are several factors affecting water erosion that may fall under four categories, namely; rainfall characteristics (rainfall intensity and duration), soil properties (particle size, in ltration, erodibility, speci c gravity and bulk density), topography (slope and vegetation cover) and surface runoff. Mohamadi & Kavian (2015) studied that among the important parameters in predicting soil erosion i.e.; raindrop size and amount, storm duration and velocity of raindrops, rainfall intensity were the main rainfall factor as an intense storm produces high kinetic energy because high-intensity storm contains a greater percentage of energy to impact on the soil. An experimental study conducted by Dong et al. (2018) tested the rainfall intensities of 60, 96, and 129 mm/hr under similar conditions and found that the high intensity produces the highest soil content in the runoff. The nding is similarly found in Almeida et al. (2021) where the highest rainfall intensity varied from 75.0 to 44.6 mm h-1 produced a maximum sediment yield (0.138 g m-2 min-1) and runoff rates (0.87 mm min-1). In a particular natural rainfall event, the intensity is rarely constant throughout the event, but it has a signi cant or a slight change in intensities. Tao et al. (2017) studied the effects of rainfall intensity variation in a rainfall event and found that it impacted the process of runoff generation without a signi cant effect on the total runoff volume. Furthermore, it is also observed that it gives a signi cant impact on the process of sediment yield and nutrient loss as well as the total sediment yield and nutrient loss accumulated meanwhile the early high rainfall intensity produced severe erosion and nutrient loss. This is agreeable with study from Alavinia et al. (2019) that applied four simulated rainfall patterns on sandy and sandy loam soil and found that there is no signi cant difference in the runoff for the two soil types. However, there were major differences in soil losses ranging from 77.6 to 82.8 g/m 2 for sandy soil and 90.1 to 134.0 g/m 2 for sandy loam soil among the different rainfall patterns and stages. B.
Wang et al. (2017) studied ve rainfall patterns on 10° soil on an experimental plot and found that soil loss is governed most by increasing-rainfall patterns while the lowest is from the constant -rainfall event. The sediment yield produced by constant-rainfall events is at around 61.8% of the average soil loss from the increasing -rainfall event. Ran et al. (2019) conducted experiments using ve rainfall patterns in 1 hr duration and had 40 mm rainfall depth on slope gradients ranging from 5° to 40° and found that the rising-falling rainfall produces the largest total runoff and soil erosion amount. They also found that at a same slope gradient, the relative difference between the total runoff and soil erosion amounts of different rainfall patterns was up to 111% and 381% respectively.
In the topography factor, Han et al. (2021) found that vegetation coverage and slope gradient signi cantly affect runoff and sediment yield, however, the effects of slope gradient on runoff and sediment yield are opposite to those of vegetation coverage. Surface runoff, other than the result of raindrop impact has been recognized as an erosive agent as it causes shear stress to the soil surface, which if it exceeds the cohesive strength of the soil would result in sediment detachment (Merritt et al., 2003). Liang et al. (2020) studied the effects of between three conditions of slope angle, rainfall intensity and vegetation cover on the erosion characteristics using indoor-simulated rainfall tests on Pisha sandstone slopes and found sediment yield is signi cantly affected by rainfall intensity and least affected by slope angle. To study the effect of runoff ow rate on soil erosion, the present work of Mbiakouo-djomo et al. (2018), aims to simulate the dynamics of soil erosion taking into account the three main parameters in uencing the phenomenon; the nature of the soil (compaction), hillslope, and the rainfall intensity. The study was conducted on the samples under the three phases of tests, namely; a phase of rain simulation, a phase of streaming simulation, and a combined phase of rain and streaming. For the phase of rain simulation, the ow rates of 0.10 and 1.20L/s were applied to samples on 0, 2, and 4.37% slopes. The result showed that the larger ow rate gives a maximum mass of soil moves about ve times increment compared to the smaller ow rate.
Through water erosion, not only sediments that are transported but various nutrients such as nitrogen and phosphorus, especially from agricultural sites may wash away in the surface runoff that can cause land degradation resulting in eutrophication. This process not only decreases the soil fertility and the ability of crops to survive but also reduces the quality of water resources for human consumption. Rainfall especially with higher intensity increases the risk of soil erosion and eventually nutrient loss. Dong et al. (2018) investigated the solute transport by using Potassium Bromide as a tracer and concluded that solute content in runoff is related to the sediment mass by showing that; a) under initial moisture content of 15% and 25%, the solute content was 1.51 and 2.63 times greater than when the initial moisture content was 5%, respectively, b) the higher the rainfall intensity applied, the higher the amount of runoff solute content and c) under slope gradient of 15° and 25°, the solute content was 1.43 and 3.51 times greater than when slope gradient was 5°, respectively. Dai et al. (2018) have studied the amount of nitrogen (TN) and phosphorus (TP) losses under natural rainfall events and found that sediment yield is the major controller of the TN and TP loss. This is because heavy rainfall produces a higher rate of soil erosion and nutrient loss where 93% of TN and 99% of TP were transported with sediments. Zhang et al. (2010) conducted eld experiments to study soil erosion and loss of nitrogen (N) from a 15° hillslope and found soil erosion caused an N loss of about 250 mg/m 2 for the bare soil plot. In general, N concentrations decreased with time and approached a steady value throughout the experiment. Rainfall is the major factor and has been studied frequently, but the effects of rainfall patterns on soil erosion and nutrient loss have rarely been investigated. Due to the reason, Tao et al. (2017) studied three types of nutrient; nitrate nitrogen, ammonia nitrogen and phosphorus losses under four different rainfall patterns and found that decreasing-type, rainfall has the most nutrient loss rate compared to the other patterns, regardless of the nutrient types.
Previous experimental and eld works studied that sediment transport is in uenced by hydraulic properties of ow, physical properties of soil, and surface characteristics (Aksoy & Kavvas, 2005). Rainfall intensity or ow rate can be considered as the second independent variable in addition to the slope for better performance, ow discharge being more de nitive than the rainfall intensity in quantifying the sediment discharge. According to Tao et al. (2017), solute transfer to the soil surface runoff and runoff erosion are in uenced by rainfall characteristics. However, the relationship of rainfall patterns on sediment as well as the nutrient loss has been rarely investigated, although both uncontrolled losses will affect the environment and eventually water resources quality. Therefore, the objective of this study was to determine the effect of different rainfall pattern conditions on the surface runoff, sediment yield and nutrient loss.

Materials And Methods
A good plan and management are needed to ensure all the progress goes smoothly so that the objectives of the study can be achieved successfully. There are six major stages involved throughout this study as shown in Fig. 2.

Advanced hydrology apparatus
Rainfall simulation in this study used an equipment provided at the university, namely H313 Hydrology Apparatus from TecQuipment. The apparatus contains a closed water circuit with a storage tank and pump. The existing runoff plot of the apparatus was 2.0m (length) × 1.0m (width) × 0.19m (depth) which provided a plot area of 2m 2 and a plot volume of 0.38m 3 (refer Fig. 3, Fig.4, and Fig. 5). The apparatus has a jacked mechanism for an adjustable slope and consists of eight nozzles, in two banks of four, with an individual valve that can be turned on or off for the supply of water. The reservoir tank capacity is approximately 220 liters.
However, for this study, some adjustments were made to the equipment to provide a more reliable method for data collection during the running of rainfall simulation. The adjustments mainly focused on the outlet chamber and surface runoff collection system as shown in Fig.   6. The existing 2 adjustable over ow pipes are removed and the whole outlet chamber area is lled with a sloping channel that diverting the runoff from the weir into both holes. The water then runs through the runoff collection drain underneath the plot area to the runoff catchment tank placed at the end of the aluminum drain. Surface runoff was measured from the height of the water collected in the tank at a prede ned time and is then converted to the volume. Other than that, to ensure the simulated rainfall mimicking the natural rainfall, the nozzle system is propped using pieces of wood so that it is not slanted following the inclined soil plot as shown in Fig. 7. However, care should be taken to not let the wood not interrupt the nozzle spray angle.

Soil properties and bed formation
Sample soil was collected from the quarry area in one of the provinces in Selangor, Malaysia. The district has more or less uniform After the soil is delivered to the site, it is rst air-dried at room temperature, whilst the big chunks were crushed to the smallest form possible as shown in Fig. 8 (A). The soil also separated from any debris before being sieved through 10 mm diameter to make it easier to work with, the sieving process is shown in Fig. 8 (B). Table 1 shows the physical and chemical properties of the soil samples used, which is classi ed as sandy loam soil with the percentages of sand (>0.05mm), silt (0.05 -0.002 mm), and clay (<0.002 mm) were 85.1%, 8.9%, and 6.0%, respectively. From the soil particle size distribution curve, the median diameter of soil particle, D 50 of 0.36 mm was obtained.
To prevent any technical problem occurs while using the equipment, the plot rst was adjusted to 7% with the aid of some wood pieces.7% slope was used because the existing capacity of the equipment along with the additional soil load is taken into consideration. Before soil lling, the plot is layered using pieces of cloth ( Fig. 8 (C)) and is lled with sieved soil layer by layer (Fig. 8 (D)) whose soil bulk density is required to be 1.5 g/cm 3 , followed by light compaction to get the leveled catchment area in the experiment tank at the marked height.
Compaction was done using long sticks as the guide for atness before compaction using a 3kg brick dropped at around 20cm height several times throughout the plot area following the method of compaction by Khaerudin et al. (2017) shown in Fig. 8 (E). The nal prepared soil plot is as shown in Fig. 8 (F).

Rainfall simulation
After the catchment area is well prepared, the rainfall simulation was started. Water is allowed to run through the nozzle system by setting the desired owrate of the ow meter in the duration of 90 minutes of rainfall for each simulation. According to Dai et al. (2018), 30 min was the typical duration for the observation of maximum rainfall intensity of natural rainfall event thus the changes of owrate were done after every 30 minutes to achieve the target rainfall pattern, namely: A (uniform-type: 8-8-8 l/min), B (increasing-type: 7-8-9 l/min), C (increasingdecreasing-type: 7-9-8 l/min) and D (decreasing-type: 9-8-7 l/min). The four rainfall patterns are shown in Fig. 9. The treatments were done in three repetitions but allowing the soil plot to fully drain for at least 24 hours before the new simulation was applied. Due to the plot is using compacted top soil with no expose to agricultural uses and no plant has been installed, fertilizer was also applied in the compaction process when necessary to ensure there are enough nutrients carried in the runoff until the rainfall simulation ends. From the catalogue, the application rate for the fertilizer is 150g/m 2 (All Cosmos Industries Sdn Bhd (n.d.)) required about 300g per 2 m 2 plot area for each application.
After started, the time of start rainfall and runoff is recorded respectively. Surface runoff is observed by taking water height in the catchment tank using a measuring ruler (refer Fig. 10 (A)) in 3-min intervals to ensure the best and presentable of result data until the runoff ended. After volume reading, 500ml runoff samples each were collected in the sample bottles prepared as shown in Fig. 10  found an increasing pattern that contributed the highest amount of soil loss. However, it is in contrast with Tao et al. (2017) that found a decreasing pattern produce the highest that is similar to this study's ndings. This is because the higher intensity applied in the rst phase of rainfall duration results in a higher capacity of raindrop hits on soil plots that cause splash erosion. The loose particles were carried out with the surface runoff shown by the gure where the early phase has the highest concentrations until it reaches stable mode through the end of the duration.

Nutrients
The nutrient concentrations namely ammonia nitrogen, nitrate-nitrogen, and phosphorus associated for each rainfall pattern are shown in Fig.   13, Fig. 14, and Fig. 15 respectively. The combination of these three types of nutrients is shown in Fig. 16. For ammonia nitrogen and nitrate nitrogen nutrient, the trends showed that most transport processes happened in the beginning 30-min of rainfall duration and consequently achieve stable mode through the end. The determined total losses for ammonia nitrogen were 3.948, 2.902, 3.536, and 4.081g; nitrate nitrogen was 3.891, 2.677, 4.032, and 3.255g; phosphorus was 1.333, 0.223, 0.2010, and 0.690g, for patterns A, B, C, and D, respectively. In general, rainfall pattern does not have a signi cant impact on the trend of nutrient losses, where the trend shows that higher concentrations at the start and eventually lowered through the end, but pattern D as compared to other patterns resulted in more severe nutrient losses.

Dai et al. (2018) studied the amount of nitrogen (TN) and phosphorus (TP) losses under natural rainfall events and found that sediment yield
is the major controller of the TN and TP loss. This is because heavy rainfall produces a higher rate of soil erosion and nutrient loss where 93% of TN and 99% of TP were transported with sediments. Zhang et al. (2010) conducted eld experiments to study soil erosion and loss of nitrogen (N) from a 15° hillslope and found soil erosion caused an N loss of about 250 mg/m 2 for the bare soil plot. In general, N concentrations decreased with time and approached a steady value throughout the experiment regardless of constant or varies-intensity rainfall conditions. Findings from this study are satisfying as the concentrations for ammonia nitrogen and nitrate nitrogen shows the same trend. Though there are not enough previous studies on the relationship of rainfall pattern to nutrient loss, however from the ndings, it is generally similarly found by Tao et al. (2017) that decreasing pattern contributes the most severe nutrient loss. The reason is that the higher rainfall intensity of the early phase has produced a higher rate of nutrient loss since the impact from high intensity released solute at the soil surface to the runoff as compared to low intensity.

Soil pro le measurement
The data in Table 2 were derived from the measurement at a distance 15cm × 15cm square grid on the soil plot surface before and after every rainfall simulation done to determine the soil pro le in the plot. Meanwhile, Fig. 17 shows the photographs of the soil plot before and after each simulation and the soil pro le plot for a respective experiment. The difference was calculated by subtracting the before and after values.
This means that the positive difference value shows that the soil is eroded meanwhile the negative value shows raised soil that is carried by runoff. The no difference value shows that there is no difference in height reading before and after the rainfall simulation that can be assumed that a particular soil area maintains the same with no eroded or raised soil involved. From these three values, positive and negative values are then added and divided by the total area of soil plot to determine the affected area in terms of soil movement process during rainfall simulation. The calculated means values of affected areas are 79.60, 68.70, 72.43, and 64.97% for pattern A, B, C and D respectively.
The lowest means value of the affected area is contributed by pattern D and highest by pattern A. 3.4 Response of total runoff, total SSC, nutrient concentrations, and soil affected area to rainfall patterns Table 3 presented the summary of experiment variables and their respective mean value. The mean is calculated from the value obtained in the three repetitions of each rainfall pattern. The variables, including runoff, sediment concentrations, and the three nutrient concentrations namely ammonia nitrogen, nitrate-nitrogen, and phosphorus. The mean values are each discussed in subchapters 3.1 to 3.3. By considering the twelve experiments, the mean calculated for runoff is 697.944 liter, for total SSC is 14,045.451mg/l; for ammonia nitrogen concentration is 3.612mg/l; for nitrate-nitrogen concentration is 3.464mg/l and for phosphorus concentration is 0.614mg/l.  Table 4 and Fig. 18 present the results of ANOVA and the mean plot for the seven particular variables studied. Variable's nutrients are combined in one plot to better observe the trend pattern among the three nutrients. ANOVA was used to evaluate differences in the runoff, sediment concentrations, nutrient concentrations, and total affected soil area between the four rainfall patterns. The outcome of the statistical test is probability named the p-value, which is compared to a threshold called the signi cance level. The performance result of the system is related if the p-value is lower than that signi cance level.
According to the results, we can conclude that there is no signi cant interaction between the seven variables and the rainfall patterns since the p-value is lower than 0.05 in none of the performance metrics. ANOVA assumes that the data come from a normally distributed population with a homogeneous variance and similar covariance and sphericity (differences between all possible pairs of groups are equal).

Conclusions
Rainfall patterns were performed in the experiments to study the effects of rainfall patterns on the sediment and nutrient loss rate. The experiment was done using Hydrology Apparatus with some adjustment on the drain system. The apparatus has a soil plot area of 2m 2 that is lled with sandy loam soil and compacted layer by layer through the top. Four rainfall patterns are chosen, namely A (uniform-type: 8-8-8 l/min), B (increasing-type: 7-8-9 l/min), C (increasing-decreasing-type: 7-9-8 l/min) and D (decreasing-type: 9-8-7 l/min) with the changes of intensity every 30 minutes that gives total rainfall duration of 90 minutes for each pattern. The simulation was done in three repetitions.  19,918.50 mg/l for patterns A, B, C, and D, respectively. Pattern D contributed to the highest amount of sediment accumulated whereby pattern C generated the lowest sediment despite the trend showed a different approach than the other 3 patterns. c. In nutrient concentrations, the trends of ammonia nitrogen and nitrate-nitrogen concentrations showed that most transport processes happened during the rst phase of rainfall duration and consequently achieve stable mode in the preceded phase through the end. The determined total losses for ammonia nitrogen were 3.986, 2.891, 3.504, and 4.601g; nitrate nitrogen was 3.934, 2.665, 4.008, and 3.259g; phosphorus was 1.346, 0.222, 0.207, and 0.679g, for patterns A, B, C, and D, respectively. In general, rainfall pattern does not have a signi cant impact on the trend of nutrient losses, where the trend shows that higher concentrations at the start and eventually lowered through the end but pattern D as compared to other patterns resulted in more severe nutrient loss. d. For the affected area of the soil movement process, the calculated means of the affected area are 79.60, 68.70, 72.43, and 64.97% for patterns A, B, C, and D respectively. The lowest mean of the affected area is contributed by pattern and highest by pattern A.
For recommendation, the simulation has been done on bare land conditions, thus further studies on the vegetation cover may need further investigation. Other than that, variation of slope gradient and different soil types may be suitable for further studies. However, in general, the relationship of rainfall patterns on nutrient loss should be investigated more in the future due it has previously been rarely investigated. Figure 1 Water erosion stages a) raindrop hits the soil b) splash impact on the soil surface (Trail Grades (and Outslope) -Trailism (n.d.)) Figure 2 The framework of methodology  Components of Hydrology and Rainfall Apparatus (Side View) Figure 5 Sketch of rainfall simulator and erosion ume (above) Plan of erosion ume and nozzle systems (below) Figure 6 Runoff catchment drainage adjustment      Determined SSC of (A) constant-type, (B) increasing-type, (C) increasing-decreasing-type and (D) decreasing-type Determined ammonia nitrogen concentrations of (A) constant-type, (B) increasing-type, (C) increasing-decreasing-type and (D) decreasingtype Figure 14 Determined nitrate nitrogen concentrations of (A) constant-type, (B) increasing-type, (C) increasing-decreasing-type and (D) decreasing-type