Hydrological characteristics using the SWAT model on several agricultural systems in the Maros karst area

In the region of karst, there are valley areas that are used by the community as agricultural land. However, drought phenomena often occur in this area, so it is necessary to know the presence of water to support agricultural processes. The aim of the research is to predict flow rates in several agricultural systems using Soil and Water Assessment Tools (SWAT). SWAT was chosen because it can simulate the model with easy-to-obtain data. Required data such as Digital Elevation Model (DEM), soil type, and land cover to form a Hydrology Unit Response (HRU). Enhanced with climate data input to run SWAT Run. The scenario is done with Land Use Split in the HRU Definition menu. The results showed that forest vegetation defined as FRSE and FRST stored water better than AGRR and RICE agricultural lands. However, it is different for underground flow, that RICE can store better. The scenario results show a decrease in the value of surface runoff by 15.3% and an increase in lateral flow by 26% from the existing condition. The underground flow continued to increase but was not significant at 4.2%.


Introduction
Karst is a term used to describe landscape conditions that tend to experience drought.This happens because the karst has an underground water system and some caves are formed by soluble rocks such as limestone, marble and gypsum [1].Even karst has more than 50% carbonate compounds and is perfectly formed in limestone with 90% calcium carbonate content.So, the greater the carbonate content, the more perfect the karst formation process is.
The phenomenon of drought in this area occurs due to a solutional process because groundwater is not able to be tied up properly so that it falls into the ground creating an underground flow [2,3,4].In some cases, karst landscapes in the hemisphere are inseparable from drought, such as in the dry tropical karst area of the Yucatan Peninsula in Mexico [5,6].In northern China over the past 40 years, environmental problems associated with karst water systems include decreased water quality, drying of springs and continuous decline of karst water level [7].In Java, the famous Gunung Sewu also experiences complete drought in the dry season [8,9].
In the karst area, there are valley areas that are used by the community as agricultural land.Especially for the Maros Regency area [10] from a finding with farmers, the author suggests that farmers can only plant rice once a year.This is due to the fact that the dam that has been built does not function properly to accommodate the water discharge because it is located on a limestone bedrock which tends to quickly allow water to pass below the surface.According to [11], All hydrological processes in the water cycle are deterministic, governed by physical laws that can be used to predict by considering the knowledge of the atmosphere, boundary layers, and the earth's surface and subsurface.However, it should also be noted that predictions can be complicated by hydrological transport phenomena, such as airflow, solutes and sediment deposits as well as the intensity of rainfall events and temperature fluctuations.
Thus, the management model needs to take into account changes in space and time.SWAT (Soil Water Assessment Tool) is one of the hydrological models that can be used.Has been described in SWAT [12], that the watershed is divided into several sub-watersheds which are then divided into smaller units (HRU) Hydrology Response Units.The HRU consists of land use, management, topography and homogeneous soil characteristics.
Three land covers are the object of the study, the first is rice fields with a monoculture farming system.Second, dry land agriculture with intercropping or rotational farming systems, the management of this land is quite diverse.Third, in secondary dryland forests, some forest plants grow side by side with plantation crops belonging to the local community

Materials and methodology
This section describes the research location and some of the data needed.Then describe the model used to assess the hydrological characteristics

Study area
The karst region of Maros Regency is geographically located between ‫"0'63ﹾ911‬ E -‫"0'05ﹾ911‬ E and ‫"0'05ﹾ4‬ S -‫"0'4ﹾ5‬ S. (figure 1.)The rock that makes up the Maros karst area is limestone as the main constituent of the Tonasa Formation, this landscape undergoes tectonic processes and igneous intrusion.The research area is in the Maros watershed which includes a karts area, administratively covering Bantimurung, Simbang and Cenrana district.
The area of the Maros watershed is 73,119 Ha [13], while the research area is around 18,266 Ha.This watershed is strongly influenced by rainfall; therefore, it fluctuates in discharge from time to time.Rainfall ranges from 2000-4000 mm/year.Rainfall above 3500 mm/year covers the most the area is 29,645.3Ha or 44.93 percent.The rainfall is spread over the middle to upstream of the Maros watershed [14], which means it covers the research area.

Data collection
Required data are in the form of primary and secondary data.The secondary data needed in this research is in the form of spatial data including, Maros Regency Administrative Map, Watershed Map Maro's research location, Land Cover Map, Soil Type Map, Maros Regency Karst Area Map and DEM data.Furthermore, climatological data in the form of rainfall data, temperature, solar radiation, relative humidity and wind speed.
Primary data includes sampling disturbed soil to obtain data on texture, organic matter and water content.Then the whole soil sample was taken to get the value of bulk density and hydraulic conductivity.

Description model
ArcSWAT is software that is integrated with ArcView (Geographic Information System), using spatially dispersed data on topography, soil, land cover, land management, and weather to predict water [15].Data input (input) and SWAT outputs are described in more detail by Arnold [12].input uses spatial data in the form of Digital Elevation Model (DEM) data as the basic data for determining the boundaries of the research area sub-watershed (watershed delineation).Data on land use, soil type and slope are data that are combined to create a Hydrological Response Unit (HRU analysis).Climate data is processed in the form of rainfall data, temperature, solar radiation, relative humidity, and wind speed (weather station).
3 Simulations were carried out to calculate the condition of water availability at the research site in several agricultural systems, namely rice fields, dryland agriculture and secondary dryland forest.The data that has been prepared is used for the calculation of the SWAT model.Simulation is intended to understand the presence of water, both on the surface and below the surface.Therefore, in this study, the outputs studied from the SWAT model are Total Water (WYLD), surface flow (SURQ), lateral flow (LATQ) and underground flow (GW_Q).After getting the value of the SWAT model results, the next step is calibration and validation.Calibration and validation are intended to test the accuracy of the output of the SWAT model.The output tested is the flow discharge (FLOW OUT) to the observation data or field measurement discharge data [16].The statistical method used in calibrating and validating is the Coefficient of Determination (R 2 ) dan Nash-Sutcliffe Efficiency (NSE).R 2 is the proportion of the total variance in the observed data that can be explained by the model.Its value ranges from 0.0 to 1.0.Higher values mean better performing models.R value 2 ≥ 0.5 is considered acceptable.
Where R 2 is evaluation model feasibility, X is observation discharge value,  ̅ is debit observational average, Y is debit model calculation.
Whereas the Nash-Sutcliffe Efficiency (NSE) shows the performance plot of observed (measured) values compared to predicted-simulation values according to 1:1.Values range from to 1, with NSE = 1 being the optimal value.The method used is Nash-Sutcliffe Efficiency (NSE) and R 2 .R 2 is the proportion of the total variance in the observed data that the model can explain.Its value ranges from 0.0 to 1.0.Higher values mean the model performs better [17][18][19][20].
The statistical method used to test the model is the model efficiency equation Nash-Sutcliffe (NSE): Where NSE is the efficiency of the model Nash-Sutcliffe, y is measured actual discharge, ŷ is discharge simulation results, Y ̅ is average rated discharge.

Land-use simulation
The simulated land use (table 1) was carried out with the tools provided by ArcSWAT.In the HRU Definition menu, there is a Land Use Refinement scenario, in which there are two scenarios, but here a Land Use Split scenario will be selected, in which the type of land use selected will be divided into one or several selected sub land uses.The land cover that will be simulated is FRST in which there are Sub Land COFF and AGRR divided into Sub Land CORN and PNUT.

Calibration and validation
The calibration and validation process uses 2015 data.It can be seen that the calibration results obtained a Nash-Sutcliffe (NS) efficiency value of 0.72 satisfying/acceptable (figure 2).Meanwhile, the validation process will show the level of uncertainty that occurs when predicting a hydrological process.
The validation results show the consistency of the model with an NS value of 0.39 (satisfactory).River discharge validation data used in 2015 (figure 3)

Water flow
The results of the SWAT analysis show that there are 22 sub-watersheds formed, but only 7 subwatersheds were selected as the sub-sub-sub-basins that represent the karst area.Sub-watershed 16 got the highest runoff value (table 2).From the SWAT results, this sub is dominated by agricultural land so it can be said to be the trigger for high runoff.The high surface runoff is directly proportional to the decrease in the underground flow value.Has been confirmed by Arsyad (2010) that vegetation affects the volume of water entering the soil and underground water reserves [21].Specifically shows the hydrological characteristics based on the land cover that agricultural land has the highest runoff value (table 3).In surface runoff forest vegetation, it looks lower, indicating the greater changes in land use such as from forest vegetation to agricultural fields causing greater changes in runoff [22].Agricultural processes are capable of significantly modifying the landscape, interfering  with soil management which alters infiltration and surface runoff characteristics and affects groundwater recharge [23].
The high runoff in paddy fields (table 2) is due to inundation carried out and very intensive land management so the infiltration value in paddy fields is very low.Setyowati in her research, explains that in paddy fields, the soil structure will be smoother, and sticky, with shallow groundwater so that it is difficult to infiltrate the soil.For lateral flow in paddy fields, it can be said to be non-existent, the solid layer formed as a steel tread makes the absence of water flow [24].
However, the base flow of paddy fields has the highest value.According to in their seminar explained that paddy fields can increase the base flow so that the supply of water through the base flow can be sustainable with the increasing area of paddy fields [25].It has also been explained that there is an increase in ground water flow every time irrigation is carried out on paddy fields, the availability of underground water will also affect the replenishment of water to the river so that it can be used for irrigation or household purposes [26].
The trend of the presence of water during the year can be seen (figure 4), that forest vegetation is more evenly distributed between surface runoff and lateral flow.The total amount of runoff can be reduced by the presence of vegetation and increase in the amount of water retained on the soil surface [22].Then forest vegetation can also maintain macro pores in the soil [27].So that the infiltration rate in forest vegetation is faster and suppresses surface runoff.Observing the trend of agricultural land, the graph shows runoff which peaks in the rainy season (figure 4).In January the runoff exceeded 300 mm, either in paddy fields (figure 5-D.) or in dry land agriculture (figure 5-A.).Dryland agriculture experienced a decrease in lateral flow so that a decrease also occurred in underground flow.Underground flows in forest and agricultural land cover appear to be more balanced, except for rice fields which reach 300 mm at their respective peaks in March.Although the peak of rain occurs in January the filling of water to the base flow will occur if all the soil pores have been filled.At first, the water seeps into the unsaturated zone and then seeps deeper into the saturated zone which becomes groundwater [28].Water will seep into the ground surface moving vertically downwards towards the groundwater table to become groundwater and groundwater then move both vertically and laterally in the groundwater system [23].
After that, the base flow will come out to the outlet and that happens when the dry season has started.So, the availability of water can last throughout the year even though the base flow has decreased because in the process the water discharge does not occur all at once, but slowly.It can be seen in September, the availability of water in the base flow in each land cover has been greatly reduced, but the rainy season has started.

Land use simulation
The simulation was carried out without significantly reducing the areas defined as forest, the area defined as FRST included coffee plantations.This is intended to prevent excessive intervention in very vulnerable karst areas.The addition of the FRSE area was also carried out in the FRST area as a form of forest vegetation improvement.The area that is simulated in the area defined as AGRR is introduced to corn and peanuts.The introduction of some of these commodities into the research area based on existing commodities was managed by the local community, although not so massively.However, it is still necessary to pay attention to facilitating agricultural processes, in their management the water content of the soil is often changed, where the moisture can increase with irrigation or decrease when the soil is drained.This is likely to have an impact on groundwater resources, with changes in the amount and timing of filling and discharge, the impact will be most noticeable in karst areas where surface water and groundwater are closely related [29].
The SWAT model simulation shows a decrease in surface runoff, followed by an increase in lateral flow but does not have a significant effect on underground flow (Figure 6).The decrease in the value of surface flow is 15.3% and the increase in lateral flow is 26% from the existing condition.Underground flow continued to increase but not significantly at 4.2%.In figure 6. Showing every rain that occurs will give an increase in flow rate.The highest discharge occurs in January following the pattern of the highest rainfall.The potential availability of water in the soil is needed in the water management process for agricultural development.The less surface runoff in an agricultural process, the better the governance is.In theory coffee-based agriculture can reduce the rate of runoff.In a thesis explained that coffee plants are included in the roux architectural model which allows the amount of canopy outpouring to be reduced and can suppress surface runoff [30].If applied to agriculture with an agroforestry system based on coffee plants, it will be better to suppress the high runoff than the monoculture system [31].Seen in the graph, coffee plants (figure 7-C) can still be matched with the area defined as FRST (figure 7-B).
Maize crops (figure 7-D.) and peanuts (figure 7-E.) introduced to the AGRR did not significantly change the water flow in the study area.However, the presence of peanuts with an intercropping system 1230 (2023) 012141 IOP Publishing doi:10.1088/1755-1315/1230/1/0121419 with corn can reduce the value of surface runoff, increase infiltration, and reduce evaporation so that it can maintain the presence of water in the soil.Peanuts in the generative phase reached the peak of the degree of canopy closure to reduce runoff [32].

Conclusion
The results of the simulation analysis show that there is a 15.3% decrease in surface runoff and a 26% increase in lateral flow from the existing conditions.The underground flow continued to increase but not significantly, namely 4.2%.The simulations carried out did not significantly change the presence of forest vegetation but only introduced one type of coffee commodity into the FRST area.Some FRST areas also experience changes in the simulation by defining them as FRSE, so that the output flow value also changes in a positive direction.
1230 (2023) 012141 IOP Publishing doi:10.1088/1755-1315/1230/1/0121414 In looking at the accuracy of the model output pattern with the results of field observations, a deterministic coefficient or linear equation is used: R 2 =

Figure 2 .
Figure 2. Regression analysis of the comparison of the observed flow rate and the model flow rate after calibration on the 2015.

Figure 3 .
Figure 3. Graph comparison of actual river flow discharge and model river flow discharge after validation (2015).

Figure 5 .
Figure 5. Graph of hydrological characteristics by land cover.

Figure 6 .
Figure 6.Comparison of hydrological characteristics of existing conditions and simulation of land use.

Figure 7 .
Figure 7. Graph of hydrological characteristics based on land use simulation.

Table 1 .
Land cover classification based on SWAT model

Table 2 .
Existing hydrological characteristics of the research area.

Table 3 .
Hydrological characteristics based on land cover.