Spatial Model of Post-Earthquake Spring Performance in the Watershed Areas

The 7.4 Mw of tectonic earthquake caused liquefaction in Pasigala on September 28, 2018, happened due to the fault movement of Palu-Koro. It affected the water availability every spring. The research aim is to determine the spatial model of water production every spring after the natural tectonic disaster, especially in Palu and Poboya watersheds-a model built based on the integration between the spatial data overlaying and the statistical regression correlation. The sites are purposively selected at six springs spots and divided into four clusters (Poboya, Uemanje, Ranjuri Beka, Mantikole). The model assessment was generated based on the springs’ performance from x variables (catchment area, land cover, aquifer, free-ground water depth, fault, number of springs users) and the y variable (water discharge). The result shows that Poboya’s performance is bad-disturbed, while Uemanje, Ranjuri, Beka, and the performance of Mantikole are disturbed. The bad performance of springs requires conserving watershed areas through forest and land conservation, tree enrichment planting, wise land management, and good water use.


INTRODUCTION
Issues related to natural disasters have become a central issue in every country in the world (Albris et al. 2020, Galata et al. 2020. It has an impact on the economy, humans, and the environment (Botzen et al. 2019, Ali et al. 2020, especially in Asian countries that are prone to earthquakes, tsunamis, and liquefaction (De Goyet 2007, Wekke et al. 2019).
The earthquakes can cause hydrological responses, such as groundwater availability in and above the ground (Wang et al. 2004, Liao et al. 2015. Stockpiling a reservoir risks an earthquake (Xuan-Nam et al. 2020) and can cause landslides and soil erosion (Tunas et al. 2020, Naharuddin et al. 2021. Each of Palu's watershed and Poboya's watershed are equipped with springs. Both were affected by the Palu-Koro movement when the earthquake event occurred. The ground movement then predicts the impact of the Poboya Springs, Ranjuri Springs in Beka village, Mantiloke Springs in Mantiloke village, and Uemanje Springs in Uemanje village. Those springs are generally located in the tectonic fault pathway. The tectonic fault structure controls the emergence of water sources and water channels under the surface. The tectonic fault structure affects the emerging water sources (Santosa 2006).
Research on the impact of the earthquake, tsunamis, and liquefaction in Palu and Donggala concerning water sources' performance has not been conducted. The research was carried out only on the focus on assessing the level of erosion hazard in some land uses after earthquakes and liquefaction (Naharuddin et al. 2020), mapping of post-disaster damage and liquefaction (Syifa et al. 2019), landslide study in the East Palu Valley after the earthquake (Mason et al. 2021). Earthquakes can cause hydrological responses, such as groundwater availability in and above the ground. This statement is in line with King et al. (2006) statement that there is a temporal change in underground water hydrology due to the earthquake. This knowledge is very important for watershed management planning and modeling of springs (Wilkerson 2008, Reghunath et al. 2009, Jakubis & Jakubisová 2019. The research on the water production in Palu's watershed and Poboya's watershed is interested in assessing water's present condition and availability at the ground and surface pathway. The research aims to determine the spatial model of springs' performance on water production post-impact of the earthquake at these sites. The research would provide spatial data and information on springs' performance. At the same time, the guideline of water resource recovery and the science and spatial technology of watershed management is also arranged as the basis for post-earthquake management planning.

Study Area
This study was collaborative research involving the Palu Poso Watershed Management Center and the Forestry Faculty of Tadulako University. It will be conducted for two years, from May to October 2019, for data collecting and June to December 2020, for spatial model arrangement.
Technically, sites were divided into four clusters. The watershed in Palu consisted of 3 clusters: Ranjuri-springs, Mantiloke-springs, and Uemanje-springs. Furthermore, the last cluster chose Poboya-springs as a part of the Poboya watershed. Geographically the location is at the coordinate point, according to Table 1.The study sites are shown in Fig. 1.

Tool and Study Material
Data were divided into primary data collected by the collaborative team, including primary data on the field condition of each spring taken from May to October 2019, and secondary data types consisting of the map. These maps were collected from a particular center or agency. National Mapping Centre (NMC) provides spatial data with a scale of 1:50,000 -1:100,000 consisting of digital maps such as Indonesian land cover, route map, river flow, and settlement area. Other data were prepared by the Palu Poso watersheds management center, such as the catchment area, land and geological map, rain intensity zonation map, and topographical and water resources.
While fault maps, water basin maps, and spatial planning maps are collected from the provincial planning agency with a scale of 1:100,000 -1,250,000. Palu's regional forestry mapping agency provides land cover data and forest area. This kind of map is available for scale 1:100,000 -1,250,000. Furthermore, those particular satellite images have been collected from the national space agency for SPOT6/7.

Research Methodology
A spatial analytical method has been implemented for this research, combined with overlaying spatial data and made a score through statistical analysis (regression-correlation). The spatial analysis, through overlying a basic map, thematic map, and SPOT image, tries to build parameter attributes and count the indicator values (Wahyuningrum & Putra 2018).
The spatial elements studied include the distribution of the fault path on September 28, 2018, geomorphological (river flow, tectonic surface pattern) and the river basin, aquifer, free land water depth, land texture, climate (the intensity of rain), topographic, land cover (vegetated or non-vegetated), water quality (discharge), catchment area  Besides, supplement data support the research, i.e., springs typology, geological, land texture, elevation, kind of land cover, land uses, land status, watershed area, accessibility, infrastructure, and community profile (number of residents).
The selection of springs locations that were examined for their spatial performance was done purposively as presented in Fig. 1 with the following considerations: (a) Poboya springs are located close to the tectonic fault lines, tectonic plains, non-forest areas, non-water basin area, typology of springs 'contact', type of land use in settlement area/ mining areas, owned land status; (b) the Uemanje springs is located a distance from the cesarean line, tectonic mountains, protected forest, non-water basin area, springs 'contact' typology, land use types in forest areas, state-owned land status; (c) Ranjuri springs is located near the cesarean line, river flow path, outside the forest state area, at the water basin area, 'depressive' springs type, residential areas land use type, land owned by the community/village property; (d) Mantikole springs is located near cesarean lines, hot springs, river flow paths, located in outside of state forest areas, in a water basin, 'depressed' springs type, agricultural land use type/ tourist areas, owned by the community/local government property. The springs' location selection is a catchment area typical of land cover, soil texture, water quantity (discharge), water quality, annual rainfall, and potential groundwater in aquifers.
The process of the research describes as follows: 1. The analysis process is based on inventory and identification data of springs.
Overlaying and delineated boundary research areas (spatial and tabular): The boundaries of the Palu and Poboya watersheds, regencies/cities, districts, villages, and forest area boundaries (status/function of the area). Spatial tabulation and analysis based on the inventory and field identification of six springs locations into the four clusters) in the Palu and Poboya watersheds.
2. Spatial modeling of springs performance.
Springs performance modeling with a spatial data overlaying approach and regression-correlation analysis was made to determine the relationships and trends among variables. The water quantity/water discharge (L.s -1 ) variable was analyzed with the cesarean moving path (kilometers), soil texture (related to infiltration/permeability), catchment area (ha), geomorphology (watershed /non-watershed in river flow paths, tectonic mountains, rainfall (mm.y -1 ), aquifer productivity, free groundwater depth, slope (% slope), vegetated land cover (% of the coverage area), number of springs users (households, worship, offices, education, health, tourism, agriculture, etc.). The results of the regression-correlation analysis become a reference for the scoring and variable scores selection. Range correlation value (R) ≥ 0.50 were selected as modelers in the spatial analysis system of springs performance. The size of each chosen variable's scoring is determined based on the coefficient determination (the level of accuracy of the regeneration model) with a value of R2 ≥ 0.25.
The springs spatial performance criteria are as follows: • Poor performance springs category: The springs' condition regarding water production has connected to a bio-geophysical and socioeconomic variable with a total weighting value of 20-46. Springs area recovery is recommended.
• Disturbed performance springs category: The springs' condition regarding water production has connected to a bio-geophysical and socioeconomic variable with a total weighting value of 47-73. Countermeasures action is recommended.
• Good performance springs category: The springs' condition regarded water production has connected to a bio-geophysical and socioeconomic variable with a total weighting value of 74-100. Prevention is recommended (The Ministry of Environment and Forestry 2017), (modified according to the research needs).

The Springs Condition
Characteristics observation of groundwater has been conducted on the spot that naturally springs. The springs indicated groundwater comes out as surface runoff. The springs condition in Palu and Poboya watersheds in four clusters are shown in Table 2.

Spatial Model of Springs Performance
The spatial model of springs performance begins with regression analysis and correlation among the determining variables of the performance of springs concerning each spring's ability to produce water (discharge) in liters/second. There are ten variables: (a) The proximity of springs from the movable tectonic fault; (b) Soil texture related to the level of infiltration; (c) water catchment areas; (d) geomorphology of river flow paths, plains/hills/tectonic mountains in the water basin/ non-water basin areas; (e) rain intensity; (f) Aquifer productivity; (g) vegetated land cover; (h) topography class; (i) Number of springs users; (j) Depth of free groundwater. The results of regression and correlation are as in Table 3.
The regression correlation analysis in Table 3 used the transformed data from  consists of null with SQRT=√xi+0,5, and other than null, has transformed with SQRT=√xi. The qualitative data such as geomorphology, aquifer productivity, and soil texture are quantified by scores 1,2,3 according to the level. The high aquifer productivity received a score of 3, while the medium and low productivity levels received scores of 2 and 1, respectively. The same treatment is also done for the soil texture, scoring 3 for fine texture, 2 for fine soil texture, and 1 for rough texture. The geomorphological water basin path and tectonic land received a score of 3. The non-water basin received 2 and 1 for tectonic hills. Table 3 shows the results of the regression analysis and the correlation between the response variable (Y) water discharge with explanatory variables (X1-X10), which are generally weak-moderate negative correlations due to the fewer water discharges (1.45 L.s -1 ). SP2 Poboya became the only site supported by an adequate water discharge from those six springs locations, which is 5,09 L.s -1 . In contrast, the SP1 Poboya and SP Mantikole (hot springs) are supported by lower discharge, 1.18 L.s -1 and 04 L.s -1 , respectively. Each SP1 Ranjuri, SP2 Ranjuri, and SP Uemanje supported the water flow of 0.70 L.s -1 , 0.18 L.s -1 , and 0.51 L.s -1 . The potential quantitative level of free groundwater (smallmedium) and small potential (<1.0 L.s -1 ) occupied the eastern part of the river basin in East Palu District, South Palu, Dolo. Medium potential (1.0-5.0 L.s -1 ) covers the entire western part of the water basin and only 40% in Sigi Biromaru and Gumbasa Districts.
The fault movement may cause less water discharge in SP Ranjuri Beka and SP Mantikole. Somehow, cesarean has led to landslides upstream of the Beka Village, disrupting the existing water flow as springs. Generally, the pattern of springs is caused by groundwater level cross topography, so the springs seem depressed and associated with the presence of fault (Santosa 2006). The topography in Mantikole was impacted by water discharge. Steep topography with alluvium rocky type (Molas Celebes Sarasin and Sarasin) and the existence of fault movement affect the rock structure disruption (Fig. 2). Santosa (2006) suggested that hot springs appear due to non-gravitational power are located in the fracture of the long earth's crust, and very deep. The water source mostly comes from rainwater that falls around it, and only a small portion comes from inside the earth (magmatic). The falling rain infiltrates the ground and is heated by magma underneath, forming a large convection current that pushes to the ground's surface.
Poboya springs cluster outside the water basin area in tectonic plains and contain two springs with different discharges, namely springs (SP) Poboya with small water discharge and SP2 Poboya with moderate water flow. Both of these SP locations are very close to the tectonic fault movement, steep topography, rock type sediment (molasa Celebes Sarasin and Sarasin) (Fig. 3). Further, SP Ranjuri Beka and Uemanje have a small water discharge located outside of the water basin area and far from the fault movement (Fig. 4). The movement disrupted the rock structure so which affects water flow. Rock cracks could decrease slope stability. The landslide may occur the rainwater seeps into the crack or when the slope is shaken. The appearance of springs is mostly influenced by the tectonic fault structure, other than the rock's constituent condition. The tectonic fault  (Rahayu et al. 2009, Santosa 2006).
The analysis of free groundwater depth (X5) shows a strong positive correlation (0.75) with water discharge and significantly affects spring supply. The level of free groundwater depth at the site was 0.61 m (<1.0 m). Therefore, the regression model explained that the shallower free groundwater from SP would better the supply of water sources. This is in line with research conducted by Zeffitni (2012) The proximity analysis of the cesarean movement on September 28, 2018, shows a strong negative correlation (-0.75) with water discharge and significantly affects the water supply to the springs. The distance of the cesarean movement at the study site is an average of 8.87 km (somewhat close), so according to the obtained regression model explains that the closer the cesarean movement distance from the SP will, the greater disturbance to the water availability at the SP. According to the Meteorological, Climatological, and Geophysical Agency, the earthquake on 28 September 2018 was located at 119.85° E: 0.18° S about 80 km north of Palu City at 10 km depth. The earthquake occurred along the Palu-Koro on the eastern side along ± 160 km in three segments, two active fault segments on the tram-Talise-Bora ± 130 km long, and one segment of the active Gumbasa-Kulawi fault ± 30 km long (Soehaimi et al. 2018, Supartoyo et al. 2018. The occurrence of a negative-moderate 'correlation between water discharge and Water Catchment Area (WCA), vegetated land, aquifer productivity, and springs users caused by less water discharge. Ideally, WCA and extensive vegetated area, high aquifer productivity, and many springs users, water discharge must also be larger to have a 'positive, strong correlation. The discrepancy between the response variable (Y) and the explanatory variable (X) in the four SP clusters in the Palu and Poboya watersheds can occur due to the fault earthquake on 28 September 2018. Each landform has characteristics of rock structure as a container aquifer groundwater (Sumartoyo 2010).
The determinant coefficient value (R2) in Table 3 shows the level of accuracy of the regression model, ranging from good (0.56) to very bad (0.01). The regression suggested that only those R2 with moderate to good category (0.29 -0.56) are suitable to build a performance model. Thus, the variable of free groundwater depth (X5), fault movement distance from springs (X3), aquifer productivity (X4), vegetated land (X9), catchment area (X1), and springs users (X10) as modelers' performance conditions in evaluation. Spatial modeling of SP performance conditions is approached by the weighting of variables X and Y. This was confirmed by Gunawan et al. (2016); weighting each parameter (factor) should be based on a correlation matrix between parameters. Table 4 shows the X and Y variables formulated as a spatial model in evaluating SP performance conditions. The model explains the water catchment area (WCA) as a reservoir for rain, as a place for vegetation to grow, and various types of land use, soil type, and groundwater potential are also included. This aligns with Gunawan et al. (2016), who stated that catchment area condition depends on several factors, including land use, slope (topography), rainfall, soil type, and groundwater potential. The results of spatial model performance for six springs locations in the four clusters are given in Table 5. Table 5 shows 'bad' performance in one site, and the other five are disturbed. The correlation regression analysis between variables Y and X with negative regression coefficient values and moderate-strong negative correlations were analyzed as the factor of springs 'bad-disturbed' water production performance. The only variable X with a great positive regression-correlation coefficient value is the level of free groundwater that averages <1 meter so that the springs can still produce water throughout the year, even with small discharge. The catchment areas have a high ability to absorb rain to fill up the aquifers as a source of water. The catchment area depends on land use, slope, rainfall, soil type, and groundwater potential (Ministry of Environment and Forestry 2017, Gunawan et al. 2016).
A spatial model shows the exceed of water utilization. Several factors are identified, such as fault movement, shrinking water catchment coverage area (8.05-68.77 ha), shrubs and grasses dominated land cover or loss of trees.
The texture and depth of plant roots determine groundwater storage capacity. The deeper the roots are, the more water is stored in the soil (Sari & Prijono 2019). The soil's adequate depth determines the groundwater storage capacity in a land-use system, the distribution of soil micropore spaces, and the balanced soil particle size between clay and sand particles. Groundwater content values and soil physical properties influence groundwater reserves. Solum thickness affects water storage capacity (Rahayu et al. 2009, Sari & Prijono 2019.
Based on the bad-disturbed performance value of modeling, thus, could be analyzed that performance   improvement and mitigation are needed through forest and land conservation, enrichment planting of trees, wise land management, and proper water utilization. Soil permeability in the secondary forest is faster than in mixed upland, monocultures, and paddy fields. Soil permeability in monocultures is higher than in mixed upland. The ability of forest areas to absorb water is faster than the rice field or even a mixed upland area. This finding is supported by Lopes et al. (2019), that the forest is a groundwater reservoir and provides a wide range of ecosystem services. It fasts in permeability and relatively fast on mixed land, has medium permeability on dry land and yard, and is slower on degraded land. Forest area consists of numerous trees and diverse upland areas provided by the better ability on water permeability. It makes   (Lisnawati & Wibowo 2010, Naharuddin et al. 2018.

CONCLUSION
From the results of spatial-statistical modeling, it is known that the condition of the springs in producing water after the tectonic earthquake in the Palu and Poboya watersheds has a "poor-disturbed" performance. This performance was caused by the use of water sources at each location of the springs that had exceeded its water production capacity, the movement of faults/faults, the water catchment area was not wide (8.05 -68.77 ha), vegetation cover dominated by shrubs and grasses and a few forest trees. Considering these conditions, it is necessary to increase the recharge area and improve evapotranspiration through the enrichment of forest and annual plants, as well as protection around springs within a radius of 200 m from various land clearing activities. To maintain the performance of springs in areas prone to tectonic earthquakes with low-moderate aquifer productivity and the presence of deep-free groundwater, it is important to conduct scientific research related to the hydrogeological aspects of the area, which is integrated with trials of developing plant species that have high water storage capacity.