Effects of Land Cover, Evapotranspiration, and Rainfall on Total Runoff in the Gumbasa River Basin, Central Sulawesi, Indonesia

Land use patterns in various forms and ways will have an impact on the environment. Indications of a decrease in the carrying capacity of the environment in a region can be seen from various disasters that occur such as floods, landslides, and droughts. Therefore, changes in land use from nonbuilt land to build land will stimulate the amount of runoff water. This study aims to determine the effect of changes in land use, evapotranspiration, and rain on total runoff. This research was carried out in the Gumbasa Watershed. Data is needed in the form of primary data (soil characteristics) and secondary data (rain, climatology, a map of the earth and a map of land use). The rainfall data used are Bora, Palolo, Wuasa, Kulawi and Bangga Bawah stations with observations for 14 years (2002-2015). Climatology data was used by Bora Station with observations for 14 years (2002-2015). The analysis was carried out namely: regional rainfall, evapotranspiration, MockWyn-UB model and linear regression. The linear line equation obtained: Y = 97.325 0.747X1 + 0.697 X2 + 0.442 X4 + 0.447 X5. This equation has been tested statistically and fulfilled the linear regression requirements. Based on the above equation, it can be said that: 1). Forest land is an excluded variable so that it is not in the equation. 2). The smaller the evapotranspiration and the greater the rainfall and the area of mixed gardens and open land, the greater the runoff, and vice versa.

Fuad Halim (2014) [2] suggests that watersheds are natural ecosystems that are bounded by ridges. Rainwater that falls in the area will flow to rivers which eventually lead to the sea or to the lake. In the Watershed, two regions are known, namely the water supply area (upstream) and the water receiving area (downstream area). These two regions are interconnected and affect the watershed ecosystem units. The function of the Watershed is as a catchment area, water storage, and distribution water.
Land use patterns in various forms and ways will have an impact on the environment. Indications of a decrease in the carrying capacity of the environment in a region can be seen from various disasters that occur such as floods, landslides, and droughts. The occurrence of flooding is basically triggered by two main things, namely (1) the less land that functions as water absorption. (2) land subsidence due to groundwater exploitation and physical development that exceeds carrying capacity [3]. Therefore changes in land use from non-built land to built land will stimulate the amount of runaway water [4]. Wahyunto (2004) [5] suggested that the impact of changes in land use from green space to built-up areas would affect the ability of water absorption by the land, and the quality of water along the watershed to cause flooding.

C. Model MockWyn-UB
Total river water runoff is calculated by the MockWyn-UB equation model [6,22,24], which is the sum of the base flow with direct runoff. Base Flow (BF) = I + ΔVn

D. Model Regression
To get the constant and correlation coefficient value of land use, evapotranspiration, rain and total runoff, multiple regression equations are used, namely: Y = a + b X 1 + c X 2 + d X 3 + e X 4 + f X 5 (6) Where: a = Constant b-f = Coefficient value for each variable Y = Total runoff, mm / year X 1 = Evapotranspiration, mm / year X 2 = Annual rainfall, mm / year X 3 = Forest area, Ha X 4 = Mixed garden area, Ha X 5 = Open land area, Ha

A. Location of Research
The research location is located in the Gumbasa River Basin (Sigi Regency and Poso District, Central Sulawesi Province, Indonesia). Geographically located at 01 0 01'-01 0 21' South Latitude and 119 0 56'-120 0 19' East Longitude. Research locations can be seen in the following map:  Table I.

Evapotranspiration
Evapotranspiration is calculated by the Penman Montieth equation, where the input is in the form of monthly climatology data (air humidity, air temperature, solar radiation time and wind speed) while the output is in the form of monthly evapotranspiration data.

Land Use
The input of the MockWyn-UB model is in the form of land use every year, while the available data is insufficient. Therefore, a regression analysis was conducted based on existing land use data so that land use data were obtained as follows:

A. Model MockWyn-UB
Based on the MockWyn-UB model, the total runoff per month (mm/month) and total annual runoff (mm/year) are as follows:  TRO To find out the good equation obtained can be known from the reflected coefficient. The closer to 100%, the better the equation will be. The coefficient of determination is shown in the adjusted R square column, which is equal to 91.1%, meaning that the equation obtained is able to predict total runoff by 91.1%, while the remaining 8.9% is obtained from other variables not examined. The independence requirement is that the independent variable has a strong correlation with the residue that can be known by looking at the value of Durbin-Watson (DW). Independence requirements are met if the DW value is close to two. In the model summary, the DW value is 2.179 meaning that the independence requirements are met b. Output Coefficient In the output coefficient (Table VI) information is obtained about variables related to total runoff, the equations obtained as well as testing the requirements for multicollinearity. Multicollinearity requirements can be tested with collinearity statistics. The tolerance value that approaches 1 indicates the absence of multicollinearity between independent variables. In the excluded variables (Table VII), it is found that the variable coefficient of forest land is very large (8.951E2) but has a tolerance value of almost zero (2.782E-9) so that the total runoff equation is ignored.  (Table VIII), information about the residual requirement can be obtained as zero. Based on Table VIII, the mean of the residue is equal to zero, meaning that the residue does not play a role in the equation obtained.  (Figure 2), normality conditions can be known, namely, the residue is normally distributed. Figure 2 shows that the residual has a normal distribution In the scatter plot output (Figure 3), linearity requirements can be known. Based on Figure 3 it appears that linearity requirements are met, meaning that the correlation between the dependent variables and free variables is linear.    (6) some conclusions can be drawn, namely: 1). Forest land (X 3 ), the variable coefficient is very large but the correlation value is almost zero so it is an excluded variable. 2). Evapotranspiration (X 1 ) as a deduction, meaning that the greater the evapotranspiration, the lower the total runoff. 3). Rain (X 2 ) as the increasing variable of total runoff, meaning that the greater the rain, the greater the total runoff. 4). Variables of mixed garden land (X 4 ) and open land (X 5 ) as increasing variables, meaning that the wider the mixed garden and open land, the greater the total runoff.

V. CONCLUSION
The equation model resulting from multiple regression analysis between variables of land use, rainfall, and evapotranspiration with total runoff is: Y= 97,325-0.747 X 1 +0.697 X 2 +0.442 X 4 +0.447 X 5 This equation has been tested statistically and fulfills linear regression requirements such as independent and dependent scatter variables around the diagonal line, histogram normally distributed, mean residue value zero, DW value approaching 2, constant variant scatter residue does not have a certain pattern and variable tolerance value independent approaching one. Based on the above equation, it can be said that: 1). Forest land is an excluded variable so that it is not in the equation. 2). The smaller the evapotranspiration and the greater the rain and the wider the area of mixed garden and open land, the greater the runoff, and vice versa.