Data Preparation Land Use Change Model and Significant Variables

61 change in the research site. The natural environment included 8 data layers, the human environment included 16 data layers, and policy included 4 data layers. Table 13. Data Layers Used in This Research Data Layer Unit Period of Data Source Dependent Variable Land Use Change 2002 – 2005 chgn0205 categorical 2002 - 2005 Land Use Change Detection Independent Variables - Natural Environment Distance from existing Forest land frstdist meters 2002, 2005, 2008 Land Use Classification Distance from existing Cropland cropdist meters 2002, 2005, 2008 Land Use Classification Distance from existing Grassland grasdist meters 2002, 2005, 2008 Land Use Classification Distance from existing Wetlands wetdist meters 2002, 2005, 2008 Land Use Classification Distance from existing Other lands othedist meters 2002, 2005, 2008 Land Use Classification Distance from existing River rivdist meters 2002, 2005, 2008 Land Use Classification Slope slope percents 2002 - 2008 SRTM-DEM 90m Altitude alt meters 2002 - 2008 SRTM-DEM 90m Independent Variables - Human Environment Continued to the next table Distance from animal husbandry animdist meters 2002 - 2008 District Government Distance from economic infrastructure econdist meters 2002 - 2008 District Government Distance from education infrastructure educdist meters 2002 - 2008 District Government Distance from environment infrastructure envidist meters 2002 - 2008 District Government Distance from fishery infrastructure fishdist meters 2002 - 2008 District Government Distance from government service infrastructure govdist meters 2002 - 2008 District Government Distance from health service infrastructure healtdist meters 2002 - 2008 District Government Distance from industry inddist meters 2002 - 2008 District Government Distance from mining area mindist meters 2002 - 2008 District Government Distance from road roaddist meters 2002 - 2008 District Government Distance from existing settlement settledist meters 2002, 2005, 2008 Land Use Classification Distance from telecommunication infrastructure teledist meters 2002 - 2008 District Government Distance from transmigration area transmdist meters 2002 - 2008 District Government Distance from transportation infrastructure transpdist meters 2002 - 2008 District Government Distance from public space pubspadist meters 2002 - 2008 District Government Population density at sub district level popdens personskm2 2002, 2005, 2008 District Government Continued to next page 62 Table 13. Data Layers Used in This Research Continue Data Layer Unit Period of Data Source Independent Variables - Policy Sub district Area sdisarea hectares 2008 District Government Forestry Spatial Plan at National Level kwsid categorical 2000 - 2008 National Government Spatial Plan at Province Level rtrwpid categorical 2000 - 2008 Province Government Forestry and Crop Spatial Plan at District Level concessid categorical 2002 - 2008 District Government Note: the period of data which are separated with coma , mean the data are available for each year, whereas the period of data which are separated with dash - mean the data are available as single data for its period. The land use change 2002 – 2005 has 26 categories regarding to 26 land use change transitions which happen in Siak District. Each land use transition has been coded into integer number 1 – 26 which expressed its unique land use transition as illustrated in Table 8 and Figure 17 of this report. The natural environment theme included distance from existing Forest Land, Cropland, Grassland, Wetlands, Other lands, and river which have been produced by applying the Euclidean distance measurement into each appropriate layer, and altitude and slope which have been derived from SRTM-DEM 90m data. Distance from Forest land frstdist Distance from Cropland cropdist Distance from Grassland grasdist Distance from Wetland wetdist Figure 24. Data Layers of Independent Variables: Natural Environment Theme 63 Distance from Other lands othedist Distance from River rivdist Altitude alt Slope slope Figure 24. Data Layers of Independent Variables: Natural Environment Theme Continue The human environment theme included distance from animal husbandry, economic infrastructure, education infrastructure, environment infrastructure, fishery infrastructure, government services infrastructure, health services infrastructure, industry, mining area, road, settlement, telecommunication infrastructure, transmigration area, transportation infrastructure, and public space, which have been also produced by applying the Euclidean distance from each layer, and population density at sub district level. Distance from animal husbandry animdist Distance from economic infrastructure econdist Distance from education infrastructure educdist Distance from environment infrastructure envidist Figure 25. Data Layers of Independent Variables: Human Environment Theme 64 Distance from fishery infrastructure fishdist Distance from government service govdist Distance from health service healtdist Distance from industry inddist Distance from mining area mindist Distance from road roaddist Distance from settlement settledist Distance from telecommunication infrastructure teledist Distance from transmigration area transmdist Distance from transportation infrastructure transpdist Distance from public space pubspadist Population density at sub district level popdens Figure 25. Data Layers of Independent Variables: Human Environment Theme Continue 65 The policy theme included sub districts area in Siak District and spatial plan of Siak District in three administrative levels that are Forestry Spatial Plan at National Level, Spatial Plan at Province Level, and Forestry and Crop Spatial Plan at District Level. In national level, Siak District is divided into 10 categories, whereas in province and district level is divided into 17 categories and 4 categories respectively. The description of the spatial plan category for each level is shown in Appendix 2. Sub district Area sdisarea Forestry Spatial Plan at National Level kwsid Spatial Plan at Province Level rtrwpid Forestry and Crop Spatial Plan at District Level concessid Figure 26. Data Layers of Independent Variables: Policy Theme After the spatial data layers for MLR modeling have been prepared, the next to be prepared was the dataset of dependent and independent variables in the form of attribute table which would be analyzed in statistical MLR model computation. The dataset of dependent and independent variables for MLR modeling have been taken from sampling point data of data layer of dependent variable which is land use change 2002 – 2005. The sampling point data has been generated by using weighted probability distribution which is provided in Hawths Tools. Each land use transition would have different number of sampling points which are associated with the values in the raster and the spatial distribution of each land use transition. The number of sampling points for each land use transition should exceed 30 points in order to fulfill the minimum requirement of sampling data in statistics computation. 66 Based on the experiment done several times for generating the sampling points, the number of sampling points which could meet the criteria of minimum 30 points for each land use transition was 10,000 points. With 10,000 sampling points, the lowest number of points that could be possessed by a land use transition was 31 points, and the highest was 1,999 points. The spatial distribution of sampling points for each land use transitions can be seen in Figure 27. Table 14. Number of Sampling Points for Each Land Use Transition Land Use Transition ID Number of Pixels Number of Points Land Use Transition ID Number of Pixels Number of Points 1 3,143,519 524 14 30,101 146 2 562,561 175 15 70,126 186 3 510,340 201 16 143,071 358 4 9,397 37 17 126 40 5 285,611 251 18 957 54 6 262,706 294 19 2,216 55 7 1,877,971 1,999 20 84,064 252 8 719,211 932 21 336 46 9 25,249 154 22 38,065 166 10 130,234 334 23 162,525 548 11 117,739 283 24 90,966 256 12 617,651 1,156 25 7,666 31 13 727,910 1,449 26 25,464 73 Continued to next table Total 9,645,782 10,000 Figure 27. Spatial Distribution of Sampling Points Furthermore, all spatial data layers which have been prepared in ERDAS Imagine raster data were spatially joined into the sampling point data which has 67 been generated in order to produce the dataset of dependent and independent variables. After the spatial join have been done to each data layer, the attribute table of sampling point data would contain land use transition IDs and its related independent variables. The attribute table of spatially joined sampling point data would be analyzed in MLR model analysis in SPSS software in order to determine the significant variables driving factors of land use change and also to find the adequate model of land use change in Siak District. In this research, the land use change model has been developed in two scenarios: 1 using all significant variables determined by the MLR model analysis and 2 using observed variables determined by the observation of existing condition in the field. Hopefully, these two scenarios may facilitate the understanding in developing land use change model using MLR model.

4.4.2 MLR Model using All Significant Variables 1

st Scenario The dataset of dependent and independent variables for MLR model analysis has been created by taking samples from each data layers which have been prepared. Each sampling point contained land use transition ID as categorical data of dependent variable and its relevant driving factors of land use change as independent variables. Then, the attribute table of sampling point data which contained the categories of dependent variable and its related values of independent variables were analyzed by using MLR model in order to determine the significant variables and produce the parameter estimates for each land use transition. The method of MLR model used in this research was Forward Stepwise which is provided in SPSS. This method is a stepwise procedure for selection or deletion of variables from a model based on a statistical algorithm that checks the importance of variables. This method is convenient for developing model which the importance of variables for its model has not been determined. MLR model analysis done has produced some important outputs, such as likelihood ratio tests for each independent variable, model fitting information, pseudo R-square, and parameter estimates. These outputs would determine the significant variables the driving factors of land use change which included on the model, determine whether the model has been developed as an adequate model 68 or not, and produce the parameter estimates coefficients of the parameters for each land use transition.

1.4.2.1 Significance Test for Independent Variables

The number of independent variables which have been included in the MLR model analysis was 28 variables which came from three themes: natural environment, human environment, and policy. Here, in the MLR model analysis the independent variables have been treated to have the same nature and it means there has been no variable which has stronger or weaker effect to the response category dependent variable, and this consideration has also been applied for each theme included in the analysis. In the MLR model analysis, the Forward Stepwise method would make selection or deletion of variables from a model by considering the Likelihood Ratio Test in each iterative process. The likelihood ratio tests check the contribution of each effect to the model. The significance level of the test should be less than 0.05 Sig.0.05 for each variable, then the effect contributes to the model, whereas variables with significance level more than 0.05 Sig.0.05 will be eliminated. Based on the likelihood ratio tests for each independent variable done in MLR model analysis, there are 24 variables from total 28 variables which are considered as significant variables of land use change in Siak District. Natural environment contributes 6 variables, human environment contributes 15 variables, and policy contributes 3 variables to the model. Otherwise, the variables are not included to the model are altitude, slope, distance from health service, and the area of sub district. The consideration of the variables which have been excluded from the model were the value of its variables did not follow the nature of logistic function of dependent and independent variables which has been applied in MLR model. In other word, the variables could not explain the land use change as dependent variable in term of MLR model used in this research. For instance, the landscape characteristics of Siak District, with altitude from 0 – 108 meters and slope from 0 – 19.9 relatively flat, and those characteristics spread evenly, could not explained the land use change which happen in Siak District, because each land use transition might have diverse landscape characteristics that could not follow