Land use probability maps based on logistic regression Model Validation

37 understanding about land use behavior as well as future land use distribution is important for planners and managers to justify a spatial planning decision and avoid inappropriate decision Pontius and Neeti, 2009. Therefore, in order to provide an appropriate explanation, the critical discussion in this research will be conducted by considering the findings and its appropriateness with some literatures. According to the change detection for the land use classification year 1991, simulated year 2000, year 2010, year 2020 and 2030, it can be seen that settlement, forest and grassland area are increasing over time, whereas estate tend to decrease and water remains constant. The trends of land use changes of the area during 1991 – 2030 are presented in following figure. Figure 12. Trends of land use change based on Scenario 1. Figure 13. Trends of land use change based on Scenario 2. 38 Figure 14. Trends of land use change based on Scenario 3. Figure 15. Trends of land use change based on Scenario 4. In case estate, the trend show that estate always decreased from time to time. Some probability could be happen, such as plants that have been grown, with a high density is often identified as a forest. In other probability, in the areas that have been harvested, the land cover looks like a bare land or grassland. It is most likely to occur, due to the habit of people in Upstream Cisadane plant fast-growing plantation and harvest it as soon as possible. However, although the area of estate decrease from time to time, estates still dominates the land use in Upstream Cisadane. Settlement has increased significantly with the annual average change 0.73 per annum. This is because people need space to build residential area and other building facility in line with the increasing of population. Grassland and forest are also increased, but not too significant. 39 In case of forest, an interesting phenomena going on where in general, forest area is decrease. In Upstream Cisadane, the area of forest is increase. Most of increased areas of forest occur outside of forest area based on forest designation map, it’s meant that reforestation happen in community forest. From fact-finding in field, the awareness of the importance of forests as a life support system encourages people to do reforestation. The analysis also shown that mature plantations may be interprets as forest, especially in the area of high-density plantation. In general, the results of these scenarios indicate that the combination of statistical analysis and CLUE-S model is valuable in representing land use behavior on upstream watershed zone and it has capacity to explain the causal factors in a complete process model. All of the significant variables derived from logistic regression can be accommodated in this model and used as a preference of scenario development. The only use of significant variables in the model with known cause- effects relationship between driving factors and land use change is one of advantages of CLUE-S model Pijanowski et al, 2000. By involving significant factors, the model could avoid spurious relationship and subjective human intervention. Besides the facts above, several important issues emerge in this modeling process, including validity of the model results, data used and analysis scale. Validity is one of important issues in land use modeling because it describes the evidence to explain complex spatial behavior behind the process. Based on the validation processing steps described in sub-chapter 2.4.2.4., the result of validation is 86.00 and categorized as fit. This value ensures that the result of this study is agree with the result of some literatures, including Verburg, et al 1999 who simulates land use scenario by using CLUE-S in Ecuador and obtains an accuracy range of 71 – 90 and Li and Yeh 2002, who predicts future land use change by using integration of neural network and cellular automata and achieves 83 validity. Pontius et al 2004 states that if the driving factors are appropriate chosen, the simulation leads to a high agreement between the simulated land use and the reality. Pontius also assures that modeling validation by using Kappa is