7. 1 Conclusions
The increasing development in South Sulawesi as a gate to the east of Indonesia will increase pressure to the population and economic which will bring the finding of this study to the most relevance. Due to this, land use change as one impact of exploitation of the natural resources impacted to the landslide evident. In South Sulawesi, the land use change was many cases located in the mountainous area, like in current research in upper Ujung-loe Watersheds.
Increasing the land use change in mountainous area affects the slope stability and landslide occurrence. It is also worthy to be mentioned that landslide susceptibility study is still limited in the study area as a result of many unreported landslide cases. Due to this, current research conducted assessment on the close correlation of land use change with landslide occurrence and the performance of LUC to produce landslide susceptibility using geographic information system and multivariate qualitative prediction model
The produced landslide susceptibility map is expected to be useful for government officials and urban planner in planning the development of the region. Geographic information systems are tools in disaster risk reduction planning which is one of the main issues of Indonesia's national development agenda to promote sustainable development and reduce the frequency of disasters and environmental degradation. Regarding landslide disaster risk, reduction and disaster mitigation are well approached by landslide susceptibility, hazard, and risk.
Based on the previous chapters and the subsequent discussions the following significant conclusions are presented:
1. Significant land use changes from 2004 to 2011 observed in the Ujung-Loe watershed that experienced a decline were no vegetation and dense vegetation classes, while those that have increased are medium vegetation and high vegetation classes. Landslides have
103 occurred 128 times during 2012 to 2014, and the most frequently occurred in 2013 and is mostly dominated by the one with the land use change from high vegetation to medium vegetation. The general land use change in Ujung Loe watershed indicates significant effect to landslides occurrence and slope instability.
2. Using land use change (LUC) as a new causative factor to produce landslide susceptibility map (LSM), LUC has a good effect. The result indicated that producing LSM with LUC was better than without LUC. Performances of each landslide model were tested using AUC curve for success and predictive rate, which had the highest value of predictive rate with LUC in both frequency ratio (FR) and logistic regression (LR) method (83.4% and 85,2%, respectively) and 80.24% of landslides validation fell in the class of high to very high. These results suggested that changing the vegetation to another landscape causes slopes unstable and increases the probability of landslide occurrence. LR method is better than FR to produce LSM. LUC affects landslide susceptibility in the study area; it was observed that the change in vegetation type to another landscape destabilized slopes. Validation of landslide susceptibility was carried out by calculating the area under the curve (AUC) of receiver operating characteristic curve (ROC). Firstly, LR shows the highest accuracy in both success and predictive rate (85.6%). Secondly, the frequency of landslides in high to a very high class of susceptibility was calculated, which indicates the level of accuracy of the method. CF returns the highest accuracy of 85.28 %.
3. Using LUC have an excellent effect to produce landslide susceptibility map. When we use FR and certainty factor (CF), LUC has the highest value on both at LUC from primary forest to open area and paddy field. In logistic regression method, LUC has the effect of landslide occurrence with significant value 0.589. Besides creating landslide susceptibility maps, this research illustrates the performance of frequency ratio (FR), certainty factor (CF) and logistic regression (LR) models as well. Two-steps of
104 validation were carried out in this study. First, performances of each landslide model were tested using AUC curve for success and predictive rate, which is more than 82 % with the highest at LR Model. In the second, the ratio of landslides falling on high to a very high class of susceptibility was obtained, which indicates the level of accuracy of the model. The CF model have highest accuracy with 85.28 % landslides fall in the range of high to very high class while in LR and FR model, it is 82.11% and 81.46%.
4. Artificial Neural Network (ANN) was the best method to produce landslide susceptibility map. The best optimization of causative factors was a combination of forward stepwise logistic regression (FSLR) - ANN with nine causative factors with AUC success rate 0.847, predictive rate 0.844 and validation with landslide fall into high and very high class with 91.30%. For this reason, this is an encouraging preliminary model towards a systematic introduction of FSLR-ANN model for optimization causative factors in landslide susceptibility assessment in the mountainous area of Ujung Loe Watershed.
7. 2 Future works
Based on the present study, further improvement could be included such as:
1. Improving the landslide inventory as the base data for landslide susceptibility assessments in the study area by using remote sensing analysis. Landslide inventory is one of the key input in landslide susceptibility mapping.
2. Further studies should also employ the high accuracy of DTM, i.e., LIDAR data to obtain better accuracy of simulation and zoning in a landslide.
105
Appendices
Appendix 1. Study area in Ujung Loe Watershed, South Sulawesi, Indonesia
106 Appendix 2. Soil map in Ujung Loe Watershed, South Sulawesi, Indonesia
107 Appendix 3. Geology map in Ujung Loe Watershed, South Sulawesi, Indonesia
108 Appendix 4. Land use/land cover in 2015 at Ujung Loe Watershed, South Sulawesi, Indonesia
109 Appendix 5. Elevation map in Ujung Loe Watershed, South Sulawesi, Indonesia
110 Appendix 6. Slope map in Ujung Loe Watershed, South Sulawesi, Indonesia
111 Appendix 7. Map of Polygon Thiessen of Rainfall in Ujung Loe Watershed, South Sulawesi, Indonesia
112 Appendix 8. Map of Population Density in Ujung Loe Watershed, South Sulawesi, Indonesia
113 Appendix 9. Graph of Rainfall data period from 2002 to 2015 in Malino Rain Gauge Station
2509
4642
3263 3239
4218 4078
2096
3719
4729
3996
319
5474
3643 4048
0 1000 2000 3000 4000 5000 6000
2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015
rainfall (mm)
Year
Rainfall by Year at Malino Station
772
626
490 396
232 151 96 41 54 115
322
674
200 400 600 800 1,000
Jan Feb Mar Apr May June July August Sep Oct Nov Dec
rainfall (mm)
Month
Rainfall Monthly at Malino Station
114 Appendix 10. Graph of Rainfall data period from 2003 - 2015 in Apparang Hulu Rain Gauge Station
427
528
280
382 348 331 343 365
117
293
401
491
100 200 300 400 500 600
Jan Peb Mar Apr Mei Juni Juli Agt Sep Okt Nop Des
Rainfall (mm)
Month
Rainfall by monthly at APARANG HULU/BATUBULERANG
2242
1436
2513 2629
2227 2213
3383 3763
2976
4121
5052 4570
3310
0 1000 2000 3000 4000 5000 6000
2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015
Rainfal (mm/year)
Year
Rainfall by Year at APARANG HULU/BATU BULERANG Station
115 Appendix 11. Graph of Rainfall data period from 2002 - 2015 in Bulo-bulo Rain Gauge Station
277 289 300
355
394 417
304
105
51
100
143
292
50 100 150 200 250 300 350 400 450
Jan Peb Mar Apr Mei Juni Juli Agt Sep Okt Nop Des
Rainfall (mm/month)
Month
Rainfall by monthly at Bulo-bulo Station
2579 3123 2940
2325 2010
4251 3752
2409
5711
2717 3081
4154
3325 2237
0 1000 2000 3000 4000 5000 6000
2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015
Rainfall (mm/year)
Year
Rainfall by Year at Bulo-bulo Station
116 Appendix 12. Map of Nine Causative Factors
117 Appendix 13. Map of Eleven Causative Factors
118 Appendix 14. Landslide susceptibility map of with and without LUC causative factors using FR, and LR method with nine
causative factor in Ujung Loe Watershed, South Sulawesi, Indonesia
119 Appendix 15. Landslide susceptibility map of with eleven causative factors using FR, CF, and LR method in Ujung Loe Watershed,
South Sulawesi, Indonesia
120 Appendix 16. Landslide susceptibility maps (LSM). (a) LSM multivariate logistic on test seventh; (b) LSM artificial neural network (ANN) with eleven
causative factor
121 Appendix 17. Landslide susceptibility maps of the best Optimized causative factor using a combination of forward stepwise (likelihood ratio)
logistic regression to eliminated causative factor and artificial neural network (FSLR-ANN) with nine causative factor
122 Appendix 18. Image of survey location
123 Appendix 19. Image of survey location
124 Appendix 20. Image of survey location
125 Appendix 21. Image of survey location
