![2. Materials and Methods The range of values of the CF is [-1, 1]. -1 indicates a maximum disfavoring effect; +1 show the strongest causative link between the class considered and waste dumping; A value close to 0 means that the prior probability is very similar to the conditional one, so it is not possible to give any indication about the certainty of the proposition.](https://image.slidesharecdn.com/integrating-remote-sensing-spatial-analysis-and-certainty-factor-model-for-waste-dumping-risk-assessment-110729170829-phpapp01/85/INTEGRATING-REMOTE-SENSING-SPATIAL-ANALYSIS-AND-CERTAINTY-FACTOR-MODEL-FOR-WASTE-DUMPING-RISK-ASSESSMENT-ppt-16-320.jpg)
INTEGRATING REMOTE SENSING, SPATIAL ANALYSIS AND CERTAINTY FACTOR MODEL FOR WASTE DUMPING RISK ASSESSMENT.ppt
- 1. Integrating Remote Sensing, Spatial Analysis And Certainty Factor Model For Waste Dumping Risk Assessment Yi-Shiang Shiu a , Meng-Lung Lin b , Yi-Chieh Chen c , Shien-Ta Fan c , Chao-Hsiung Huang c , Tzu-How Chu a a Department of Geography, National Taiwan University, Taiwan b Department of Tourism, Aletheia University, Taiwan c Spatial Information Research Center, College of Science, National Taiwan University, Taiwan IEEE IGARSS 2011 VANCOUVER, CANADA
- 2. Outline 1. Introduction 2. Materials and Methods 3. Results and Discussion 4. Conclusions
- 3. 1. Introduction Waste dumping is one of the main pollution causing land deterioration and resource depletion. With the growing awareness of environmental issues, environmental protection authorities in many countries have also implemented various waste controlling measures to deter waste dumping (Nemerow and Agardy, 2008; Tam and Le, 2009) Inspectors’ routine patrols are necessary to increase the deterrence. However, these tasks are also manpower and time-consuming.
- 4. 1. Introduction Construction waste dumping Mixed construction and industrial waste dumping Industrial and household waste dumping
- 5. 1. Introduction Risk assessment could be helpful to save the environmental protection resource. Risk assessment has been widely used to investigate the probability of risk occurrence and evaluate risk level for hazard prevention (Zhang et al, 2009). Decision support system is often applied to help managers make the best decision in risk assessment Artificial intelligence (AI) is frequently used as the core of decision support system. The AI-based applications of risk assessment include Flood risk assessment (Tian et al, 2010) Surface coal mine safety analysis (Lilic et al, 2010) Environmental impact assessment of transportation facilities (Liu and Yu, 2009)
- 6. 1. Introduction Certainty factor (CF) can be used to evaluate the reliability of the rules induced from the decision support system (Sinha and Zhao, 2008). Rules evaluation with CF could be helpful for the safety assessment of industrial equipment (Kumar et al, 2009), level crossing surveillance (Tao and Lin, 2008) and thermal power plant (Zhang et al, 2009). Geographic information system (GIS)-based risk assessment with CF model can be beneficial for landslide susceptibility mapping (Binaghi et al, 1998; Lan et al, 2004; Aboye, 2009)
- 7. 1. Introduction To save the environmental protection resource, risk assessment integrating remote sensing and GIS could be helpful to predict and map potential waste dumping area. This study proposed waste dumping risk assessment based on certainty factor model Spatial analysis was used to generate spatial factors relative to waste dumping. Remotely sensed imagery was used to map real-time waste dumping distribution and help update the waste dumping risk map. 45 waste dumping and 45 no waste dumping cases are used to validate our result.
- 8. Outline 1. Introduction 2. Materials and Methods 3. Results and Discussion 4. Conclusions
- 9. 2. Materials and Methods GIS data and spatial analysis This study used 8 factors to accomplish waste dumping risk assessment. Based on the interviews and questionnaires with the inspectors in Environmental Protection Bureau (EPB) from 7 counties in Taiwan, we generalized the first 7 factors relative to waste dumping.
- 10. 2. Materials and Methods GIS data and spatial analysis The 7 factors are distances to: Dikes Rivers Idle lands Factories Roads Sea Residential/commercial areas The distances in the 7 factors were computed using Spatial Analyst module and then reclassified into 5 classes (i.e. very close, close, moderate, far and very far).
- 11. 2. Materials and Methods Remote sensing data and hybrid classification The 8th factor is the suspected waste dumping area mapped with FORMOSAT-2 imagery and hybrid classification. General Specification of FORMOSAT-2 Launch Year 2004 PAN 0.45~0.90μm MS 0.45~0.52μm （ Blue ） 0.52~0.60μm （ Green ） 0.63~0.69μm （ Red ） 0.76~0.90μm （ Near Infrared ） Remote Sensing Ground Resolution PAN （ Black/white ） Image 2 meters MS (color) Image 8 meters Image Swatch 24 kilometers
- 12. 2. Materials and Methods Remote sensing data and hybrid classification Suspected waste dumping mapping is not an easy task because waste dumping usually consists of various materials which show high spectral heterogeneity in satellite imagery. To overcome high spectral heterogeneity and overlap, hybrid of supervised and unsupervised classification could be helpful (Turner and Congalton, 1998). Unsupervised clustering is useful to stratify input images and cluster the manually collected training data into spectrally homogeneous subclasses for the use in the subsequent supervised classification (Bauer et al, 1994; Stuckens et al, 2000; Tømmervik et al, 2003; Lo and Choi, 2004). Hybrid approaches could be helpful to overcome difficulties in delineating appropriate training samples for complex study areas such as mountainous (Kuemmerle et al, 2006) or waste dumping areas.
- 13. 2. Materials and Methods
- 14. 2. Materials and Methods CF model The CF model was used to combine all 8 factors and generate waste dumping risk map. The CF at each pixel is defined as the change in certainty that a proposition is true (i.e. an area is waste dumping prone) from without the evidence to given the evidence at each pixel for each factor (Binaghi et al, 1998): ‘ No evidence’ means prior probability of having waste dumping cases in the study area ‘ With evidence’ means the conditional probability of having waste dumping cases given a certain class of a causative factor.
- 15. 2. Materials and Methods CF was calculated with the equation : where CF ij is certainty factor given to class i of factor j ; f ij is the waste dumping density within the class i of factor j ; f is the waste dumping density within the entire map.
- 16. 2. Materials and Methods The range of values of the CF is [-1, 1]. -1 indicates a maximum disfavoring effect; +1 show the strongest causative link between the class considered and waste dumping; A value close to 0 means that the prior probability is very similar to the conditional one, so it is not possible to give any indication about the certainty of the proposition.
- 17. 2. Materials and Methods The combination of all CF layers could be the basis for the waste dumping risk assessment. The CF layers were then combined pairwise. The combination of two CF’s, x and y , due to two different layers of information, is expressed as z , given by (Binaghi et al, 1998):
- 18. Outline 1. Introduction 2. Materials and Methods 3. Results and Discussion 4. Conclusions
- 19. Dike Rivers Residential /commercial areas Idle lands Factories Roads Sea
- 20. Suspected waste dumping area mapped with FORMOSAT-2 imagery and hybrid classification.
- 21. 3. Results and Discussion The result of image classification: Commission error: 32% Omission error: 34% Overall accuracy: 70% Validating the result with the 45 waste dumping and 45 no waste dumping cases, CF model predicted 75.56% of the waste dumping cases in the very high potential area.
- 22. 3. Results and Discussion Waste dumping risk map generated with CF model.
- 23. 3. Results and Discussion This study validated the proposed waste dumping risk assessment with 45 waste dumping cases in 7 counties in Taiwan. Because the spectral characteristic of waste dumping is easily confused with bare soil, there would be a lot of errors when only using satellite imagery. Combining satellite imagery and other GIS data with CF model can help delineate the area with the highest waste dumping potential level.
- 24. Outline 1. Introduction 2. Materials and Methods 3. Results and Discussion 4. Conclusions
- 25. 4. Conclusions Integrating remote sensing, spatial analysis and CF model for waste dumping risk assessment could be reliable. Potential problems The selection of the method to reclassified the distances of 7 factors is subjective; The limit of multispectral satellite imagery. Future applications include planning routes for environmental protection inspectors and helping inspectors to concentrate the patrol areas, which could save manpower significantly.
- 26. Thank you for your attention!!