36th Asian Conference on Remote Sensing 2015 (ACRS 2015), Quezon City, Philippines, Volume 3, pp. 2037-2050, 2015
3D BUILDING GIS DATABASE GENERATION FROM LIDAR DATA AND FREE ONLINE WEB MAPS AND ITS APPLICATION FOR FLOOD HAZARD EXPOSURE ASSESSMENT Jojene R. Santillan, Meriam Makinano-Santillan, Linbert C. Cutamora, Jesiree L. Serviano CSU Phil-LiDAR 1, College of Engineering and Information Technology, Caraga State University, Ampayon, Butuan City, Philippines E-mail:
[email protected]
KEY WORDS: 3D buildings, database, LiDAR, flood, exposure, vulnerability ABSTRACT: Identifying buildings that are exposed and vulnerable to flooding is important in flood disaster preparedness, risk assessment, and mitigation. In most cases, the availability of a 3D building database where each building is attributed in terms of name, type (e.g., residential, commercial, government, educational, etc.), and height (among many other attributes) makes the required analysis fast, efficient and informative. In this paper, we highlight the usefulness of a 3D building database generated from LiDAR data and free online web maps in assessing the exposure and vulnerability of buildings to flooding through a case study made for Cabadbaran River Basin, Mindanao, Philippines. The results of this case study consist of a series of maps and statistics showing exposure and vulnerability of buildings to flooding that can be utilized by Local Government Units and the communities in the river basin in their flood disaster risk reduction and management strategies. 1.
INTRODUCTION:
Flooding is one of the most destructive natural disasters in the Philippines. It can cause loss of lives and damages to properties as well as to infrastructures like buildings, roads and bridges. Excessive quantity of rainfall brought by tropical storms is the most common cause of flooding, just like what happened in Metro Manila during the passing of Tropical Storm Ondoy (Cheng, 2009), and in various provinces in Mindanao when Tropical Storms Agaton and Seniang caused rivers to overflows (NDRRMC, 2014; 2015). Although the frequency of flood-related disasters has grown in recent years, the tools to model and understand flood risks have also increased in number, had become more sophisticated, and are readily available for use. Examples of these are flood models that allow hydrodynamic scenario simulation of flood water propagation, as well as in the assessment of flood damage (Vojinovic and Tutulic, 2009). The hazard maps produced by these flood models are used as important inputs for assessing the exposure and vulnerability of localities to various flood scenarios, usually conducted using Geographic Information System (GIS) tools and techniques (Fedeski and Gwilliam, 2007). An important part of this GIS-based assessment is the identification of the type, location, and height of buildings that are exposed to various scenarios and levels of flood hazards. One practical application is to identify in advance the specific buildings and households that can be affected by an expected flooding scenario in order to prevent casualties through evacuation, or to lessen the impact through conduct of flood mitigation activities. In this situation, the availability of a 3D building database where each building is attributed in terms of name, type (e.g., residential, commercial, government, educational, etc.), and height (among many other attributes) can make the required hazard and vulnerability assessment fast, efficient and informative. LiDAR-derived products such as Digital Surface Models (DSMs) have been widely utilized to map earth features such as buildings, roads and bridges (Priestnall et al., 2000; Wang and Schenk, 2000). The availability of height information provided by LiDAR data makes it very suitable in generating a 3D building database. Several approaches have been developed in extracting 3D building information from LiDAR data (Rottensteiner, 2003; Sohn and Dowman, 2007). Building information such as geometry, position and height are primarily generated. Using LiDAR data, buildings may be distinguished from vegetation objects by evaluating shape measures, or the heights of the features (Wang and Schenk, 2000). In a LiDAR DSM, buildings can be considered as detached objects rising vertically on all sides with a minimum height above the bare earth. In terms of geometry, buildings footprints are bounded by distinct edges of regular shapes relative to natural objects. Most building surfaces can be approximated by simple geometric shapes (i.e., triangles, squares, and rectangles). One limitation of using purely LiDAR data such as a DSM in building extraction is the difficulty to correctly delineate the shape of the building due to the presence of nearby vegetation that covers some parts of the building. In this case, the use of high resolution images is a useful supplementary dataset to aid or complete the extraction process. Another limitation is the inability to determine the type of the building, and this sometimes requires field inspection or the use of secondary spatial datasets such as maps.
2037
36th Asian Conference on Remote Sensing 2015 (ACRS 2015), Quezon City, Philippines, Volume 3, pp. 2037-2050, 2015
In this paper, we generated a 3D GIS database of buildings in Cabadbaran River Basin, Mindanao, Philippines through analysis of various datasets that included 1-m resolution LiDAR DSM and Digital Terrain Model (DTM), high resolution images in Google Earth, and free online web maps such as Wikimapia. The use of Google Earth images and Wikimapia to address the limitation of using the DTM and DSM during the building type classification. We also present in this paper how the 3D buildings database was used as exposure datasets for the flood hazard assessment of the river basin that included assessment of the vulnerability of the buildings to the flooding. 2.
THE STUDY AREA
Cabadbaran River Basin (Figure 1) is located in Agusan del Norte, Mindanao, Philippines. It covers a major portion of Figure 1. The Cabadbaran River Basin, Agusan del Norte, Philippines, shown here in a Digital Surface Model. Cabadbaran City. It has an approximate drainage area of 215 km2. The Cabadbaran river basin is one of the areas affected by Tropical Storms „Agaton‟ and „Seniang‟ last January 2014 and December 2014, respectively. Flooding due the heavy rains brought by these tropical storms caused widespread damages in agriculture and infrastructures within the river basin, especially in Cabadbaran City. The presence of buildings and the occurrence of flooding make the river basin an ideal case study area for integrated 3D building database generation and flood hazard assessment.
3.
MATERIALS AND METHODS
3.1 Datasets Used We used the 1-meter resolution LiDAR-derived Digital Surface Model (DSM) and Digital Terrain Model (DTM) for extracting the building features within the river basin (Figure 2). The LiDAR DSM and DTM were acquired and processed by Data Pre-Processing Component of the University of the Philippines – Diliman Phil-LiDAR 1 project. The two DEMs were provided in ESRI GRID format with Universal Transverse Mercator (UTM) Zone 51 North projection and the World Geodetic System (WGS) 1984 as horizontal reference. Both DEMs have the Mean Sea Level (MSL) as vertical datum. For areas covered with dense vegetation, Google Earth images were utilized to improve the precision in extracting the buildings. These high resolution satellite images were also used to re-check and compare the building polygons extracted from the LiDAR DSM. Free online web maps such as Wikimapia (http://wikimapia.org/) and Google Map (https://www.google.com.ph/maps) (Figure 3) were used to gather information such as name and type of the buildings within the river basin. We used the provided table from UP-Diliman Phil-LiDAR 1 project containing a summary of different types of buildings with corresponding codes for the building type attribute. Flood depth maps generated by Caraga State University through the CSU Phil-LiDAR 1 project were used as input in flood hazard and building vulnerability assessment. These flood depth maps represent maximum depth of flooding due to rainfall events with varying intensity and duration (i.e., return periods of 2, 5, 10, 25, 50 and 100year). These flood depth maps were transformed into flood hazard maps by categorizing the flood depths into hazard levels as follows: low (1.5 m depth). These maps are shown in Figure 4 and Figure 5.
2038
36th Asian Conference on Remote Sensing 2015 (ACRS 2015), Quezon City, Philippines, Volume 3, pp. 2037-2050, 2015
Figure 2. The LiDAR DTM and DSM for a portion of Cabadbaran River Basin.
Figure 3. Screenshots of the online web maps: Google map (left) and Wikimapia (right) that were utilized as references for building attribution.
2039
36th Asian Conference on Remote Sensing 2015 (ACRS 2015), Quezon City, Philippines, Volume 3, pp. 2037-2050, 2015
Figure 4. Flood hazard maps of Cabadbaran river basin for 2-, 5-, & 10-year rainfall return periods.
2040
36th Asian Conference on Remote Sensing 2015 (ACRS 2015), Quezon City, Philippines, Volume 3, pp. 2037-2050, 2015
Figure 5. Flood hazard maps of Cabadbaran river basin for 25-, 50-, & 100-year rainfall return periods.
2041
36th Asian Conference on Remote Sensing 2015 (ACRS 2015), Quezon City, Philippines, Volume 3, pp. 2037-2050, 2015
3.2 Building Extraction from LiDAR data Buildings were manually digitized from the LiDAR DSM using ArcGIS 10.1 software (Figure 6). The footprints of the buildings were traced using the polygon feature type. In addition, we also used the high resolution images from Google Earth as reference for checking and comparing the results of the extracted features. Since some buildings were covered by dense vegetation and their shapes were indistinguishable in the DSM, we examined its corresponding Google Earth image to identify the shape of the feature. The extracted buildings were saved as a GIS Shapefiles. 3.3 Height Estimation Attribution
and
Feature
The heights of the buildings were calculated using the average of the base elevations and top elevations for each footprint (Figure 7). The base and top elevation values were extracted from the DTM and DSM, respectively.
Figure 6. The manually digitized buildings (in blue color) in Cabadbaran river basin overlaid in the Digital Surface Model.
The extracted buildings were attributed using the data obtained from the Wikimapia and Google Map. Information such as the type (Table 1) and building names (for nonresidential types) were obtained from these sources.
Figure 7. Building height estimation using the LiDAR DSM and DTM.
Table 1. Summary of the types of the building with corresponding codes. Building Type Residential School Market/Prominent Stores Agricultural & Agro-Industrial Medical Institution Barangay Hall Military Institution Sports Center/Gymnasium/Covered Court Telecommunication Facilities Transport Terminal (Road, Rail, Air, and Marine) Warehouse Power Plant/Substation NGO/CSO Offices Police Station Water Supply/Sewerage Religious Institution Bank Factory Gas Station Fire Station Other Government Offices Other Commercial Establishments
Code RS SC MK AG MD BH ML SP TC TR WH PP NG PO WT RL BN FC GS FR OG OC
2042
36th Asian Conference on Remote Sensing 2015 (ACRS 2015), Quezon City, Philippines, Volume 3, pp. 2037-2050, 2015
3.4 Flood Hazard and Building’s Vulnerability Exposure Assessment We conducted GIS overlay analysis of the building footprints and 3D flood hazard maps of the basin to identify which buildings are exposed to various levels of flood hazards (e.g., low, medium, high). In addition to this, we also characterized the degree of flood exposure of buildings by comparing their heights with the simulated flood depths in determining the vulnerability of the buildings in the river basin. If a building is located in a location where flood depth is less than 0.10 m, then it is coded as “Not vulnerable”. If the flood depth at the building‟s location is 0.1 m to less than 0.25*h (h is building height), then the vulnerability is “Low”. On the other hand, if the flood depth is > 0.25h and ≤ 0.5h, then the vulnerability is medium. If the flood depth is > 0.5h , then the vulnerability is high. 4.
RESULTS AND DISCUSSION
4.1 Buildings Database Figure 8 shows a snapshot of the buildings extracted from the LiDAR DSM. There were a total of 9,086 identified buildings. Figure 9 provides a summary of the results in attribution of the extracted buildings. Statistics shows there were 8,605 residential buildings identified, and it comprises 95% of the total buildings within the river basin.
Number of Building
Figure 8. The extracted buildings in Cabadbaran River Basin classified according to type.
10000 9000 8000 7000 6000 5000 4000 3000 2000 1000 0
8,605
6
4
22
1
12
13
4
172
28
2
13
179
18
Building Type
Figure 9. The number of buildings in Cabadbaran River Basin according to type.
2043
4
3
36th Asian Conference on Remote Sensing 2015 (ACRS 2015), Quezon City, Philippines, Volume 3, pp. 2037-2050, 2015
4.2 Exposure of Buildings to Flood Hazards Figure 10 shows the number of buildings under various hazard levels of flooding due to rainfall events with varying return periods. The locations of these exposed buildings are shown in Figures 11-12. Statistics show that as the rainfall return period increases (which also means increase in rainfall intensity and duration), the number of buildings affected by flooding also increases. Consequently, the number of buildings that are not flooded decreases. In all rainfall scenarios considered, majority of the buildings appears to be not flooded. For flood-affected buildings, more buildings are exposed to „low‟ flood hazard levels than those in „medium‟ and „high‟ hazard levels. This result means that majority of areas within the river basin where buildings are located are relatively not prone to flooding; and if there is flooding, the level of hazard is low.
Buildings Exposed to Different Flood Hazard Levels 7,000 6,422
6,000
2-year 5-year 10-year 25-year 50-year 100-year
5,880 5,460
5,000
4,988 4,720
No. of Buildings
4,407
4,000 3,190 3,241 3,225 2,961 2,735
3,000
2,314
2,000 1,248
1,000 334
451
632
791
959
166 206 16 20 33 117
Not Flooded
Low
Medium
High
Flood Hazard Level
Figure 10. Number of buildings in Cabadbaran River Basin exposed to various hazard levels due to flooding caused by rainfall events of different return periods.
4.3 Vulnerability of Buildings to Flood Hazards Figure 13 shows the number of buildings in Cabadbaran City proper that were categorized according to their vulnerability to flooding caused by rainfall events of different return periods. The locations of these vulnerable buildings are shown in Figures 14-15. It can be observed that majority of the buildings are not vulnerable to flooding, especially for flooding due to rainfall events of 2, 5, 10, and 25 return periods. For flood affected buildings, more buildings are in „low‟ vulnerabilities, with increasing number as the rainfall return period increases. The generated statistics also show that buildings exposed to medium and high flood hazard levels does necessarily mean they will also have medium and low vulnerability. Looking at the graphs (Figure 10 and 13), the total number of buildings under medium and high hazard exposure are higher than the total number of buildings in medium and high vulnerabilities. This implies that even if a building‟s location has medium or high flood hazard levels, a building‟s vulnerability can be lesser if its height is much higher than the depth of flooding. All of these results, however, only used height as basis for assessing a building‟s vulnerability. The type of building material and other factors were not considered.
2044
36th Asian Conference on Remote Sensing 2015 (ACRS 2015), Quezon City, Philippines, Volume 3, pp. 2037-2050, 2015
Figure 11. Flood hazard exposure levels of buildings in Cabadbaran City proper for a 2-, 5-, & 10-year rain return period flood event.
2045
36th Asian Conference on Remote Sensing 2015 (ACRS 2015), Quezon City, Philippines, Volume 3, pp. 2037-2050, 2015
Figure 12. Flood hazard exposure levels of buildings in Cabadbaran City proper for a 25-, 50-, & 100-year rain return period flood event.
2046
36th Asian Conference on Remote Sensing 2015 (ACRS 2015), Quezon City, Philippines, Volume 3, pp. 2037-2050, 2015
7,000
Vulnerability of Buildings to Flooding Caused by Rainfall Events of Different Return Periods 6,658
2-year
6,142
6,000
5-year
5,708
10-year
5,230
No. of Buildings
5,000
4,929
25-year
4,610
50-year 3,963 3,745 3,527
4,000
100-year
3,170 2,817
3,000
2,334
2,000
1,000 301 370 182 256 79 107
15
20
26
73 111 143
Not Vulnerable
Low
Medium
High
Flood Vulnerability Type
Figure 13. The number of buildings in Cabadbaran River Basin vulnerable to flooding caused by rainfall events of varying return periods.
2047
36th Asian Conference on Remote Sensing 2015 (ACRS 2015), Quezon City, Philippines, Volume 3, pp. 2037-2050, 2015
Figure 14. Vulnerability of buildings in Cabadbaran City proper for a 2-, 5-, & 10-year rain return period flood event.
2048
36th Asian Conference on Remote Sensing 2015 (ACRS 2015), Quezon City, Philippines, Volume 3, pp. 2037-2050, 2015
Figure 15. Vulnerability of buildings in Cabadbaran City proper for a 25-, 50-, & 100-year rain return period flood event.
2049
36th Asian Conference on Remote Sensing 2015 (ACRS 2015), Quezon City, Philippines, Volume 3, pp. 2037-2050, 2015
CONCLUSIONS AND FUTURE WORKS In this paper we showed that a 3D building database extracted from LiDAR data (with each building attributed in terms of type and height) can be a valuable dataset in assessing the exposure and vulnerability of buildings to flooding. The results of this case study made for Cabadbaran River Basin highlights the use of this database to generate statistics as well as in creating maps that can show the spatial distribution of buildings exposed to low, medium and high hazard levels of flooding caused by rainfall events with return periods of 2, 5, 10, 25, 50 and 100-year. The height information derived for each building also allowed fast generation of statistics and maps showing the building‟s vulnerability to flooding. Although the vulnerability assessment was purely based on the building height, the information generated from the analysis can be very useful in flood disaster preparedness and mitigation. One practical application would be identifying those buildings (and informing their occupants) that can be of danger when a particular rainfall event of specific return period is expected to occur. Since the maps and statistics of those buildings exposed and vulnerable to flooding were already generated according to rainfall return period, it is already easy to identify those locations and conduct appropriate measures such as early evacuation to prevent casualties. The next phase of this study is to expand the analysis where the type and material of buildings will be considered in the exposure and vulnerability assessment. Also, we will improve and consider physical basis in computing a building‟s vulnerability level. Currently, the vulnerability levels adopted in the analysis were only based on building height and flood depth, and the criteria used were initially assumed by the authors due to absence of proper reference material during the time when this study was conducted.
ACKNOWLEDGEMENT This work is an output of the Caraga State University (CSU) Phil-LiDAR 1 project under the “Phil-LiDAR 1. Hazard Mapping of the Philippines using LiDAR” program funded by the Department of Science and Technology (DOST). The SAR DEM and the LiDAR DTM and DSM used in this work were provided by the University of the Philippines Disaster Risk and Exposure for Mitigation (UP DREAM)/Phil-LIDAR 1 Program. REFERENCES Cheng, M.H., 2009. Natural disasters highlight gaps in preparedness. The Lancet, 374( 969), pp. 1317-1318. Fedeski, M., Gwilliam, J., 2007. Urban sustainability in the presence of flood and geological hazards: The development of a GIS-based vulnerability and risk assessment methodology. Landscape and Urban Planning, Vol. 83, No. 1, pp. 50-61. NDRRMC, 2014. NDRRMC Updates Sitrep No. 33 re: Effects of Tropical Depression Agaton. National Disaster Risk Reduction and Management Council. Retrieved February 1, 2014 from http://www.ndrrmc.gov.ph/. NDRRMC, 2015. SitRep No. 22 re Effects of Tropical Storm SENIANG". National Disaster Risk Reduction and Management Council. Retrieved January 10, 2015 from http://www.ndrrmc.gov.ph/. Priestnall, G., Jaafar, J., Duncan, A., 2000. Extracting urban features from LiDAR digital surface models. Computers, Environment and Urban Systems, Vol. 24, No. 2, pp. 65-78. Rottensteiner, F. , 2003. Automatic generation of high-quality building models from lidar data. Computer Graphics and Applications, IEEE, Vol. 23, No. 6, pp. 42-50. Sohn, G., Dowman, I., 2007. Data fusion of high-resolution satellite imagery and LiDAR data for automatic building extraction. ISPRS Journal of Photogrammetry and Remote Sensing, Vol. 62, No. 1, pp. 43-63. Vojinovic, Z., Tutulic, D., 2009. On the use of 1D and coupled 1D-2D modelling approaches for assessment of flood damage in urban areas. Urban Water Journal, Vol. 6, No. 3, pp. 183-199. Wang, Z., Schenk, T., 2000. Building extraction and reconstruction from LiDAR data, International Archives of Photogrammetry and Remote Sensing, Vol. 33, No. B3, pp. 958-964.
2050
