• No results found

Methodologies for Cataloging Attribute and Damage Information in GIS

CHAPTER 9: CONCLUSIONS AND RECOMMENDATIONS

9.4 Methodologies for Cataloging Attribute and Damage Information in GIS

Framework

The third objective of this dissertation was to develop a methodology to collect and catalog building attribute and damage information in a GIS framework. Chapter 4 presented

177

guidelines for systematic collection of field data for robust analysis and the state-of-the-art in damage assessment using remote sensing imagery, GIS datasets, GPS and other advanced technologies. The advantages of incorporation of remote sensing damage assessments were discussed and practices employed in remote sensing damage detection for earthquakes, windstorms and inundation flooding were reviewed, with presentation of state-of-the-art damage assessment techniques for high velocity flood events given in Chapter 5. The contribution of the many constraints affecting damage assessment study area was discussed and the utility of storing and managing data in a GIS framework was demonstrated through specific examples of manually- and automatically-classified building attributes.

Specific case study examples were given for Hurricanes Katrina and Ike and sources of publicly available pre- and post-event GIS and remote sensing data for these two storms were provided. Rapid assessments were completed for each of the case studies through the collection of GPS-synchronized high definition video with the VIEWS™ platform. The benefits of both rapid and traditional damage assessments were discussed and the Hurricane Ike case study incorporated both types of data collection through the implementation of an individual building survey form tailored to match assessment characteristics of the initial WF Damage Scale. Building attribute and damage information was recorded for the two case studies, yielding a detailed building database of almost 2,000 residential buildings. The benefits of a database management system were demonstrated through the extraction of pertinent attribute and damage information for each storm that yielded meaningful information highlighting the vast differences in construction and performance of buildings for the two study areas.

An overview of the physical hazard environments in Hurricanes Katrina and Ike indicated that simple comparison of building performance for the two storms was not appropriate. The effort spent in collecting, cataloging and managing this data, however, has yielded a dataset with

178

significant applications to advancing our understanding of the interaction of storm surge with the built environment. Chapters 6 and 7 presented remote sensing damage assessment methodologies that would have little meaning without field collected damage information. Further, the data has vast applications in the future development of the storm surge building damage model presented in Chapter 8. The two very different hazard environments and building inventories will ensure development of assessment and modeling methodologies that can be better applied to multiple hurricane events.

9.5 Literature Review of Remote Sensing Methodologies for Assessing Damage to

Buildings from High Velocity Flood Events

As discussed in Chapter 4, remote sensing damage detection methodologies are applicable only for the specific damage signatures for which they were developed. The majority of remote sensing assessments for flood events were found to focus on a determination of the extents and depth of flooding. To quantify flood effects on buildings, these methodologies further utilize standard depth-damage or depth-loss functions such as those presented in Chapter 2 to calculate flood damage or loss from the assessed depth. Because the damage mechanisms associated with high velocity flooding are significantly different from those associated with simple inundation flooding, a review of existing damage assessment methodologies for high velocity flood events was conducted. The review was organized using the concept of a spatially- tiered damage assessment and the techniques were presented at the regional, neighborhood and per-building levels.

Regional damage assessment methodologies consisted of automated processes that determined the general areas of impact from high velocity flood events. The majority of the regional assessment techniques reviewed utilized multi-temporal pixel-based methods to identify areas that exhibited a change in spectral return between pre- and post-event imagery. All of the

179

regional events reviewed reported good results in identifying areas impacted by tsunami or hurricane storm surge events.

Neighborhood level damage assessment techniques were reviewed for the 2004 South Asian Tsunami and 2005 Hurricanes Katrina and Rita. Neighborhood level analyses more precisely identify the impacted areas or provide a measure of damage severity through either neighborhood-level characterization of damage or identification of salient indicators from remote sensing imagery that define locations of severe damage. For the South Asian Tsunami, the primary neighborhood level damage indicator was the scoured land surface, defining areas most severely impacted by the tsunami. For Hurricanes Katrina and Rita, the debris line left by destroyed buildings as storm surge receded was identified as the most prominent indicator of areas of increased damage. In the case of Hurricane Rita, FEMA found that the extents of storm surge damage were limited by the location of the debris line for the majority of sites investigated that had a debris line.

Review of per-building level damage assessment methodologies revealed that neither a consistent goal in the number of damage categories differentiated nor a standardized approach for damage assessment has been established. Per-building assessments currently employed detect either collapse/non-collapse conditions or define more intermediate states of damage. Object-based visual assessment of collapse/non-collapse generally yielded good results. Studies that attempted to define more intermediate states of damage were also able to detect collapse categories, but could not accurately detect building damage for lower damage states, especially when the roof was undamaged.