Quantitative estimation of building damages is extremely important after a disastrous event. An integrated remote sensing and GIS-based analysis approach can provide timely information on building damages. This preliminary analysis can help both national and international humanitarian actors to coordinate and plan aid response as well as need assessment operations. Remote sensing-based building damage assessment is generally performed using qualitative (e.g. photointerpretation) or quantitative methods (e.g. image classification). Pre- and post-event satellite images once acquired can be analysed and interpreted by expert analysts able to categorise assessed building according to different damage classes.
Tropical Cyclone Harold 20 made landfall on Luganville in Vanuatu on the 7th of April 2020, causing extensive damage to buildings and infrastructure. Government decision-makers at National Disaster Management Office need damage statistics quickly to address emergency response needs.
In this exercise you are going to use a Very High-Resolution Satellite image taken on the 9th of April 2020 and Compare Against an Image from 07 September 2019, to detect the damages.
By going through this exercise, you will learn to apply GIS methodologies and tools to:
|File Name||Data Type||Source|
|Raster\Lug_07Sep2019_WV2_R1C1.tif||Raster (Post-Disaster)||World View 2, Maxar|
|Raster\Lug_09Apr2020_WV2_R1C1.tif||Raster (Pre-Disaster)||World View 2, Maxar|
A Geodatabase is a data storage and management framework developed by ESRI. It is a relational database containing spatial and non-spatial objects. The primary elements of a Geodatabase are feature classes, rasters and tables. A Feature Dataset is a container for feature classes (point, line, and polygon) that are thematically related and share the same coordinate system and geographic extent. Geodatabases offer many advantages for GIS analysis such as: centralized data storage, support for advanced feature geometry, and more accurate data entry and editingthrough the definition and use of attribute domains.
For this exercise, you will create a personal geodatabase which will contain a point feature class corresponding to those building you will be assessing the damage level. To improve the overall quality of your dataset and to make the data entry process faster during the assessment you will create a number of different domain values to define a set of “legal attributes” which you will use to populate your geodatabase.
Attribute domains are rules that describe the legal values of a field type (a field will not accept a value that is not in that domain). Using domains helps ensure data integrity by limiting the choice of values for a field. To create new domain values:
The “New Feature Dataset Wizard” will ask you to select a coordinate system to set the spatial reference for the “Damage_Features” feature dataset:
We are going to create two attribute fields for capturing Damage Class and Analysis Confidence. Then both fields must be linked to the domains created in the previous section.
Now you have a Geodatabase for performing damage analysis.
Why did we set the Damage field data type as short integer?
Damage assessment maps are primary based on satellite analysis using pre- and post-disaster imageries. Photo interpretation methods are generally used to identify damages from imagery. Damage assessment maps aim to provide an overall view of affected areas, including a rapid assessment of the extent of damage. These maps are very useful for planning, coordination and setting priorities for both disaster relief and need assessment activities. A detailed inventory of all areas affected by disasters may require large-scale satellite derived mapping products for which high and very high-resolution imagery becomes necessary.
For this exercise, you will perform building damage assessment using a pre-defined Area of Interest (AOI) with Worldview 2 satellite image acquired after the TC Harold on the 9th of April 2020 and reference data from same satellite senor captured on the 07 September 2019.
This is the data layer (empty feature class) you will be editing during the damage assessment.
What is the spatial resolution of pre and post disaster images?
The above step will be super useful when we start digitising the buildings.
A “Create Features” window will be visible where you have options to digitise both Damaged and Not damaged buildings
You will notice the “correct symbols” you used in the create feature window have appeared in the building damage attribute table.
Now you can continue the above process to digitise all the damaged and not damaged buildings in the AOI. To do a fast assessment you can digitise all buildings from pre disaster imagery and add damage attributes in bulk.
Sometimes during post disaster damage assessments, we end up with points or polygons of different buildings, however this kind dataset cannot convey the information of damage distribution. It is a common practice to prepare damage distribution surface by interpolation or kernel density technique. Both methods have their limitations and it is up to the analyst to select appropriate technique.
In the first part of the exercise, you are going to prepare a damage intensity surface. In the second part, you will be making your own thematic map using damage intensity surface. The main challenge of this exercise is to figure what information you would like to convey with your map. We will also use symbology and labelling to convey this information in a clear, coherent, and efficient manner. The final output will be a pdf map that will be presented and discussed.
The new “Damage_Z” attribute field will be used to obtain new integer values corresponding to the different damage level classes:
– No visible Damage (Damage_ID) = 0 (Damage_Z)
– Possible Damage (Damage_ID) = 10 (Damage_Z)
Use “Field Calculator” to assign values listed above for “Damage_Z”
Note that “Damage_Z” are values that are assigned to carry more importance towards damaged buildings and can be changed to match a specific damage scale as needed.
Where would you say is the highest concentration of destroyed buildings based on the interpolation surface?
Try running kernel density function to demonstrate building damage intensity. Which result do you prefer & why?