Analysis of structural-damage predictions caused by an earthquake to identify areas with high damage levels

1. A method comprising:

identifying features associated with buildings within a region, the features comprising, at least, an age of building, a material of building, and number of storeys of the building;

training, by one or more processors, a machine learning program that learns from the identified features and earthquake damage data to generate a model for predicting damage caused by an earthquake based on an amount of shaking caused by the earthquake, the training including earthquake damage data for a plurality of earthquakes and values for the identified features, the training appraising relationships between the features and the damage caused by the plurality of earthquakes;

estimating, by the one or more processors using the model, a block damage value, from a plurality of damage values, for blocks in the region caused by the earthquake;

identifying, by the one or more processors, a critical damage area (CDA) within the region, the CDA comprising a plurality of blocks geographically clustered having the block damage value above a threshold damage value from the plurality of damage values; and

causing, by the one or more processors, presentation of the CDA within a map of the region.

2. The method as recited in claim 1 , further comprising:

determining parameters for identifying the CDA, the parameters comprising one or more of minimum size of the CDA, maximum size of the CDA, maximum number of CDAs within the region, and maximum distance between blocks in the CDA.

3. The method as recited in claim 2 , wherein identifying the CDA further comprises:

identifying blocks in the region with block damage values above the threshold damage value;

selecting a plurality of blocks that meet the parameters for identifying the CDA; and

creating the CDA with the selected plurality of blocks.

4. The method as recited in claim 1 , further comprising:

analyzing values of the features for buildings in the CDA and buildings in an area next to the CDA;

determining which features show a significant difference between the values of the features for the buildings in the CDA and the buildings in the area next to the CDA; and

presenting the determined features in a user interface as factors that explain the block damage values for blocks in the CDA.

5. The method as recited in claim 4 , wherein determining which features show a significant difference comprises:

analyzing each of a plurality of features used by the machine learning program, wherein analyzing each feature further comprises:

calculating one or more statistical values for the values of the feature in the buildings in the CDA;

calculating one or more of the statistical values for the values of the feature in buildings in the area next to the CDA; and

identifying the feature as showing the significant difference based on a comparison between the calculated statistical values for the buildings in the CDA and the buildings in the area next to the CDA.

6. The method as recited in claim 5 , further comprising:

sorting the features showing a significant difference based on an estimation of a contribution of each feature for explaining damage differences.

7. The method as recited in claim 1 , wherein presentation of the CDA further comprises:

presenting an area occupied by the CDA in the map of the region with a color different from any color in a rest of the map.

8. The method as recited in claim 1 , wherein presentation of the CDA further comprises:

presenting an area occupied by the CDA in the map of the region surrounded by a line type different from other lines in a rest of the map.

9. The method as recited in claim 1 , wherein the features are classified as built environment features, natural environment features, and shaking features.

10. The method as recited in claim 1 , wherein the features include one or more of type of structure, amount of shaking, type of soil, structural parameters for a building, population density, damage state, elevation, and spectral displacement.

11. A system comprising:

a memory comprising instructions; and

one or more computer processors, wherein the instructions, when executed by the one or more computer processors, cause the one or more computer processors to perform operations comprising:

identifying features associated with buildings within a region, the features comprising, at least, an age of building, a material of building, and number of storeys of the building;

training a machine learning program that learns from the identified features and earthquake damage data to generate a model for predicting damage caused by an earthquake based on an amount of shaking caused by the earthquake, the training including earthquake damage data for a plurality of earthquakes and values for the identified features, the training appraising relationships between the features and the damage caused by the plurality of earthquakes;

estimating, using the model, a block damage value, from a plurality of damage values, for blocks in the region caused by the earthquake;

identifying a critical damage area (CDA) within the region, the CDA comprising a plurality of blocks geographically clustered having the block damage value above a threshold damage value from the plurality of damage values; and

causing presentation of the CDA within a map of the region.

12. The system as recited in claim 11 , wherein the instructions further cause the one or more computer processors to perform operations comprising:

determining parameters for identifying the CDA, the parameters comprising one or more of minimum size of the CDA, maximum size of the CDA, maximum number of CDAs within the region, and maximum distance between blocks in the CDA.

13. The system as recited in claim 12 , wherein identifying the CDA further comprises:

identifying blocks in the region with block damage values above the threshold damage value;

selecting a plurality of blocks that meet the parameters for identifying the CDA; and

creating the CDA with the selected plurality of blocks.

14. The system as recited in claim 11 , wherein the instructions further cause the one or more computer processors to perform operations comprising:

analyzing values of the features for buildings in the CDA and buildings in an area next to the CDA;

determining which features show a significant difference between the values of the features for the buildings in the CDA and the buildings in the area next to the CDA; and

presenting the determined features in a user interface as factors that explain the block damage values for blocks in the CDA.

15. The system as recited in claim 14 , wherein determining which features show a significant difference comprises:

analyzing each of a plurality of features used by the machine learning program, wherein analyzing each feature further comprises:

calculating one or more statistical values for the values of the feature in the buildings in the CDA;

calculating one or more of the statistical values for the values of the feature in the buildings in the area next to the CDA; and

identifying the feature as showing the significant difference based on a comparison between the calculated statistical values for the buildings in the CDA and the buildings in the area next to the CDA.

16. A non-transitory machine-readable storage medium including instructions that, when executed by a machine, cause the machine to perform operations comprising:

identifying features associated with buildings within a region, the features comprising, at least, an age of building, a material of building, and number of storeys of the building;

training a machine learning program that learns from the identified features and earthquake damage data to generate a model for predicting damage caused by an earthquake based on an amount of shaking caused by the earthquake, the training including earthquake damage data for a plurality of earthquakes and values for the identified features, the training appraising relationships between the features and the damage caused by the plurality of earthquakes;

estimating, using the model, a block damage value, from a plurality of damage values, for blocks in the region caused by the earthquake;

identifying a critical damage area (CDA) within the region, the CDA comprising a plurality of blocks geographically clustered having the block damage value above a threshold damage value from the plurality of damage values; and

causing presentation of the CDA within a map of the region.

17. The machine-readable storage medium as recited in claim 16 , wherein the machine further performs operations comprising:

determining parameters for identifying the CDA, the parameters comprising one or more of minimum size of the CDA, maximum size of the CDA, maximum number of CDAs within the region, and maximum distance between blocks in the CDA.

18. The machine-readable storage medium as recited in claim 17 , wherein identifying the CDA further comprises:

identifying blocks in the region with block damage values above the threshold damage value;

selecting a plurality of blocks that meet the parameters for identifying the CDA; and

creating the CDA with the selected plurality of blocks.

19. The machine-readable storage medium as recited in claim 16 , wherein the machine further performs operations comprising:

analyzing values of the features for buildings in the CDA and buildings in an area next to the CDA;

determining which features show a significant difference between the values of the features for the buildings in the CDA and the buildings in the area next to the CDA; and

presenting the determined features in a user interface as factors that explain the block damage values for blocks in the CDA.

20. The machine-readable storage medium as recited in claim 19 , wherein determining which features show a significant difference comprises:

analyzing each of a plurality of features used by the machine learning program, wherein analyzing each feature further comprises:

calculating one or more statistical values for the values of the feature in the buildings in the CDA;

calculating one or more of the statistical values for the values of the feature in the buildings in the area next to the CDA; and

identifying the feature as showing the significant difference based on a comparison between the calculated statistical values for the buildings in the CDA and the buildings in the area next to the CDA.