1. Data methods
This first version of exposure assessment prototype provides a framework for exploring the exposure of physical assets and population to two types of natural hazards: heat and flooding.
This page describes the different datasets and methods used for computing amenity and population exposure indicators to those two hazards.
The data workflow used for assessing amenities' exposure to heat and flooding hazard is composed of two main steps:
- Extracting the necessary data from global databases based on cities' boundaries. The collected datasets are stored in AWS buckets.
- Computing exposure indicators using city-wide collected data and storing the outputs tables to be queried by the exposure dashboard.
The table below describes the different datasets for computing amenities' exposure indicators:
| Dataset | Provider | Format | Data description | Temporal coverage | Spatial resolution | Uses |
|---|---|---|---|---|---|---|
| Administrative boundaries | geoBoundaries or Open Street Map depending on data availability | geojson |
Polygons of cities' boundaries at different administrative levels | NA | Depending on the city | The boundaries are used for extracting amenities and hazards data at the city levels. |
| Amenities | Open Street Map using overpass API | geojson |
Locations and descriptions of physical assets. | NA | NA | Physical assets are extracted for each city. Exposure indicators are computed based on the hazards levels at the amenities' locations. |
| Amenities sectors | Global Covenant of Mayors' Common Reporting Framework | csv |
Sectors categorization of urban amenities | NA | NA | Exposure indicators are aggregated at the sector level to identify the most exposed sectors within the cities. |
| Land Surface Temperature | Landsat using Google Earth Engine | GeoTiff |
Estimation of average land surface temperature during hottest months based on Landsat 4,5,7 and 8 computed within Google Earth Engine. | 2013-present | 30m | Land Surface Temperature values are used as proxies for estimating heat value and computing population and assets, exposure to extreme heat hazard. |
| Flooding | WRI-Aqueduct | GeoTiff |
Modeled inundation depths | historical, 2030, 2050, 2080 | 30" × 30" | Estimating assets' and population exposure to flood hazard |
| Population | WorldPop | GeoTiff |
Estimated population by age and gender groups | 2020 | 100m | Estimating population exposure to flood and heat hazard. |
Two data sources are used for collecting the administrative boundaries regarding the selected cities: the geoBoundaries database and OpenStreetMap depending on data availability at the defied administrative level. While the geoBoundaries database provide a structured API for collecting national and subnational boundaries at various levels, it doesn't include all sub-national levels. OpenStreetMap boundaries have been used for collecting those of Vitacura and Makati.
Amenity data is provided by Open Street Map and collected by the mean of the Overpass API. OpenStreetMap's free tagging system allows the map to include an unlimited number of attributes describing each feature. These tags are represented in key: value structure. The key, is used to describe a topic, category, or type of feature (e.g., highway or name). The value provides detail for the key-specified feature. For example, school features can be tagged with key = amenity and value = school.
For each city, we collected the location and attributes of available amenities situated within the city boundary using two feature keys: key = amenity and key = building.
| city | collected amenity data |
|---|---|
| Hobart (Australia) | Hobart amenities.geojson |
| Vitacura (Chilie) | Vitacura amenities.geojson |
| Makati (Philipines) | Makati amenities.geojson |
In order to aggregate exposure indicators by sectors, each amenity is associated with a sector name based on the sector taxonomy provided by the Global Convenant of Mayers Common Reporting Framework. The amenity sectors mapping table between OSM taxonomy and GCoM framework is provided in this link.
Land Surface Temperature estimation within the city boundary is based on Landsat data provided with high spatial resolution (30m). In order to collect this data, we used an open source library provided by Sofia L. Ermida based on the Google Earth Engine (Link to the artilce). We extracted for each city an average heat value for a 3 months period corresponding to the hot season based on the last 4 years. A Google Earth Engine (This google earth engine script) generates the three months hottest period based on NEX-GDDP data.
Here is an example of parameters used for collecting LST data for the city of Marroco:
var satellite = 'L8';
var date_start = '2018-03-01';
var date_end = '2021-04-30';
var month_start = 6;
var month_end = 9;
var use_ndvi = true;The extraction of Land Surface Temperature data by city is based on a Google Earth Engine script (collect_heat_city.js). It requires two google earth engine dependencies based on the work of Ermida et al. on estimating LST from Landsat series: Landsat_LST.js and cloudmask.js.
Floods are considered as one of the most dangerous and costly natural hazards with important social and physical impacts. Due to economic growth and rapid urbanization, more people and physical assets are located in flood prone areas and exposed then to flooding risk. Reducing social and physical impacts of floods and building effective flood risk management and adaptation strategies require accurate and high resolution flood risk data.
Aqueduct Floods is an online platform that measures riverine and coastal flood risks under current baseline conditions and future projections in 2030, 2050, and 2080. The hazard is represented by inundation maps showing flood extent and depth for floods of several return periods (2, 5, 10, 25, 100, 250, 500, and 1000 years) at a resolution of 5 * 5 arc minutes. This equates to roughly 10 km * 10 km pixels at the equator.
In Aqueduct Floods, flood risk is calculated for 2010, 2030, 2050, and 2080. Future changes in climate and socio-economic conditions are represented using three scenarios by combining representative concentration pathways (RCP) and shared socioeconomic pathways (SSP).
Aqueduct Floods include two hazard components, riverine and costal floods. Both hazards are represented by global scale layers of inundation depth at 30" * 30" resolution, with different layers representing inundation depths for different annual average probabilities of occurrence.
The raw data is available in this web page as raster files. To know more about the methodology used for producing these hazards maps, please read this technical note.
The population data at the city level is collected from **WorldPop **database which provides high resolution estimation of population spatial distribution. Full details on the methods and datasets used in constructing the WorldPop data, along with open access publications, are provided on the WorldPop website. In brief, recent census-based population counts matched to their associated administrative units are disaggregated to ~100x100m grid cells through machine learning approaches that exploit the relationships between population densities and a range of geospatial covariate layers.
For each pilot city, we extracted the raster file corresponding to population data within the city boundary. This dataset contains breakdown of estimated population by age and gender groups. These socio-demographic groups are used in order to measure the exposure of vulnerable population: elderly (age > 65 y.o), children (age < 18 y.o), and women.
The collected data is stored in AWS S3 buckets and is publicly available to download:
| topic | city | data download link |
|---|---|---|
| Amenity | Hobart | Hobart_amenity.geojson |
| Amenity | Makati | Makati_amenity.geojson |
| Amenity | Vitacura | Vitacura_amenity.geojson |
| Land Surface Temperature (LST) | Hobart | Hobart_LST.geotiff |
| Land Surface Temperature (LST) | Makati | Makati_LST.geotiff |
| Land Surface Temperature (LST) | Vitacura | Vitacura_LST.geotiff |
| Population | Hobart | Hobart_pop.geotiff, Hobart_pop_children.geotiff, Hobart_pop_elderly.geotiff, Hobart_pop_women.geotiff |
| Population | Makati | Makati_pop.geotiff, Makati_pop_children.geotiff, Makati_pop_elderly.geotiff, Makati_pop_women.geotiff |
| Population | Vitacura | Vitacura_pop.geotiff, Makati_pop_children.geotiff, Makati_pop_elderly.geotiff, Makati_pop_women.geotiff |
| Flooding | Vitacura | Vitacura_riverine_rp001000_1980.geotiff |
| Flooding | Makati | Makati_riverine_rp001000_1980.geotiff |
| Flooding | Hobart | Hobart_riverine_rp001000_1980.geotiff |
Amenity exposure to heat () may be quantified simply as the heat value (as collected from Land Surface Temperature data) at the amenity location. Since our goal consists of identifying the most exposed amenities within each city, we need to define an indicator that may inform us about within city variability of exposed amenities.
One of the challenges with heat exposure assessment methods consists of defining the appropriate heat threshold () to use, that can be adapted to various geographical contexts. Since we are attempting to provide an interactive tool providing users the possibility to interact with the collected data, we propose to consider the heat threshold as a flexible variable that the user may change through a dashboard interface. Hence, the user may explore the impact of different heat thresholds on the amenity exposure assessment. By default, the (
) is set as the average heat value of all amenities located within the city boundary.
Amenity exposure indicator is proposed at three levels: individual amenity level, sector level, and city level.
| Level | Indicator name | Indicator description | Formula |
|---|---|---|---|
| Individual amenity | Amenity exposure to heat ( |
Amenity exposure to heat ( |
|
| Sectors | Sector exposure to heat | Sectors' exposure to heat ( |
|
| Sectors | Percent of exposed amenities by sector | We calculate fir each sector the number of amenities with exposure value ( |
|
| City | Amenity average heat | The average heat value of amenities located within the city boundary |
In order to estimate amenities' exposure to flooding, inundation depths values are collected based on amenities' locations. Similarly to heat exposure, exposure assessment is based on a user defined threshold that can be parametrized depending on users' requirements. Amenities are then classified into two categories:
- Exposed amenities: where inundation depth at the amenity location is higher than the inundation depth threshold
- Not exposed amenities: where inundation depth at the amenity location is lower than the inundation depth threshold
Amenity exposure to flooding is also aggregated at the sector level by calculating the percent of exposed amenities by sector for each city.
Population exposure to heat is computed at the population layer resolution from WorldPop database (100m*100m). As a factor of both population density and heat value, population exposure indicator is defined by weighting the population count values by the heat values.
- A first step consists of generating a binary raster layer to identify areas that are exposed to heat depending on the user defined heat threshold (
).
- The obtained binary exposure raster layer is multiplied by the population count to generate a population exposure level layer. Hence, areas with heat values lower than the defined threshold are characterized by null exposure level.
- The population exposure layer is classified into 3 exposure categories:
- Low exposure areas: cells with null population exposure levels (heat value lower than selected threshold)
- Moderate exposure areas: cells with population exposure levels lower than median population exposure value within the city
- High exposure areas: cells with population exposure levels higher than median population exposure value within the city
The estimation of population exposure to flooding is computed by calculating the number of persons living in area exposed to flooding hazards. Similarly to population exposure to heat, a threshold value corresponding to inundation depth (in meters) is defined by the user to determine if the area is considered or not exposed to high flooding risk. As shown in the figure below, a composite layer is derived by weighting population count distribution layer with the flooding exposure layer.
