OpenStreetMap aided important post hurricane mortality study Puerto Rico

A recent study published in the New England Journal of Medicine is making the news- Mortality in Puerto Rico after Hurricane Maria

The study puts the number at around 4,600 deaths which may be attributable to the hurricanes (Irma and Maria) which damaged Puerto Rico’s infrastructure. The storms left a significant portion of the island without electricity, potable water, and communications for an extended period of time. Access to supplies and movement was hampered by landslides, damaged roads and bridges.

While there is no dispute in the sharp uptick in overall deaths in Puerto Rico in the months immediately following the storms, linking the deaths to the storm has been a contentious issue. In the days and weeks following the storm hospitals were either functioning with reduced capacity or none at all. Government resources understandably were directed to recovery efforts, so counting the dead was not a top priority. The way that the government tallies the dead does not help either as storm related deaths are counted only if they are certified as directly caused the storm – dying after getting hit by flying debris during the storm counts. Yet, an elderly diabetic person who’s insulin spoiled because of a lack of refrigeration and died because his glucose got out of whack is considered to have died of complications from diabetes. A person with a respiratory infection may have gone without early treatment because of damaged roads and died as a result of complications – died of respiratory disease not because of the storm. No communications meant no 911. Such conditions, also takes a toll on mental health, so suicide rates also increased during this period.

171205-9296-1
Months after the hurricane, certain roads were still susceptible to landslides and communities lacked electricity, running water and communications.

OpenStreetMap data aided this study in ways not possible only a few years ago. On a recent radio interview, Domingo Marqués, one of the study’s authors, said that without the map density data this study would not have been possible¹. Thanks to a worldwide push lead by the Humanitarian OpenStreet Maps Team and other contributors holding map-a-thons and working individually – sufficient map data for Puerto Rico was largely available by the time the study needed it. For this study in particular, OSM information was used for selecting the sample.

From the study:
Sampling buildings using OpenStreetMap
Households within barrios were identified using OpenStreetMap (OSM) layers for structures identified as “buildings”. For each randomly selected barrio, we iteratively downloaded structure information using the OSM overpass API, calculated centroids for structures identified as buildings, and randomly sampled 35 locations. We generated geospatial PDFs for each barrio level with an OSM base layer, a barrio boundary and the sampled building points. The geospatial PDFs were loaded on Samsung Tab A 7” Android devices and displayed using PDFMaps. Enumerators were trained to load maps, identify their position and navigate using these geospatial PDFs.

The study thanked OSM contributors and serves as another example to OpenStreetMap’s usefulness.

171021-9034
Map-a-thon’s like this one held in San Juan, Puerto Rico helped improve OpenStreetMap data

As somber as this mortality study is, it can give us hope for better responding for catastrophes in the future by understanding how these deaths occurred. Traditional hurricane preparedness centered towards seeking shelter away from areas prone to flooding. Analysis of the causes of these fatalities along with OpenStreetMap may change this thinking. A location may not be prone to flooding, but may be still vulnerable because of landslide cutting off the only access road. Medicine, Potable Water and other supplies could be pre-positioned prior to the storm’s arrival and tailored on demographic figures to better serve communities which may have a hard time evacuating. After a storm, rapid post disaster data analysis can optimize relief resources to people in need. Temporary clinics could be set quickly up after a storm in critical locations to tend people suffering from chronic diseases or respiratory disease in order to avoid any complications which could lead to death.

OpenStreetMap and the continuing contribution its volunteers in drawing and identifying buildings and other physical features will hopefully play a role in many other studies and applications in understanding this disaster and preparing for future ones worldwide.

¹ 5/30/2018 AM 810 Fuego Cruzado interview

Finding where a photo was taken using Google Vision

Scanning photos taken by grandparents on a trip to Europe in 1960 is quite fascinating. Seeing the places they visited, some which I have also had the opportunity to visit as well, is a nice way to remember them. However, the photos presented a bit of a challenge as while some had written captions mentioning the location they were taken, most lacked this information.

Modern digital cameras embed metadata into an image file with details such as the date/time taken and geolocation. These bits of information make it easier to know where a photo was taken. So with these analog photos- I have only a black and white image of an unknown place.

1960’s photos meet Artificial Intelligence

Luckly, modern image recognition techniques can identify landmarks. Google’s Vision API can identify an image’s location by using AI smarts to compare them to a vast image dataset. As more images are fed, Google Vision “learns” more about landmarks and the objects contain within the images. While the API is targeted for application developers, Google provides a web site in which you can upload an image and it returns this information as well.

Google Vision API Website
The Google Vision API provides a way of identifying landmarks from photos. In this example, the service correctly identified the photo as taken at the Buen Retiro Park in Madrid, Spain – https://cloud.google.com/vision/

In mobile devices, Google Lens also uses this technology to recognize landmarks and present relevant information.

Google Lens
Google Lens uses Google Vision to identify landmarks from images.

 

Adding location information to the image file metadata

After positively identifying the landmark and geographical location where the photo was taken using Google Vision, I turn to GeoSetter for adding the missing geolocation information as well as captions to the image file metadata. Adding this information allows other applications and services to use it as well.

 

GeoSetter
Adding geolocation information to an image file using GeoSetter – http://www.geosetter.de/en/main-en/

 

OneDrive
Applications or services as OneDrive can use geolocation data to display the location on a map. The caption is also displayed showing the name of the monument as it was identified with the help of Google’s Landmark Recognition engine.
2018-05-22 (1).png
Geotagged photo shown in Windows Photo Gallery – the added metadata of the location taken is shown as a Geotag.

Related:

Moving away from Windows Photo Gallery Geotags

Windows Photo Gallery “Geotags” was a nice feature for adding location information to photos. It is no longer supported and has caused some angst among some users. So I am writing this post to provide a way to access and maintain the geo location information in your photos.

First, a quick explainer on how geotags are saved in image files and used to work in Windows Photo Gallery (WPG) – As you may know, photo files such as JPG can store additional information called metadata. The best analogy for understanding what metadata is by imagining a printed picture with written information written on the back. There are multiple ways (formats) of writing this information known as EXIF, IPTC, XMP. Think of it as paper forms with fields (such as date, time, caption, locations) to fill in, which are added to a photo.

WindowsPhotoGallery_Geotag
A photo taken at the Seattle Central Library shown in Windows Photo Gallery. Placing the mouse over the geotag would reveal all the geotag information.

The fields which Windows Photo Gallery uses for geotags are: IPTC Core and Extension Location Fields (Location, City, State/Providence, Country) and EXIF GPS Latitude, GPS Longitude.

Here is the general logic behind Windows Photo Gallery’s Geotags (in order):

1. If a photo already contains Location, City, State/Province and Country fields filled in, WPG would use the information it has to display as the Geotag. It would display the IPTC Extension fields as the first choice (if available) and the IPTC Core “legacy” fields (if available) as the second option.

2. If a photo has GPS Latitude and Longitude, but no City, State/Province, or Country information in the fields described in (1), then WPG would consult a Microsoft web service to determine the “Geotag” based on latitude and longitude, this process is called reverse geocoding. It will not however, write the “Geotag” information back to the file, unless they are edited manually by the user.

3. If you, the user, edit a Geotag in WPG manually, it will be saved back to the file and that information becomes the “geotag”. However, here it is where it gets a bit tricky- There are multiple Location, City, State/Province, Country fields among the different metadata standards (IPTC and within XMP). WPG uses the “IPTC Extension- Location Created” fields to store the “Geotags” info within the photo.

IMG_3394
Windows Photo Gallery “Geotag” rename dialog box. When edited, the information is saved back to the file to the IPTC Extension “Location Created” fields.

Here is an example of the metadata from one of my photos. You will note multiple fields which may seem to contain redundant information, but the ones WPG would write to are just four which start with “Location Created”. These fields are detailed in the 2008 IPTC Extension Spec, further discussion on how this information is handled by certain applications can be found on the Metadata Working Group Spec.

Location Created City Seattle
Location Created Country Name United States
Location Created Province State Washington
Location Created Sublocation Seattle Central Library

Given that Microsoft decided to discontinue WPG last year, the service which determines the geotags is no longer working, so holdout users may find these fields empty or have trouble adding new geotags. Hence, exiftool, Lightroom or other photo management applications will only be able to show you geotag information if it was written within the file.

GeoSetter to the Rescue

As a workaround, I rely on a freeware application called GeoSetter (which internally uses exiftool). Geosetter uses a service from Geonames.org, in a similar fashion which WPG did, to aid in finding location info as well it allows users to edit them manually. I exclusively have been using GeoSetter for some time, which unlike WPG it will always save the information back to the photo file. GeoSetter also allows one to edit the latitude and longitude info, which I was unable to do so in WPG.

GeoSetter_Edit_Location
GeoSetter’s Edit Data dialog allows users to edit EXIF GPS as well as IPTC Location fields.

However, as good as the Geosetter is, there is a slight catch – Geosetter by default does not read nor write to the newer 2008 IPTC Extension format which Windows Photo Gallery writes to. This is not just an issue with Geosetter, other photo management tools, like XnView MP, still primarily use the IPTC Core “legacy” Location fields, instead of the newer IPTC Extension Location fields. The IPTC Photo Metadata User Guide, explains the reasoning of creating the IPTC Extension fields. Luckily, Geosetter is versatile enough to provide a means for a workaround.

To make GeoSetter read any existing WPG Geotag – You will need to copy any saved “geotag” information from the IPTC Extension, to the equivalent IPTC Core fields using an exiftool command. In a previous post I described how to use exiftool for accessing Windows Photo Gallery. You will only need to run this step once on your photos from the command line. I advise you to make a backup copy of your files before executing it.

exiftool *.jpg -"Location<LocationCreatedSubLocation" -"City<LocationCreatedCity" -"Province-State<LocationCreatedProvinceState" -"Country<LocationCreatedCountryName" -r -overwrite_original
-r means that it will execute recursevly, so that all subfolder are included.
-overwrite_original means that exiftool will not create a backup of the file.

Again, this will only work on file which WPG saved the geotag information to the file, otherwise the fields will be empty.

To configure GeoSetter write back information to the newer format when an edit is made– In the GeoSetter application menu, go to File | Settings. In the Settings dialog, go to the “ExifTool” tab. Ensure that the “Use Additional Exiftool Commands after GeoSetter command” checkbox is enabled and add this text to the text field underneath:

-execute -"LocationCreatedSubLocation<Location" -"LocationCreatedCity<City" -"LocationCreatedProvinceState<Province-State" -"LocationCreatedCountryName<Country" -"LocationCreatedCountryCode<CountryCode"

 

GeoSetter_Exiftool_Settings
GeoSetter’s Settings dialog. Adding additional exifool commands allow custom actions such as copying field values to another. In this case location information is copied to IPTC 2008 Spec fields.

 

If you are still a die hard Windows Photo Gallery fan, you will find that with these tips the geotag information populated with what you was entered by GeoSetter. While the reverse geocoding capability may be gone, you can still use WPG sort by GeoTag and Search functionality as before. Still, looking forward, I would avoid editing geotags in WPG and rely instead on GeoSetter as a replacement for this functionality in order to avoid the problems previously described.


References:

Related Posts:

​Crowd-sourced maps assist in Hurricane Maria relief efforts: How volunteers are helping put Puerto Rico on the map

As Hurricane Maria‘s winds and rain battered our home in San Juan, among the many thoughts that bounced in my head in those long hours was wondering about the people living in the mountainous regions of the island. The winding roads, heavy foliage, cliffs, bridges and terrain susceptible to landslides could make it the worst place to be in such a storm. Many small communities on those mountains would become isolated for days.

The storm left us without power, water and cellphone service. In the following days, I managed to connect to one of the few hotspots available and got an email indicating that The Humanitarian Open Streets Maps team was responding to the disaster per the request of the Red Cross. As a local OpenStreetsMap user, I new what that meant. Yet never imagined how it would take shape and how vital that information would become to the relief efforts.

OpenStreetMaps, or OSM as the name implies, is an open and publicly available geographical database which anybody can use and edit. Think of it as the Wikipedia for maps. Governments, Non-governmental organizations and companies such as AppleFoursquare and Yahoo use OSM. You may have seen OSM data in a map on your phone without even knowing it.
For some time I have tinkered as a contributor to OSM in my spare time. A small local group of volunteers have been able in getting some local government municipalities to contribute data to OSM. Most of it may be out of date so OSM volunteers review and even sometimes perform field work to ensure accuracy and quality. Local organizations such as Foundation for Puerto Rico have sponsored mapping activities to under-served communities. Initiatives like these have led to good PR road information in OSM. While larger cities may have good building descriptions, rural areas especially the mountain regions in Puerto Rico may have very little geographical information available.

When a major disaster strikes and a large scale response effort needs to be executed knowing where those building structures are located is essential. That’s when the Humanitarian OSM team can quickly turn the focus of mapping volunteers to analyse satellite imagery on a location in need. To my surprise, universities stateside sprang into action organizing Map-A-Thons for Puerto Rico. In these map-a-thons events, experienced OSM users recruit newbies and teach them the basics to put them to work in outlining buildings and drawing roads over satellite imagery. I even encouraged my brother who lives in NYC to attend one of the events. An avalanche of data ensued and its importance cannot be underestimated. When someone’s home is outlined on a map, it means something. It means that it exists, it is there and that a someone may live there.

I visited the Red Cross Operations Center in San Juan and witnessed something truly amazing. A flurry of activity with volunteers from around the world gathering to help our island. I had chance to talk with a Red Cross worker working the maps data and the folks responsible for prioritizing the relief efforts. In such a situation, it only beckons a human being to ask “How can I help?”, which I did. I was told the Red Cross was printing large format maps elsewhere and they needed that capability there. I quickly contacted some friends which where able to get them a large format printer to the ops center. As best as I could contacted the local OSM volunteers and GIS professionals, which could aide the Red Cross.

Red Cross volunteers in San Juan, PR hold up a map of Puerto Rico
Red Cross volunteers in San Juan, PR hold up a map of Puerto Rico

The Red Cross is providing disaster relief right now, as well as employing highly capable geographical information systems (GIS) tools with volunteers on the field to catalog which roads are accessible or obstructed. That obstruction could be a landslide, fallen trees or a collapsed bridge. The information is shared with FEMA and local agencies which in instances in coordination with the military determine the course of action for rendering aid.

VIDEO: Rutgers ‘map-a-thon’ aids relief efforts in Puerto Rico

Disaster relief volunteers from various organizations are being sent to every corner of the island, including hospitals, shelters, and elderly homes. Volunteers coming from afar, not familiar with Puerto Rico may need map data to navigate a post hurricane disaster zone where street signage may have been blown by the fierce winds.

So one can imagine a repeating story playing out in Puerto Rico. That of an elderly person living in the mountainous region of the island, who may be alone, without power, enduring sweltering heat, with no clean water and in need. Yet thanks to a someone in another part of the world who drew a simple outline of his house, a relief worker knows that there is a home there, how to get there and a knock on the door can happen.


You can help Puerto Rico and the Virgin Islands in various ways. Participating in a Mapathon, or donating to the Red Cross are some.

If you have a computer, an internet connection and some time you can assist in mapping Puerto Rico.